To Boldly Go Where Few Powering the Transition Virtual Tours for Ammonia

ChemEs Have Gone Before p. 14 to Net Zero p. 42 Plant Safety p. 50

AUGUST 2024
AUGUST 2024

Chemical
Engineering
Progress

An AIChE Publication aiche.org/cep

SPECIAL SECTION:

AI and Digitalization
SPECIAL SECTION: AI AND DIGITALIZATION

Discover the expanding applications

of this rapidly evolving technology
VOL.120/NO.8

202408_Outside_Front_Cover_FNL.indd OF_Cover1 7/22/24 4:39 PM

202408_Inside_Front_Cover2.indd IF_Cover2 7/23/24 10:23 AM
 Chemical
Engineering
Progress
An AIChE Publication aiche.org/cep

Features August 2024 • Vol. 120 No. 8

4 UPDATE 4

17 SPECIAL SECTION:
AI AND DIGITALIZATION
18 The Rise of Artificial Intelligence
19 A Chemical Engineer’s Introduction to ChatGPT
27 Implementing Artificial Intelligence in
Process Safety Studies 17
36 The Industrial Metaverse for the Process Industries

42 ENERGY
Powering the Transition to Net Zero with
Electric Cracking Technology
This article provides an update on a technology demonstration of an
electric cracking furnace and considers future integration options into
petrochemical sites.
50

50 SAFETY
Employing Virtual Tours to Speed Up Plant Turnarounds
Drones and 360-degree cameras have made it relatively easy to
create virtual plant tours. For one ammonia production site in
particular, virtual tours ensured a safe and cost-effective turnaround.

Departments

3 EDITORIAL 12 NEW PRODUCTS Annual Meeting and Co-located

AI and Machine Learning: Conferences
14 PROFILE
Proceed with Caution • Robert A. Brown Named as AIChE’s
To Boldly Go Where Few ChemEs
Schowalter Lecturer for 2024
9 CATALYZING COMMERCIALIZATION Have Gone Before
• Center for Hydrogen Safety Opens
CdTe-Perovskite Tandem Photovoltaics
16 CAREER CONNECTION Nominations for Zalosh Safety Awards
Achieve Record Efficiency
How to Determine and Advocate for • Virtual Career Fair, Sept. 10
10 AIChE JOURNAL HIGHLIGHT the Compensation You Deserve
62 CEP MARKETPLACE
Enabling Flexible Biomass Supply
59 INSTITUTE NEWS
Chains with Mobile Modules 64 CALENDARS
• Alper to Deliver 2024 Acrivos
11 PROCESS SAFETY BEACON Professional Progress Award Lecture
Corrosion — Another Hidden Threat • Registration Opens for 2024 AIChE

(Publication Number 101-920) (Print ISSN 0360-7275, Online ISSN 1945-0710) is published monthly by the American Institute of Chemical Engineers (AIChE), 120 Wall St., 23rd Floor,
New York, NY 10005. All correspondence should be sent to the Editor-in-Chief at the address above, or by email to emilyp@aiche.org. The statements and opinions in this magazine
reflect the views of the contributors and not of AIChE, which assumes no responsibility for them. Subscription rates for nonmembers: North America $254/yr; International $456/yr (air
service included). Back issues are available from AIChE Customer Service (1-800-AIChemE). Individual copies: AIChE members $25 (plus postage outside North America), Nonmembers
$35 (plus postage outside North America). Claims for missing issues must be filed with AIChE Customer Service (customerservice@aiche.org) within three months by North American
subscribers, six months by international subscribers. Claims made after that may be fulfilled as supplies permit. Periodicals postage paid at New York, NY, and additional mailing offices.
Copyright 2024 by AIChE. Postmaster: Please send changes of address to AIChE, 120 Wall Street, 23rd Floor, New York, NY 10005. Return Undeliverable Canadian Addresses to: Station
A, P.O. Box 54, Windsor, ON N9A 6J5 Copying restriction and permissions: AIChE authorizes the photocopying of individual articles from CEP for the personal use of nonmembers
by libraries and other users registered with the Copyright Clearance Center (CCC) Transactional Reporting Service, provided that a fee of $30 per article is paid directly to CCC, 222
Rosewood Dr., Danvers, MA 01923. Fee code: 0360-7275/00 $30. This consent does not extend to other kinds of copying, such as that for purposes of general distribution, for advertising or
promotion, for creating new collective works, or for resale. To request permission, go to www.aiche.org/permissions. To purchase an electronic reprint of an article published after 2001,
contact Karen Simpson at kares@aiche.org. Each issue of CEP is indexed regularly by Engineering Village and Applied Science & Technology Index.

202408_TOC.indd 1 7/22/24 2:24 PM

202408_Virtual Careert Fair.indd 2 7/25/24 11:53 AM
 EDI TO RIAL

Chemical
Engineering
Progress
AI and Machine Learning:
Proceed with Caution
120 WALL STREET
ne of my favorite movies growing up was Terminator 2: Judgment Day.
O
NEW YORK, NY
10005
aiche.org For those unfamiliar with the Terminator movies, they center on human-
kind’s resistance to the rise of a malevolent artificial intelligence (AI) entity
EDITOR-IN-CHIEF AIChE
Emily Petruzzelli CUSTOMER SERVICE known as Skynet. Some of the largest pushback against the burgeoning use of
emilyp@aiche.org 1-800-AIChemE AI is prompted by such media portrayals. But the reality is that AI has a long
ASSISTANT
(1-800-242-4363) way to go before it can take over the world.
EDITORS EXECUTIVE Compared to computers, humans excel at ignoring obviously misleading
Gordon Ellis DIRECTOR information when learning about new topics. For chatbots based on large-
gorde@aiche.org Darlene Schuster
darls@aiche.org
language models (LLMs), such distinctions are difficult. LLMs are only as reli-
Melanie Mesropian
melam@aiche.org
able as the data they are fed. For example, in a popular Forbes article published
EDITORIAL in May, the author collated several examples of Google’s AI providing hilariously
Colby Kaimanu Tinsley ADVISORY BOARD
colbt@aiche.org Saadet U. Acikgoz misleading results to simple queries. For one user who wanted to stop the cheese
Honeywell UOP from sliding off of his pizza, Google’s AI Overview suggested adding “about
COLUMNISTS
Jason Balich Manish Aggarwal 1/8 cup of non-toxic glue to the sauce to give it more tackiness,” drawing from a
jason.balich@ Aspen Technology, Inc. Reddit troll post from over a decade ago.
wolfgreenfield.com Joseph S. Alford* Google is not alone when it comes to their AI product generating absurd
Darius Mortazavi Automation
Consulting Services responses. I was playing around with OpenAI’s GPT-4o — the current default
darius@thecolumn.co
Heriberto Cabezas*
ChatGPT model for free users — and found that it can struggle with basic
Kate Williamson
kate@scientechresumes.com Széchenyi István Univ. comprehension. For example, I asked it “Which is bigger? 4.9 or 4.11,” and it
Vishwamitra Research answered: “When comparing version-like numbers, 4.11 is bigger than 4.9.” I’m
PRODUCTION
MANAGER
Institute no math whiz, but even I know that 4.9 is greater than 4.11. However, by the way
Karen Simpson T. Bond Calloway, Jr.* that the program responded, we can tell that it interpreted a simple math problem
kares@aiche.org Univ. of South Carolina as a comparison of version-like numbers — for example, different versions of
ART DIRECTOR
Laura Dietsche, P.E.* the same software. If I asked a human this (admittedly, poorly worded) question,
Dow
Debbie Slott they would likely answer 4.9 or request clarification on the context.
Moiz Diwan
DIRECTOR OF AbbVie, Inc. I asked Meta AI (another chatbot) a similar question and received a correct
PUBLICATIONS Ignacio E. Grossmann*
answer of “4.9 is larger than 4.11.” Because the model didn’t initially give me
AND BUSINESS Carnegie Mellon Univ. any of the rationale behind the answer, I asked a follow-up question of “Why?”
DEVELOPMENT
Michelle Bryner Loraine A. Huchler, P.E. for more insight. The model replied, “In this case, 4.9 is larger than 4.11 by
michelleb@aiche.org MarTech Systems 0.78.” In reality, we know that 4.9 is larger than 4.11 by 0.79.
Sanjeev Kapur While it is quite easy to break out a calculator and verify the math in this
Apex PetroConsultants, LLC case, I wonder if simple errors like this can bleed into other areas of the LLMs,
John O’Connell* which can color responses and lead to further inaccuracies. For instance, I asked
Univ. of Virginia
ChatGPT, “What has a greater atomic mass: Cerium or Praseodymium?” The
California Polytechnic,
San Luis Obispo answer it gave, Cerium, was incorrect — although it did tell me the correct
Venkat Pattabathula atomic masses for both elements (140.12 amu for Ce and 140.91 amu for Pr).
SVP Chemical Plant Services For now, relying on LLMs to do your job will almost certainly lead to prob-
Gavin P. Towler, CEng* lems. Still, these tools can be extremely helpful for specific tasks, like finding
Honeywell Performance related references for a niche topic of interest or summarizing action items from
Materials and Technologies
a meeting transcript. Our Special Section on AI and Digitalization this month
Bruce Vaughen, P.E.*
CCPS
examines some of the more practical aspects of using AI. Articles review how
* AIChE Fellow
and where chemical engineers can use LLMs effectively (pp. 19–25), how AI
is being used for HAZOP revalidation studies (pp. 27–34), and how AI-driven
digital twins represent a first foray into the industrial metaverse (pp. 36–41).
Although the AI systems of today are not yet ready to lead a robot uprising
against humankind, their use in engineering processes and our day-to-day lives is
very much inevitable in the coming decade.
Emily Petruzzelli, Editor-in-Chief

August 2024 aiche.org/cep 3

202408_Editorial Draft 1.indd 3 7/23/24 1:16 PM

 U PDAT E

New Method of Gene Editing Allows Complex Changes in a Single Step

esearchers have discovered a gene- tion can insert genetic sequences as binase that is guided by a non-coding
R editing technique that allows for
cutting, pasting, and swapping around
well as excise them or invert them; it
can also allow for the manipulation of
RNA, which the researchers have
dubbed bRNA, for “bridgeRNA.” The
genetic sequences in a single step. longer genetic sequences than other bRNA consists of two loops of nucleo-
The mechanism, which they have methods, up to hundreds of thousands tides. One is a target-binding loop that
dubbed “bridge recombination,” takes of base pairs. recognizes the target site in the DNA.
advantage of a naturally occurring CRISPR-Cas9 editing requires the The other is a donor-binding loop that
genetic phenomenon — transposable recruitment of the cell’s own DNA- recognizes the DNA sequence to be
elements — which are gene sequences repair machinery to fix the cuts made moved into that target site. These two
that can jump from one portion of the by the gene editor, according to Hsu. loops can be independently pro-
gene to another. (They’re sometimes This introduces opportunities for error grammed to direct the bridge recombi-
called jumping genes.) Researchers and off-target effects. Bridge recombi- nase to any sequence of DNA that the
led by Patrick Hsu, a bioengineer nation takes place in one step, making researchers desire, and then facilitate
at the Arc Institute and the Univ. of mistakes less likely. the plugging in, cutting out, or swap-
California, Berkeley, and Hiroshi “It takes us beyond the DNA and ping around of customizable stretches
Nishimasu, a structural bioscientist at RNA cutting capabilities of CRISPR of genetic material.
the Univ. of Tokyo, determined how and to a broader suite of capabilities Nishimasu led a team to understand
these transposable elements relocate for DNA and RNA design,” Hsu says. the structure of the recombinase and
themselves and learned that the pro- However, the technology has so bRNA using cryo-electron microscopy,
cess can be adapted and programmed far only been tested in Escherichia which uses cryogenic temperatures to
for targeted gene editing. coli, which is a prokaryotic bacterium, freeze tiny samples in place for inves-
The method could have advan- meaning it lacks the membrane-bound tigation. They found that the bridge
tages over other gene-editing methods organelles found in animal cells recombination system is unique among
that have been developed, such as (including human cells). Using bridge all other RNA-guided systems due to
RNA interference and CRISPR-Cas9. recombination for human gene editing its uncommon two-loop structure.
RNA interference is a method that “is going to take time and develop- The recombination reaction first
allows genes to be silenced, whereas ment, and that’s something we’re cleaves to the top of the two DNA
CRISPR-Cas9 uses bacterial mole- actively exploring,” Hsu says. strands in the genome and allows
cules to slice a genome and insert new The bridge recombination tool is for the exchange of genetic mate-
genetic sequences. Bridge recombina- made up of an enzyme called a recom- rial. This process creates a temporary

▶ The new bridge recombination method of gene editing

creates a temporary four-armed T shape called a Holliday
junction during the exchange of genetic material. Image
courtesy of Visual Science.

4 aiche.org/cep August 2024

202408_Update Draft 2.indd 4 7/22/24 4:35 PM

 four-armed T shape called a Holliday move the process from bacteria into DNA to be mixed and matched, the
junction, which is then smoothed back more complex organisms, researchers researchers hope the technique could
into two continuous lines of DNA are optimistic about the possibilities. be used not just to snip individual
by the cleavage of the bottom strand In bacteria, Hsu believes, the technique genes, but to do chromosome-scale
of DNA and the exchange of genetic could be used to accelerate metabolic genome engineering.
information in that strand. Unlike in engineering, inserting new enzyme “Instead of today, where we’re
CRISPR-Cas9, which slices the two pathways so that bacteria can produce saying for gene and cell therapy, ‘We
DNA strands in the genome in one valuable commercial compounds. In have a broken gene, let’s just put in
snip, there is never a double-strand humans, the technique could eventu- a gene,’ we want to be able to design
DNA break that requires repair. ally be used in gene therapies — for entire sections of the genome,” Hsu
“This is an incredibly complicated example, to insert missing genes such says, comparing the process to mov-
mechanism that would normally need as the dystrophin gene in muscular dys- ing from assembly-level computer
multiple different steps of enzymatic trophy patients, or to cut out damaging programming to Python. “Instead of
engagement and then host DNA genetic repeats that lead to Hunting- single bases at a time, or individual
repair,” Hsu says. ton’s disease, a brain disease that pairs of bases, you could change
The researchers released a tool on causes uncontrollable muscle move- 100,000 bases or more.”
the Arc Institute website for bridge- ment. CRISPR-Cas9 is currently being Durrant, M. G., et al., “Bridge RNAs Direct
RNA design, using computational investigated for treating these kinds of Programmable Recombination of Target and
Donor DNA,” Nature, doi: 10.1038/s41586-024-
methods to determine which sequences genetic diseases, but bridge recombina- 07552-4 (June 26, 2024).
are likely to lead to high recombina- tion may ultimately be safer, Hsu says. Hiraizumi, M., et al., “Structural Mechanism of
tion efficiency. Finally, because bridge recombina- Bridge RNA-Guided Recombination,” Nature,
doi: 10.1038/s41586-024-07570-2 (June 26,
While there is still work to do to tion allows for enormous chunks of 2024).

SEPTEMBER 24, 2024 | HOUSTON, TX

Facility Siting and the
New EPA RMP Rule
Don’t miss this half day in-person
workshop. Attendees will leave with
practical strategies for implementing
facility siting compliance under the new
EPA RMP Rule with BONUS resource
tool ‘take-aways’. Scan Me

Dr. Georges Melhem Devon Downs

ioMosaic Corp Farley & Partners LLP

1-844-ioMosaic | www.ioMosaic.com

August 2024 aiche.org/cep 5

202408_Update Draft 2.indd 5 7/22/24 4:35 PM

 U PDAT E

Biochar Breakthrough Could Yield Cheap Hydrogen Production

new method of using biomass “Based on a preliminary techno- anode and the reaction won’t happen;
A and solar energy has broken
a barrier in the amount of voltage
economic analysis, our cost comes in
cents, not dollars, for a kilogram of
if they are too negatively charged, they
will cake the electrode and the reaction
needed to produce green hydrogen hydrogen,” Singh says. will rapidly dwindle to nothing. Singh
via electrolysis. Basic water electrolysis requires and his colleagues tried tackling this
For the first time, researchers have 1.23 V of electricity at minimum, problem with agitation to no avail.
developed a process that operates at a Singh says. In an electrolyzer, H2 pro- Agitation created too smooth a flow to
total cell potential of less than a volt; duction occurs at the cathode, while clear the electrode surface, Singh adds.
in fact, the new method requires a water splitting (i.e., oxygen evolution) Then they tried using a sparger to
minimum of about a fifth of a volt of occurs at the anode. The anodic water bubble nitrogen or argon through the
electricity to split water, enabling sub- splitting or oxygen evolution reaction solution. “It created a frothy, bubbling
stantial H2 production at only 0.5 V. is sluggish and requires high over- mixture,” Singh says, which was key
That translates to only about 10 kWh potentials. Adding carbon as a reactant to the reaction. “It was pushing away
of electricity per kilogram of hydro- at the anode vastly reduces that energy all the particles and flushing it out of
gen produced, which is a sixth of the requirement. This is because the car- the electrode surface, and then we
electricity requirements for the current bon binds favorably with the oxygen started getting current.”
state-of-the-art, claims study leader from the split water to make carbon The team used platinum-titanium
Meenesh Singh, an associate professor dioxide. Theoretically, this process mesh as an anode for the oxidation
of chemical engineering at the Univ. of should take only about 0.21 V, but in reaction and copper for the cathode. In
Illinois, Chicago. practice, Singh says, no one had gotten addition, the team raised the tem-
Because electricity costs are the below 1 V, no matter what source of perature of the reaction from room
main driver of the cost of hydrogen carbon they tried. temperature to 80°C and got to work
production, the new method could The problem, Singh says, has optimizing their carbon source. They
reduce the cost of green hydrogen from been in achieving the ideal amount of turned to biochar, partially combusted
the current $6 to $7 per kg to less than contact between carbon particles and organic waste, as a sustainable source
the U.S. Dept. of Energy’s 10-year goal the anode. If the carbon particles are of carbon. The researchers tested cow
— set in 2021 — of $1 per kg. positively charged, they will repel the manure, corn husk, hemp waste, paper
waste, and sugarcane bagasse biochar,
and activated the carbon by mixing the
biomass with a sulfuric acid solvent.
Cow manure was the most active
carbon source, followed by paper
waste. The activity is likely the result
of the fine particle size in the biochar
slurry made from these materials. The
other materials made for less-smooth
slurries, Singh claims, though they
might perform as well as cow manure
if processed by grinding or ball milling
during the production step.
Singh used an off-the-shelf silicon
solar cell as a green energy source for
the biochar-assisted electrolysis and
found that the overall efficiency of
the system was 35%. Singh envisions
using the carbon dioxide generated in
▲ Biochar — i.e., partially combusted organic waste — could be the key to sustainable hydrogen production.
In a unique study, researchers combined biochar originating from cow manure with solar-energy-driven water
the process to make industrially valu-
electrolysis to produce hydrogen at unprecedented efficiencies. Image courtesy of Meenesh Singh, Univ. of able chemicals.
Illinois, Chicago. But scaling up the method will be

6 aiche.org/cep August 2024

202408_Update Draft 2.indd 6 7/22/24 4:35 PM

 challenging. “This is a very unconven-
tional electrochemistry,” Singh says. Scientists Investigate
As the carbon source is a slurry, new Wastewater-Based Fertilizers
electrolyzers will need to be developed
that can handle slurry transport, not new nationwide analysis of bio- and toxicity,” Prasse says. The new
just liquid or gas. The researchers’
industry partner, Orochem Technolo-
A solids — the byproducts of waste-
water treatment often used as fertilizer
study is a first step toward evaluat-
ing risk, he says. While other studies
gies Inc., is currently working on the — finds a plethora of unregulated of biosolid samples have turned up
development of such an electrolyzer. organic compounds, including some concerning organic chemicals in the
On the chemistry side, the carbon parti- that may have potential impacts on past, there had not been a large-scale
cles within the biochar tend to become human health. study seeking to catalogue the many
encapsulated in a layer of ash during It’s not clear yet whether there is possible organics that might be found
the reaction process, Singh says, which any negative impact of using biosolids, in these products.
can stall the reaction over time. “We cautions study author Carsten Prasse, “The idea behind this is to come up
are working to improve the stability of an assistant professor of environmental with a priority list of compounds that
the reaction to 100% consumption,” he health and engineering at Johns Hop- might be — and that ‘might’ is very
says. “That is a current hurdle.” kins Univ. For one thing, researchers important — that might be problematic
Kani, N., et al., “Sub-Volt Conversion of don’t yet know the quantities of these and have to be assessed in detail in the
Activated Biochar and Water for H2 Production organics or if they enter the environ- future,” Prasse says.
Near Equilibrium Via Biochar-Assisted Water
Electrolysis,” Cell Reports Physical Science, doi:
ment in any significant way. Biosolids are treated sewage
10.1016/j.xcrp.2024.102013 (June 19, 2024). “Risk is a combination of exposure sludge, largely consisting of the

SHARPEN YOUR
PROFESSIONAL
SKILLS
INSTRUCTOR-LED COURSES ELEARNING COURSES
Process Safety Boot Camp | VIRTUAL Topic areas include:
September 9-13
• Process Safety
HAZOP Studies/Advanced PHA Concepts • Hydrogen Safety
Combo Course | VIRTUAL
September 23-27 • Process Intensification

Process Safety Boot Camp | HOUSTON, TX • Chemical Engineering Practice

October 21-24 • Chemicals & Materials and more!
Overview of Risk Based
Process Safety | VIRTUAL GROUP TRAINING
October 30-31 Contact sales@aiche.org for a quote.

Learn more at aiche.org/academy Powered by

© 2024 AIChE 9548_24 07.24

August 2024 aiche.org/cep 7

202408_Update Draft 2.indd 7 7/22/24 4:35 PM

 U PDAT E

biomass of bacteria that are used to of chemical features in samples. They impacts, depending on the level
degrade waste. About 3.76 million dry then compared the chemical profiles of exposure. They also found the com-
m.t. were produced as of 2022, accord- detected with an EPA database of com- mon chemical intermediate p-cresol,
ing to the U.S. Environmental Protec- pounds and their hazards to determine which can be dangerous if inhaled,
tion Agency (EPA). Of that, 56% was which organics might have impacts on ingested, or absorbed through the skin.
recycled as fertilizer (the rest was human health and the environment. However, the researchers don’t
incinerated or landfilled). About half The researchers tested 16 samples yet know how high a concentration
of biosolid fertilizer is used on agri- from 13 different treatment plants of these compounds is present in the
cultural land, while the rest ends up in in nine different U.S. cities as well biosolids. If the totals are negligible,
places like parks and golf courses. as three Canadian cities. They only the presence of these organics may not
The EPA regulates heavy met- further analyzed compounds that have any health or ecological impacts.
als in biosolids as well as pathogens. occurred in 80% of the samples, with At the same time, the study did not
Class A biosolids must be pathogen- the aim of focusing on organics with highlight every potentially danger-
free, while Class B biosolids have widespread distribution. ous chemical. PFAS, for example, are
reduced pathogens and must be There were 92 organic compounds known to persist nearly indefinitely in
used only in locations where there is present in 80% of the samples. For 58 the environment and have been linked
limited risk of human contact, such of those, the study was the first report to immune system damage, low birth
as mining land reclamation. There is of these organics in U.S. or Canadian weight, and other health problems.
no regulation of organics in biosolids, biosolids. It’s too early to rank these PFAS were not flagged in the new
however, despite rising concerns chemicals by risk, but of those in research, because they appeared in
about the presence of harmful com- the EPA database, there was ample only 70% to 75% of samples, Prasse
pounds. In April 2022, for example, evidence of toxicity to aquatic animals adds, not the 80% which marked the
Maine banned all land application of and to mammals, as well as of accumu- team’s cutoff.
biosolids because of high levels of lation and persistence in the environ- The next step, Prasse suggests, is
the forever chemicals PFAS (per- and ment. Common types of chemicals to evaluate the potential dangers of
polyfluoroalkyl substances) found in included pharmaceuticals, fragrances, these organics in real-world scenarios.
soil, water, and agricultural products and preservatives. He and his team are already launching
of farms that used biosolid fertilizers. For example, the researchers investigations into whether these com-
Prasse and his colleagues cast a found evidence of the antifungal pounds are taken up into plants grown
wide net, using a method called non- ketoconazole, which can have health in biosolid-amended soil. They’re also
targeted analysis to detect thousands effects on humans as well as ecological working to quantify the concentrations
of these organics in biosolid samples.
Using biosolids as fertilizer is
cheaper than incineration or landfill-
ing, the latter of which can still cause
contamination issues from leaching,
Prasse adds. The research isn’t an
argument for dropping this source of
fertilizer, he says, but rather the start
of a discussion about whether more
safety measures are needed.
“We are not saying we shouldn’t
do this,” he states. “We are just saying
there might be some things we have
to look out for. If there are certain
compounds present, how can we deal
with them?”
Newmeyer, M. N., et al., “Combining Non-
targeted Analysis with Computer-Based Hazard
Comparison Approaches to Support Prioritization
▲ Biosolid-based fertilizers originate from treated sewage sludge from wastewater treatment plants like this of Unregulated Organic Contaminants in Bio-
one. New analysis of these fertilizers has detected the presence of unregulated organic compounds that may solids,” Environmental Science & Technology,
have ramifications for human health. doi: 10.1021/acs.est.4c02934 (June 25, 2024).

8 aiche.org/cep August 2024

202408_Update Draft 2.indd 8 7/22/24 4:35 PM

 CATALY Z I NG C OM M ERCIALI Z ATION

CdTe-Perovskite Tandem Photovoltaics Achieve Record Efficiency

chieving energy sustainability improve single-junction solar cell effi- of CdTe-perovskite tandem cells for
A necessitates efficient, cost-
effective renewable energy solutions
ciency is ongoing, another approach is
to use tandem solar cells.
commercial production.
The demonstrated efficiency of
such as photovoltaics, which convert Combining CdTe and metal-halide the 4-T CdTe-perovskite tandem cells
solar energy to electricity. A growing perovskites is especially promising for provides an alternative to silicon-based
share of commercial solar cells are achieving high conversion efficiency perovskite tandem cells and opens
made with polycrystalline cadmium at lower costs. CdTe is known for its new avenues for more affordable and
telluride (CdTe) thin-film technology. near-optimal bandgap, high absorption efficient solar energy solutions. Future
CdTe solar cells of at least 1 cm2 have coefficient, and low-cost manufactur- improvements will include fabricating
demonstrated up to 21.1% conversion ing compared to silicon photovoltaic films on both sides of the glass sub-
efficiency; however, improving their technology. Perovskites offer excel- strate, eliminating the need for the liq-
efficiency further has been difficult. lent light absorption, low-temperature uid, and simplifying the manufacturing
In photovoltaic devices, the junc- thin-film fabrication, high power process. A joint invention disclosure
tions and their bandgaps determine conversion efficiency, and flexibility with CSU and IITB is in progress,
the range of light wavelengths that of bandgap tuning, which is essential with companies having expressed
can be absorbed and converted into for tandem cell integration. Despite interest in its commercialization.
electricity. Single-junction cells absorb tandem solar cells being an estab- This work builds on earlier col-
a specific range of wavelengths limited lished concept, limited research has laborations between CSU and Univ.
by their single bandgap, while tandem explored CdTe. of Texas at Austin (UTA) researchers
cells utilize multiple bandgaps to Researchers at Colorado State with the Center for a Solar Powered
absorb a broader spectrum of light Univ. (CSU), funded partly by the U.S. Future (SPF2050), an NSF Industry-
more efficiently. While research to National Science Foundation (NSF), University Cooperative Research Cen-
and the Indian Institute of Technology ter, which in 2019 demonstrated the
Bombay (IITB) have demonstrated a potential of combining wide bandgap
Anti-Reflection Coating four-terminal (4-T) CdTe-perovskite perovskite materials with CdTe in a
Soda-Lime Glass tandem solar cell with an unprece- tandem architecture. This foundational
Indium-Doped Tin Oxide dented power conversion efficiency of research highlights the crucial role of
PZ: PFN-Br
24.2%. Previously, two-terminal (2-T) academic collaboration in advancing
Perovskite Absorber
PCBM/BCP
CdTe-perovskite tandem cells had tandem solar cell technology.
Intrinsic Tin Oxide
demonstrated less than 10% efficiency. “The demonstration of a successful
Indium-Doped Tin Oxide Previous efforts on 2-T tandems CdTe-perovskite tandem photovoltaic
Anti-Reflection Coating faced optimization and reliability device is an exceptional example of the
challenges. Using a 4-T instead of a potential outcomes that can be achieved
Soda-Lime Glass
2-T device configuration avoids the from cooperative research centers such
Fluorine-Doped Tin Oxide

MgxZn1-xO Buffer
need to match currents through each as SPF2050. The synergistic research
absorber layer and reduces fabrication being pursued by CSU along with UTA
CdSexTe1-x/CdTe
Graded Bandgap
complexities. and IITB is poised to have a strong
Absorber In this pioneering work by the impact on future academic as well
Tellurium CSU-IITB team, 18.3% efficient near- as commercial developments long-
Graphite/Ni Electrode infrared-transparent perovskite solar term for all forms of perovskite-based
cells were stacked on 19.56% efficient research,” says Ray Lewandowski, co-
▲ Researchers at Colorado State Univ. in collabora- CdTe cells to form a 4-T tandem chair of the Industrial Advisory Board
tion with Indian Institute of Technology Bombay
demonstrated 24.2% efficiency for a four-terminal structure. Refractive optical losses for SPF2050.
two-junction CdTe-perovskite tandem solar cell. between the tandem cells were largely
Perovskite and CdSexTe1-x/CdTe layers absorb the mitigated using an index-matching This research was funded in part by the
light to generate electrical charge while other layers U.S. NSF Industry-University Cooperative
facilitate extraction of electricity and the glass
liquid to achieve a tandem efficiency Research Center (IUCRC) for a Solar Powered
substrate provides structural resilience. of 24.2%, demonstrating the potential Future (SPF2050).

This article was prepared by the U.S. National Science Foundation in partnership with CEP.

August 2024 aiche.org/cep 9

202408_NSF Draft 1.indd 9 7/22/24 4:17 PM

 A I C hE J O U R N A L H I G H LI G H T JOUR NAL
An Official Publication of the American Institute of Chemical Engineers
chemical engineering research and development August 2024

Enabling Flexible Biomass Supply VI

EW

Chains with Mobile Modules

RE
ER e
PE ich
View this journal online at wileyonlinelibrary.com D /a
AN om
N al.c
IO ntr
SS tce
MI rip
UB usc
E S an
L IN mc.m
ON

iomass feedstocks are emerging sources of raw materials To address this challenge, the authors first adopted
B for chemical production with potentially lower green-
house gas (GHG) emissions than traditional feedstocks.
the rolling horizon planning framework to simplify the
multi-period decision-making with a two-stage stochastic
Deploying low-carbon chemical technologies that use programming (TSSP) formulation. The rolling horizon
renewable feedstocks offers promising benefits. However, approach reduces the time dimension by focusing on only
successful implementation requires more considerations a fraction of the entire planning period at a time (e.g., three
beyond the facility boundary, including supply and transport months as opposed to one year) but iteratively re-solving the
limitations. With the ongoing effort to decarbonize chemical planning with updated information. Next, generalized Bend-
production and fierce competition, it is essential for chemi- ers decomposition (GBD) was implemented to divide the
cal companies to account for high uncertainties and their overall large-scale MINLP model into two types of problems
impacts on the biomass supply chain. that are easier to solve sequentially:
A notable feature of biomass feedstocks is their temporal • First, a mixed-integer linear programming (MILP) mas-
and spatial variability since each crop only grows in certain ter problem was formulated to decide module installation,
areas and is harvested during a specific season. Unlike most movements, and restoration after possible disruptions.
current petrochemical plants that run on a steady supply of • Second, several nonlinear programming (NLP) sub-
raw materials with relatively consistent properties, bio- problems were defined for the site operations, shipment,
refineries have to address unstable feedstock quantities and and backorder in each uncertain scenario. Once the module-
compositions as a result of the variable temporal-spatial related integer variables were fixed, all subproblems were
patterns in supply. Moreover, ever-changing demands and decoupled and solved efficiently.
unforeseen disruptions add to the complexity of supply chain The proposed solution strategy effectively reduced the
planning and product management decision-making. computational time of a large-scale supply chain planning
To combat these issues, modular manufacturing has been problem that included 220 counties as supply regions within
established as an effective approach to improving the flex- Missouri and Illinois with a 12-month planning period.
ibility of chemical production, especially in the presence of In addition to the optimal supply chain design and opera-
frequent supply or demand shifts. Placing modular produc- tion, the proposed rolling horizon model also keeps track of
tion units close to feedstocks reduces the need for transporta- the lifecycle assessment (LCA) uncertainties of biomass-
tion between supply, facility, and market, and it decreases based chemical production. Understanding fluctuations and
associated GHG emissions and operational costs. obtaining reliable uncertainty data on material or energy
In the AIChE Journal article, “Design and Operation consumption in chemical plants poses a significant chal-
of Modular Biorefinery Supply Chain Under Uncertainty lenge due to the high costs associated with measurement.
Using Generalized Benders Decomposition,” Yuqing Luo Consequently, LCA practitioners typically have to resort to
and Marianthi Ierapetritou (Univ. of Delaware) formulated qualitative expert judgment (e.g., the Pedigree method) as a
a two-stage stochastic programming model for a biorefinery placeholder for input distribution parameters. These qualita-
supply chain with movable production units. This supply tive analyses inevitably fail to reflect the actual magnitude of
chain planning optimization model captured the module LCA result variabilities and lack the ability to guide com-
placement and material flows within the Missouri-Illinois parisons or decisions.
region, where an abundant supply of corn stover, willow, and The authors utilized the environmental performance of
poplar is available in different seasons. the biomass supply chain in each scenario of the stochastic
In the modular supply chain optimization model, a large planning model to quantify variations in LCA results. Unlike
number of decisions increases the model size significantly, most traditional LCA uncertainty evaluation methods, well-
especially the discrete decision variables associated with unit characterized probability distributions of supply, demand, and
installation or movement at each production site. Further- disruptions were propagated to the GHG emission outputs in
more, nonlinear relationships exist in the cost functions and the supply chain optimization model. This quantitative under-
plant operation, adding to the model complexity. Overall, the standing of LCA uncertainties results in accurate downstream
resulting mixed integer nonlinear programming (MINLP) carbon accounting and enables robust decision-making.
model is challenging to solve by standard solvers due to In summary, the supply chain optimization model
the vast number of locations, time periods, and uncertain provides insights into improved economic performance and
scenario combinations involved, particularly in analyzing the LCA uncertainties, both of which are vital to promoting
Missouri-Illinois region across different seasons. sustainability in the emerging biomass economy.

10 aiche.org/cep August 2024

202408_Journal Highlight Draft 1.indd 10 7/22/24 4:16 PM

 This Issue Sponsored by

Messages for Manufacturing Personnel

http://www.aiche.org/CCPS/Publications/Beacon/index.aspx www.iomosaic.com

Corrosion — Another Hidden Threat August 2024

(a) Did You Know?

• Corrosion can occur inside and outside of process equip-
ment and on support structures.
• Corrosion is a reaction between a material, usually metal,
and its environment. The most familiar is the corrosion of iron
or steel, which forms iron oxides, or rust.
• There are many mechanisms for corrosion. Figure 1 shows
only two.
• Most corrosion mechanisms are slow and take years to
cause equipment failure. However, under some conditions, cor-
rosion can be surprisingly fast.
(b) • Corrosion rates are typically stated as milli-inches per
year or micrometers (µm) per year (1 milli-inch = 25.4 µm).
When reviewing corrosion data, it is important to know which
units were used to measure the corrosion rate.
• Concrete can be corroded by acidic materials. This can
degrade the effectiveness of containment systems for tanks,
piping, and loading/unloading operations.
• Not all corrosion involves metals. Gaskets, O-rings, and
other non-metal parts can fail from material attack.

What Can You Do?

• When making rounds, watch for signs of corrosion, such
as discolored insulation, damaged concrete, or stains on equip-
▲ Figure 1. Corrosion has been the cause of many past process safety incidents. ment, piping, or structures.
Image (a) adapted from U.S. Chemical Safety and Hazard Investigation Board (CSB)
Report No. 2012-03-I-CA. • Watch for places where insulation has been damaged and
water can saturate the insulation or fireproofing.
• Material dripping from insulated lines may indicate that
I n Figure 1a, an older section of piping had corroded, but a decision was
made to continue operating until the next inspection. The piping failed,
releasing hot, flammable liquid that formed a vapor cloud. It ignited and
the insulation has been damaged, but it could also be a leak.
Treat all drips with care and report them to your supervisor.
Do not try to identify the leak without proper personal protec-
caused a large fire. There were no fatalities. tive equipment (PPE).
In Figure 1b, a spherical tank in liquefied natural gas (LNG) service was • When opening piping and equipment, examine the
being hydro-tested. Water, which has a specific gravity of 1.0, was used as the gaskets and O-rings. If they show signs of attack, such as dis-
test material, but it is over twice as dense as LNG, which has a specific gravity coloration or cracks, point this out to your supervisor. This may
of 0.45. The legs of the sphere were fireproofed, and no one recognized that indicate that the gasket or O-ring material is not suitable for the
corrosion had occurred under the fireproofing. The additional load of the water present service.
caused the legs to fail. One person was injured, and another was fatally injured.

Corrosion — it’s inside, outside, and all around your plant.

©AIChE 2024. All rights reserved. Reproduction for non-commercial, educational purposes is encouraged. However, reproduction for any commercial purpose without express written consent of
AIChE is strictly prohibited. Contact us at ccps_beacon@aiche.org or 646-495-1371.

August 2024 aiche.org/cep 11

202408_Beacon [2].indd 11 7/22/24 1:22 PM

 N E W PR O D U C T S

BIOPROCESSING ing pressure, and the largest being the

Animal-Origin-Free TR080, which has a 7.5-kW motor and
Formulation Enables Faster a 15-bar pumping pressure.
T-Cell Therapy Manufacturing Lutz Pumps, Inc.
www.lutzpumps.com

Dry Claw Vacuum

Pump Features
Easy-to-Maintain Design

by compressed air, which improves

its reliability and safety. It requires
71% less compressed air than compa-
rable systems, reducing process costs.
Without the need for electric cords
The CLAWVAC CP B dry claw
or motors, this vacuum features an
The Gibco CTS OpTmizer One vacuum pump uses stainless steel
inherently safe design that can be used
Serum-Free Medium (SFM) is an claws and a coated and corrosion-
in hazardous or wet environments.
animal-origin-free (AOF) formulation resistant vacuum chamber to provide
Operating at less than 88 dBA, this
designed for clinical and commercial low-maintenance performance across
vacuum is quieter than most compa-
cell therapy manufacturing to deliver a wide range of hazardous or demand-
rable vacuums on the market.
increased scalability and performance ing applications. The design features a
Guardair Corp.
of T-cell expansion. The medium deliv- separated gearbox to increase cooling
www.guardair.com
ers high cell viability and improved and prevent oil from contaminating the
growth while maintaining the early processes. This pump is sufficiently
memory T-cell phenotype. Addition- Wobble Plate Pumps Transfer durable to operate in processes involv-
ally, the formulation helps mitigate Viscous Fluids ing particles, vapors, and contaminated
risk by reducing opportunities for gases. Flushing kits, liquid separators,
contamination or inconsistency inher- and inlet filters can additionally be
ent to animal-origin components. This affixed to the unit to further prevent
ready-to-use, one-part medium assists damage and contamination.
cell therapy manufacturers in improv- Leybold
ing overall drug function for both www.leybold.com
autologous and allogeneic therapies.
Thermo Fisher Scientific Inc. HEAT TRANSFER
www.thermofisher.com Compressor Is Suitable for Large
Industrial Heat Pumps
FLUIDS AND SOLIDS Wobble plate pumps use a wobble In industrial applications, heat pumps
HANDLING plate to actuate pistons, giving them provide a sustainable alternative to tra-
Vacuum Provides High the ability to move difficult-to-flow ditional gas boilers, as they eliminate
Performance with Inherently materials. The TR series of wobble on-site carbon emissions. The BOCK
Safe Design plate pumps is capable of pumping up HGX56 CO2 T is a six-cylinder semi-
The PowerQUAD pneumatic to 30,000 L of highly viscous liquids hermetic reciprocating compressor
vacuum’s novel design features a per hour, including oils, lubricants, for natural and low-global-warming-
dual-stage vacuum, multi-venturi- varnish, sealants, and silicones. The potential (GWP) refrigerants that is
generating engine that produces higher series includes three model sizes, with designed for large-capacity indus-
vacuum head and flow than alternative the smallest being the TR015, which trial heat pumps. This unit achieves
technologies. The vacuum is powered has a 3-kW motor and 10-bar pump- water temperatures up to 90°C. The

12 aiche.org/cep August 2024

202408_New products [1].indd 12 7/23/24 11:08 AM

 compressor’s large size reduces the Propane Dehydrogenation
number of equivalent smaller com- Catalyst Improves Selectivity
pressors required for large industrial
heat pumps, refrigeration, and district
heating systems.
Danfoss
www.danfoss.com

INSTRUMENTATION
Sensor Integrates
Adjustable Timer Check, have been added to the CATOFIN 312 is a propane dehydro-
manufacturer’s line of helium quality genation (PDH) catalyst that demon-
assurance products. Smart-Spray is strates enhanced performance com-
an addition to the manufacturer’s pared to previous catalyst offerings.
helium spray gun that adjusts spray The catalyst provides greater selectivity
volume and monitors helium flow via and a 20% longer life than comparable
a color display, removing the need for past catalysts, reducing operational
error-prone manual adjustments and costs. The performance of this catalyst
alerting personnel to leaks. I•Boost is can be analyzed via the manufacturer’s
a signal processing filter that accel- software, CLARITY, which provides
erates the leak signal, shortening performance data, advanced visualiza-
response times. Spray-Check is an tion, analysis, and monitoring tools.
easily installed test leak to help set up Clariant
and test the monitoring system. www.clariant.com
Inficon
The M30 capacitive sensor has www.inficon.com ENVIRONMENTAL,
a built-in adjustable timer from HEALTH, AND SAFETY
30 sec to 10 min, granting operators MATERIALS AND Single-Use Nitrile Gloves
more control over bulk storage tank CHEMICALS Degrade Quickly in Landfills
filling cycles. The housing is IP67- Sustainable Steel Reduces The ShieldON EcoSeries of nitrile
rated watertight and dust-resistant. The Carbon Footprint gloves is a sustainable alternative to
sensor is flush-mounted and capable of The production of steel is a highly traditional single-use gloves in medi-
both in-contact and contactless sensing carbon-intensive process despite steel cal, laboratory, and industry applica-
through plastic or glass vessels. The being a highly recyclable material. tions. These gloves include organic
20-turn potentiometer allows accurate Now, Green Steel Stainless+ offers additives that attract microbes that
adjustments of the sensing range from competitive mechanical properties and naturally disintegrate the material,
2 mm to 20 mm. The sensor operates heat and corrosion resistance while degrading up to 67% in landfills after
on 20–250 V AC/DC and can work at being made with over 95% recycled 23 months and leaving only biogas,
temperatures up to 70°C. material. The manufacturer’s electric water, and inert soil. The series features
Rechner Electronics Industries Inc. arc furnace (EAF) technology converts EcoBlue, EcoGreen, and EcoBlack
www.rechner.com scrap and post-consumer material into gloves and provides the same neces-
a steel with 75% lower Scope 3 emis- sary durability and resistance as com-
Helium Leak Testing Devices sions and a carbon footprint of around parable alternatives. The gloves are
Maintain Process Purity 1.0 tCO2e per ton of steel supplied. Food and Drug Administration (FDA)-
Helium leak tests are crucial for The material’s composition, includ- approved and meet American Society
maintaining purity in semiconductor ing the amount of ferroalloys, can for Testing and Materials (ASTM)
manufacturing, as well as solar cell be adjusted to meet an application’s D5526 and ASTM D5511 standards for
and flat panel display production. mechanical requirements. biodegradability and performance.
Three intelligent technologies, the Swiss Steel Group United Global Trading Corp.
Smart-Spray, I•Boost, and Spray- www.swisssteelgroup.com www.unitedglobaltradingcorp.com

August 2024 aiche.org/cep 13

202408_New products [1].indd 13 7/23/24 11:08 AM

 PR O F I LE

To Boldly Go Where
Few ChemEs Have Gone Before
hemical engineer. Surgeon. Astronaut. Finding suc- moved to California in 1994 for his residency at the Univ. of
C cess in any one of those professions is a noteworthy
accomplishment. By exploring all three of these career
California, San Francisco (UCSF), where he specialized in
orthopedic oncology.
paths, Robert L. Satcher Jr., MD, has managed an incred- While Satcher was in residency at UCSF, his research
ible achievement. In 2009, Satcher spent 11 days in space, was focused on the response of the musculoskeletal system
completing an orbit around the Earth on the Space Shuttle to mechanical inputs. However, his principal investigator
Atlantis and serving as the crew’s medical doctor onboard (PI) was also involved in another project at NASA’s Ames
the International Space Station (ISS). “Going to space was a Research Center (ARC). Satcher had the opportunity to visit
very rewarding experience,” says Satcher. ARC for his research and while there, he had the opportunity
Satcher was born in Hampton, VA, and lived there to meet and talk with astronauts. “It was kind of serendipi-
through junior high school. He and his family then moved to tous,” Satcher recalls. “That led to me becoming interested
Denmark, SC, and his parents worked at Voorhees College in applying to become an astronaut.”
(now Voorhees Univ.), a historically Black university. Dur- After his residency and fellowships in orthopedic oncol-
ing high school, Satcher was interested in math, chemistry, ogy concluded, Satcher began working at Northwestern
and physics, therefore chemical engineering became an Univ. as an assistant professor in the Dept. of Orthope-
obvious choice for him when picking a college major. He dic Surgery in 2001. During his time there, he applied to
was accepted into the Massachusetts Institute of Technology NASA’s astronaut training program and was accepted. In
(MIT) for chemical engineering, in addition to receiving a 2004, Satcher moved to Houston, TX, to undergo astronaut
scholarship from DuPont that required him to intern at sev- training, which he completed in two years.
eral of their locations during the summer. “All of that solidi- Astronaut training is not for the faint of heart. Satcher’s
fied my interest in chemical engineering,” recounts Satcher. training included instruction in shuttle and International
During his undergraduate years, Satcher worked on vari- Space Station (ISS) systems, scientific and technical brief-
ous research projects that were more life-science oriented. ings, and physiological training. To prepare for spacewalks,
“The more I did that,” he said, “the more I realized that if I astronauts train at the Sonny Carter Training Facility,
wanted to study these problems, where would they be appli- which houses the world’s largest indoor pool, the Neutral
cable?” This was when Satcher began to find an interest in Buoyancy Laboratory. Sized at 200 ft long, 100 ft wide,
medicine and improving medical care through research. Still and 40 ft deep, full-size replicas of the space shuttle and
wanting to continue his chemical engineering education, he areas of the ISS are submerged in the pool. Satcher trained
was accepted into an MD-PhD program, a combined cur- for spacewalks under water as it simulates the absence
riculum between of light and weightlessness that one can expect in space.
MIT and Harvard Astronauts also receive aircraft training on T-38 jets, as
Medical School. well as wilderness survival training.
“Chemical “From when I first started training to when I flew was
engineering has about five years,” Satcher says. “The astronaut office assigns
such a broad foot- you to particular jobs that are supporting roles while you’re
print with a heavy waiting to be assigned to the flight.”
involvement in In 2009, he was assigned to the STS-129 mission aboard
the life sciences,” the Space Shuttle Atlantis to the ISS — in total, an 11-day
says Satcher. mission. This mission would make him the first Black male
“From the physician to serve as the crew’s medical officer.
engineering sub- The STS-129 mission was one of the final construction
specialties, chemi- missions for the ISS and the 31st shuttle mission to the sta-
cal engineering is tion. The Space Shuttle Atlantis was set to launch on Nov.
probably the one 16, 2009, from NASA’s Kennedy Space Center in Florida.
that produces the With a crew of six members on board, including Satcher,
▲ Figure 1. All six crew members of STS-129, including most doctors.” Atlantis launched at 2:28 PM (Figure 1). The shuttle was
Satcher (back, right), get ready to make their way to
launch pad 39A to prepare for liftoff on Space Shuttle After he received commanded by Charles O. Hobaugh, a veteran in space
Atlantis. Image courtesy of NASA. both degrees, he exploration, who had piloted two missions prior. Less than

14 aiche.org/cep August 2024

202408_Profile InDesign 2.indd 14 7/22/24 11:28 AM

 ▲ Figure 2. On his second spacewalk, Satcher worked
on the truss of the International Space Station (ISS).
Image courtesy of NASA.

two years later, the Space Shuttle Atlantis would complete for two more years, working support jobs at the astronaut
its 33rd and final mission, thus concluding the American office before moving back toward the clinical and academic
Space Shuttle program. worlds. In 2011, he joined the Univ. of Texas M.D. Ander-
On board the ISS, Satcher was the mission specialist son Cancer Center as a surgical oncologist and professor.
responsible for monitoring the health of his crewmates, tak- Today, he treats patients who have musculoskeletal system
ing care of any medical issues, and ensuring that everybody cancers. His current research is on metastatic disease of
was appropriately adjusting to microgravity and the space the bone with a focus on kidney and prostate cancers —
environment. “I would check in on everybody daily and then an extension of his PhD research, which combined fluid
check in with the flight medical doctors who were in Hous- mechanics with life sciences. “My current research is kind
ton,” he remembers. of inherent to chemical engineering as it continues to play a
The routine ailments experienced by the crew members very active role in what I do,” notes Satcher. “I look at how
were typically space motion sickness and back pain due to the cells in the musculoskeletal system and in bones in par-
the microgravity’s effect on the spine. “We didn’t have any ticular respond to external factors, such as colonizing cancer
serious medical problems when I was on board,” Satcher cells. We’re also interested in how they’re responding to the
says. If that were to happen, he recalls, the strategy was to mechanical environment in bones.”
stabilize the individual and then transport them back to Earth, Aside from his family and two kids, Satcher considers
which would take three hours in case of an emergency. his time at NASA his greatest achievement to date, as it has
Satcher took part in two of the three spacewalks of allowed him to interact with more people than anything else.
the STS-129 mission, resulting in approximately 12 hours As a medical doctor and researcher, he serves on advisory
of extravehicular activity (EVA) (Figure 2). During these panels for NASA. “What they’re grappling with now is how
EVAs, he worked on the maintenence of the Canadian to take care of people when they do space flight missions to
robotic arm and reconfigured the truss of the ISS. Satcher the moon and beyond,” says Satcher. He also takes part in
remembers the mission being “busy yet fun.” He met public appearances.
several other international astronauts on the ISS. “It was After going through three major career phases, Satcher
a special experience,” recounts Satcher. “In addition to leads a busy life, balancing his research along with overseeing
the work you’re doing, just the relationships you establish procedures in the operating room. He is also the founder of a
when you’re there are quite meaningful.” Overall, the mis- startup company called OrbitalMed, Inc., which is focused on
sion left a lasting impact on his life, checking off the box commercial astronaut healthcare.
of becoming an astronaut. “I really got to do everything I If given the opportunity, he would enjoy being an active
imagined that I wanted to do going there,” he says. astronaut again on a future mission. “Going back to space is
The mission concluded as they landed back on Earth on very appealing,” says Satcher. “It would be a lot of fun to do,
Nov. 27, 2009. After the mission, Satcher remained at NASA so if that opportunity arose, I would do it again.”

August 2024 aiche.org/cep 15

202408_Profile InDesign 2.indd 15 7/22/24 11:28 AM

 CAREER C ON N EC TION

How to Determine and Advocate for the

Compensation You Deserve
Kate Williamson ■ Scientech Résumés
egotiating your salary is a crucial part of the hiring be vague: “The salary range specified in the job posting coin-
N process, yet questions about salary expectations often
induce fear and anxiety across the career spectrum. Your sal-
cides with my expectations and is within the market range in
[city, state].” Defer this conversation until you understand the
ary impacts your financial health, career growth, and overall role better, setting you up for salary discussions after the first
job satisfaction. Despite its importance, many job seekers or second interview. You could frame this as: “My priority is
are unprepared to navigate these conversations and secure finding the right fit where I can add value to an organization
fair compensation. This column covers strategies to help you whose mission and values align with mine. As such, I’d pre-
identify and clarify your salary needs and empower you to fer to understand more about the role and expectations before
advocate for the compensation you deserve. discussing specific numbers. Can we revisit this topic later?”
People avoid salary negotiations for various reasons, Another option is to share your calculated salary range.
from fear of rejection and lack of preparedness to cultural For example: “Based on my research, experience, and the
factors and job security concerns. Not negotiating has imme- market rate for this role in [city, state], I am targeting a base
diate and long-term consequences. A lower starting salary pay between [$X and $Y].” Base pay is an important distinc-
means lower future raises, bonuses, and benefits, signifi- tion because it is only one component in a compensation/ben-
cantly reducing lifetime earnings and retirement savings. For efits package. Always add the following caveat: “That said,
instance, a 5% increase on an initial salary of $75,000 nets the range might shift once I have a more in-depth understand-
an additional $3,750 annually. An average annual raise of ing of all the job requirements, expectations, and benefits.”
3% could amount to an extra $178,000 over 30 years. What about applications that require you to input a
What is the bare minimum salary you could accept to desired salary? Give a range based on your salary research,
make ends meet? Calculate your average monthly expenses ensuring the lower end is at least 10% above your current
over six months. By looking at six months, you can see clear or targeted salary so that you start negotiations happy if the
patterns in your spending, rather than analyzing one month’s company makes an offer on the lower end. For example, if
expenses in isolation. Start with necessities: housing, grocer- you are a chemical engineer targeting $105,000 to $125,000,
ies, insurance, utilities, clothing, and debts. Include discre- you would list a range of $115,000 to $125,000. If only a
tionary spending, like regular coffee purchases. Multiply single number is allowed, choose a mid-to-high value, such
your monthly average expense by 12 to get your “bare mini- as $115,000. If text input is permitted, write “Negotiable”
mum” annual salary. Your “livable” salary accounts for taxes for flexibility. Given the low success rate of online applica-
and benefits, typically about 30% of the average paycheck. tions (less than 7%), prioritize networking and informational
Double your “bare minimum” salary to account for these fac- interviewing. Create a company target list and build con-
tors, as well as emergency savings and discretionary funds. nections with people who can offer valuable salary insights,
What ideal salary meets all your current and future which, in turn, will help you to respond more confidently
financial needs? Consider retirement and long-term plans, to salary questions. Combined with a strong referral, this
such as buying a home. Take your annual “livable” salary and approach is more likely to lead to meaningful conversations
multiply it by 20%. This buffer helps you aim higher during and better negotiation outcomes.
your negotiations with your potential employer. What if the salary range is not posted? Use the above
Once you have answered these questions, research market approach and flip the script. For example: “I believe it’s
salaries for your career target(s) using past AIChE Salary essential to align expectations, and I would like to discuss
Surveys (CEP, June 2023, pp. 21–31) and platforms like salary in the context of the value I can bring. Can you share
Glassdoor, Payscale, Salary.com, Indeed.com, and LinkedIn the salary range for the role?” This bold but strategic move
Salary. Gather data from three to five platforms and average addresses the topic upfront and prevents any mismatches in
the results to understand your market value while accounting salary expectations later in the process.
for discrepancies and variations in sources and methodologies. Each approach has advantages and disadvantages; what
Create your salary discussion plan. Salary discussions works best may vary by situation. This framework will help
can happen at any stage. If asked about salary early, such as you advocate for the compensation you deserve.
on the application or during the initial screening call, provide Kate Williamson is a former R&D technician and product develop-
a range based on your research rather than a specific number, ment engineer with a master’s degree in applied chemistry. In 2016, she
established Scientech Résumés, a highly specialized career management
giving you flexibility later. You have an advantage if the firm that works exclusively with STEM professionals. Learn more at
salary range is in the job posting. Answer the question, but www.scientechresumes.com.

16 aiche.org/cep August 2024

202408_Career Connection InDesign 1.indd 16 7/22/24 12:01 PM

 SPECIAL SECTION:

AI and Digitalization
Discover the expanding applications
of this rapidly evolving technology

The Rise of Artificial Implementing Artificial Intelligence

Intelligence p. 18 in Process Safety Studies p. 27

A Chemical Engineer’s The Industrial Metaverse for the

Introduction to ChatGPT p. 19 Process Industries p. 36

202408_Supplement_Cover Revised.indd 17 7/23/24 12:58 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

The Rise
of Artificial Intelligence
A
rtificial intelligence (AI) has seen a rapid expansion in ing article explores how AI can be implemented to help
its capabilities, especially in recent years. As defined enhance process hazard analyses (PHAs). The authors of
by Amit Gupta in “Introduction to Deep Learning: “Implementing Artificial Intelligence in Process Safety
Part 1” (CEP, June 2018, pp. 22–29), artificial intelligence Studies” (pp. 27–33), suggest that the use of AI may be
is the “capability of a machine to imitate intelligent human most beneficial in the revalidation of hazard and operabil-
behavior.” Likewise, machine learning (ML) describes a ity (HAZOP) studies. This article discusses how AI can
computer’s ability to learn without explicit instructions. use existing plant data, as well as management of change
AI has the potential to completely reshape the way we (MOC) reports, to re-evaluate risks and revalidate causes,
live, work, and engage with technology. We are constantly consequences, and safeguards.
interacting with AI/ML, from voice assistants such as Siri or To wrap up, the final article reviews “The Industrial
Alexa, to image and music curation based on our personal Metaverse for the Process Industries” (pp. 36–41). “The
preferences through services such as Pinterest or Spotify. industrial metaverse is the concept of a digital world to
Unsurprisingly, AI and ML have found their way into many mirror and simulate real machines and factories,” write
technical industries, including chemical engineering. Aglave et al. of Siemens Digital Industries. One of the
From problem-solving to real-time monitoring and pre- most well-known applications of the industrial metaverse
dictive diagnostics, AI has a wide range of applications and within the CPI is the implementation of digital twins,
uses in the chemical process industries (CPI). However, which can digitally recreate and simulate equipment on
before rushing to employ AI wherever possible, it is crucial the factory floor.
to fully understand the technology and its limitations in By leveraging real-time data along with simulated
order to safely and effectively implement it. This special data from the digital twin, engineers can swiftly predict
section discusses how the CPI is making its first forays into and determine measures that need to be taken in order to
AI and ML. Each article covers the implications of utiliz- ensure that their operations run as safely and smoothly as
ing this ever-growing technology, showcasing currently possible. This article highlights the extent of the capabili-
available platforms, the various pros and cons, and ways ties of digital twins, and touches upon the challenges of
to safely implement AI into current operations through the implementing the industrial metaverse as a whole in cur-
lens of digitalization. rent facilities.
Kicking off this special section is the article “A Chemi- This special section gives a glimpse into what AI is
cal Engineer’s Introduction to ChatGPT” (pp. 19–25) capable of in the world of chemical engineering. Fully
by Sean J. Dee and Sahand Faraji of Exponent, Inc. AI understanding how the technology works is the first step
chatbots are one of the most talked about and notable toward unlocking its potential and safely implementing it
developments of AI technology. Users can seek the help of in CPI facilities. AIChE conferences — such as the 2024
these AI chatbots for nearly any situation, even chemical Global Conference on Process Safety and Big Data taking
engineering and process safety. place Oct. 29–30 in Germany, jointly hosted by the Center
However, many experts remain justifiably skeptical for Chemical Process Safety (CCPS) and the European
of their performance, accuracy, and data security when it Process Safety Centre (EPSC) — are a great place to start.
comes to the chatbot’s ability to answer technical ques- This conference in particular features presentations on a
tions. As the title of the article suggests, the authors discuss wide spectrum of topics from industry experts, including
the basics of these large-language models (LLMs), specifi- AI applications in process safety, the impact of digitaliza-
cally OpenAI’s ChatGPT, and their strengths and limita- tion on cybersecurity, and more.
tions in CPI applications.
Continuing the theme of process safety, the follow- Melanie Mesropian, Assistant Editor, CEP

18 aiche.org/cep August 2024

202408_AI Special Section_ Letter InDesign.indd 18 7/22/24 4:43 PM

 A Chemical
Engineer’s
Introduction
to ChatGPT
Sean J. Dee, P.E. ■ Sahand Faraji ■ Exponent, Inc.

As OpenAI’s ChatGPT gains popularity across different professions,

chemical engineers have questions about if (and how) the tool
works in practice.

C
hatbots have captured the imagination of chemical the technology’s limitations, as well as the strategies for
engineers as they debate how artificial intelligence optimizing its output. This article aims to provide an intro-
(AI) technologies can be leveraged to enhance duction to these aspects of chatbots. The authors have used
industry practices. Many companies are exploring the a variety of chatbot technologies, but this article focuses
deployment of AI-based tools such as OpenAI’s ChatGPT on OpenAI’s ChatGPT. It reviews the recent updates and
or Microsoft Copilot, evaluating whether these technolo- changes to ChatGPT announced by OpenAI in May 2024,
gies can improve the quality and efficiency of their employ- compares paid and free ChatGPT plans, and discusses
ees’ workstreams. As chemical engineers begin to experi- key questions facing the deployment of chatbot technolo-
ment with these tools, they uncover interesting insights into gies such as ChatGPT. Finally, two illustrative case studies
the strengths, weaknesses, and potential applications of showcase the potential applications of the technology and its
the technology. performance as a process safety tool.
However, users in the engineering space have reported a
broad spectrum of first impressions, ranging from impressed How does ChatGPT work?
to underwhelmed to highly skeptical. These tools are often ChatGPT is based on a technology referred to as a large
“black box” technologies, and users without deep knowl- language model (LLM) (1). As shown in Figure 1, when
edge of the underlying research may struggle to recognize a user types an input prompt into ChatGPT, the technol-

August 2024 aiche.org/cep 19

202408_AI Special Section_Dee [1].indd 19 7/23/24 12:59 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

ogy breaks the string of text into smaller pieces of data output from the analysis is the most likely next token in the
called tokens, which can be as short as one character or as sequence of text. This process is then iterated to continually
long as one word. The model then analyzes the relation- append words onto the output. In practice, ChatGPT builds
ships between these tokens based on an internal analysis its output one word at a time based on its internal architec-
framework referred to as its architecture, which is typically ture and the context provided by the text in the input prompt.
a transformer model. This architecture uses a mechanism Increasing and refining the parameters within the archi-
called “attention” to weigh the relevance of different tokens tecture is a key feature of improving the performance and
in the context of the input. functionality of the technology. The number of parameters
The architecture in ChatGPT is OpenAI’s generative in OpenAI’s GPT architecture has expanded over time. In
pre-trained transformer (GPT) (2). This architecture is a 2018, the first iteration of the GPT architecture included
type of artificial neural network that is inspired by bio- approximately 117 million parameters, while current
logical neural networks where layers of interconnected versions boast 375 billion parameters in GPT-3.5 and
nodes (e.g., neurons connected by synapses) work together 1.7 trillion in GPT-4 (4–6).
to relay and analyze information. OpenAI’s GPT uses Before training (1), the tool may provide random outputs
weighting factors and biases, referred to as parameters, to for the next word. After training on a text dataset and devel-
characterize the relationships between tokenized data from oping parameters to model and predict patterns, the technol-
the input prompt. These parameters are developed by train- ogy is better equipped to predict more accurate outputs. For
ing GPT on large datasets of text to help the technology example, “around,” “right,” or “back” could be alternate
recognize underlying patterns and relationships between output tokens for the example prompt shown in Figure 1.
words within the dataset. There are often multiple acceptable outputs when predicting
GPT is trained using both supervised and unsupervised the next word in a text sequence. As a result, the output from
machine learning (3), with the difference being the use of the technology may vary even when prompted with the exact
labeled input and output datasets in supervised training same input text.
(e.g., providing a series of spam emails will help the tool According to OpenAI (1), there are three primary
recognize spam-specific patterns) and unlabeled datasets for sources of data used for model training: publicly avail-
unsupervised training. After training, the architecture can able information on the internet, licensed information from
also be further fine-tuned through reinforcement learning third parties, and information provided by users or human
from human feedback (RLHF), where human trainers rank trainers. OpenAI states that the training dataset is filtered
or evaluate the technology’s outputs to further refine the for inappropriate content (e.g., hate speech, adult content,
internal parameter set. spam, personal information) and does not include informa-
In the example prompt in Figure 1, the key relation- tion behind paywalls. Through training, the architecture
ships would likely be between the words “instead,” “turned” develops and modifies its parameter set, and once training
and “turning,” and “left,” which establish the next token is completed, the architecture no longer has access to its
would likely be a directional word other than “left.” The training dataset. In other words, when a user interacts with
Tokens

Instead
Artificial Neural Network

of
Input Prompt

turning Output Token

“Instead of
turning left,
left right
she turned
_______”
,

she

turned

◀ Figure 1. The architecture of ChatGPT

analyzes the relationships between the words
Output Prompt in an input prompt based on an artificial neural
Instead of turning left , she turned right
network that is developed through training on
large datasets. These relationships are used to
determine the output.

20 aiche.org/cep August 2024

202408_AI Special Section_Dee [1].indd 20 7/22/24 4:46 PM

 ChatGPT, the tool generates text based on its architecture’s many companies, governments, and regulators are address-
understanding of the relationships between tokens in the ing this issue as they develop and update policies, regula-
input prompt and the context it has learned during training. tions, and guidance.
It does not search, find, or retrieve specific pieces of infor- Beginning in April 2023 (8), ChatGPT users were given
mation in its training data to generate its outputs, which is a the option to opt out of allowing their conversations to be
common misconception of those unfamiliar with the under- used for model training. According to OpenAI, when a user
lying technology. Instead, it uses patterns and knowledge opts out of this setting, new conversations generated in the
embedded in the model during training to produce coherent platform are not used for model training. The conversation
and contextually relevant responses. will still appear in the user’s chat history unless the user
Fundamentally, ChatGPT is a conversational text- deletes it. ChatGPT also has a “temporary chat” feature
generating tool designed to mimic human language. located in the interface’s model selection drop-down menu.
However, as the technology has developed over time, it is Temporary chats are also not used for training but do not
apparent that the applications and potential benefits extend appear in the user’s chat history (although a copy of the
beyond simply predicting the next word in a sentence. chat may be maintained by OpenAI for 30 days for safety
reasons). OpenAI also offers Team and Enterprise plans that
What is GPT-4o? exclude data from training as a default setting and offer addi-
In May 2024 (7), OpenAI announced several key tional data security measures (i.e., administrative console
changes to ChatGPT, including the release of the newest and controls and custom data retention windows).
version of its architecture, GPT-4o, which stands for “omni” Several lawsuits have been filed alleging copyright law
or “omnimodel.” GPT-4o touts an enhanced ability to pro- violations associated with OpenAI’s use of copyrighted
cess across various inputs, including text, speech (audio), materials for training and ChatGPT’s outputs containing
and vision (images). This processing reportedly occurs copyrighted content (9). In response, OpenAI has pointed
within a single architecture, a significant improvement over to fair use policy, a legal doctrine focused on allowing
prior versions that required multiple models for processing the use of copyrighted materials without permission from
inputs that were not text-based. In previous models, audio the copyright holder for transformative purposes such
had to be transcribed to text; the text was then processed in as research or teaching (10, 11). OpenAI’s terms of use
GPT-4 and then converted to an audio output using text-to- state that the output is owned by the user, and the user is
speech processing. responsible for ensuring both inputs and outputs do not
OpenAI states that GPT-4o also significantly improves violate applicable laws or infringe upon the rights of oth-
analysis involving non-English-based languages. Demon- ers (12). Users should be cautious and mindful of these
strations of the new GPT-4o model highlighted applications considerations when using ChatGPT and OpenAI’s tools.
based on conversational speech prompts and visual prompts “ChatGPT can make mistakes” is a warning statement
(e.g., interacting with the camera or the visual display of that is clearly placed at the bottom of the user interface,
a computer screen). OpenAI also notes that since GPT-4o regardless of whether the user has selected the GPT-3.5,
is the first model with end-to-end modalities across text, GPT-4, or GPT-4o model. While the potential capabilities of
vision, and audio, the company is still uncovering the poten- the tools can be impressive, it is important to remember that
tial capabilities and limitations of GPT-4o. outputs should be verified when accuracy is important. This
is particularly true in engineering applications, which may
Concerns regarding data security, copyright, require a deep understanding of scientific and mathematic
and accuracy principles that may not be accurately represented in the large
Data security, copyright, and accuracy are three very text database used to train ChatGPT.
important concerns regarding the use of ChatGPT. LLMs
like ChatGPT can be trained on the data generated by Free and paid plan options
users interacting with the platform to improve the model’s The corporate structure surrounding OpenAI is unique
accuracy and capabilities. However, many users and busi- (13). OpenAI started as a nonprofit with the goal of building
nesses may be concerned about how their data will be safe artificial general intelligence (AGI) — artificial intel-
used or whether training on sensitive data (i.e., personal or ligence that is equivalent to, or exceeds, human intelligence
confidential trade secrets) could result in that information — for the benefit of humanity. In 2019, a “capped-profit”
being provided in a subsequent ChatGPT output to other entity (OpenAI LP) was created to help accelerate invest-
users. OpenAI provides information on these topics on its ment. Today, OpenAI is a combination of the nonprofit and
website, and a summary of key features is provided here. capped-profit entities. As a result, offerings from OpenAI
However, appropriate caution should still be exercised, and (like ChatGPT) may include a free version to provide AI

August 2024 aiche.org/cep 21

202408_AI Special Section_Dee [1].indd 21 7/22/24 4:46 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

to the masses and paid options with additional features and The Enterprise plan opens unlimited access to restricted
functionality to generate capital to support further techno- tools, additional security controls, custom data retention
logical development. policies, and priority support. OpenAI does not publish rates
OpenAI currently offers four plans for ChatGPT (14). for the Enterprise plan, and their website refers interested
The features and offerings of the plans change over time, but users to their sales team.
this section presents a brief overview of the options available
at the time of this writing. Case study examples with GPT-4o
Historically, the paid subscription plans provided users The following two case studies are provided as examples
access to advanced tools and features, like GPT-4, attach- of hypothetical applications for a chatbot technology. The
ment upload, and the GPT Store, that were not available case study input prompts were given to ChatGPT with the
to users on the Free plan. However, OpenAI announced in GPT-4o model in May 2024.
May 2024 that it was expanding ChatGPT capabilities for Case study 1: Reviewing regulatory standards. A safety
Free users, giving them access to the newly released GPT-4o professional at a chemical company is asked by maintenance
architecture, file upload and vision capabilities, data analy- staff if their Occupational Safety and Health Administration
sis, and the GPT store (7). (OSHA) 40-hour hazardous waste operations and emergency
Paid subscriptions offer expanded limits for the advanced response (HAZWOPER) training fulfills the requirements to
tools. For example, a Plus user on the $20 per month plan perform confined space entry of an atmospheric storage tank
can submit 40 messages every three hours to GPT-4 and 80 at the facility (Figure 2). The following input prompt was
messages every three hours to GPT-4o (14). These limits provided to ChatGPT:
may be adjusted dynamically by OpenAI based on available “You are an environmental, health, and safety profes-
capacity in the platform. Both Free and Plus have unlimited sional. If a worker goes through a 40-hour HAZWOPER
access to the GPT-3.5 architecture when respective limits training course, is the worker considered appropriately
on other architectures are reached. Additionally, Plus users trained according to OSHA regulations to perform confined
are given early access to new features, the ability to build, space entry on an atmospheric storage tank? If you need
create, and use customized GPTs, and access to specialized more information before responding, please ask a question
GPTs, such as DALL·E, for image creation. for clarification. Include references to the relevant citations
In addition to Free and Plus plans, OpenAI also offers in the OSHA regulations for both ‘HAZWOPER’ and ‘con-
Team and Enterprise plans. The Team plan includes higher fined space’ in your response.”
message limits for GPT-4, GPT-4o, and tools like DALL·E, In response, GPT-4o noted that HAZWOPER training
and excludes data from training as a default setting. The (29 CFR 1910.120) provides a broad overview of safety
Team plan also provides a workspace where GPTs can be topics for workers involved in the cleanup of hazardous
created and shared with other Team members and users substances. Confined space training (29 CFR 1910.146) cov-
with administrative console control. The Team plan is ers hazards, equipment, and rescue procedures for confined
billed per user at either $25 per month (annually) or $30 per spaces. GPT-4o then provided a summary of key points and
month (monthly). concluded that a worker with 40-hour HAZWOPER training

◀ Figure 2. In the first hypothetical case study,

engineers asked ChatGPT whether a 40-hr
hazardous waste operations and emergency
response (HAZWOPER) training course met the
Occupational Safety and Health Administration’s
(OSHA’s) requirements for performing confined
space entry. The large language model (LLM)
identified that personnel would need additional
training on confined-space safety, and the
HAZWOPER module would not be sufficient.
Image generated with OpenAI’s DALL·E.

22 aiche.org/cep August 2024

202408_AI Special Section_Dee [1].indd 22 7/22/24 4:46 PM

 would not be considered appropriately trained for confined and typically violates the ASME BPVC, because the code
space entry. Workers would require both trainings if they states there should not be obstructions between the valve
were performing tasks that combined hazardous materials and the vessel it protects.
and confined spaces. While this discussion includes several key themes
Case study 2: Mitigating corrosion of relief valves. An related to corrosion, it failed to identify a rupture disc as a
engineer at an acid production facility is asked for ideas on potential design solution based on the initial input prompt.
how to extend the service life of a relief valve on an acid Rupture discs are often used to protect relief devices
reactor, which is a rated pressure vessel that operates at high from corrosion (15).
pressure and temperature (Figure 3). The initial proposal This example highlights an important limitation of
from maintenance staff is to install a check valve between
the reactor and the relief valve to prevent the reactor’s cor- SEAN J. DEE, P.E., PhD, CFEI is a Senior Managing
Engineer in the Thermal Sciences practice at
rosive vapors from entering the relief header. The engineer Exponent, Inc. (4580 Weaver Parkway, Suite
is concerned this modification would violate the American 100, Warrenville, IL 60555; Email: sdee@
exponent.com). He applies the fundamentals of
Society of Mechanical Engineers (ASME) Boiler and Pres- chemical engineering to help clients address
sure Vessel Code (BPVC). The following input prompt was the risks associated with accidents, losses,
and injuries. His core areas of expertise include
provided to ChatGPT: chemical process safety, fire and explosion
“You are the engineer at a facility that produces a cor- investigation, alternative energy, and high-purity
oxygen systems. He services clients in many diverse industries,
rosive acid product in a pressure vessel at high temperature including insurance, legal, and chemical processing. He earned
and pressure. The pressure relief valve on the pressure vessel his PhD in chemical engineering from the Univ. of California,
Berkeley, where his thesis focused on the catalytic produc-
is corroding in service. Can you give me three options for tion of biofuels from biomass. He also earned a BS in chemical
modifications to the system that could be implemented to engineering and a BA in sociology from Case Western Reserve
Univ. He is a certified fire and explosion investigator and a profes-
reduce the corrosion of the pressure relief valve and extend sional engineer licensed in multiple states. He is also an elected
the service life of the relief valve? Could we install a check Director of AIChE’s Process Safety Division (formerly the Safety &
Health Division).
valve between the vessel and the relief valve? Or would that
violate the ASME Boiler and Pressure Vessel Code?” SAHAND FARAJI, PhD, is a Senior Associate in
the Thermal Sciences practice at Exponent,
GPT-4o’s first suggestion was to upgrade the relief Inc. (15615 Alton Parkway, Suite 350, Irvine, CA
valve to corrosion-resistant materials such as Hastelloy, 92618, Email: sfaraji@exponent.com). He has
expertise in chemical engineering and industrial
Inconel, or other high-nickel alloys. Second, it proposed processing, hydrogen and clean energy produc-
applying a protective coating or liner to the relief valve, tion, nanomaterial synthesis, petrochemical and
process safety, and experimental combustion
such as Teflon, ceramic, or epoxy, to provide a barrier processes. He earned his PhD in chemical
between the corrosive vapors and the metal surface. Third, engineering from the Univ. of California, Irvine,
where he studied the synthesis and evaluation of nanomaterials for
it discussed a recirculating barrier fluid system, where combustion technologies. Prior to attending graduate school, he
water or a non-reactive oil could create a barrier between received a MSc in chemical engineering from Istanbul Technical
Univ. in Turkey and a BSc in chemical engineering from Univ. of
the valve and the corrosive acid. The response also stated Tabriz in Iran.
that installing a check valve would not be recommended

▶ Figure 3. In the second hypothetical case

study, engineers asked ChatGPT how to address
a corroding pressure relief valve under acidic
conditions. The LLM analyzed the American
Society of Mechanical Engineers (ASME) boiler
and pressure vessel code to come up with a
broad but not exhaustive list of possible solu-
tions. Image generated with OpenAI’s DALL·E.

August 2024 aiche.org/cep 23

202408_AI Special Section_Dee [1].indd 23 7/22/24 4:47 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

ChatGPT. Recall that ChatGPT is a

(a)
conversational tool that builds its output
Prompt 1: Can you provide me a picture with Prompt 2: I like the person on the left. Could
based on the relationships between the
one person dressed in a hazmat suit and you make the person on the right look more text in the input prompt. It is likely that
another dressed as a construction worker like a general-duty construction worker
with a rescue or fall harness. I’d like both (hard hat, overalls, jeans, and boots with no
the training database had very strong
workers to be depicted standing in front of a respiratory protection, etc.)? relationships for text regarding corro-
large diameter and height atmospheric sion and the use of corrosion-resistant
storage tank.
materials, protective coatings, and/or
barriers. In practice, the problem dis-
Prompt 1 Image Prompt 2 Image cussed in the input prompt is a combi-
nation of two issues — corrosion and
pressure relief — which both need to
be addressed in parallel in the proposed
solution. Research on prior models of
ChatGPT (GPT-3 and GPT-4) suggests
that the architecture struggles with
problems that require advanced plan-
Prompt 3: This is better. Can you make both Prompt 4: Thank you — one last tweak.
ning or problems where a seemingly
workers approximately the same height, Can you put the harness on the worker on discontinuous leap forward is needed to
remove the coveralls from the worker on the the right and no harness on the person on progress toward the solution of a task
right, and add a fall harness to the worker on the left?
the right? (16). It is possible this complex relation-
ship between the competing objectives
of protecting the relief valve against cor-
Prompt 3 Image Prompt 4 Image
rosion and protecting the pressure vessel
from overpressure required a so-called
“Eureka” idea to arrive at a rupture disc
as a solution.
Using DALL·E images. One of
the advanced tools OpenAI offers is
DALL·E, which can create images
from text prompts within the ChatGPT
user interface. DALL·E is a 12-billion
(b) parameter version of GPT-3 that was first
Prompt 1: Can you generate a picture of a Prompt 2: Okay — can you make the
released in 2021 and has subsequently
pressure relief valve on top of a large pressure relief valve yellow in that been updated with DALL·E2 and
chemical reactor? image, and can you also make it heavily DALL·E3 to improve the resolution,
corroded/discolored so that it looks like it
needs replacement? accuracy, and realism of the generated
images (17). The outputted images are
owned by the user, including the right to
Prompt 1 Image Prompt 2 Image
reprint, sell, and merchandise the image,
subject to OpenAI’s Content Policy and
Terms (18). OpenAI’s guidelines for
DALL·E also state that users are encour-
aged to proactively disclose the use of AI
in their work and cannot mislead others
about AI’s involvement (19).
Figures 2 and 3 were generated with
▲ Figure 4. The images for the hypothetical case studies (Figures 2 and 3) were generated using DALL·E. (a) DALL·E. In general, several iterations
The first case study required four successive prompts; early outputs were fed back into DALL·E to further refine are required to refine the image output to
the image. (b) The image for the second hypothetical case study was also generated using DALL·E, and two of
the outputs that DALL·E generated are shown here. Prior prompts in the second case study (not shown here)
match a prompt’s intent. Figure 4 demon-
attempted to improve the accuracy of DALL·E’s depiction of a pressure relief valve, to varying levels of success. strates this process. In the pressure relief
Images generated with OpenAI’s DALL·E. valve case study, DALL·E required five

24 aiche.org/cep August 2024

202408_AI Special Section_Dee [1].indd 24 7/22/24 4:47 PM

 prompts before the prompts shown in Figure 4b to refine the their architecture, training, datasets, and context provided in
tool’s depiction of a relief valve. These prompts included the input prompt.
uploading example pictures of pressure relief valves into These caveats are not meant to diminish the significance
the user prompt, and DALL·E searching for photographs of of the more recent iterations of ChatGPT. Researchers are
relief valves across multiple search engines and stock photo openly questioning whether GPT-4 represents an early but
websites. Ultimately, these iterations had varying degrees of incomplete form of AGI, which has led to interesting debates
success in improving the accuracy of the image. regarding definitions, measures, and assumptions regarding
the concept of “intelligence” (20). This does not mean that
Closing thoughts ChatGPT has independent thoughts, emotions, or motiva-
This article should not be interpreted as an endorsement tions, which some may consider central to a definition of
of ChatGPT or other AI tools for any application; however, intelligence. However, research has shown that GPT-4
significant progress is being made in LLM technology that is demonstrates emergent behaviors and capabilities in other
widening the technology’s applicability in process safety. In areas of intelligence, including reasoning, problem solv-
the authors’ experience, the GPT-4 model, released in 2023 ing, abstract thinking, complex idea comprehension, and
to Plus users, displayed stark improvements in performance learning (within a user interaction session). While Chat-
over the Free user’s GPT-3.5 model. With the recent expan- GPT was originally developed to predict the next word in a
sion of the functionality of ChatGPT’s free platform, many sentence, the resulting technology has demonstrated much
casual users will likely also experience these performance broader capabilities. As researchers and developers continue
improvements when experimenting with OpenAI’s more to improve the technology, its limitations may become less
advanced tools. However, users must remember that LLMs apparent. For these reasons, it is important that users, and
are text-generation tools designed to construct outputs based engineers in particular, have a fundamental understanding
on the relationships between tokenized data. They can still of how these tools work and recognize the importance of
provide inaccurate responses based on the limitations of verifying the accuracy of their output.

Literature Cited
1. OpenAI, “How ChatGPT and Our Language Models are 10. O’Brien, M., “ChatGPT-Maker Braces for Fight with New York
Developed,” https://help.openai.com/en/articles/7842364-how- Times and Authors on ‘Fair Use’ of Copyrighted Works,” Seattle
chatgpt-and-our-language-models-are-developed (accessed Times, https://www.seattletimes.com/business/chatgpt-maker-
May 31, 2024). braces-for-fight-with-new-york-times-and-authors-on-fair-use-of-
2. Delua, J., “Supervised Versus Unsupervised Learning: What’s the copyrighted-works (Jan. 9, 2024).
Difference?” IBM, https://www.ibm.com/think/topics/supervised- 11. OpenAI, “OpenAI and Journalism,” https://openai.com/index/
vs-unsupervised-learning (accessed May 31, 2024). openai-and-journalism (Jan. 8, 2024).
3. Open AI, “Introduction,” https://platform.openai.com/docs/ 12. OpenAI, “Terms of Use,” https://openai.com/policies/terms-of-use
introduction (accessed May 31, 2024). (Nov. 14, 2023).
4. Schriener, M., “GPT-4 Architecture, Datasets, Costs, and More 13. OpenAI, “Our Structure,” https://openai.com/our-structure
Leaked,” The Decoder, https://the-decoder.com/gpt-4-architecture- (accessed May 31, 2024).
datasets-costs-and-more-leaked (July 11, 2023). 14. OpenAI, “Pricing,” https://openai.com/chatgpt/pricing (accessed
5. Marr, B., “A Short History of ChatGPT: How We Got to May 31, 2024).
Where We Are Today,” Forbes, https://www.forbes.com/sites/ 15. Towler, G., and R. K. Sinnott, “Chemical Engineering Design:
bernardmarr/2023/05/19/a-short-history-of-chatgpt-how-we-got-to- Principles, Practice and Economics of Plant and Process Design,”
where-we-are-today/?sh=47e0d6dd674f (May 19, 2023). 2nd ed., Chapter 10, Elsevier, Amsterdam, Netherlands (2013).
6. Emmanuel, C., “GPT-3.5 and GPT-4 Comparison: Exploring the 16. Bubeck, S., et al., “Spark of Artificial General Intelligence: Early
Developments in AI-Language Models,” https://medium.com/@ Experiments with GPT-4,” Microsoft Research, https://arxiv.org/
chudeemmanuel3/gpt-3-5-and-gpt-4-comparison-47d837de2226 abs/2303.12712 (accessed Feb. 27, 2024).
(accessed May 31, 2024). 17. OpenAI, “DALL·E: Creating Images from Text,” https://openai.
7. OpenAI, “Introducing GPT-4o and More Tools to ChatGPT Free com/index/DALL·E (accessed July 3, 2024).
Users,” https://openai.com/index/gpt-4o-and-more-tools-to-chatgpt- 18. OpenAI, “Can I Sell Images I Create with DALL·E?” https://help.
free (May 13, 2024). openai.com/en/articles/6425277-can-i-sell-images-i-create-with-
8. OpenAI, “New Ways to Manage Your Data in ChatGPT,” https:// DALL·E (accessed July 3, 2024).
openai.com/index/new-ways-to-manage-your-data-in-chatgpt (Apr. 19. OpenAI, “How Should I Credit DALL·E in My Work?” https://
25, 2023). help.openai.com/en/articles/6640875-how-should-i-credit-DALL·E-
9. Grynbaum, M. M., and R. Mac, “The Times Sues OpenAI and in-my-work (accessed July 3, 2024).
Microsoft Over A.I. Use of Copyrighted Work,” New York Times, 20. Bubeck, S., “First Contact,” presented at Massachusetts Insti-
https://www.nytimes.com/2023/12/27/business/media/new-york- tute of Technology, Cambridge, MA, https://www.youtube.com/
times-open-ai-microsoft-lawsuit.html (Dec. 27, 2023). watch?v=qbIk7-JPB2c (Mar. 22, 2023).

August 2024 aiche.org/cep 25

202408_AI Special Section_Dee [1].indd 25 7/22/24 4:49 PM

 April 6 – 10, 2025
Hilton Anatole • Dallas, TX

CALL FOR ABSTRACTS NOW OPEN!

You are invited to submit an abstract for presentation to share research and knowledge at
the 2025 AIChE® Spring Meeting & 21st Global Congress on Process Safety, covering a
wide range of subjects relevant to the current needs of industry.

TOPICAL CONFERENCES 21ST GLOBAL CONGRESS ON PROCESS

• 25th Topical Conference on Gas Utilization SAFETY FEATURES
• 28th Topical Conference on Refinery • 14th Process Safety Management
Processing Mentoring (PSM2) Symposium
• 37th Ethylene Producers’ Conference • 27th Process Plant Safety Symposium
(PPSS)
• Energy Transition
• 40th Center for Chemical Process Safety
• Emerging Technologies in Clean Energy (CCPS) International Conference
• Heat Exchangers • 59th Annual Loss Prevention Symposium
• Hydrogen Safety Topical Conference (LPS)
• Industry 4.0 Topical Conference • Perspectives on Process Safety from
• Kister Distillation Symposium Around the World
• Leadership Development- Professional
Skills to Catalyze your Career YOU CAN ALSO SUBMIT AN ABSTRACT TO
ONE OF THE FOLLOWING DIVISIONS:
• Process Intensification
• Chemical Engineering & the Law
• Waste Plastics Topical Conference • Computing and Systems Technology
• Environmental
• Fuels and Petrochemicals
• Management
ABSTRACT • Process Development
DEADLINE: • Separations
Nov. 4, 2024 • Transport and Energy Process

For more information and instructions on how to submit an abstract,

please visit aiche.org/spring or aiche.org/gcps
FOLLOW US: #AIChESpring #GCPS © 2024 AIChE 9531_24 • 07.24

202408_Spring.indd 26 7/25/24 11:52 AM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

Implementing
Artificial Intelligence
in Process Safety
Studies
Ambalavanan Balasubramanian ■ Kuwait Oil Company
Srinivas Ganti ■ Technip Energies
Haarish Dharan ■ NewEngen

AI can enhance risk assessments — starting with process hazard

analysis (PHA) revalidation and expanding to other studies — provided
that proper digitalization initiatives are in place.

A
rtificial Intelligence (AI) has been advancing rapidly, tions of such an undertaking. This article also proposes a
and its capabilities can be leveraged for effective roadmap to help the CPI integrate AI into its PSM studies
and optimal utilization of resources in the chemical with ease and efficiency.
process industries (CPI), especially the oil and gas indus-
try. These industries need to conduct several process safety Process safety studies: Current capabilities
management (PSM) studies as part of their risk management Hazard identification is an important step in risk man-
strategy, which requires significant resources, including agement. Accordingly, the process industries, especially
competent personnel, time, and money. oil and gas companies, invest time, effort, and money to
This article explores how different PSM studies can perform process safety studies to identify hazards, manage
benefit from AI. The successful implementation of AI risk, and ensure safety and asset integrity. These studies
depends on the availability of quality labeled data and plant- may include:
specific training of the AI model. In the CPI today, most • hazard and operability (HAZOP) studies
plant data is not available in a labeled format and cannot • layers of protection analysis (LOPA)
be used directly in machine learning algorithms. Therefore, • failure mode and effects analysis (FMEA)
after carefully reviewing the data availability and resource • hazard identification (HAZID) studies
requirements for PSM studies, this article suggests starting • safety-critical equipment (SCE) identification
with rule-based AI implementation as an initial step, fol- • quantitative risk assessment (QRA)
lowed later by the application of machine learning and deep • fire and explosion risk assessment (FERA)
learning concepts. • fire and gas (F&G) detector mapping
Implementing AI in PSM studies can be a game- • occupied building risk assessment (OBRA)
changer, and this article discusses the necessity and limita- • pre-startup safety review (PSSR).

August 2024 aiche.org/cep 27

202408_Ambalavanan Draft 1.indd 27 7/23/24 1:00 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

This article classifies process safety studies into three and protect people, assets, and the environment.
categories — workshop, desktop modeling, and site verifi- PHAs demand a structured thinking process by a multi-
cation — based on the location and resources required, as disciplinary team to identify all potential hazards and recom-
shown in Table 1. During workshop studies, a team compris- mend risk reduction measures. In order to achieve a quality
ing experts from different disciplines within the company outcome from a PHA, a large, talented workforce should
will brainstorm systematically to identify potential hazards be allowed to work for an adequate duration. Sometimes,
and suggest appropriate risk-control measures. An experi- companies find it difficult to allocate quality manpower for
enced facilitator will guide the team through this process. PHAs for long time periods. In a time where quality man-
Predominantly, companies will delegate the desktop model- power is scarce, implementing AI in PHA studies is urgent.
ing studies to an independent third-party consultant who is
a subject matter expert (SME). Pre-startup safety reviews PHA basics
(PSSR), compliance audits, and inspections are usually A PHA is a method used to identify, evaluate, and
checklist-based activities which require site verification to control hazards and risks in complex processes. Risks that
understand the reality of the facility. could adversely affect people, assets, and the environment
Desktop modeling studies. The majority of the cal- are assessed. PHAs involve qualitative or semi-quantitative
culations for desktop modeling studies are performed by assessments where knowledgeable individuals discuss tech-
software, where SMEs have to validate the inputs and infer nical process details to identify hazards. The goal of a PHA is
the results. In the recent past, industry software providers to determine whether sufficient safeguards are in place.
have put in enormous efforts to improve the user experi- Types of PHA methods include:
ence and allow for seamless interaction between software • HAZOP: A widely recognized PHA method, it system-
products. To name an example, Phast and Safeti by DNV use atically examines deviations from the design intent
an application programming interface (API) that allows the • What-if analysis: involves brainstorming scenarios to
user to input a process model created in Aspen HYSYS or explore potential hazards
data formatted in a specific way in Microsoft Excel. Simi- • FMEA: focuses on identifying failure modes and their
larly, F&G detector mapping software Detect3D by Insight consequences
Numerics performs detector optimization to suggest the • LOPA: quantifies risk based on layers of protection
position of F&G detectors using generic algorithms. • Fault tree analysis (FTA): analyzes system failures
With the rapid pace of software development, intra- using a tree-like structure
software communication will be enhanced, thereby reduc- • Event tree analysis (ETA): models possible events and
ing human errors and time requirements and improving the their outcomes
efficiency and accuracy of desktop modeling studies. • Bow tie analysis: visualizes hazards, causes, and
Site verification studies. Engineers visit the site as part consequences.
of studies like PSSR, PSM Audit, etc. Complete automation The most widely used PHA method in the CPI is
of such studies is neither advisable nor practical. However, HAZOP studies. By successfully applying AI to HAZOP
robots, drones, and 360-deg. cameras are being used to studies, the same automation technique can be applied to
gather data from remote or hazardous areas. AI-enabled other workshop studies such as LOPA, bow-tie, and what-if
software analyzes this data to perform inspections. studies, among others.
Workshop studies. Process hazard analyses (PHAs) are
essential for ensuring safety in complex processes. These Background on AI in HAZOPs
analyses involve thorough discussions among experts to The concept of AI has been evolving since the 1950s
identify potential hazards, evaluate the effectiveness of exist- (Figure 1). Any technique that enables computers to mimic
ing safeguards, and recommend risk reduction measures. By human intelligence is called AI, while machine learning
focusing on technical details, PHAs help prevent incidents and deep learning are subsets of AI. Machine learning

Table 1. Process safety studies can be divided into three categories.

Type of Process Safety Study Location and Resources Examples
Multidisciplinary team performs the study HAZOP, LOPA, What-if, HAZID, FMEA,
Workshop
in one room SCE Identification
Independent subject matter experts per-
Desktop Modeling QRA, FERA, F&G Mapping, OBRA
form the study in an office
Engineers perform the study at the site
Site Verification PSSR, Audit, Inspection
in question

28 aiche.org/cep August 2024

202408_Ambalavanan Draft 1.indd 28 7/23/24 1:00 PM

 ▶ Figure 1. Both artificial Pioneering work at 1974: IChemE 1977: U.K. Late 1980s:
intelligence (AI) and Imperial Chemical one-week Chemical 2001: IEC 61882 2016:
Use of
safety course Industries Edition 1 IEC 61882
hazard and operability Industries by software for
Trevor Kletz et al. including Association Edition 2
(HAZOP) studies have HAZOP publication
recording
been around for many
years. However, until 1950 1960 1970 1980 1990 2000 2010
recently, AI has not been
integrated into HAZOP
studies. Artificial Intelligence Machine Learning
Deep Learning

Machine Learning

Supervised Unsupervised Reinforcement

Regression:
Clustering: Ridge
Linear, Polynomial
SVD, PCA, K-Means Regression ◀ Figure 2. This figure
Decision Tree Deep Learning showcases AI techniques and
Random Forest
models. Regression, decision
trees, and random forests can
be deployed to identify causes
Classification: Association Analysis: Lasson Input Output in a HAZOP. K-nearest neigh-
KNN (Consequence Appropri, FP-Growth Regression
Layer
Hidden Hidden
Layer bors (KNN) can define con-
Judgement), Trees, Logistic Layer 1 Layer 2 sequences, and naïve Bayes
Regression, Naïve Bayes Bayesian
Hidden Markov Model
Regression
can perform risk ranking. Deep
(Risk Ranking), SVM learning enables advanced
cause identification.

concepts evolved during the 1980s, and in 2010, deep

Semi-Supervised
learning was developed. In machine learning and deep Learning Topics
learning, the algorithm learns and performs without being Approach
explicitly programmed.
Topic
While a few companies, such as HAZOP.AI, have ven- Modeling
Documents

tured into AI-driven HAZOP, it remains a relatively uncom-

mon practice. AI is not widely adopted in HAZOP due to a Frequency
few key limitations: of Words
• low availability of quality industrial data in a digital
format ▲ Figure 3. This figure shows the process of converting textual plant data into
• lack of appropriate AI tools, such as effective machine machine learning patterns using a semi-supervised approach: collect and pre-
process text, transform it into numerical representations, and manually label a
learning algorithms, to fit the context of PHA analysis small subset. Use semi-supervised algorithms (e.g., self-training) to propagate
• a requirement to train the AI module by providing labels and extract features. Train the model with both labeled and pseudo-labeled
appropriate feedback in order to produce accurate and qual- data, then validate and evaluate it with metrics like accuracy. Deploy, monitor, and
ity results. update the model to maintain performance, leveraging both labeled and unlabeled
data efficiently.
Data. Quality labeled data is the first and foremost
requirement for AI implementation. thinking characteristic of human cognition. As a result, deep
AI tools. Machine learning enables computers to learn learning, with its advanced and distinct modeling capa-
without coding, using algorithms that are designed with bilities, is seen as a more suitable approach for tasks that
the assumption that data is structured in specific ways and demand a deeper understanding and interpretation akin to
adheres to certain rules. For instance, naïve Bayes, a type of human reasoning (Figure 2).
generative learning algorithm, operates on probabilistic data, Training of AI modules. AI requires domain-based train-
while the K-nearest neighbors (KNN) algorithm relies on the ing; in the case of integrating AI in process safety studies,
assumption of proximity, meaning that similar datapoints are the AI would need to be trained on process plant PHAs. One
considered to be close to each other. of the critical bottlenecks to implementing AI in PSM stud-
Therefore, while machine learning algorithms can ies is that most of the data are textual and unlabeled. Some
effectively implement safeguards and predict outcomes, methods already identified by industry experts (1, 2) explain
identifying causes often requires the kind of unconventional how to use text-based data (Figure 3).
Article continues on next page

August 2024 aiche.org/cep 29

202408_Ambalavanan Draft 1.indd 29 7/23/24 1:00 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

Humans play a crucial role in semi-supervised learning available data is validated. Companies can achieve quality
algorithms — a type of machine learning that combines a HAZOP studies if they are ready to spend money and sched-
small amount of labeled data with a large amount of unla- ule appropriately.
beled data during training. Humans can be involved in the However, during revalidation, the operations, mainte-
following processes: nance, engineering services, and technical services teams
• labeling a small subset of data are required to evaluate the current operation and track
• model design and parameter tuning changes in the plant. Sparing key personnel for revalidating
• quality assurance and validation the HAZOP while the plant is running creates a challeng-
• iterative improvement ing situation for management. Hence, the use of AI can
• ethical oversight help overcome the unavailability of key personnel during
• integrating domain knowledge PHA/HAZOP revalidation.
• handling anomalies and outliers.
Overall, human involvement is essential in guiding, eval- Implementing AI in revalidation HAZOP studies
uating, and improving semi-supervised learning algorithms, Implementing AI in a revalidation HAZOP study should
ensuring that they are effective, ethical, and aligned with the only begin after quality labeled data has been gathered and
specific needs of the task. There are two ways to train the AI trained personnel have been assembled. Afterward, follow
model using human intelligence: the plan of action below for embarking on this project.
• Train computer engineers who can develop realistic 1. Select quality HAZOP reports to train the AI module
models and algorithms on plant processes. 2. Integrate existing plant data into the AI module
• Train plant personnel on functionalities, capabilities, 3. Select a suitable AI tool; begin with rule-based
and limitations of the different AI modules, who then can algorithms
train the AI module. 4. Move to advanced AI algorithms.
These steps are described in more detail here.
Design HAZOP vs. revalidation HAZOP studies 1. Select quality HAZOP reports. Training the AI module
A HAZOP study is performed during the front-end engi- will require high-quality HAZOP reports that demonstrate
neering design (FEED) stage and the detailed engineering successful outcomes of team collaboration. If such reports
stage. In addition, as per the U.S. Occupational Safety and are not available, it is recommended that a HAZOP session
Health Administration (OSHA) and other regulatory bodies, be conducted and a report be assembled, which can then be
the PHA study must be revalidated every five years. utilized to train the AI module.
During the design stages, HAZOP studies are conducted 2. Integration of existing plant data. Many companies
with the participation of licensors, contractors, vendors, have started integrating and accumulating data in one place
project management consultants, and company employees. for various benefits. AI implementation may demand a
This ensures that the design is thoroughly reviewed and the few more integrations, like management of change (MOC)
records and inspection reports, to be reflected on the intel-
DCS Database Other Databases
ligent piping and instrumentation diagram (P&ID). Higher
Control
Loop
Safeguard
Status Incident
MOC Reliability-
Related
quality and updated information yield accurate results
Records
Functioning Investigations Info (Figure 4).
Alarms and
Operator Digitalization efforts have been essential in giving
Actions Maintenance Intelligent
Actions During Records P&ID inputs to the HAZOP worksheet. For example, changes
Record Start-ups
in the P&ID and findings in inspection and corrosion
reports will help the AI module identify new causes and
consequences. Likewise, distributed control system (DCS)
trends and alarm history will help the module revalidate
the safeguards.
The AI algorithm will need this plant data to evalu-
ate changes in the causes, consequences, and safeguards
Management Operations For AI
Information System Team of the process, and update the risk ranking of the HAZOP
(Figure 5).
▲ Figure 4. Data gathered from the distributed control system (DCS), mainte- 3. Use rule-based systems before advanced AI algo-
nance and reliability records, incident investigations reports, management of
rithms. The initial step toward implementing AI in the CPI
change (MOC) records, and intelligent piping and instrumentation diagrams
(P&IDs), etc., are accumulated in one place. This will act as a data warehouse for is to start with rule-based systems. Rule-based systems use
AI modules. a set of predefined rules to make decisions. This approach

30 aiche.org/cep August 2024

202408_Ambalavanan Draft 1.indd 30 7/23/24 1:00 PM

 Stay Within
◀ Figure 5. Available plant data
Operation Change like MOC records, DCS history,
Operating Limits
Recognize (From DCS History) and inspection and corrosion
Inspection and
Change Corrosion (I&C) reports are integrated into
MOC Reports Sustain Barriers the AI module to identify the
(DCS History Input) causes, consequences, safe-
guards, and risk ranking, as well
as provide recommendations.

Causes Consequences Safeguards Risk Ranking Recommendation

offers a straightforward, easy-to-adapt beginning point for Design Limits

AI implementation due to its inherent simplicity. Rule- (e.g., Design Pressure)

based implementation serves as a painless starting point,

providing a foundation for building more-complex and
sophisticated AI systems.
Assumptions like causes, consequences, and safeguards
Safe Operating Limits
considered in a HAZOP can be revalidated against real-time (e.g., High-Pressure Alarm)
feedback (through digitalization). Some practical examples
of implementing rule-based systems include:
• Causes: First, the software needs to establish safe
operating limits and equipment design limits (Figure 6). ▲ Figure 6. The rule-based system uses a set of predefined rules to make deci-
Then, it will use inputs such as the history of abnormal sions. For example, after a set of safe operating limits is established, the system
can use inputs like a new working pressure to revise causes and consequences in
operating conditions, inspection and corrosion team reports
the HAZOP.
— such as a reduction in maximum allowable working
pressure (MAWP) — or information about new pressure (a)
sources added through MOC records to revise the causes Relief Pressure = 50 psi
to Atmosphere
within the HAZOP.
• Consequences: During the initial HAZOP, imagine that
inlet Source A of a pressure vessel has a maximum pressure
of 50 psig, as shown in Figure 7a. The cause of blocked Source A

discharge was identified during the initial HAZOP. Since Maximum Upstream MAWP = Cause: Oulet
Valve Blocked
the vessel is designed for 50 psig, there are no dangerous Pressure = 50 psi 50 psi

consequences. The consequence outcome and ranking in

the HAZOP were substantiated by validating the results in
mathematical modeling software like Phast. However, dur- Consequence: No Dangerous Consequence
ing revalidation five years later, MOC records identified that
one more inlet was added to this vessel. This new pump’s (b)
Relief Pressure = 50 psi
maximum shutoff pressure was 100 psi. Since the MAWP is to Atmosphere
50 psi, a blocked outlet valve could potentially have a conse-
quence (Figure 7b). With the introduction of AI, the impact
Source A
of a new pressure source can be seamlessly transferred into
MAWP = Cause: Oulet
Aspen HYSYS and subsequently modeled in Phast to re- 50 psi Valve Blocked
evaluate the consequence outcomes, which can be used to
update the consequence ranking in the HAZOP.
• Safeguards: Imagine that a level control valve was con- Source B
Consequence: Potential
Maximum Upstream
sidered a safeguard in a HAZOP record from 2017. Based on Pressure = 100 psi Vessel Overpressure and Release of
the DCS history in 2022, it is evident that this level indicator Flammable Material
Revalidation: Dangerous Consequence
controller (LIC) is always automated, and if the valve does
not change, then AI software will keep the safeguard as it is. ▲ Figure 7. (a) Since the vessel has a maximum allowable working pressure
(MAWP) of 50 psi, and the inlet has a maximum pressure of 50 psi, there are no
Similarly, a sand detector is considered a safeguard against
dangerous consequences due to a blocked outlet valve. (b) During revalidation,
corrosion in the HAZOP record from 2017. Based on the MOC records showed that one more inlet was added to the vessel, potentially
DCS history in 2022, sand detectors were found to be faulty causing an overpressure scenario.

August 2024 aiche.org/cep 31

202408_Ambalavanan Draft 1.indd 31 7/23/24 1:00 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

Table 2. Revalidation of safeguards can be performed based on real-time data using AI.
Is it still considered a valid
Safeguards From Earlier HAZOP Real-Time Data safeguard during the
revalidation HAZOP by AI?
Level loop in auto mode will regu-
Level control loop is in auto mode and there
Level control late level by opening/closing the
is no physical change to level control valve, Yes
valve level control valve automatically to
which remains in operation as designed
prevent overflow
Sand detectors are faulty on many occa-
Sand detector A sand detector will be present No
sions (availability <40%)
Level alarm Based on MOC records, it is found that there
low (LAL) Operators have 30 min to take is one more outlet, which will reduce the
No
indication with action available time for the operator to respond
operator action from 30 min to 15 min

complete, and the team is comfortable using AI for the

Select the quality Select a suitable AI Tool
revalidation HAZOP, then the team can start implementing
HAZOP reports AI in other PSM studies as well. The team should strive
for continual improvement by providing feedback to the AI
module. Team members can utilize the lessons learned and
Develop the internal team
Start with PHA (HAZOP)
revalidation to gain to provide feedback and
train AI tools to explore the remaining areas of PSM studies.
confidence train the AI module The overall roadmap for implementing AI in revalidation
HAZOP studies and beyond is shown in Figure 8.
Strive for continual Utilize the lessons learned
improvement by providing and train AI tools to
Overall strategy for implementing AI in PSM studies
feedback to the AI module explore other PSM studies In earlier sections, it was demonstrated that implementa-
tion of AI for HAZOP revalidation is the best starting point
▲ Figure 8. This roadmap details the recommended steps for implementing AI for implementing AI in PSM studies. The overall strategy
within PSM studies, beginning with HAZOP revalidation.
for extending AI to all PSM studies is highlighted in this sec-
on many occasions. In this case, the rule-based AI module tion. Key steps include:
will remove the sand detectors from the list of safeguards • embracing digitalization
and re-evaluate the risk ranking. • transitioning from rule-based to machine learning and
Using rule-based AI, HAZOP safeguards can be revali- deep learning
dated based on DCS history, healthiness of instruments, • using explainable AI
change in operating philosophy, MOC records, etc. Exam- • building a skilled internal team
ples of validating different safeguards based on real-time • fostering collaboration between human intelligence
data using AI are provided in Table 2. and AI.
4. Moving to advanced AI algorithms. Rule-based AI and Digitalization. Data is crucial for AI. Every industry or
simple machine-learning algorithms are good starting points processing plant requires its own historical data to develop
for implementing AI in revalidation HAZOP studies. Even and train AI when needed. Even if there are no immediate
though they are relatively simple and easy to adapt, rule- plans to implement AI, it is recommended to begin digita-
based and simple machine-learning algorithms have inherent lization processes such as implementing smart P&IDs, and
limitations. They are highly dependent on the quality of data. centralizing DCS data, records, and incident investigations.
For example, if the selected PHA report fails to identify a Process digital twins can also be a valuable tool. By prepar-
cause, then during PHA revalidation, the rule-based algo- ing in advance, a facility can quickly implement AI when the
rithm will not be able to identify the cause. time is right.
Hence, it is suggested that after successful implementa- Transitioning from rule-based to machine learning and
tion of rule-based AI and simple machine-learning algo- deep learning. We recommend starting with rule-based
rithms, efforts should be made to utilize the benefits of deep system implementation, as described earlier in the article.
learning algorithms, which are able to think like a human For example, AI can be used to revalidate the safeguards and
and beyond the available training data to identify all the pos- consequences based on certain rules. When the rules become
sible hazards scenarios. complex, upgrade to a simple machine-learning model.
Recommended roadmap. After these four steps are Using simple machine-learning algorithms, the AI can be

32 aiche.org/cep August 2024

202408_Ambalavanan Draft 1.indd 32 7/23/24 1:00 PM

 trained to identify causes in the HAZOP based on certain • building skills in the workplace, through identify-
operational consequences like off-spec product. When your ing skills gaps, targeting learning to roles and individuals,
model reaches its limit, upgrade to a more complex one. and more
To perform further in-depth analysis, the AI implementa- • consider setting up an AI center of excellence, which is
tion project team can employ sophisticated algorithms like a dedicated unit that has the goal and vision for implement-
random forest and deep learning. ing AI, identifying new case studies, managing external
Utilizing explainable AI. One of the most significant partners, and sharing best practices.
issues with AI today is the lack of trustworthiness. Although This approach can help ensure that your AI initiatives
we receive results from AI, we often don’t understand how are strategic, efficient, and effective. A suggested process for
those results are determined. For this reason, a new field of building a skilled team is depicted in Figure 10.
research has emerged called “explainable AI” or XAI (3, 4). Fostering collaboration between human intelligence and
The XAI report aims to explain how AI results are obtained, AI. Human and artificial intelligence should work together to
which factors are considered, and which are not (Figure 9). reap the maximum benefits. Table 3 compares the strengths
XAI will assist in making an informed decision. Addi- of human intelligence vs. AI.
tionally, it helps in addressing legal arguments. Hence, it
is important to consider a vendor’s XAI capabilities when Conclusions
selecting an AI implementation. The use of AI can greatly enhance process safety studies,
Building a skilled internal team. To implement AI on a provided that proper digitalization initiatives are in place
large scale, it is crucial to have a competent internal team that and high-quality data are available. To start this journey, it
can handle the technical, operational, and ethical challenges is recommended that the facility employ AI in a revalidation
of AI. Building such a team can be a challenging but reward- HAZOP study. Eventually, this approach can be extended to
ing task. The process involves several steps, including: other PHA techniques and PSM studies as well.
• assessing internal capabilities, including data availabil- Developing a competent internal team and combining
ity, talent pool, infrastructure, and culture human intelligence with AI through explainable AI will
maximize the potential for success of implementing AI in
Plant-Based Data Black Box process safety.
Output
Input
• Raw Data
Unacceptable Article continues on next page
Risk User
• API doesn’t
Interface understand
Data why Identify required competencies related to AI
• Safeguards
• Variables

Plant-Based Data Explainable AI Competency-Based

Output Recruit Staff Train Staff
Input Selection
Unacceptable User
• Raw Data
Risk understands
• API
why and Observations, Assess
Interface
what has Tests, Exams, etc. Competence
Data
• Safeguards been
• Variables considered
Competency-Based Maintain/Enhance
Training Competencies
▲ Figure 9. In some of the AI modules, the correlation between input and output
is a “black box.” In this case, users will not understand how the output is derived
from the inputs. Explainable AI, a recent development, helps the user understand ▲ Figure 10. This flow chart details the process of developing a competent inter-
how the output is derived from the set of input data. nal team for better integration, understanding, and utilization of AI in PHA studies.

Table 3. A comparison of AI vs. human intelligence

Parameter Artificial Intelligence Human Intelligence
Computers are much faster than humans and can Humans take a considerable amount of time to com-
Time efficiency
complete the tasks quickly plete tasks
Humans learn how to manage a wide variety of skills
Influences The conclusion is always based on rules and logic
during their lifetime
Humans have the ability to make decisions based on
Decision Making The time needed to teach a system is high
past experiences
The quality and amount of data decide the quality of
Data Humans depend on more than data
the output

August 2024 aiche.org/cep 33

202408_Ambalavanan Draft 1.indd 33 7/23/24 1:00 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

AMBALAVANAN BALASUBRAMANIAN is a
certified functional safety expert and a certified
automation cybersecurity specialist. He is a
consultant for the Process Safety Management
Team at Kuwait Oil Company, focusing on safety

Expand and
instrumented systems (SIS) and safety lifecycle
activities. He is also actively involved in develop-
ing digitalization strategies and implementing

Build on Your
artificial intelligence in process safety studies.
He has a degree in electronics and instrumenta-
tion engineering.

SRINIVAS GANTI, a process safety engineer, has

predominantly worked in India and the Middle Professional
Network
East and is currently based in Kuwait. With a
degree in chemical engineering, he possesses
extensive expertise in various process studies,
and currently he is working as lead design
engineer — technical safety at Technip Energies.
His proficiency extends to performing quantita-
tive risk assessments (QRAs), fire and explosion
risk assessments (FERAs), hazard identification
(HAZID) studies, hazard and operability (HAZOP) studies, safety
integrity level (SIL) studies, and fire and gas (F&G) mapping studies, Join one of our Forums
among others. Ganti has a vibrant background in process safety
studies across a wide spectrum of industries, including upstream,
midstream, and downstream oil and gas companies, fertilizers, phar-
maceuticals, chlor-alkali, organic chemicals, cement, metallurgical,
and food industries. Notably, he received accolades for his pivotal
role in implementing process safety management (PSM) at one of
the world’s largest liquefied natural gas (LNG) terminals and the
Middle East’s largest greenfield refinery. Additionally, Ganti actively
shares his expertise as a trainer, having successfully trained over
3,000 professionals on process and occupational safety topics.

HAARISH DHARAN is an Analytics Engineer at

NewEngen, specializing in data science and
data engineering. His expertise encompasses
hands-on experience with the Google Cloud
Platform (GCP) and proficiency in Python and
SQL. His interest in analytics has been defined
by a passion for leveraging technology to
unravel data complexities, enabling informed
decision-making and driving business growth.
He demonstrates success in collaborative
settings, where diverse teams are engaged to address intricate
challenges and devise impactful solutions. This inclination towards
collaboration has inspired an exploration of how his skills can be
applied in various industries, including the oil and gas sector.

Literature Cited
1. Ahadh, A., et al., “Text Mining of Accident Reports using
Semi-Supervised Keyword Extraction and Topic Modeling,”
© 2024 AIChE 9072_24 • 05.24

Process Safety and Environmental Protection, 155, pp. 455–465

(Nov. 2021).
2. Ekramipooya, A., et al., “Application of Natural Language
Processing and Machine Learning in the Prediction of Deviations
in the HAZOP Study Worksheet: A Comparison of Classifiers,”
Process Safety and Environmental Protection, 176, pp. 65–73
(Aug. 2023).
3. Phillips, J., et al., “Four Principles of Explainable Artificial
Intelligence,” NISTIR 8312, National Institute of Standards
and Technology (NIST), https://doi.org/10.6028/NIST.IR.8312
(Sept. 2021). Learn more and join at
aiche.org/communities
4. “What is Explainable AI?” IBM, https://www.ibm.com/topics/
explainable-ai (accessed June 21, 2024).

34 aiche.org/cep August 2024

202408_Ambalavanan Draft 1.indd 34 7/24/24 9:11 AM

 Explore the
transformative
impact of data
on Process
Safety.

OCTOBER 29 - 30, 2024

Leonardo Royal Hotel Frankfurt | Frankfurt am Main, Germany

You are invited to attend the 2024 Global Conference on Process Safety and Big Data, taking place
on October 29-30 at Leonardo Royal Hotel Frankfurt in Frankfurt am Main, Germany. We encourage
all professionals from industry, academia, and research institutions to submit their abstract on the
advancement of process safety and big data, for oral or poster presentations.

SESSION TOPICS INCLUDE:

• Business Cases for Digitalization • Predictive Models and Tools for Maintenance and
• Practical Examples/Case Studies Asset Integrity
• Impact of Digitalization on Cybersecurity • Drones/UAVs and Robotics in Support of Safety
• Big Data Analytics for Process Safety • Applying AI in Process Safety — A Case Study
• Data Pre-Processing • Tools to Avoid Human Errors
• Data Lakes / Intelligent Dashboards / KPIs • Use of large language models and NLP in
improving process safety
• Standards and Methods
• Barrier Health Monitoring/Dynamic Risk
• Augmented/Virtual reality
• Permit to Work (Digitalization, Big Data)

Submit an abstract, register and Interested in sponsoring and/or

learn more about #BigData2024 exhibiting at this event?
aiche.org/ccps/bigdata Contact sales@aiche.org

FOLLOW US: #BIGDATA2024

CO-ORGANIZED BY
© 2024 AIChE 9547_24 • 07.24

202408_CCPS_BigData.indd 35 7/16/24 4:48 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

The Industrial
Metaverse for the
Process Industries
Ravindra Aglave ■ Andrea Sassetti ■ Iiro O. Esko ■ Siemens Digital Industries

The industrial metaverse digitally mirrors the process units and

equipment in an industrial facility. Digital twins — the best-known
application of the industrial metaverse — are already transforming how
engineers design and interact with processes.

V
ideo game technology gets more impressive every What is the industrial metaverse?
year, giving us semi-immersive experiences of The industrial metaverse is a world that is always on and
imaginary worlds. Likewise, flight simulators — designed to simulate equipment in an industrial environ-
which are immersive, realistic, and partially physics-based ment. It can be described through the following capabilities:
simulated training environments — have been around for • it helps visualize a system through a virtual world and
decades. Advances in technology now allow the creation gain insights in an immersive environment
of a complete metaverse — a digital reality that mimics • it continuously and interactively evaluates, simulates,
the physical reality. The industrial metaverse is the concept and predicts the behavior of a system
of a digital world that mirrors and simulates real machines • it integrates physical and virtual asset monitoring,
and factories, buildings and cities, grids, and transportation analysis, and management in a closed loop
systems. By seamlessly integrating technologies like cloud • it allows real-time collaboration, including reviews and
and edge computing, artificial intelligence (AI), and digital instant changes.
twins, the industrial metaverse can optimize processes and In this digital environment (Figure 1), problems can be
drive sustainable practices, ultimately shaping the future of found in the real world, duplicated in the digital world, and
real-world systems. the resulting solution implemented in reality. Better yet,
An industrial metaverse is driven by industry needs, digital twins can help engineers uncover problems virtually
current and upcoming technological capabilities, and orga- before they arise in the real world. All of this can happen in
nizational or individual potential for solving engineering, a virtual but connected platform, enabling a whole new level
operations, and safety issues. of collaboration. The industrial metaverse allows people
This article discusses the industrial metaverse’s applica- to work together across countries and continents as if they
tion in the chemical process industries (CPI), primarily focus- were together in the very same room, in front of the same
ing on the best-known technology of the industrial metaverse: machines or objects.
digital twins. Finally, the article highlights the challenges of In the industrial metaverse, engineers can try out new
introducing high-performance computing to current facilities ideas easily and swiftly. Virtual reality acts as an additional
and upskilling the workforce as it stands today. layer of support for people who are working hands-on, on-

36 aiche.org/cep August 2024

202408_AI Special Section_Esko InDesign 2.indd 36 7/22/24 4:44 PM

 site. The metaverse is a virtual realm where operators and tion is proving to be a critical tool as these entities explore
engineers can travel into the past and into the future to better options to improve their processes and systems.
understand problems and processes to find optimal solutions. The exploitation of data, computation, and algorithms
Digital twin basics. The industrial metaverse might has exploded over the past decade. Bigger volumes of data,
sound like science fiction, but it is, in part, already put into as well as high-performance computing and advanced AI
practice across various processes. One of the key technolo- and machine learning algorithms, are changing the way
gies of the industrial metaverse is already very advanced process engineering is performed. A combination of data
and used every day: the digital twin. Digital twins combine and algorithms can be used to develop models for predictive
the virtual and physical worlds to speed up the planning of analytics. The digital twin is one such model, which virtually
plants and process units. By connecting the digital twins to models a specific part of a process or unit operation (3).
their real-world equivalents and leveraging data from both, Depending on the specific aspect or process that needs
engineers can improve operations over the entire lifecycle. to be modeled, more than one type of digital twin may be
A partnership between Siemens and NVIDIA is taking needed, for example: a twin of the process unit, a twin of
digital twins to the next level, creating digital twins that allow the production plant, and a twin of the performance of the
interaction in real time (Figure 2) through photorealistic product while operating in the real world. Such models can
renderings that look and behave like the real equipment (1). be data-based, physics-based, or a hybrid of the two.
For example, the digital twin in Figure 2 simulates a mixing Digital models are limited by their inputs and outputs,
process; operators can adjust the agitation speed in the real since they must be quantifiable, measurable values. Not all
process based on analysis from the digital twin. data provide useful information for process analytics, and,
The digital twin simulates behavior as if it were to occur likewise, not all insightful metrics can be measured. To go
in the real world. For example, if an increase or decrease from data-based to more useful models, physics-based mod-
in temperature is simulated, the digital twin will react to it els that rely on the fundamental principles of mass, energy,
immediately and in the same way as the real asset on the
factory floor or a process plant would. It can also simulate
exactly what happens when an individual part fails or a new
component is installed.

Digital twins as the basis for the metaverse

The transition to a low-carbon world will require
coordination across an industrial enterprise, as well as
maximizing the potential for engineering innovation. Using
new materials and technologies that reduce the number of
manufacturing operations can help companies to be more
▲ Figure 2. The purpose of the industrial metaverse is to solve real-world prob-
energy efficient. Many businesses and governments have lems based on physics-based simulations, automation, control, and performance
already pledged to tackle climate change and are setting evaluation. For example, a digital twin of a tank and agitator can help the user
aggressive goals to meet net-zero targets (2). Digitaliza- better understand current or potential problems that may arise in a mixing process.

Immersive Decision Support

Immersive User Experience
Contextualized immersion to align cross-enterprise design, build, and operational challenges

Generative Artificial Intelligence

Analytics Layer
Ever-evolving design space exploration that strips away bias and drives convergent modeling
The Industrial Metaverse

Cloud Native Reality Capture

Reality Capture Layer Drone scans, photogrammetry, ultrasonic, X-ray, and other sources of visual data for operational
assets and construction progress capture
Computer-Aided Design (CAD) and CAD Aggregation
Designed Reality Model-based systems engineering that aligns and augments cross-discipline design and
change impact analysis
Active Integrating Gateway and REST APIs
Federated Enterprise Data Pre-configured, end-to-end process templates to connect product lifecycle management with common
enterprise systems, including enterprise resource planning (ERP) and manufacturing execution systems (MESs)
Hardware and Edge Applications
Industrial Edge Open, ready-to-use edge computing platform. It encompasses edge devices, edge applications,
and connectivity through an integrated app and device management infrastructure
▲ Figure 1. The industrial metaverse has a wide variety of capabilities that are applicable to various fields within the chemical process industries (CPI).

August 2024 aiche.org/cep 37

202408_AI Special Section_Esko InDesign 2.indd 37 7/22/24 4:44 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

and momentum balances should be implemented (4). or turbulent, and they can include Newtonian and highly
The implementation of a digital twin is an encapsulated viscous non-Newtonian fluids. Chemical reactions may be
(i.e., standalone) software object or model that mirrors driven by electrothermal and electromagnetic forces. Adding
a unique physical object, process, or system. Data from further complexity, these interactions are scale-dependent.
multiple digital twins can be aggregated for a composite Hence, equipment design traditionally has been based on
view across several real-world entities, such as a power rules of thumb, experience, and some level of modeling and
plant and its related processes. For chemical engineers, simulation. However, to remain competitive in the market
this aggregation is often a continuous process manufactur- and consistently produce quality products with higher yield,
ing plant, a fleet of plants, or a process unit. Integrating conversion, and purity, the process industries are actively
and connecting the digital and physical world is therefore researching and developing digital twins.
done through digital twins and aims to create a holistic and In the CPI, physics-based digital twins are gaining
complete industrial metaverse. prominence in virtual equipment development. A digital
Digital twins can also be used to predict and optimize twin needs to be capable of predicting the behavior of its
performance of the physical asset by leveraging various real-world counterpart. These predicted behaviors could
simulations in the virtual world prior to implementing the be operational performance, efficiency, and productivity,
right setup in the real world. While the benefit of simulation as well as reliability and structural integrity. The purpose
is obvious, on the other side of the coin, the simulation mod- depends on the aims of the digital twin’s application. In such
els are sometimes large and resource-consuming to execute, cases, performance must be predicted through methods such
and then not able to provide information in real time. as predictive engineering (5). Using a simulation to create an
Until recently, digital twins could not be deployed at the “executable” digital twin, engineers can determine optimal
operational level, as they could be too complex or slow to equipment designs, performance, or run-time optimization
react to the speed of operation. This is changing quickly as before a unit goes into production.
more computing power is becoming available on the indus- The executable digital twin is a reduced physics-based
trial edge, at offices, and in the cloud. simulation model that is simple enough to be run in near
Executable digital twins. Processes in the CPI are com- real time, while still being accurate enough to represent the
plex. Using simulation to understand this complexity creates physical asset. They can be self-adapting and optimizing
opportunities to deploy insights to your company’s advantage. by continuously gathering data from the field (Figure 4).
Physics-based simulation models at varied length scales allow They can be leveraged by anyone at any point of the
engineers to capture this complexity and use it to explore product’s lifecycle on any certified device, from edge to
design and operating spaces to find optimum conditions. cloud, without the need for heavy simulation software and
Real-world problems may range from nanometer scales infrastructure. This concept is in use today and is allowing
of length to several meters. Similarly, a phenomenon may industries to move toward something that can be described
last from a few femtoseconds to hours. The need to investi- as an industrial metaverse.
gate problems across all lengths and time scales (Figure 3) is
one of the essential requirements for physics-based simula- Beyond digital twins
tion. An engineering team may need a fit-for-purpose tool The photorealistic, physics-based, and real-time digital
for a given scale. If multiple tools are required, integration twin is not the only component of the industrial metaverse,
must be seamless while offering a simple-to-use workflow. but it is possibly the most crucial one. The digital twin is at
This naturally adds to the level of detail required from the the heart of the industrial metaverse; however, a 3D digital
industrial metaverse. twin based on mechanical CAD authoring alone won’t suf-
Process industries involve highly complex physicochem- fice to fully support the metaverse if there is no connection
ical phenomena in terms of multiphase flows, chemical reac- to physics- or first-principles-based models with real-life
tions, heat transfer, and mass transfer. Flows can be laminar constraints or limiting factors. AI assistants that help make
pm nm µm mm m Length Scale

Quantum Molecular Coarse Grained Meso Continuum Equipment Plant

◀ Figure 3. Capturing
physical and chemical
behavior across the
entire range of length
scales involved in a
process is essential
Discovery Engineering in creating a robust
digital twin.

38 aiche.org/cep August 2024

202408_AI Special Section_Esko InDesign 2.indd 38 7/22/24 4:44 PM

 sense of the data across multiple interoperable platforms This virtual world will not be built by just one company
as sources of information are also essential components or a few partners. It is the users who will create it; it is their
of the industrial metaverse (Figure 5). Open and inter- reality and experience, their needs and innovations that will
operable systems that can exchange data and information shape it. Openness and interoperability are therefore the
between systems and different vendors is a key require- basic prerequisites for building the industrial metaverse.
ment. Machine-learning-enabled systems learn as more At the same time, people and companies will need easy
information is made available, whether that information and flexible solutions to develop and cultivate the metaverse.
be measured or simulated. Additional technologies and Large corporations can perhaps afford to develop tailor-
innovations, such as computing power, faster networks, made, individual solutions. Everyone else needs hardware
more powerful AI, and more advanced virtual reality and software solutions that are easy to adapt and preferably
technologies, will be critical in implementing the industrial work as plug-and-play. They will also need a digital business
metaverse. A good overview can be found in a recent MIT platform (e.g., Siemens Xcelerator) where they can find
Technology Review Insights publication (6). solutions, such as a curated portfolio of software and IoT-
AI is a key technological component in bringing speed enabled hardware.
and transparency into organizations that balance information
overloads with workforce constraints. AI assistants working Operational considerations and challenges
on industrial metaverse data can be packaged as industry- Building up current technologies to create an industrial
specific solutions, such as in anomaly detection applications, metaverse is viable and already happening. Through digital
condition monitoring services, or low-latency edge process- twin use cases for process industry scale-up and metaverse
ing units analyzing image or timeseries data. Local and prototyping, the industrial metaverse will become more
remote teams benefit from AI guidance and recommendations understandable to those companies who wish to implement
for improving performance, energy efficiency, and asset man- it. Still, adoption depends on specific industry tendencies
agement. An AI-driven industrial metaverse helps profession- and on the professionals acting in an industry value chain
als decide where to implement AI capabilities and at the same spanning from academia to commercial operations and com-
time more available data makes AI work better. pany leadership.
As an industry example, an AI for a drug repurposing pro- The World Economic Forum has published a landing
gram brings decision transparency and end-to-end automation page on this topic (7) and engages stakeholders to focus on
to a previously manual process. Organizational knowledge two key themes, “Metaverse Governance” and “Economic
can be collated by AI to form an actionable recommendation
system for fast-paced, cost-aware, and on-demand insights on
which therapeutic agents, compounds, and drugs promise the ML/Data Science/ Simulation Files—
Historical Data Machine Order 3D and/or Process
most potential for a company. Raw data to train the AI may Reduction (MOR) Models
come from molecular simulation, formulations, digital twins,
production processes, market demand, and, when available,
from a complete industrial metaverse.
An AI-native approach to the industrial metaverse sup- Executable Digital Twin
ports faster insights, assisted workflows, strategic leveraging Data
Physics-Based
Real Plant
of physical data, efficient information exploration, and more Prediction • Virtual Sensor(s) Models
• Model Predictive Control
tailored solutions. The industrial metaverse benefits from AI,
just like AI benefits from the industrial metaverse, and both
generate value for organizations and individuals.
The technologies listed in Figure 5 and many other inno- ▲ Figure 4. Data flows between virtual models (the executable digital twin and
physics-based models) and physical operations (the plant).
vations are on their way — all of which will pave the way
for a fully realized industrial metaverse. For example, the Real-life Meet industry needs,
internet-of-things (IoT) already connects more than 10 billion Digital Twins Synchronized reduce downtime, and
adapt to supply
Models
devices, a number that is set to triple in this decade. And chain uncertainties
with the rollout of 5G and integrated space-air-ground
networks, high-connectivity, and low-latency networks will Talent, Production,
Interoperable AI-enabled
be available around the world. From virtual and augmented Platforms Systems Customer, Supply
Chain
reality glasses to edge and cloud computing, all the way to
blockchain and AI, more and more technologies are coming ▲ Figure 5. In addition to digital twins, several other components will be essential
together that will enhance the industrial metaverse. to the industrial metaverse.

August 2024 aiche.org/cep 39

202408_AI Special Section_Esko InDesign 2.indd 39 7/22/24 4:44 PM

 SPECIAL SEC TION: AI AN D DIGI TALIZ ATION

Social Value Creation.” A lack of clear vision, strategy, or environments ranging from cloud to office to plant, as
governance — similar to a missing imperative for global well as in the knowledge levels of the current workforce.
collaboration in understanding the industrial metaverse A multi-disciplinary development approach is needed to
— can lead to potential barriers for the adoption of this bridge this kind of a gap.
technology and must therefore be considered. In addition to It is easy to imagine how collaboration in a virtual world
thinking about economic possibilities and considerations, the that has knowledge of the limitations of the physical world
authors encourage readers to explore the technical side of would make CPI operators, designers, and process engineers
the concept, possibly through commercial vendor initiatives faster and more effective. However, the current information
such as the Universal Scene Description (OpenUSD) from architecture is not capable of supporting a fully realized
NVIDIA (10) or an independent nonprofit consortium such metaverse. Likewise, the majority of industry professionals
as The Metaverse Standards Forum (11) that has established are not capable of supporting an industrial metaverse today.
over ten working group topics. To move the industrial metaverse closer to fruition, it
There may still exist a gap between reality and what could be useful to consult the collaborative information tech-
is required from an industrial metaverse. This gap can be nology/operational technology (IT/OT) teams in companies
present at current CPI facilities in networking capabilities, that did not exist ten years ago. It could also prove useful to
at computing capacity near production, latencies between consult network technology vendors, or any of the market
analyst and consulting companies that have modern and
Building an grounded reference architectures available for integrating
Industrial Metaverse Prototype traditional plant OT with modern IT (12).
For workforce development, some countries have pro-
A metaverse prototype is available for industry and
academia to visit at the MxD Innovation Institute
on Goose Island in Chicago, IL (Figure 6). This specific
grams available for upskilling industry professionals toward
the requirements relevant to building and defining the indus-
metaverse prototype contains components covering or trial metaverse. Public and federal programs are available for
including advanced robotics simulation, virtual automation some of the fundamental components of the industrial meta-
commissioning, material deformation analysis, computa- verse in addition to private initiatives. This includes innova-
tional fluid dynamics, industrial electrical design, and more
tion institutes running under Manufacturing USA contribut-
as it evolves, grows, and matures. Building this industrial
metaverse prototype of aggregated digital twins required ing to manufacturing workforce development (13) involving
overcoming constraints such as lack of connectivity in robotics, automation, AI, and many CPI-relevant topics.
industrial networks, lack of interoperability across tech-
nologies, and lack of skills in the workforce (8, 9). Closing thoughts
MxD advances economic prosperity and national secu- An industrial metaverse is driven by industry needs,
rity by strengthening U.S. manufacturing competitiveness
through technology innovation, workforce development, current and upcoming technological capabilities, and
and cybersecurity preparedness. In partnership with the organizational or individual potential for adopting new
U.S. Dept. of Defense (DoD), the MxD Innovation Institute approaches to engineering, operations, and safety issues.
serves as an ecosystem to solve critical manufacturing Earlier this year at the AIChE Spring Meeting (14), the
challenges by accelerating digital adoption, empowering a authors spoke about their individual experiences, reflecting
skilled workforce, and modernizing supply chains. MxD is
also the National Center for Cybersecurity in Manufactur-
on their own potential for driving this technology forward.
ing as designated by DoD. In a similar way, each reader can reflect on their own
experiences and find that they probably are already doing
plenty to help create and form an industrial metaverse that
matches the specific needs of the CPI.
During the AIChE Spring Meeting, another thought
leader spoke from his individual experience; Michael
Lefenfeld, President and CEO of Hexion, Inc., delivered an
impactful message about innovation that may be relevant to
future technologies such as the industrial metaverse. “Open
innovation democratizes ideations, shortens time-to-market,
and reduces cost. It encourages co-creation with academia,
vendors, manufacturers, start-ups, and established compa-
nies. It creates distributed, multifunctional, collaborative
▲ Figure 6. At the MxD Innovation Institute, a work-in-progress industrial meta- forms of innovation. It allows teams to leverage a larger pool
verse prototype is being developed. of ideators, expand resources and experience sets, and brings

40 aiche.org/cep August 2024

202408_AI Special Section_Esko InDesign 2.indd 40 7/22/24 4:44 PM

 new perspectives to the table,” he said (15). “Putting digital
Literature Cited
first enables advanced technologies and analytics to improve
product throughput, reduce formulation cost, reduce siloes, 1. Siemens AG, “Siemens and NVIDIA Expand Collaboration
on Generative AI for Immersive Real-Time Visualization,”
and improve collaboration. It plays a key role in discovery Siemens, press.siemens.com/global/en/pressrelease/siemens-and-
and execution phases, virtualizing insights, accelerates lab nvidia-expand-collaboration-generative-ai-immersive-real-time
work, and in improving prototyping and testing. It acceler- (Mar. 18, 2024).
ates product innovation and market expansion and opens 2. Friedrich, J., et al., “Net Zero Targets: Which Countries Have
doors to products and problem-solving services that were Them and How They Stack Up,” World Resources Institute,
https://www.wri.org/events/2021/6/net-zero-targets-which-
previously impossible” (15). Even though Lefenfeld’s pre- countries-have-them-and-how-they-stack (June 2, 2021).
sentation primarily spoke about innovation and competitive- 3. Pfeiffer, B., et al., “Evolution of a Digital Twin for a Steam
ness, it captured many of the same themes that will be neces- Cracker,” presented at the 24th IEEE International Conference
sary to make the industrial metaverse concept successful. on Emerging Technologies and Factory Automation, Zaragoza,
As the CPI begins to embrace the industrial metaverse, Spain (Sept. 10–13, 2019).
ethical questions surrounding professional and social inclu- 4. Pantelides, C., “Deep Knowledge in the Process Industries,”
presented at the 2nd European Forum on New Technologies —
sion and lack of proper AI education will become more Digitalisation In Chemical Engineering, Frankfurt, Germany
prominent. Food for thought may be procured from, for (Mar. 1, 2019).
example, the Markkula Center for Applied Ethics (16) (part 5. Straw, M., et al., “Predictive Digital Twin for Performance and
of Santa Clara Univ.), or from numerous articles and thought Integrity,” presented at the Offshore Technology Conference,
leaders who have published works on and continue to study Houston, TX (May 2–5, 2022).
6. MIT Technology Review Insights, “The Emergent Industrial
the impact of access to digital technology in human experi-
Metaverse,” MIT Technology Review, https://www.technology
ences, learning, and inclusion (17). review.com/2023/03/29/1070355/the-emergent-industrial-
metaverse (accessed June 20, 2024).
RAVINDRA AGLAVE is the Director for Chemical
7. World Economic Forum, “Defining and Building the
& Process Industry in the simulation business Metaverse,” World Economic Forum, https://initiatives.
at Siemens. In this role, he is responsible for weforum.org/defining-and-building-the-metaverse/home
bringing new modeling and physics knowledge (accessed June 20, 2024).
into simulation tools that can be deployed in the
industry. Aglave joined Siemens in 2012. He has 8. Siemens, “Industrial Metaverse: Hype or Hope?” YouTube,
more than 20 years of experience in engineering www.youtube.com/watch?v=gRHP2IiNXKo (Nov. 28, 2023).
analysis for performance evaluation, troubleshoot-
ing, and problem-solving backed by simulation 9. Siemens, “The Industrial Metaverse,” Siemens, www.siemens.
of various kinds. He has previously held research com/global/en/company/digital-transformation/industrial-
and engineering positions at BASF (Germany), Bechtel (Houston, metaverse.html (accessed June 20, 2024).
TX) and Callidus/UOP Honeywell Technologies Inc. (Tulsa, OK). He
earned his PhD in natural sciences from the Univ. of Heidelberg, 10. Omniverse, “OpenUSD,” NVIDIA, https://www.nvidia.com/
Germany and his MS from the Univ. of Mumbai, and BS in chemical en-us/omniverse/usd (accessed June 20, 2024).
engineering from the Univ. of Pune in India.
11. “The Metaverse Standards Forum,” Metaverse Standards Forum,
ANDREA SASSETTI started his adventure in the https://metaverse-standards.org (accessed June 20, 2024).
digital enterprise in early 2017 initially as an 12. DeBeasi, P., “Reference Architecture for Integrating OT and
innovator and now as an Innovation Manager
within the Siemens Digital Enterprise Lab at Modern IT,” Gartner Research, https://www.gartner.com/en/
MxD in Chicago, IL. Previously, he led a software doc/792965-reference-architecture-for-integrating-ot-and-
development team in the Siemens Manufacturing modern-it (accessed June 20, 2024).
Execution System segment. His background is in
the manufacturing execution space, but his focus 13. Manufacturing USA, “Manufacturing Workforce Develop-
has grown to cover product lifecycle manage- ments,” Manufacturing USA, https://www.manufacturingusa.
ment, manufacturing operation management, com/key-initiatives/manufacturing-workforce-development
automation, and industrial metaverse. He has deep knowledge of the
technologies used in digitalization and the value of applying them
(accessed June 20, 2024).
across industries. Sassetti earned a degree in software engineering 14. Esko, I. O., and A. Sassetti, “A View into Industrial Metaverse,”
at the Univ. of Genoa. presented at the 2024 AIChE Spring Meeting, New Orleans, LA
IIRO O. ESKO is an Industry Manager and Digital (Mar. 24–28, 2024).
Transformation Architect at Siemens. His mandate 15. Lefenfeld, M., “Engineering for Innovation & Growth: Creating
at this company is to maintain an understanding World Class R&D and Manufacturing Organizations Focused
of industry needs on currently available technolo-
gies, and on how to best enable people to use on Accelerating Growth,” presented at the 2024 AIChE Spring
technologies intuitively in design, realization, and Meeting, New Orleans, LA (Mar. 24–28, 2024).
optimization tasks to reach business goals. Esko 16. Markkula Center for Applied Ethics, “Ethics,” Santa Clara
currently works in Houston, TX, having transferred
there from Sao Paulo, Brazil, in 2019. He is a native Univ., www.scu.edu/ethics (accessed June 20, 2024).
of Finland, where he graduated from Tampere 17. NCDIT Division of Broadband and Digital Equity, “What is
Univ. of Applied Sciences (TAMK) with a BS in chemical engineer- the Digital Divide?” North Carolina Department of Information
ing before moving to Brazil and working for 10 years for a software
company acquired by Siemens. Esko is an MBA candidate at the Technology, https://www.ncbroadband.gov/digital-divide/what-
Bauer College of Business at the Univ of Houston. digital-divide (accessed June 20, 2024).

August 2024 aiche.org/cep 41

202408_AI Special Section_Esko InDesign 2.indd 41 7/22/24 4:44 PM

 EN E R GY

Powering the
Transition to Net Zero
with Electric Cracking
Technology
Tobias Sinn ■ Martin Hofstaetter ■ Rainer Kemper ■ Gunther Kracker ■ Linde Engineering

This article provides an update on a technology demonstration of an

electric cracking furnace and considers future integration options into
petrochemical sites.

T
he petrochemical industry accounts for approximately ing technology, the main energy supply for the entire plant
5% of global carbon dioxide emissions (1), with is provided by burning the plant’s methane fraction byprod-
steam cracking processes responsible for a significant uct and natural gas in the cracking furnaces. In addition,
share of this output. Steam crackers generate as much as the required utility steam is generated from waste heat. A
25% of the greenhouse gas (GHG) emissions of the Euro- substantial portion of a steam cracker plant’s Scope 1 GHG
pean chemical industry; globally, steam crackers account emissions are released into the atmosphere via the furnace
for hundreds of millions of tons of GHG emissions per stacks. Therefore, the electrification of cracking furnaces
year (2–4). In light of this, innovative technologies must be represents an opportune starting point for significantly
developed and rapidly made accessible in order to achieve reducing GHG emissions and achieving the sustainability
carbon neutrality and counteract the negative climate goals set out in the 2015 Paris Climate Agreement.
impact of GHG emissions. The electrification of cracking furnaces differentiates
More than 90% of today’s CO2 emissions in modern itself from other steam cracker decarbonization methods
steam cracking plants are owed to the high energy require- mainly through its avoidance of CO2 generation. Other
ments of this process (3). The endothermic conversion that decarbonization methods based on the elimination of the
takes place in the cracking furnaces requires high tempera- generated CO2 emissions, such as fluegas CO2 capture and
tures — for example, the coil or reaction tube outlet has an blue hydrogen firing, all depend on CO2 capture and the
operating temperature of roughly 850°C and a maximum subsequent storage and/or utilization of the captured CO2.
tube metal temperature of up to 1,100°C. The additional In contrast, electric furnaces operate without any additional
energy necessary for the separation section, which separates expenses for CO2 capture and handling. They are particu-
the raw cracked gas into commercial high-value chemicals larly suited for locations with high availability of renewable
(HVCs), is also significant. or low-carbon power supply and long-term low electric
With the currently established conventional steam crack- energy costs.

42 aiche.org/cep August 2024

202408_Energy Sinn Draft 1.indd 42 7/22/24 4:58 PM

 The first part of this article presents the electric crack- applied directly to the process tubes/cracking coils inside
ing furnace technologies that three companies are working the reactor. Joule heating inside the tube wall, caused by the
in partnership to demonstrate (5). Two alternative electri- electric resistivity of the cracking coil material, provides the
cal heating concepts for steam cracker furnaces have been required thermal energy for the chemical reaction.
developed and are currently being evaluated in a large-scale Extensive pilot testing activities and numerical simula-
demonstration plant. tions with different calculation approaches were undertaken
The second part of the article highlights the main differ- to confirm the feasibility of both heating technologies and
ences in the heat duties of an electrically heated cracking fur- to establish a sound knowledge and data basis for the design
nace compared to those of a conventional fired furnace. Said of a demonstration plant and future commercial plants.
differences may significantly impact a plant’s energy balance The technoeconomic analysis conducted by the partners
but, at the same time, reveal a large potential for increas- showed that both heating concepts have their own advan-
ing the plant’s overall efficiency. Selected heat integration tages depending on the specific use case, although they offer
schemes, from the less-complex to the more intricate, enable comparable energy efficiency/power consumption. This is
the adaptation of the eFurnace periphery to project-specific the reason why both approaches are demonstrated alongside
boundary constraints. Special equipment designs for high- each other in one demonstration plant (Figure 2). The sur-
temperature heat exchange from reactor effluent to reactor rounding elements of the existing steam cracker into which
feed are disclosed, and the article provides specific details on the demonstration plant is integrated are shown in the faded
the technical design. background of Figure 2.
The demonstration plant at the BASF Verbund site in
Technology demonstration Ludwigshafen, Germany, has the main goal of demonstrat-
For the core part of the electric furnace, the companies ing continuous olefin production using electric energy as a
BASF, SABIC, and Linde in partnership have developed two heat source, which will serve as a reference for an industrial-
alternative technical solutions, whose principles are shown scale plant. The plant is currently in commissioning/test
in Figure 1. operation and will be operated under industrial conditions.
Both heating concepts of the STARBRIDGE™ technol- Such test conditions will demonstrate the reliability of the
ogy are based on resistive heating, in which electric power materials of construction and custom-made components
is supplied to the resistive heating material via solid electri- for use in this type of high-temperature reactor. Mechanical
cal conductors (unlike inductive heating). The difference completion of the demonstration plant was achieved in Janu-
between the two concepts lies in the location of the resistive ary 2024. The test operation will last for several years, and
heating material within the furnace. the plant will be accessible for reference visits.
In a furnace with indirect heating, an electric current is The demonstration plant has the following key features:
applied to distinct heating elements that are brought to the • Both heating technologies (direct and indirect) will be
necessary temperature by Joule heating. The supply of the demonstrated in separate coilboxes.
thermal energy required in the process tubes is subsequently • Up to 6 MW of renewable power can be drawn from
achieved by primarily radiative heat transfer from the heat- the 6 kV grid and used for preheating the feed and sustain-
ing elements and the insulated walls of the coilbox.
In a furnace with direct heating, an electric current is
Indirect Heating Direct Heating
Raw Raw
Olefins Olefins
Naphtha Aromatics Naphtha Aromatics

850°C 850°C
eFurnace eFurnace

Resistive heating elements are Coils exposed to electric current,

installed at the walls of the furnace; which generates resistive heat
heat transfer to coil by radiation directly in the tube walls

▲ Figure 1. The two diagrams illustrate the indirect (left) and direct (right) heating
principle for electrified steam cracker furnaces and represent the core of Linde’s ▲ Figure 2. A 3D model snapshot of the eFurnace demonstration plant at BASF
STARBRIDGE technology. Ludwigshafen is shown here.

August 2024 aiche.org/cep 43

202408_Energy Sinn Draft 1.indd 43 7/22/24 4:58 PM

 EN E R GY

ing the cracking reaction. technology to full commercial deployment. Some fields of
• 4 m.t./hr of naphtha feed can be processed. knowledge and experience gained from the demonstration
• The cooled cracked gas is routed to the existing BASF plant include:
steam cracker for the commercial production of HVCs. • EPC: computer-aided engineering tools, vendor
• Cracking coils in both coilboxes are of commercial qualification, acceptance testing, assembly procedures, and
dimensions. fits/tolerances
The indicated capacity and the layout of the demo unit • commissioning/startup: sequences and procedures,
were selected to enable reliable upscaling. Major equipment ramping rates
is already close to commercial-size dimensions, or the scal- • technology validation: demonstration plant operation
ing factor is below 10, which is considered in the industry as and upscaling
avoiding significant upscaling risks. • operations: runtimes and yields, transient behavior
The demonstration plant has been designed with maxi- • maintenance: procedures and effort.
mum flexibility in mind. The two furnace boxes can be To ensure the successful implementation of the novel
run in parallel or one at a time, while the other may be in heating technologies, the companies have devised a com-
standby, decoking, or shutdown mode. The decoking process prehensive testing program. Starting in 2024, the program
is carried out similarly to the traditional fired cracking will span several years and encompass a wide range of
furnaces, utilizing a mixture of steam and air to burn coke. objectives. At its core, the program seeks to prove the
Figure 3 provides a functional sketch of the demonstration effectiveness of the two heating concepts. This effort will be
plant’s architecture. supported by simulation models further validated by experi-
The integration of the demonstration plant into the infra- mental results generated during the program.
structure of one of the existing BASF Ludwigshafen steam In addition to proving the effectiveness of the heating
crackers is a key feature. But it also means that the demon- concepts, the program will also generate valuable opera-
stration plant’s engineering, procurement, and construction tional experience. Company personnel will gain expertise
(EPC) process, as well as the plant’s operation, must comply in the operation and maintenance of the heating systems.
with the strict regulations and permitting processes of a This experience will be extremely useful in scaling the
commercial production facility. Consequently, all common technology to commercial application. Moreover, the testing
industry-standard design reviews have been conducted, program will identify ways to optimize and further improve
documented, and followed up on, based on best practices the heating technologies as part of the scale-up.
from conventional process plant execution projects.
In addition, the standard procedures were supplemented Impact of transitioning from fired to electric furnaces
with a newly devised multi-disciplinary review format, Conventional fired furnaces are integral to the opera-
namely an “electric hazard and operability (HAZOP)” tion of steam cracking plants due to their role in hosting the
study, which was conducted with support from certified cracking reaction and their contribution to the plant’s energy
experts. This review specifically focused on novel aspects system. The furnace is divided into a radiant section and
of the implemented technologies and their innate, high a convection section, with 35–50% of the total firing duty
number of electric components. Furthermore, the demon- transferred to the coils in the radiant section and residual
stration plant did undergo the complete certification process fluegas heat being recovered in the convection section
for a commercial unit. (Figure 4, left). Modern fired cracking furnaces use 90–95%
The EPC execution, as well as the operation of the dem- of the fired duty and recover heat from the hot fluegas in the
onstration plant, are providing a wealth of knowledge and convection zone. The recovered heat is used for process and
information that is invaluable in bringing the STARBRIDGE steam heating and preheating of combustion air.

▶ Figure 3. In the demonstration plant, the Cracked Gas

two furnace boxes (direct and indirect heat- High-Value Chemicals
Cracked Gas Cracked Gas Separation in
ing) can be run in parallel or one at a time. Cooling Cooling Existing SC

Direct Furnace Box Indirect Furnace Box

Feedstock and
Utilities from
Existing Steam Feedstock and Utilities
Cracker (SC) Indirect from Existing SC
Direct Preheating
Preheating

Electricity Supply

44 aiche.org/cep August 2024

202408_Energy Sinn Draft 1.indd 44 7/22/24 4:58 PM

 The heat of the cracked gas is recovered in quench support the system if required.
exchangers downstream of the radiant coils. One or more The use of supplementary electrical power for preheating
quench exchangers are utilized to rapidly cool the process purposes enables the decoupling of the preheating and reac-
gas below a critical temperature level to effectively halt the tion sections, as preheating is no longer constrained by the
cracking reactions while simultaneously cooling the gas available fluegas heat. This advantageous feature, combined
for downstream processing. The quench exchanger typi- with the operational flexibility of the coil heating, represents
cally vaporizes (super-) high-pressure boiler feed water (of a significant advantage of the eFurnace technology, allowing
a typical pressure range between 60 bar and 125 bar abso- optimization of furnace run length, product yields, and feed
lute pressure), and in some configurations also preheats the flexibility to the fullest extent.
boiler feed water. The generated steam is superheated in the The combination of eFurnace and heat integration has a
convection zone by the fluegas and distributed to the cracker higher efficiency by eliminating fluegas heat losses, leading
plant’s central steam system for use in heat exchangers and to a decrease in the required heat duty for a constant capac-
steam turbines. ity (see the right side of Figure 4). Remaining losses occur
An electrically heated cracking furnace generates heat mainly in heat exchangers and in the electrical infrastructure,
for the cracking reaction using Joule heating, either directly particularly transformer losses. Compared to a conventional
in the cracking coils or indirectly in heating elements that fired furnace, the total heat losses of the eFurnace and the
radiate heat to the cracking coils. Since an eFurnace can heat integration unit are substantially lower. However, the
transfer at least 95% of the consumed electrical energy to energy surplus from the furnace unit to the plant in the form
the process fluid in the form of heat, waste heat recovery of superheated high-pressure steam decreases in the eFur-
from vent streams becomes unnecessary. Consequently, nace case, as part of the cracked gas heat is utilized (directly
heat for feed and dilution steam heating, boiler feed water or indirectly) for feed and dilution steam heating purposes.
heating, and high-pressure steam superheating needs to
be supplied in a different way. Thus, for full functional- Design aspects for an electrical furnace and
ity, an eFurnace is coupled with a heat integration unit, heat integration system
which replaces the former convection zone and provides Aspects like the overall plant energy balance, the furnace
the required heating (Figure 4, right). With the absence of configuration, and the feed type and flexibility are determin-
fluegas, the system’s inherent heat sources are limited to ing factors in the design of a heat integration system for pre-
cracked gas heat only. Electrical power can be sourced to heating the feed and quenching the cracked gas (Figure 5).
Quenched Fluegas Loss (6%) Quenched
Cracked Gas Cracked Gas
100% Loss (1%) 100%
VHP Steam Steam VHP Steam
Superheat (11%) Steam Superheat (3%)
80% 80%
▶ Figure 4. This figure compares the Feed and Feed and
heat duties of a conventional fired Dilution Dilution
Total Fired Duty

furnace and an electrically heated Steam Steam 60%

Heat Duty
60%
Heating (38%) Heating (38%)
furnace with a heat integration unit.
Fluegas
(left) A conventional fired furnace Loss
recovers heat from the hot fluegas Loss 40% (2.4%) 40%
Casing Heating and Heating and
in the convection zone. Recovered (1%) Endothermic Endothermic
Cracking Cracking (44%)
heat is used for furnace feed and Reaction (44%) 20% 20%
steam superheating. (right) An
electric cracking furnace requires a
heat integration unit to provide feed Tail Convection
Gas Section 0% Feed and Steam 0%
heating and high-pressure steam eFurnace Box
Radiant Section Superheating
superheating.

◀ Figure 5. This sketch

Heat Integration Unit Energy
Separation Section
illustrates how a heat integration
eFurnace Furnace Highly Adaptable to unit connects the eFurnace box
Feed Boundary Conditions with the separation section,
recovering heat from a hot
cracked gas for furnace feed
Feed Preheating Cracked Gas
Furnace preheating, as well as facilitating
Quench
Effluent Quenched the generation of steam and
Cracked Gas energy/heat provision to the
separation section.
Article continues on next page

August 2024 aiche.org/cep 45

202408_Energy Sinn Draft 1.indd 45 7/22/24 4:58 PM

 EN E R GY

up to heavy feedstocks such as atmospheric gas oil (AGO)

Feed-specific coking mechanisms must be or chemically recycled plastic in the high boiling range.
Just as convection section designs are feed-type dependent,
considered when choosing the operating the design of an eFurnace feed preheating system must also
temperature range to maximize furnace run consider the unique characteristics of the feedstock.
length between decoking intervals. As such, the operating temperature range of the cracked
gas quench is dictated by the feedstock. Feed-specific cok-
ing mechanisms must be considered when choosing the
The use of cracked gas heat for preheating the feed and operating temperature range to maximize furnace run length
generating steam must align with the plant’s overall energy between decoking intervals. Figure 7 illustrates the sensitiv-
requirements, mainly considering process steam consumers ity of furnace and quench run length as a function of quench
and turbine drives. The changes in the energy balance for a outlet temperature on the cracked gas side for different
single fired furnace replacement in an existing plant may still types of feedstocks. As depicted in the diagram, heavier
be within the plant’s operational margins, whereas multiple feedstocks are subject to a greater reduction in furnace run
furnace replacements may require simultaneous modifica- length as the quench outlet temperature decreases. At lower
tions in the separation section. temperatures, heavy cracked gas components (originating
The lack of excessive heat from a convection section from cracking high-boiling feedstocks) are susceptible to
and the availability of electrical power from renewable increased condensation (6). Minimum cracked gas outlet
sources increase the incentive to utilize energy recovered temperatures could therefore limit the extent of heat trans-
from cracked gas cooling only for process heating instead of ferred from cracked gas to furnace feed in some scenarios
powering pumps and compressors. The superior efficiency of with heavy feedstocks.
electrically powered drives for rotating equipment, typically
over 90%, compared to conventional condensing steam tur- Selected design approaches for different levels of
bines — typically only 30–40% — encourages the adoption cracked gas heat integration
of such systems to further improve the overall plant effi- This section proposes three novel approaches to design-
ciency, which is particularly interesting for grassroot plants. ing a heat integration unit, using naphtha as a feedstock.
As for the feed flexibility, both electric heating tech- • The first approach — conventional cracked gas heat
nologies are equally suitable, covering conventional gas or recovery — is suitable for a cracker with a high steam
liquid feedstock but also more sustainable feedstock alterna- demand on the separation section, e.g., as in a conventional
tives such as bio-feedstock and plastic waste pyrolysis oil, naphtha cracker (Figure 8a).
enabling further GHG emission reductions on the Scope 3 • The second approach, which has a more advanced heat
side. The customization of an electric furnace to a given feed- integration, is suitable for a cracker with a moderate separa-
stock scenario therefore mostly takes place at the periphery tion section steam demand, e.g., a naphtha cracker with
of the coilbox, particularly in the heat integration system. partially electrified compressor drives (Figure 8b).
Figure 6 illustrates feed-specific boiling curves — start- • The third approach represents an intense heat integra-
ing with full-range naphtha in the lower temperature range tion scheme and is suitable for a low steam demand in the
650
600
HVGO 600
Low Quality Heavy Feedstocks (AGO, HVGO)
Outlet Temperature, °C

500 NG Cond.
ASTM Temperature, °C

B 550
400 High Quality
AGO
500
300 NG Cond.
A
450
200 Full-Range Naphtha
Naphtha 400
100
Ethane
350
0 0 10 20 30 40 50 60
0 20 40 60 80 100
Run Length, days
Volume %
▲ Figure 7. To ensure the minimum required furnace run length, it is mandatory
▲ Figure 6. Typical boiling curves for liquid cracker feedstocks, including different to consider the design requirements associated with different feedstocks. Specifi-
types of natural gas condensates (NG Cond), atmospheric gas oil (AGO), and heavy cally, it is crucial to design the quench in a manner that ensures the system stays
vacuum gas oil (HVGO). within feed-specific operating temperature ranges.

46 aiche.org/cep August 2024

202408_Energy Sinn Draft 1.indd 46 7/22/24 4:58 PM

 separation section, e.g., a naphtha cracker with full com- such steam in steam turbines is not thermodynamically rec-
pressor drive electrification (Figure 8c). ommended as the turbine produces little power compared to
Approach 1. The first approach corresponds to a conven- the electrical duty spent for superheating the steam. Instead,
tional cracked gas quench section with one single quench electrifying the compressor drive and using the cracked gas
stage comprising a linear quench exchanger cooler. Unlike for process heating is far more efficient.
Approach 3, conventional quench exchangers and TEMA- Approach 2. A more advanced heat integration scheme
type heat exchangers are applied in Approach 1. Very high is shown in Figure 8b. The quench section is now split
pressure (VHP) steam is generated on the shell side and into three stages. Hot reactor effluent leaves the eFurnace
routed to a steam drum. The same steam can then be used to coilbox, enters a first conventional quench exchanger, and
vaporize and superheat the hydrocarbon feedstock. generates steam. The rapid quenching of cracked gas below
This allows feed preheating up to 300°C just by steam a critical temperature stops secondary reactions that are
heating, which corresponds to approximately 40% of responsible for coke formation on the quench exchanger’s
the preheating duty. At the same time, 45% of the high- tube walls that results in loss of product yield. In the
temperature heat of the furnace effluent is thus transferred to example of naphtha, this corresponds to a primary quench
the furnace feed. outlet temperature of approximately 600°C. Accordingly, the
The remaining duty from 300°C up to the coil inlet length of the primary quench exchanger (PQE) is far shorter
temperature (~600°C) is covered by an electrical heater. than in the first example.
Electrical heating is preferably applied at elevated tempera- Cooling below 600°C can be achieved with less stringent
tures that exceed conventional heating media such as steam requirements on cooling rate and residence time, enabling
or thermal oil. 55% of the cracked gas heat is available as the use of a countercurrent feed/effluent exchanger in a sec-
a net surplus of VHP steam, which carries energy to the ondary position of the quench train. The selected operating
separation section. The steam surplus accounts for approxi- temperature window of the exchanger prevents pre-cracking
mately 0.5 tons of VHP steam per ton of naphtha feed. The of hydrocarbon molecules on excessively hot surfaces as
specific power demand of the coilbox and heat integration well as condensation coking of high-boiling components on
unit accounts for 1.26 MWh per ton of naphtha. excessively cold surfaces.
Due to the absence of hot fluegas, superheating of Poor heat transfer coefficients are associated with
produced high-pressure steam in an eFurnace might require gaseous fluids on both the cracked gas and feed side, which
electrical heating. Whether superheating is needed depends will require a large heat exchange surface or, alternatively, a
on the purpose of the steam use — for driving steam turbines high pressure drop. To counter this, various ideas have been
(yes) or for process heat consumers (no). However, using implemented to achieve a technically feasible exchanger

Energy (Steam/Power) (a) (b) Steam (c) Steam

Steam

Boiler Feed Boiler Feed Boiler Feed

Feed Water Water Water
Vaporization and SQE
Cracked Gas FEX- Cracked Gas Cracked Gas
Preheating TQE
SQE

FEX-
PQE
PQE eHeater eHeater eHeater
Hydrocarbons Dilution
Steam Superheated Superheated Superheated
Feed Feed Feed

1 2 3
tonSteam 1 2 3
tonNaphtha 0.5 0.2 0.01
Keep Steam Production; Reduced Steam Production; Minimized Steam Production;
MWh
tonNaphtha 1.26 1.06 0.85 Increased Power Demand Reduced Power Demand Minimum Power Demand

▲ Figure 8. The heat integration unit can be designed to meet site-specific requirements. (a) In Approach 1, conventional cracked gas heat recovery occurs in the
primary quench exchanger (PQE). (b) In Approach 2, the quench is split into three sections: the PQE, feed/effluent quench exchanger in the secondary position (FEX-SQE),
and tertiary quench exchanger (TQE). (c) Approach 3 represents an intense heat integration scheme where the feed/effluent quench exchanger is in the primary posi-
tion (FEX-PQE).

August 2024 aiche.org/cep 47

202408_Energy Sinn Draft 1.indd 47 7/22/24 4:58 PM

 EN E R GY

design with optimum performance. 25–35% is transferred via the feed/effluent quench stage.
First and most important, the differential temperature is This translates into a specific net steam surplus of 0.2 tons
maximized by accepting a cracked gas outlet temperature of VHP steam per ton of naphtha, which corresponds to
around 450°C. This limits the amount of transferred heat and approximately 25% of the cracked gas heat. The specific
requires a third quench stage downstream of the feed/efflu- power consumption is reduced to 1.06 MWh per ton of
ent exchanger — an unconventional approach, particularly naphtha due to more advanced heat integration.
for liquid feedstock cracked gas cooling. Approach 3. A higher cracked gas heat integration above
Second, the feed/effluent exchanger (FEX) utilizes a 80% can only be achieved by means of a special gas/gas
straight tube heat exchanger design. A countercurrent flow heat exchanger design. Approach 3, shown in Figure 8c,
configuration routes the feed on the shell side in a no-tube- comprises two cooling stages: a feed/effluent quench
in-window design, preventing hot spots and heat exchange exchanger in the primary position (FEX-PQE) followed by a
in stagnation zones. The cracked gas enters and exits the conventional quench in the secondary position (SQE).
tubes of the exchanger via inlet and outlet cones to avoid The FEX now faces stringent requirements in its primary
tube blockage due to larger coke particles present in the position: The reactor effluent downstream of the cracking
cracked gas. Tube design parameters are similar to conven- coils is at the high temperatures (~800–850°C) at which
tional TLEs. secondary side reactions continue to take place. For optimal
This exchanger enables the direct transfer of high performance and to prevent rapid coking, the residence time,
temperature heat into the feed stream up to temperatures of cooling rate, and pressure drop on the cracked gas side must
around 450°C without compromising product selectivity as match conventional quench exchanger designs.
the pressure drop of the device is not increased. Feed-side Achieving such performance with maximized heat trans-
preheating can be implemented in such a way that pres- fer rates in a gas/gas heat exchanger requires sophisticated
sure drop is not higher than in a conventional convection measures that go far beyond simple exchanger design tactics
zone bundle. such as increasing differential temperature. One key element
The third stage is a conventional tertiary quench is the addition of heat transfer enhancement structures that
exchanger (TQE) that generates steam to the same steam promote turbulence or increase the heat exchanger’s surface
drum as the primary quench exchanger and recovers the area. Application of turbulence-promoting structures on the
residual high-temperature heat. gaseous feed side allows the process designer to increase the
As the FEX outlet temperature is still below the desired rate of heat transfer from the outside surface of the cracked
coil inlet temperature, a final heating step in an electrical gas tube to the flowing feed. Methods like fluid impingement
heater is necessary — similar to Approach 1 but at a reduced or high-shear-inducing geometries have been highly effec-
duty. Moreover, this electrical heater allows the facility to tive for optimal cracked gas quenching (7). A collaborative
compensate for operational fluctuations and to guarantee a effort between Linde, SABIC, ARVOS, and BASF resulted
constant furnace inlet temperature. in promising FEX designs based on said technologies. How-
This scheme allows the process unit to integrate 75–80% ever, these benefits can only be achieved at the expense of
of the high-temperature cracked gas heat. Of that heat, additional pressure drop on the feed side.
Apart from the challenges related to heat transfer, the
(a)
risk of pre-cracking feed hydrocarbons in the vicinity of an
excessively hot exchanger tube surface (hot end) as well as
the risk of condensation coking of cracked gas components
on a too-cold exchanger tube surface (cold end) are other
▶ Figure 9. In a gas/ pressing issues.
gas feed/effluent (b)
exchanger (FEX),
Dividing the device into co-current and countercurrent
a countercurrent stages instead of a single pure countercurrent stage helps
flow design — as to resolve these issues (Figure 9). However, this limits the
shown in (a) — risks maximum achievable feed preheating temperature due to the
pre-cracking feed
hydrocarbons and
limitation on the degree of maximum integrated heat. A final
condensation coking electrical heating step is required (as in Approach 1 and 2)
of cracked gas efflu- (c) and decouples furnace operation from feed preheating.
ent. Instead, a co- and Approach 3 allows the process unit to integrate >90%
countercurrent flow
design — as shown of the available high-temperature cracked gas heat. Of that
in (b) and (c) — is 90%, approximately 45% is transferred via the primary FEX.
recommended. This translates into a specific net steam surplus of 0.01 tons

48 aiche.org/cep August 2024

202408_Energy Sinn Draft 1.indd 48 7/22/24 4:58 PM

 of VHP steam per ton of naphtha, which corresponds to a provide commercial offers for the industrial-scale deploy-
maximum of 5% of the cracked gas heat export. The specific ment of the STARBRIDGE technology by the end of 2024.
power demand of the coilbox and heat integration accounts In the meantime, feasibility studies for integration of the
for only 0.85 MWh per ton of naphtha. eFurnace and process periphery, as well as the balance-
of-plant engineering, are in execution for various clients.
Conclusion and outlook These studies aim to develop the most effective approaches
The electrification of cracking furnaces is set to play a for scaling and implementation to realize the customer’s
key role in the transformation of the steam cracking industry net-zero targets.
toward net-zero GHG emissions. The large-scale demon-
stration of two applicable electrical heating technologies in TOBIAS SINN is a senior process design engineer with Linde Engineer-
an industrial plant setting is a significant milestone on this ing (Email: tobias.sinn@linde.com). He studied at the Technical
Univ. of Karlsruhe and Technical Univ. of Catalonia in Barcelona and
path. A successful demonstration can prove that renewably obtained an engineering diploma (equivalent to a master’s degree)
produced electricity can be used in the future as the main in mechanical engineering with a focus on energy and environmen-
tal technology. He joined Linde Engineering in 2011 and has more
energy source in steam crackers while still ensuring reliable than 12 years of experience in olefin technologies as a process
long-term furnace operation for base chemical production. design engineer for large olefin plants with a focus on the separation
section. Projects include design of new plants as well as complex
The demonstration plant is testing the direct and indirect revamp projects and investigations into alternative olefin routes.
eFurnace technologies in a real plant setting throughout Since 2020, he developed with his team progressive low-emission
plant designs, starting from the concept evaluation phase through
2024 and beyond. Insights and learnings gained from this to the detailed design of innovative plant components. Sinn was
operation will be used to optimize the first commercial involved in the project execution of the first eFurnace demonstra-
tion unit and in the development of advanced heat integration
eFurnace installation. concepts for future upscaled eFurnace applications. Recently,
A heat integration technology portfolio, featuring novel Sinn has extended his process engineering expertise and is also
involved in the design of world-scale blue hydrogen plants. He is a
feed/effluent exchanger designs, has been developed for co-inventor of several patents and patent applications in the field of
plant integration, providing a high flexibility to meet site- low-emission and electric steam cracking furnaces.

specific needs. This broad range of heat integration technolo- MARTIN HOFSTAETTER joined Linde Engineering in 2014 and has
more than 10 years of experience in olefin technologies as a process
gies is suitable for all types of feeds, offering customized engineer for large olefin plants with a focus on the furnace section
energy balance flexibility as well as operational flexibility. (Email: martin.hofstaetter@linde.com). Projects include design of
new plants as well as complex revamp projects and start-up sup-
In terms of commercialization, Linde Engineering will port. Since 2018, he has been closely involved in the conceptual
design and development of the electric cracking furnace, starting
from the initial development phase and accompanying the project as
Literature Cited technology implementation manager toward the design, procure-
ment, construction, commissioning, and start-up of the world’s first
1. Boulamanti, A., and J. A. Moya, “Energy Efficiency and GHG demonstration plant. He is co-inventor of several patents in the
Emissions: Prospective Scenarios for the Chemical and Petro- field of electric cracking furnaces, as well as coauthor of numerous
chemical Industry,” European Commission, JRC Science for publications and conference contributions in that field. Hofstaetter
Policy Report, p. 14, Table 3, https://publications.jrc.ec.europa. graduated with a master’s degree in energy and process engineer-
ing from the Technical Univ. of Munich.
eu/repository/bitstream/JRC105767/kj-na-28471-enn.pdf (2017).
2. Lange, J. P., “Towards Circular Carbo-Chemicals – The Meta- RAINER KEMPER is a technology advisor in the Sustainable Hydro-
morphosis of Petrochemicals,” Energy & Environmental Science, carbon Solutions product line at Linde Engineering in Munich,
14, pp. 4358–4376 (2021). Germany. He has 35 years of experience with Linde, and he has
assumed various roles and responsibilities in this period related
3. Amghizar, I., et al., “Sustainable Innovations in Steam Cracking: to the design of petrochemical plants, including Head of Process
CO2 Neutral Olefin Production,” Reaction Chemistry & Engi- Engineering, Head of Systems Engineering, and Head of Conceptual
neering, 5, pp. 239–257 (2020). Design. His professional experience covers the design of new olefin
plants, revamping of olefin plants, as well as new technology devel-
4. Middleton, J., “Decarbonisation of Steam Crackers,” Decar- opments. Kemper holds a Dipl.-Ing. degree (equivalent to a Master
bonisation Technology, https://decarbonisationtechnology.com/ of Science degree).
article/1/decarbonisation-of-steam-crackers (Nov. 2021).
GUNTHER KRACKER, PhD, is the Head of Product Management &
5. Zellhuber, M., et al., “Electrification of Steam Cracker Furnaces
Cost Estimation for the Sustainable Hydrocarbon Solutions product
– Technology Demonstration and the Route to Commercializa- line at Linde Engineering (Email: gunther.kracker@linde.com). He
tion,” AIChE Paper Number 27f, Ethylene Producers’ Confer- has worked at Linde Engineering since 2004, and his experience
ence, 2023 Spring Meeting and 19th Global Congress on Process in the petrochemical industry includes serving as the lead process
Safety (2023). engineer for major steam cracker grassroot and revamp projects.
From 2013 to 2020, Kracker led the process engineering for petro-
6. Cai, H., et al., “Coke Formation in Steam Crackers for Ethylene chemical plants team. After 2017, his team of process engineers also
Production,” Chemical Engineering and Processing, 41, included the process engineering group for cracking furnace design.
pp. 199–214 (2002). Since 2021, he has led product management and cost estimation
7. Broekhuis, R., “Electrification of Cracking Furnaces,” AIChE for petrochemical plants, and from 2023 onward his responsibility
extended to natural gas plants and petrochemical plants in the new
Keynote 29a, 3rd Topical Conference on Heat Exchangers, 2024 Sustainable Hydrocarbon Solutions product line. Kracker earned his
Spring Meeting and 20th Global Congress on Process Safety MA and PhD degrees in process engineering at the Univ. of Technol-
(2024). ogy in Graz, Austria.

August 2024 aiche.org/cep 49

202408_Energy Sinn Draft 1.indd 49 7/22/24 4:58 PM

 S AF E T Y

Employing
Virtual Tours
to Speed Up
Plant Turnarounds
Venkat Pattabathula, P.E. ■ SVP Chemical Plant Services
Mike Antony ■ Arunkumar Murugan ■ Fitiri

Drones and 360-degree cameras have made it relatively easy to create

virtual plant tours. For one ammonia production site in particular,
virtual tours ensured a safe and cost-effective turnaround.

F
or facilities looking to speed turnarounds, 360-degree and other augmented and mixed-reality applications readily
photography is gaining popularity as the cost of accessible to the chemical process industries (CPI).
cameras decreases and image processing capabilities A virtual tour allows users to virtually walk through a
improve. It used to take months to fully photograph and 3D specific facility — like an ammonia plant — using a series
scan a chemical process facility; with continued improve- of still images, 360-deg. images, and videos. This technol-
ments in 360-deg. imaging technology, this process can now ogy gives engineers and planners the ability to inspect a
be completed in a matter of days. This has made virtual tours developing or ongoing process without having to travel or be

50 aiche.org/cep August 2024

202408_Safety_VP [1].indd 50 7/22/24 4:33 PM

 at the location physically. ment and safety muster points, identify the isolation
This article discusses how 3D scans of CPI facilities points and equipment that require special permits, and
are created and presents examples that show how virtual receive training before entering the plant. Virtual tours
tours can assist in planning and executing construction can be used for operation walk-throughs, work order
projects, plant modifications, and safe turnarounds. An planning, and day-to-day communication. Operation
illustrative case study demonstrates how virtual tours walk-throughs allow employees to navigate the plant virtu-
helped one ammonia plant decrease cost overruns during ally, identifying areas that may require special attention
a turnaround. or maintenance.
Though physical plant visits are essential, virtual tours
Applications in the CPI often function as an intermediate stage for offsite person-
Normal operations. During periods of normal opera- nel. They can offer new and unique functionalities, like
tion, virtual tours can be implemented as a training tool integrated distance measurements between equipment and
for operators to familiarize themselves with plant equip- piping elevations. Work order planning can be done more

August 2024 aiche.org/cep 51

202408_Safety_VP [1].indd 51 7/22/24 4:33 PM

 S AF E T Y

efficiently using a virtual tour, as it accurately represents the turnaround. This often involves opening and inspecting
plant’s layout and equipment. Day-to-day communication the equipment before starting the repair work. Repairs
can be improved by using virtual tours to identify specific frequently require the involvement of the original equip-
locations and equipment. ment manufacturer (OEM) or specialized outside contrac-
Turnarounds. Virtual tours are beneficial during tors. The virtual tour can be a productive tool during this
plant turnarounds for contractor training; contractors process, as it can record various stages of the repairs, guide
can navigate the plant virtually before arriving onsite, future planning, and aid in developing a standard operat-
allowing them to familiarize themselves with their work ing procedure (SOP) for repair and training. Virtual tours
area and identify potential issues. Virtual tours can help are additionally used to document the opening of rotating
ensure crane and overhead movement are done safely equipment and the closing of casings and to monitor and
and efficiently. take inventory of spare parts.
Making scaled drawings from virtual tours is also In addition to documenting reaction vessel repairs,
critical when planning and preparing for turnaround photography can be used to document cooling tower and
work. During turnarounds, virtual tours can document storage tank repairs. For cooling towers, virtual tours
plant modifications, routine maintenance, first-out jobs, can document scaffolding, fan blade removal and repair,
heat exchanger cleaning, catalyst changeouts, and rotat- vibration switch replacement, flow nozzle replacement, the
ing and static equipment inspection and repair, among tower’s interior before and after repair, and the status of the
other applications. cooling water circulation pump. For tank inspection, virtual
• Plant modifications. During a turnaround, plant modi- tours can document the opening of manways and flanges,
fications may be required to address any issues that have safety system placement, tank clearing and washing, weld
been identified prior to the turnaround. Virtual tours can inspection, visual and specialty inspections, repair, and
help document the plant’s current state before any modifi- tank closing.
cations. This documentation provides a reference point for • Catalyst changeouts. Catalysts are used in many
identifying discrepancies or issues that must be addressed chemical processes and may require replacement during
during the turnaround. Once modifications are complete, turnarounds. Virtual tours can document the location and
virtual tours can help ensure the changes meet the plant’s condition of a reactor before and after catalyst replacement,
original design. ensuring that it has been replaced correctly.
• First-out maintenance. First-out maintenance tasks are Digital tours can also help monitor different stages of
the critical-path activities that require a plant shutdown to be a reactor’s turnaround, including scaffolding construction,
completed. They are also high priority and need to be sched- reactor opening, the unloading and loading of catalyst, and
uled and completed before other routine activities. During reactor closing.
turnarounds, first-out maintenance activities can include
equipment repair and miscellaneous hot and cold work. Vir- Equipment
tual tours can be used to document these activities, providing Virtual tours (often called virtual walkthroughs) of
a comprehensive visual representation of the work done. any CPI facility can be constructed with 360-deg. cameras
This documentation can ensure that maintenance activi- and drones. Multiple high-resolution images are shot and
ties have been completed and identify any issues requiring stitched together to provide a seamless high-definition
further attention. representation of the plant. The creators of virtual models
• Heat exchanger cleaning. Heat exchangers are critical frequently use the following equipment to build a 3D scan of
components of ammonia plants and other CPI facilities, a chemical plant:
and cleaning them is crucial during turnarounds. Virtual • 360-deg. cameras. These cameras capture images
tours can be used to document the current visual condition from all directions and stitch them together to create a
of heat exchangers before and after cleaning. This docu- 360-deg. view. They are relatively easy to use and offer
mentation can ensure the completion of cleaning activities high-quality results. However, they require multiple
and identify any issues that may require further attention. shots to capture the entire plant and may not capture
Digital tours also help personnel monitor the opening of detailed close-ups.
heat exchangers, removal of tube bundles from the shell, • 4k 3D cameras. 3D cameras with multiple lenses
cleaning, and closing. and infrared can capture the environment in true 3D.
• Equipment repair. Since large rotating equipment This creates a genuinely realistic depth perception as
cannot be repaired while the plant is in operation and each the transition between the points is as seamless as real
repair requires significant resources and downtime, it is life. Since the data captured is in 3D, this is a time-
customary to change parts and perform repairs during consuming process, but it often produces a significantly

52 aiche.org/cep August 2024

202408_Safety_VP [1].indd 52 7/22/24 4:33 PM

 high-resolution image (Figure 1). ammonia plant turnarounds. The turnaround progress shoot
• Drones. Drones can capture images from different can be initiated on a set timetable or after meeting certain
angles and heights, providing a more comprehensive view of milestones, such as the completion of reactor catalyst load-
the plant (Figure 2). They can also capture detailed close- ing. Progression shots can be captured during various stages
ups and offer flexibility when capturing hard-to-reach areas. of turnaround maintenance, including those listed in the
However, they require skilled operators and may be affected previous applications section of this article.
by weather conditions. In addition to progression shots for active turnarounds
• Laser scanning. Laser scanning uses a laser to capture and plant construction, a few other types of shoots are com-
detailed 3D images of the plant. It provides highly accurate mon when creating virtual tours for operational facilities.
measurements and detailed images. However, it can be time- Custom shoots. Custom shoots are as-built shoots with
consuming and expensive. specific and custom objectives. Custom shoots are made to
meet customers’ specific objectives; this could include train-
Shoot types ing operators in a part of the plant’s operation or shooting an
To be used most effectively during plant construction entire plant with 360-deg. and 3D views to facilitate plant
or turnaround, creating a virtual tour is not always a one- walkthroughs and contractor training before a turnaround.
and-done process. Instead, photos are taken at specific points Panoramas. An aerial sightline panorama captures
in the project to document progress. Thus, the virtual tour panoramic images facing away from the building’s
evolves as the turnaround or construction project continues. exterior at a central point(s) in each elevation. This
Progression shots. Progression shots are captured at represents the outward sightline a person would experi-
regular intervals of construction or during a turnaround to ence from certain floors and units within the building after
track its progress. These progression shots can be used for completion. The client chooses cylindrical panoramas
planning, crane movements, and contractor safety meet- to capture primarily left-right views. They would favor
ings. Generally, progression shots are captured at important correct panning views or spherical panoramas to capture
milestones such as: total panoramic views (i.e., left, right, up, down). The
• site survey at the beginning of the project panoramas are linked to elevation plans and individual
• site preparation floorplans, as necessary, to represent documentation
• foundation construction adequately (Figure 3).
• structural steel erection Slideshows. Slideshows are frequently used to compile
• piping design and fabrication and present images from virtual tours, as they help illustrate
• equipment installation miscellaneous occurrences, plant conditions, or project
• pipe installation progress. Slideshows can include aerial shots and straight-
• electrical and instrumentation installation down satellite views.
• pipe testing
• insulation and coating Methodology for turnarounds
• commissioning and startup If a facility is planning a turnaround and wants to
• finalization and handover. employ virtual tours to help with the transition, partnering
The same method can be used to shoot the progression of with a company that can capture high-res images for virtual

▲ Figure 1. High-resolution images offer a high level of detail and allow engineers ▲ Figure 2. Aerial photographs taken by drones are often used in the composition
and operators to zoom in on specific pieces of equipment. of virtual tours.

August 2024 aiche.org/cep 53

202408_Safety_VP [1].indd 53 7/22/24 4:33 PM

 S AF E T Y

tours is often the first step. The partner company will number within a plot or equipment layout plan (Figure 3)
receive the facility’s requirements and prepare a detailed • back-end editing of images and videos
work scope and methodology. From there, the following • conversion to augmented reality and virtual
steps often take place: reality images
1. The partner company will consult with the client to • website development specific to the project
identify project-specific needs and appropriate documenta- • giving authorized plant personnel access to the final
tion required for the project. document. Authorized plant users will receive their user-
2. Facility personnel will edit and revise plan drawings to name and password, allowing them access to specified
make them ready for site use. documentation. The customer administrator will specify
3. The partner company will analyze the baseline sched- what parts/levels of access the user should receive.
ule to evaluate and estimate appropriate photo set intervals
and durations for each project, and determine the optimal Considerations for construction and
photograph locations/perspectives (i.e., hotspots) based on plant modification
plot plans, equipment layout, etc. During construction, 3D scanning helps analyze prog-
4. The partner company will finalize the camera ress, compare before and after images, and encourage col-
and other hardware requirements, as well as software laboration among remote and onsite teams. This allows for
requirements. easy comparison of the plant’s current state to the original
5. Together, the partner company and the facility will design, indicating any discrepancies or issues that must be
finalize the site workforce requirements as they pertain to the addressed. Virtual tours can also quantify the completion of
virtual tour. specific tasks or projects.
6. A site kick-off meeting will be held, hosted by the Collaboration. Enabling multiple project team members
facility team. to regularly receive updates on field conditions without
7. The facility team will lead a safety orientation accessing the job site can vastly improve team collaboration,
for personnel. and an online platform accessible to all members removes
8. The partner company will devise the exact scope and budget obstacles. Cloud-based access to site conditions may
schedule of the shoot. even eliminate the need for many team members to be physi-
After the initial planning and organizing have taken cally present at a project site. The use of 360-deg. photos can
place, the personnel from the partner company will take the generate considerable time and cost savings, as they grant
appropriate photos, scans, and videos as agreed on during the ability to remotely review and resolve potential issues
the planning stage. After the shooting, the editing and com- in real-time and perform quality inspections of current field
position process will require: conditions. They additionally help reduce the number of
• linking each photo set to the appropriate equipment tag tasks that would traditionally require a site visit. The ability

◀ Figure 3. Panoramic photographs that

capture a 360-deg. view of the facility are
frequently linked to the level or location
within a floor or process plan from which
they were taken.

54 aiche.org/cep August 2024

202408_Safety_VP [1].indd 54 7/22/24 4:33 PM

 to annotate and comment on pictures helps all parties to construction where access to the job site is limited to essential
communicate effectively (Figure 4). With better information workers or stakeholders, some of whom are unable or unwill-
and clearer and more efficient communication, less time is ing to travel due to the job site location.
wasted on unproductive tasks, which helps the project stay With project team members working offsite, communi-
on track and within budget. cation and obtaining information in managing RFIs becomes
Requests for information (RFI). According to a 2013 more difficult. RFIs must be tightly focused to avoid the
Navigant Construction Forum survey of 1,362 projects, there three common problems that arise from so many requests:
are an average of 9.9 RFIs for every $1 million in con- higher project costs, lengthy project delays, and reduced
struction costs (1). An RFI is a formal mechanism through quality. With 360-deg. photography, workers can review
which a contractor can pose a question to the facility owner progress remotely, and issues can be quickly identified,
or construction manager; this process is a routine form of isolated, and resolved without risking quality or adding too
communication in construction projects today to resolve much delay.
ambiguities and clarify necessary details. The average project Scheduling. The construction industry is continually
had 796 RFIs, and each required an average of eight hours demanding improved costs, shorter timelines, and greater
of administrative and technical effort to receive, review, log, efficiencies. Virtual tours can help achieve these goals by
and respond to. This translated to an average project cost of increasing schedule accuracy. Potential scheduling con-
$859,680 (1). The challenge of responding to RFIs has the flicts, delays, or coordination issues can be identified as
potential to become even more significant in the new era of soon as possible. A 3D scan of a site creates a better under-

▲ Figure 4. Images, 360-deg. photographs, and virtual tours can be easily shared among team members, allowing for rapid communication. Images can also be annotated
for clearer collaboration.

August 2024 aiche.org/cep 55

202408_Safety_VP [1].indd 55 7/22/24 4:33 PM

 S AF E T Y

standing of the facility, which can help workers recognize there were 1,008 fatal construction-related injuries in 2020
potential issues that may not have been identified without caused by fires, falls, exposure to harmful substances, and
visual accompaniment. Having access to complete and transportation incidents. The baseline for construction site
accurate job site conditions available to the entire team in a safety risk management is following all Occupational Safety
360-deg. visual environment allows for a better and faster and Health Administration (OSHA) guidelines and regula-
decision-making process. By creating a historical progress tions. Virtual tours of ammonia plants, as well as other CPI
dataset, the project team can update future activities based facilities, can be used to identify the equipment locations
on past productivity. with electrical area classifications, as well as to train opera-
The schedule can be tied to budget and payment applica- tors and contractors.
tions. Using the current schedule and percentage of job Mitigating project management risks. In the CPI, risk
completion, the project team can forecast upcoming activi- doesn’t only refer to safety. Anything that can prevent a
ties and subsequent contractor payments. project from being completed to spec, on time, and on
budget is a risk. Poor planning and budgeting, resource
Risk management management issues, and time management issues are just
Risk management in construction and turnaround, some of the risks that can impact project outcomes. By fol-
especially in ammonia plants, may include process safety, lowing project management best practices, a project team
project management, financial, environmental, and legal can mitigate these risks and increase the facility’s chances
risks. Additionally, the cybersecurity of the virtual tour data of completing the project successfully. One best practice
is a critical consideration. is to digitalize your project management workflows using
Process safety. Despite significant safety improvements cloud-based software. By moving away from manual pro-
over the past century, construction remains a dangerous cesses and cumbersome spreadsheets, project managers can
industry. According to the U.S. Bureau of Labor Statistics, save time, reduce their workload, and identify risks before

Yara Pilbara Virtual Tour

T he Yara Pilbara facility, located on the Burrup Penin-
sula in Australia, creates ammonia for fertilizers. To
efficiently prepare for and execute an upcoming turn-
around, the facility contracted Fitiri to implement a virtual
tour using 360-deg. cameras and drones (Figure 5). This
involved scanning the entire facility at the ground level,
as well as taking aerial photographs. In addition, the
digital recreation of the Yara Pilbara facility included five
separate elevation levels.
With the help of virtual tours, Yara could efficiently
collaborate with internal and external teams, saving sig-
nificant time and money on travel expenses and commu-
nication. Some of the challenges during the turnaround ▲ Figure 5. The developers behind the creation of Yara Pilbara’s virtual tours
used drone photographs like this one to create a thorough, multi-level digital
included equipment handover, simultaneous operations
recreation of the facility.
(SIMOPS), and plant startup delays due to delays in main-
tenance activities on equipment like the syngas compres-
sor, ammonia refrigeration compressor, and reformed gas
waste heat boiler. By properly allocating manpower and
effectively planning using virtual tours, Yara was able to
overcome these challenges.
The virtual tour helped identify the equipment
locations with electrical/hazardous area classification
(Figure 6). Another advantage of using a virtual tour
was reducing the time delay, which in turn assisted
in controlling cost overruns. The facility was able to
successfully complete its turnaround with no major
safety issues or cost overruns, and it experienced
only a minor delay compared to the committed turn- ▲ Figure 6. This ground-level photograph, which shows electrical equipment,
around duration. is part of the virtual tour of the Yara Pilbara facility. This virtual tour allowed the
facility to reduce its turnover time and avoid cost overruns.

56 aiche.org/cep August 2024

202408_Safety_VP [1].indd 56 7/22/24 4:33 PM

 they get out of hand. For example, virtual tours enable proj- tion with certificates issued by cybersecurity firms. Within the
ect managers to complete inspections and respond to RFIs database, password data may be hashed and salted to impede
without going onsite. brute force or reverse lookup attacks. The application should
Managing financial risks. Now more than ever, many run in a load-balanced, high-availability cluster with auto-
stakeholders are focused on mitigating the financial risks matic failover and high redundancy to reduce risk.
impacting turnaround projects. Macroeconomic issues like
inflation and supply shortages have an outsized impact on Final thoughts
the construction industry, where delays are likely to com- As the famous saying goes, a picture is worth a thou-
pound and lead to cost overruns. While project managers sand words; nothing can convey the details of the plant to
have little control over the economy, there are ways to man- an offsite stranger better than a virtual tour. High-resolution
age financial risk and create more wiggle room when dealing photography and 3D plant scans help eliminate last-minute
with these uncertainties. surprises and avoid costly mistakes. A virtual tour is easy to
One cost that’s often higher than necessary is rework. view and can be accessed through any web browser, so the
The best way to mitigate the financial risk of rework is with user need not undergo complicated training or install any
project progress photography. By photographing every part additional software.
of the build at every stage of turnaround, the project team Virtual tours provide a valuable tool for improving
can avoid issues that necessitate tearing up slabs or walls. safety and reliability during turnaround services and normal
Managing environmental risks. Environmental risk in operations. By capturing 360-deg. video, virtual tours can be
ammonia plant turnarounds involves two unique risk areas: used for a wide range of activities, including plant modifica-
risks that the project poses to the environment and risks that tions, routine maintenance, first-out jobs, heat exchanger
the environment poses to the project. There is nothing project cleaning, catalyst changeouts, and rotating and static equip-
managers can do to stop the weather; however, they can ment inspection and repair. As the technology continues to
take steps to mitigate the damage caused by flooding, wind, improve, virtual tours will continue to provide a compre-
moisture, and other environmental factors. Keeping consistent hensive visual representation of work done and ensure that
and comprehensive documentation is a best practice. Visual plants operate as originally intended.
documentation makes it easier and faster to assess damage
and file insurance claims following a natural disaster. VENKAT PATTABATHULA, P.E., is the Principal Consultant at SVP
Managing legal risks. Any construction project will Chemical Plant Services. Prior to this position, he was the global
involve many parties, from the clients and general contractors ammonia technology manager for Incitec Pivot Ltd., where he
supported manufacturing facilities in Australia and North America.
to the subcontractors and tradespeople employed throughout His specialties include process design, project development,
the build. The sheer number of intersecting interests makes commissioning, plant operation, process safety management,
and manufacturing excellence programs. Pattabathula has been a
construction a legally risky field. Contract disputes are a member of AIChE since 1989 and has been an elected member of
major source of legal risk in construction. To manage this AIChE’s Ammonia Safety Committee since 2005. He is a chartered
professional engineer of Engineers Australia and a registered
risk, it is essential to use specific, attorney-approved contract professional engineer of Queensland. Pattabathula earned an
language, retain access to legal counsel, and maintain com- MTech degree in chemical engineering from the Indian Institute
of Technology.
prehensive visual documentation of all onsite work.
MIKE ANTONY is the process application engineer at Fitiri. He has over
Data security. The virtual tour’s data security is critical 40 years of extensive experience in education, agriculture, informa-
for process plants, as unauthorized access to the virtual tour is tion technology, and chemical engineering. In the agricultural field,
he specializes in regenerative farming and sustainable aquaponics.
similar to unauthorized entry into the plant facility. The own- In information tech, he specializes in improving operational safety
ers and composers of virtual tours should ensure that users and productivity in the chemical process industries (CPI), mobile
technology, industrial internet of things (IIOT), virtual reality, artificial
can only access the system through their sign-in credentials intelligence (AI), and blockchain. In the field of chemical engineering,
and view the areas that are assigned to them. The level of he specializes in hydrogen, methanol, ammonia, and urea, as well
as nitrogen, phosphorus, and potassium (NPK) fertilizers. He holds
access should be role-based, where some personnel have a postgraduate degree in chemical engineering from the Indian
strictly read-only access while others have read/write rights. Institute of Technology.
All data in transit should be secured with 256-bit secure ARUNKUMAR MURUGAN is the Director of Software Engineering at
sockets layer and transport layer security (SSL/TLS) encryp- Fitiri. He focuses on improving human factors in the workplace using
digital transformation. He has extensive knowledge in developing
enterprise-grade software systems using web, mobile, cloud, and
machine learning/artificial intelligence (ML/AI) technologies. He
Literature Cited has also worked with ammonia plant operators, supervisors, and
managers. Murugan additionally contributed to the development of
1. Hughes, N., et al., “Impact and Control of RFIs on Construc- PlantMS, a comprehensive plant operations management solution.
tion Projects,” Navigant Construction Forum, McLean, VA He holds a postgraduate degree in information systems and an MBA
(Apr. 2013). from Lamar Univ.

August 2024 aiche.org/cep 57

202408_Safety_VP [1].indd 57 7/22/24 4:33 PM

 From the Boardroom to the Control Room

SAVE THE DATE

Sept. 24-26, 2024

Join us this September to learn from peers and industry experts from all levels of manufacturing on how
advanced manufacturing is transforming the process industries. This robust program will feature two tracks
focused on technical innovations to improve plant efficiency and effectiveness, as well as next-gen
business trends in plant development, sustainability, and resiliency.

Session Topics Include: AMPc Program Committee Members Include:

• Building the Plant of the Future • Chair: Sachin Padhye, Infosys Consulting
• Making your Existing Plant/Processes • Ibrahim Al Syed, Celanese
More Sustainable • Stefanos Baratsas, ExxonMobil
• Supply Chain & Supply Chain Resilience • Katelyn Kelsey Dubois, Dow
to Support Advanced Manufacturing
• Nitish Mittal,
and Processing
ExxonMobil
• Data Implementation for enabling
• Ignasi Palou-Rivera,
High Performance Computing in
RAPID®
Advanced Manufacturing
• Mike Rinker,
• Uses of Generative AI in Advanced Manufacturing
U.S. Department
and Processing
of Energy
• Control Room Design

September 24-26 • Houston, TX • aiche.org/AMPc

PRESENTED BY

© 2024 AIChE 9357_24 • 5.24

202408_AMPC.indd 58 7/19/24 1:52 PM

 I N STI T U TE N E WS

Alper to Deliver 2024 Acrivos Professional Progress Award Lecture

al S. Alper, the Kenneth A. Kobe Professor will discuss how advances in metabolic engineer-
H of Chemical Engineering at the Univ. of
Texas at Austin (UT Austin), will present the
ing and synthetic biology are improving the valo-
rization and remediation of waste, and presenting
American Institute of Chemical Engineers’ new routes for environmental stewardship and
(AIChE’s) Andreas Acrivos Professional Progress sustainable industrial bioproducts.
Award Lecture for 2024. As the 2023 recipient Alper is the Principal Investigator in UT
of the Acrivos Award for Professional Progress Austin’s Laboratory for Cellular and Metabolic
in Chemical Engineering, Alper will discuss his Engineering, where his research group applies
research during the 2024 AIChE Annual Meeting and extends the approaches of related fields such
on Oct. 29 in San Diego, CA. as synthetic biology, systems biology, and protein
The Andreas Acrivos Award for Professional Progress engineering. His research is documented in more than 150
in Chemical Engineering recognizes the contributions of articles and eight book chapters, and he has delivered nearly
a chemical engineer in their early career. The award is 200 invited lectures at institutions worldwide.
endowed by the AIChE Foundation and named in honor of A Fellow of the American Institute for Medical and Bio-
Andreas Acrivos, Professor Emeritus at The City College of logical Engineering and a member of the National Academy
New York, who pioneered the field of fluid dynamics and is of Inventors, Alper has organized conferences for AIChE’s
an influential leader in the chemical engineering profession. Society for Biological Engineering and International
The associated lecture is a highlight of each year’s AIChE Metabolic Engineering Society. He is a chemical engi-
Annual Meeting — a foremost educational forum for neering alumnus of the Univ. of Maryland, College Park,
chemical engineers working in research and development. and the Massachusetts Institute of Technology, where he
(See related article below.) earned his BS and PhD, respectively. He was a postdoctoral
In his Professional Progress Award lecture, entitled research associate at the Whitehead Institute for Biomedical
“Engineering Biology to Develop a Taste for Waste,” Alper Research and at Shire Human Genetic Therapies.

Chemical Engineering Reimagined: Registration Opens for

2024 AIChE Annual Meeting and Co-located Conferences
his Fall, chemical engineering The San Diego meeting site boasts
T researchers, visionaries, and
early-career professionals will share
several co-located AIChE-sponsored
conferences. These include:
their ideas and chart a course for • Conference on Constraint-
the profession’s future when AIChE Based Reconstruction and Analysis
holds its 2024 Annual Meeting in San (COBRA; Oct. 24–26)
Diego, CA, Oct. 27–31. The associated Annual Student • Bioengineering and Translational Medicine Confer-
Conference is set for Oct. 25–28. ence (Oct. 27–28)
Returning to the Golden State for the first time since • Cell Therapies and Immunotherapy Conference
the Fall 2016 conference in San Francisco, this year’s (Oct. 28–30)
AIChE Annual Meeting will focus on the next steps for • Food Innovation and Engineering Conference
the evolving chemical engineering profession — exempli- (FOODIE; Oct. 31–Nov. 1).
fied by a collaboration between members of AIChE’s Fel- Registrants for the Annual Meeting can take advantage
lows Council and representatives of the Science History of package pricing by also registering for one or more of
Institute, who are organizing a discussion about the roles those adjacent specialty conferences at a reduced rate.
that chemical engineers are destined to play in the unfold- Expanded coverage of the Annual Meeting’s featured
ing energy transition. events, major lectures, and the technical program will
Registration for the 2024 AIChE Annual Meeting is appear in the September issue of CEP. For up-to-date
now open, with reduced rates for the professional confer- information about the Annual Meeting, its program, and
ence in effect through Sept. 16. registration, visit www.aiche.org/annual.

August 2024 aiche.org/cep 59

202408_InstituteNews_July21.indd 59 7/22/24 11:26 AM

 I N STI T U TE N E WS

Robert A. Brown Named as AIChE’s Schowalter Lecturer for 2024

obert A. Brown, President Emeritus, Profes- while underscoring the central importance and
R sor of Engineering, and Professor of Com-
puting and Data Sciences at Boston Univ., will
interrelationship of teaching and research, with
a focus on the interconnections spanning disci-
present AIChE’s William R. Schowalter Lecture plines and programs. His key academic initiatives
on Oct. 30 at the 2024 AIChE Annual Meeting in have included the establishment of the National
San Diego, CA (Oct. 27–31). Emerging Infectious Diseases Laboratory and
The lectureship is endowed by the AIChE the Rajen Kilachand Center for Integrated Life
Foundation and named for William Schowalter, Sciences and Engineering, among others.
Professor Emeritus of Chemical Engineering at Prior to joining Boston Univ., Brown served as
Princeton Univ. — a pioneer in fluid mechanics provost and Warren K. Lewis Professor of Chemi-
and an influential academic leader. Reflecting cal Engineering at the Massachusetts Institute of
Schowalter’s contributions to chemical engineering, the Technology, where he was on the faculty from 1979 to 2005.
lecture’s subject matter alternates between topics related to He has published more than 250 papers in areas related to
fluid mechanics — typically delivered by a speaker from mathematical modeling of transport phenomena in materials,
academia, and topics of general interest to the profession — and he served as executive editor of the journal Chemical
typically presented by a speaker from industry. This year’s Engineering Science from 1991 to 2004.
lecture focuses on topics of general interest. Brown is a member of the American Academy of Arts
In Brown’s lecture, entitled “Organizing Universities in and Sciences, the National Academy of Sciences, and the
the Century of Convergent Research,” he will discuss how National Academy of Engineering — the latter of which
emerging societal needs and advancing technology call for granted him its Simon Ramo Founders Award for his con-
and enable interdisciplinary collaborations in research and tributions to engineering and his leadership in transforming
problem solving, and necessitate that research institutions disciplines in institutions. He earned BS and MS degrees in
restructure their traditional approaches in order to adapt to chemical engineering at the Univ. of Texas at Austin, and a
these opportunities and demands. PhD in chemical engineering from the Univ. of Minnesota.
Brown served as President of Boston Univ. from 2005 Brown has a forthcoming book, and he is a frequent
to 2023. During his tenure, he emphasized the continued contributor to the literature on leadership of modern
improvement of academic quality across the university, research universities.

Center for Hydrogen Safety Opens Nominations for Zalosh Safety Awards
he Center for Hydrogen Safety (CHS) — an AIChE AIChE’s Center for Chemical Process
T technical community and a global organization leading
the safe advancement of hydrogen as the next-generation
Safety (CCPS), he contributed exten-
sively to the annual Loss Prevention
fuel — has created a slate of prizes that will spotlight Symposium, and he was the principal
significant contributions to safety practices within the author of the first edition of the CCPS
hydrogen industry. book “Guidelines for Safe Handling of
The Robert Zalosh Hydrogen Safety Excellence Awards Powders and Bulk Solids.”
will recognize contributions in three categories of achieve- The individual Zalosh Hydrogen
ment: individual, project team, and organization. Nomina- Safety Excellence Awards will rec-
▲ Zalosh
tions for the inaugural awards are due Oct. 31, 2024, with ognize people who have made major
the first prizes targeted for presentation at a CHS or AIChE contributions to hydrogen safety during their careers. The
event in 2025. team project prizes will honor groups that have completed
The awards are named for Robert Zalosh (1944–2023), particularly impactful initiatives. The awards for organi-
an AIChE Fellow and Professor Emeritus at Worcester zations will highlight academic institutions, companies,
Polytechnic Institute, and commemorate his lasting impact government entities, or other organizations that have dem-
on hydrogen and fuel cell safety. Zalosh was an incident onstrated a commitment to hydrogen safety.
investigator for the U.S. Chemical Safety Board and a To file a nomination for an individual, project team, or
member of the Hydrogen Safety Panel that assisted the U.S. organization, visit the CHS website at www.aiche.org/CHS.
Dept. of Energy and the Center for Hydrogen Safety. Within For more information, email CHSawards@aiche.org.

60 aiche.org/cep August 2024

202408_InstituteNews_July21.indd 60 7/22/24 11:26 AM

 ing
Connecting
COMMUNITIES
EmpowReErSS Transforming
FUTURES
PROG

Access a breadth of resources and Sharpen your professional skills and keep pace
expertise you need at every stage of with the evolving skills required in your industry.
aiche.org/ili/academy
your professional journey
Find your community, share your interests, and pursue
Advance your career with access to resume-building technical, professional, or societal development.
assistance and workshops that prepare you for interviews, aiche.org/community
job fairs, and job boards. aiche.org/careerdiscovery &
careerengineer.aiche.org Discover our Affinity Program featuring comprehensive
insurance solutions for health, life, or professional liability
Expand your network and connect with like-minded and exclusive discounts across various services and
individuals through meetings and conferences. products relevant to your professional and personal needs.
aiche.org/events
Stay in the know with access to curated technical content.
aiche.org/publications Join or Renew your Membership Now

© 2024 AIChE 9521_24 • 07.24

Virtual Career Fair will Connect Chemical Engineers and Employers, Sept. 10
ob seekers in the chemical engineering arena, along Fair offers many of the familiar elements of an in-person
J with representatives from many of the companies and
industries that employ them, will discuss job opportuni-
employment fair, but eliminates the restrictions imposed
by geography and travel expenses. “AIChE members, in
ties and make employment connections when AIChE particular, will benefit from priority placement at the fair,
presents the latest event in its Virtual Career Fair series. with members’ résumés given the most prominent posi-
The online fair is slated for Tuesday, Sept. 10, 2024, tions in the fair’s résumé database,” Young notes.
from 4:00 PM to 7:00 PM EDT. Job seekers, regardless of their employment status or
Chemical engineers at all career stages — includ- location, can create a profile, submit their résumé, read
ing new graduates — are invited to register for the job postings at virtual company booths, and sign up for
event, where they will interact with employers from a a live chat with recruiters. Employers can register for a
cross section of the global chemical process industries. virtual exhibit booth, post available jobs, review candi-
The Virtual Career Fair is hosted by AIChE’s Institute date profiles, and chat with job seekers in real time.
for Learning and Innovation (www.aiche.org/ILI), as a Engineers who wish to register for and participate
component of its lifelong career support for chemical in AIChE’s Virtual Career Fair will find instructions at
engineers across sectors. www.aiche.org/careerfair. Employers can use the same
According to Wendy Young, Senior Director of the registration platform to post their jobs and set up Virtual
Institute for Learning and Innovation, the Virtual Career Career Fair booths.

August 2024 aiche.org/cep 61

202408_InstituteNews_July21.indd 61 7/22/24 11:26 AM

 EQUIPMENT ADVERTISERS’ INDEX
HEAT EXCHANGERS
2024 AIChE Annual Meeting BC
www.aiche.org/annual
▼

Liquid Cooled

2025 AIChE Spring Meeting and 21st GCPS 26

www.aiche.org/spring; www.aiche.org/gcps
▼

Air Cooled
Aon IBC
www.aicheinsurance.com
FOR GASES & LIQUIDS!
Talk Directly with Design Engineers!
Blower Cooling Vent Condensing Armstrong Chemtec Group IFC
www.armstrong-chemtec.com

(952) 933-2559 info@xchanger.com

Global Conference on Process Safety & Big Data 35
www.aiche.org/ccps/bigdata
Join the Team
that’s Doing a Intelligen 63
World of Good www.intelligen.com

ioMosaic 5
www.iomosaic.com

Xchanger 62
https://xchanger.com

Kathy Chen
Senior Engineer – ADVERTISING AND RECRUITMENT
Retired
Bethlehem Steel
Tania Babiuk
The AIChE Foundation gratefully acknowledges
Kathy Chen for her gift to the AIChE Legacy Email: tanib@aiche.org
Society ensuring the future of the profession.
Chemical engineers make a world of difference.
Learn more and get involved at
www.doingaworldofgood.org IN THE NEXT ISSUE
Lifecycle Thinking for Next-Generation Chemical Engineering
The thinking underpinning lifecycle assessment (LCA) has proven to be flexible
and adaptable enough to guide analyses of emerging technologies, social
impacts, and carbon emissions accounting approaches. Advancements in the
application of LCA and its principles will be essential for chemical engineers to
understand and leverage to meet the challenges of the energy transition.
© 2024 AIChE 9037_24 • 1.24

62 aiche.org/cep August 2024

202408_Marketplace.indd 62 7/22/24 4:59 PM

202408_Intelligen.indd 63 7/16/24 4:56 PM
 CALEN DAR S

Oct. 14–16, 2024 • International Conference on Apr. 6–10, 2025 • AIChE Spring Meeting &
CRISPR Technologies 21st Global Congress on Process Safety
AIChE MEETINGS Bahia Resort Hotel • San Diego, CA Hilton Anatole • Dallas, TX
For information and registration
details, visit www.aiche.org/ Oct. 24–26, 2024 • Conference on Constraint- May 5–8, 2025 • Offshore Technology Conference
conferences or call Customer Service Based Reconstruction & Analysis (COBRA) NRG Center • Houston, TX
Omni San Diego • San Diego, CA
at 1-800-242-4363 or 1-203-702-7660
(outside the U.S.). May 13–15, 2025 • Commercializing Industrial
Oct. 25–28, 2024 • AIChE Annual Student Biotechnology
Conference Massachusetts Institute of Technology • Cambridge, MA
Aug. 6–7, 2024 • International Mammalian San Diego Convention Center • San Diego, CA
Synthetic Biology Workshop (mSBW)
Boston Univ. • Boston, MA Oct. 27–28, 2024 • Bioengineering & Translational
Aug. 6–7, 2024 • Chemical Ventures Conference
Medicine Conference
Hilton San Diego Bayfront • San Diego, CA
AIChE ACADEMY
mHub • Chicago, IL
Public Courses
Oct. 27–31, 2024 • AIChE Annual Meeting
Sept. 4–5, 2024 • CCPS Workshop on Process San Diego Convention Center • San Diego, CA For more information, and to register,
Safety Metrics: API-RP-754 Implementation visit www.aiche.org/academy.
Grand Hyatt Al-Khobar • Saudi Arabia
Oct. 28–30, 2024 • DIERS Fall Meeting
Houston, TX Virtual Aug. 13–14, 2024 •
Sept. 9–12, 2024 • Safety in Ammonia Plants &
Related Facilities Symposium AIChE Career Discovery Workshop
Manchester Grand Hyatt • San Diego, CA Oct. 28–30, 2024 • Cell Therapies & Course # CHCRDVTL • Instructor: V. Patrick
Immunotherapy Conference
Manchester Grand Hyatt San Diego • San Diego, CA
Sept. 10, 2024 • CCPS Canadian Regional Meeting Virtual Sept. 9–13, 2024 •
The BOW, Ovintiv Auditorium • Calgary, Alberta, Canada Process Safety Boot Camp
Oct. 29–30, 2024 • Global Conference on Course # CH900VTL • Presented by CCPS
Process Safety & Big Data
Sept. 10, 2024 • AIChE Virtual Career Fair Leonardo Royal Hotel • Frankfurt am Main, Germany
Virtual Visit www.aiche.org/careerfair Virtual Sept. 17–18, 2024 •
AIChE Career Discovery Workshop
Oct. 31–Nov. 1, 2024 • Food Innovation & Course # CHCRDVTL • Instructor: V. Patrick
Sept. 10–12, 2024 • Optogenetic Technologies & Engineering (FOODIE) Conference
Applications Conference San Diego Convention Center • San Diego, CA
Boston Univ. • Boston, MA Virtual Sept. 23–27, 2024 • Combo Course:
Nov. 12–14, 2024 • International Conference on HAZOP Studies AND Advanced Concepts for
Sept. 10–12, 2024 • tcbiomass 2024 Microbiome Engineering Process Hazard Analysis (CH157VTL & CH754VTL)
Westin Hotel • Itasca, IL Course # CH759VTL • Presented by CCPS
Tufts Univ. • Boston, MA
Sept. 18–20, 2024 • CCPS Latin American Virtual Oct. 7–10, 2024 • Chemical Engineering
Conference on Process Safety Nov. 26–27, 2024 • CCPS Global Summit on
Process Safety for Non-Chemical Engineers
Hilton Garden Inn • Barranquilla, Colombia Course # CH710VTL • Instructors: K. Jacob, S. Jacob
Mumbai, India
Sept. 20–25, 2024 • Global Industrial Microbiology Oct. 8–10, 2024 • Houston, TX •
Congress & Metabolic Engineering Summit Dec. 2–6, 2024 • International Process
Safety Week SuperChems™ for DIERS Software
Shanghai, China
Virtual Visit www.aiche.org/ccps/conferences-events User Training
Course # CH174 • Instructor: E. Kumpinsky
Sept. 24–26, 2024 • Advanced Manufacturing &
Processing Conference (AMPc) Dec. 4, 2024 • 2024 AIChE Gala
The Chifley Houston • Houston, TX The Pierre Hotel • New York, NY Oct. 15–17, 2024 • Houston, TX • Combo Course:
Flow of Solids AND Pneumatic Conveying
Oct. 3–5, 2024 • International Conference on Dec. 9–11, 2024 • Battery & Energy Storage (CH032 & CH033)
Conference Course # CH757 • Instructors: E. Maynard, H. Purutyan
Epigenetics & Bioengineering
Hotel Casa Amsterdam • Amsterdam, Netherlands New York Univ. Kimmel Center • New York, NY
Oct. 21–24, 2024 • Houston, TX •
Oct. 9–10, 2024 • SPE/AIChE Joint Workshop: Jan. 5–8, 2025 • International Conference on Process Safety Boot Camp
Flow Assurance Best Practices for Today, Biomolecular Engineering Course # CH900 • Presented by CCPS
Challenges for the Future The Royal Sonesta Houston Galleria • Houston, TX
The San Luis Resort • Galveston, TX Virtual Oct. 22–23, 2024 • Project Management
Feb. 16–18, 2025 • AfroBioTech Conference for Chemical Engineers
Oct. 10, 2024 • CCPS South East Asia Meeting Georgia Institute of Technology • Atlanta, GA Course # CH138VTL • Instructors: G. Carr, G. Englehardt
FM Global Centre • Singapore
Feb. 20–21, 2025 • International Congress on Virtual Oct. 30–31, 2024 • Overview of Risk-Based
Oct. 10, 2024 • CCPS Trinidad & Tobago Meeting Sustainability Science & Engineering (ICOSSE) Process Safety
Pointe-à-Pierre Staff Club • Trinidad and Tobago Univ. of Auckland • Auckland, New Zealand Course # CH925VTL • Presented by CCPS

64 aiche.org/cep August 2024

202408_Calendar_July21.indd 64 7/22/24 10:43 AM

 Boost your family’s
life insurance
protection and
save 30% as an
AIChE member

You choose your cost!

MEMBER A wide range of monthly costs lets you pick one

INSURANCE
SOLUTIONS comfortable for you!
Plan for What Matters Most  Up to $1,000,000 Term Life coverage available at a 30% savings1
for AIChE members and their spouses/domestic partners
 Owned by you, not your employer – coverage will be
there for you and your family even if you change jobs
 Group rates exclusively for AIChE members
 Apply online at your convenience – no agent
visit necessary

See your cost options now 2

ShowMeMyCost.com/AIChEtl
1
AIChE Group Term Life Insurance currently includes a 30% premium credit. Although future credits are dependent upon actual claims experience
and cannot be promised or guaranteed, the Program has been granting premium credits for years.
2
Along with policy features, eligibility, renewability, exclusions and limitations.
Underwritten by New York Life Insurance Company, 51 Madison Avenue, New York, NY 10010 on Group Policy Form GMR.
AIChE Insurance is administered by Affinity Insurance Services, Inc.; a licensed producer in all states (TX 13695); (AR Ins. Lic. 100106022); in CA
& MN, AIS Affinity Insurance Agency, Inc. (CA 0795465); in OK, AIS Affinity Insurance Services, Inc.; in CA, Aon Affinity Insurance Services, Inc.,
(CA 0G94493), Aon Direct Insurance Administrators and Berkely Insurance Agency and in NY, AIS Affinity Insurance Agency.

CA Insurance License #0G94493 AIChE-TL-7324

NYL-5034607 (04/24)

202408_Inside_Back_Cover3.indd IB_Cover3 7/16/24 4:52 PM

 October 27 – 31, 2024
San Diego Convention Center
Hilton San Diego Bayfront

CHEMICAL ENGINEERING REIMAGINED

The 2024 AIChE® Annual Meeting will bring together top academic and industry thinkers
to reimagine the continuing evolution of chemical engineering as a field and profession.

FEATURED SPEAKERS & EVENTS

John M. Prausnitz 2024 Andreas Acrivos

AIChE Institute Lecture Q Award for Professional
IGNACIO E. GROSSMANN Progress in Chemical
Rudolph R. and Florence Dean Engineering Lecture Q
University Professor, HAL ALPER
Department of Chemical Kenneth A. Kobe Professor of
Engineering Chemical Engineering
Carnegie Mellon University The University of Texas at Austin

William R. Schowalter Lecture Q IACChE’s James Y.

ROBERT A. BROWN Oldshue Lecture
President Emeritus, Professor of JUAN INGA
Engineering, and Professor of Computing & Director
Data Sciences
Inga Group
Boston University

Q Supported by the AIChE Foundation

REGISTRATION IS OPEN! aiche.org/annual/joinus

FOLLOW US: #AIChEAnnual

© 2024 AIChE 9252_24 • 07.24

202408_Outside_Back_Cover4.indd OB_Cover4 7/16/24 4:04 PM