Modelling and Simulation

In Process Industry

A proven framework for applying

process simulation models
for solving real-world problems.

1
 INTRODUCTION

How to successfully apply process

modelling and simulation to solve
real-life challenges?
Process modelling and simulation offer substantial
cost savings in validating or optimizing industrial
processes. As new equipment and research often
require significant investments, the validation process
for new solutions has to be both fast and bulletproof.

Process simulation is a useful and This white paper outlines a

powerful tool, which has many systematic approach for solving
application areas in the process real-world problems in the pro-
industry, such as: research, cess industry that require the use
control, optimization, feasibility of process simulation models to
studies, design, troubleshooting, get to the solution.
de-bottlenecking, environmental
impact, knowledge buildup, Modelling means the
training, safety. construction of an accurate
Process simulation models give mathematical representation
the ability to predict the outcome of a process.
of certain actions and therefore Simulation is making
increase profits, safety and/or calculations of how changes
knowledge of the process. in the model should affect the
modelled process.

2
 Benefits of
Modelling and
Simulation

The list of applications for process Speed: Modelling and simulation

models is almost endless. Over the allow you to calculate changes in
last 50 years, there has been a complex systems in a few seconds.
widespread use of models for Flexibility: Alternative choices can
predicting both steady-state and be safely tested before making an
dynamic behavior of processes. investment.
This was principally in the chemical
process industries but, more Predictability: The whole lifecycle
recently, these techniques have of a plant or a piece of equipment
been applied into other areas such can be analyzed beforehand with
as minerals, pharmaceuticals, and simulations.
bio-products. Education: Improved knowledge
of your process for your process
design team, plant engineers and
Dynamic models analyze the
process operators.
time dependency of the
phenomena in a process. Optimization: Simulation results
can steer a process to significantly
Steady-state models focus on increase its efficiency, for example
the outcomes of a process. by reducing energy consumption
or minimizing waste.

It doesn’t matter whether you want to build the model yourself,

with or without help, or if you just want to use models that were
built for you: Somehow, you’ll have to get from your process to
a model of your process, such that you can use this model to
answer questions on the process.

3
 Typical Application Areas Emergency response
training

Startup and
shutdown for normal Routine operations
operations training

Developing,maintaining
Train operators and
and expanding process
engineers
knowledge

Education
Examining and
Feasibility analysis of Improve process
Troubleshooting improving control
novel designs safety
strategies

Model based Optimization using Process Improve process Less product

structural and
innovations
parametric changes
Improvement performance contamination

Quantifying emission
Environmental Detection of
Less Contamination rates for a specific Impact hazardous operating
design regimes

Dispersion Estimating acute

Characterizing social
predictions for air Waste minimization accident effects (fire,
and economic
and water releases Financial impact explosion)
Impact

Revenue increase Cost Reduction

Optimizing
Reducing Reducing Predictive Reducing
reactor Making more
resources and personnel costs
downtime performance product maintenance
utilities

4
My team and I regularly have
discussions about a wide variety
of application areas, with many
people, from plant managers to
operators to researchers. These
people often have great ideas for
improvements, but somehow, it’s
very hard to take it to the next
level and actually get started.

Obviously, before you can even get

started, there clearly is a need for
sympathetic persons in
management and plant operations
who know and appreciate the
power of simulation models, and
who will demand their use for study
of their own particular plants.

5
 One of the main challenges that must be overcome is
procrastination in one form or another. You’ll probably
recognize this from you own line of business:

Busy, busy, busy…

More often than not, we are so
caught up with our daily
routines that we hardly find
the time to stop and think
about seemingly less urgent
matters. This obviously also
applies in the process industry,
where the main emphasis will
be (and should be) on daily
operations, maintenance, and
safety.

But typically, also questions arise like

➢ “is the knowledge level of operators and engineers sufficient?”

➢ “is the plant running efficiently or safe enough?”
➢ “Can we debottleneck or optimize our processes?”
➢ “Is this new idea I have actually feasible?”

Typically questions in which a process model could play a significant role.

6
Eeeh…
You have a great idea for an The easy way out, obviously, is to
improvement, but modelling is just postpone or to skip all tog-
not one of your daily tasks and you ether.
consider yourself far from a
This does not have to be the case
modelling expert. You reason that
if a clear, stepwise method is
you’ll probably get minimal or no
adhered to, such that you can see
support to get this implemented.
results quickly and see the added
You, your coworkers, or boss
value of process modelling and
might think it will be too difficult
simulation.
or too expensive, so prefer to
stick with something they know.

The construction of a process model, either by you, with you, or for

you, involves a number of operations, which can be broken down into
a number of principle steps. I will introduce you to a systematic
modelling methodology and will show you how we successfully apply
this for our customers.

7
 I would like to make sure that this whitepaper makes sense to you. You
can answer the following questions to determine if this approach is right
for you and your company.

➢ Are you into process ➢ Do you see the added value

engineering - either as an of process modelling and
engineer, manager, researcher, simulation?
operator, teacher, or student? ➢ Do you expect that a process
➢ Do you have approval, budget model would help you with
and time to either make a the challenge or
model yourself, let a model be improvement idea you’re
built, or get help to make a facing?
model?

If you answer the above questions with "yes", then this approach
is suitable for guiding you to a solution quickly.

8
My name is Mathieu Westerweele
(1973), and in my professional life I
am into equation-based process
modelling. Besides that, I like to
spend time with my family (I have
four kids), play strategic board
games and love to play piano,
trumpet, or trombone. Preferably
improvising because that’s a real
dynamic challenge.

In 2005 I co-founded Mobatec and

together with my team we solve
Real-world challenges for companies (knowledge buildup, optimization,
design, control, learning/training, safety) in which process models play a
significant role to get to the solution.

Over the years I have built a lot of dynamic process models for many
different purposes, such as research, Operator Training Simulators, teaching,
control, PLC testing, real time simulating, optimization, engineering. I have
spent many many hours debugging these models and ran into all
imaginable errors. As a result, I can solve any problem related to dynamic
process models quickly and efficiently.

9
Step-by-Step Method
A model is an imitation of reality and a process model is a particular form of
representation. In the process of model building, we are translating our real-
world problem into an equivalent mathematical problem which we solve and
then attempt to interpret. We do this to gain insight into the original real-
world problem, such that we can solve it.

Real World Mathematical

problem 1 Problem

4
2

Mathematical
Interpretation 3 Solution

Process models are fundamentally mathematical in nature. They attempt to

capture, in the form of equations, certain characteristics of a process for a
specific use of that model. Hence, the purpose (i.e. the original real world
problem) is very much a key issue in model building.

Model development is inherently iterative in its nature. Usually, you have to

return to and (partially) repeat earlier steps, during the development of the
model. No one gets it right the first time! In fact, we never get a perfect
model, just one that is usable.

10
 Real World Mathematical
problem 1 Problem

Before starting to set up a process ➢ What are the time

model, the problem definition characteristics of the
should be clearly stated. This def- process model (steady-state
ines the process, the modelling versus dynamic)?
goal, and the validation criteria.
Obviously, you should first answer
the question “Will a process mod- 1B: Identify Controlling Factors
el help me with the problem I’m
The most important or essential
facing?” with yes.
characteristics required in the
model for the stated modelling
goal need to be assessed:
1A: Problem Definition
➢ What governing phenom-
You have to deal with the task of
ena or mechanisms are
translating the real problem to
there in the system?
one represented in mathematical
➢ What data on the process
terms. Some of the key issues that
are available and what is the
must be dealt with here are:
quality and accuracy of the
➢ What do you understand data?
about the real-world ➢ What are the process in-
problem? puts, states, outputs, and
➢ What is the end goal, what disturbances?
are the criteria for success?
➢ How accurate does the
model have to be?

11
1C: Evaluate Problem Data data and parameter values to-
gether with their uncertainties
Most models of real processes are
or precision.
of the “grey-box” type, i.e. a mix of
knowledge based “white-box m- At this step, you may find out
odels” and data driven “black-box that there are neither suitable
models”. parameter values found in the
literature nor measured data to
Therefore, you will almost always
estimate them. This situation
need to use either measured
may force you to reconsider
process data directly or
your decisions in steps A and B
estimated parameter values in
and to return there to change
our models. It is very important to
them.
consider both measured process

12
 Mathematical
Problem

Mathematical
Solution

The mapping of the process into a The resulting model should be

model does not only depend on similar to the real process in
the chosen theory, but also on terms of its important properties
the conditions under which the for the intended use. In other
process is being viewed. The words, the model should describe
mapped characteristics vary thus or reflect somehow the properties
not only with the applied theory of the real system relevant to the
but also with the conditions. modelling goal.

Different tasks focus on different On the other hand, a model will

characteristics and require differ- never be identical to the real
ent levels of information about process system. It should be
these characteristics. For exam- substantially less complex and
ple, control would usually be hence much cheaper and easy to
achievable with very simple dyn- handle, so that the analysis can be
amic models, whilst the design of carried out in a convenient way.
a reactor often requires very
detailed information about this
particular part of the process.

13
2A: Construct the model procedure, such that the model
can be solved for the unknown
At this point it is important that
values of the variables repre-
you make up your mind whether
senting the process. In all cases,
you want to construct the model
you must ensure that the model is
completely by yourself, or with
well posed such that the "excess"
support, or if you want someone
variables or degrees of freedom
else to build the model for you.
are satisfied. Numerical problems
You and/or the person(s) helping or lack of solution techniques may
you to build the model also need prevent you from using a
to decide what kind of process particular type of process model
model (simple or complex, data- and can lead to additional
driven or knowledge based, simplifying assumptions to obtain
steady-state or dynamic) would a solvable model.
best suit the problem at hand.

Only then can you actually start

2C: Verify / Debug the model
constructing and solving the
solution
model.
Having a solution is just the start
There are many different routes
of the analysis. Verification is
that can be chosen to get from
determining whether the model
your process to a model of that
is behaving correctly. Is it coded
process, which obviously depend
correctly and giving you the
on which form of model would be
answers you intended? This is not
best suited for the end goal. Entire
the same as model validation
books have been written on this
where we check the model
subject and it goes beyond the
against reality. You need to check
scope of this guide to dive into
carefully that the model is
that.
correctly implemented. Struct-
ured programming using "top-
2B: Solve the model down" algorithm design can help
here as well as the use of modular
Having generated some math- code which has been tested
ematical description of the real- thoroughly. This is particularly
world process, it is now necessary important for large-scale models.
to find or implement a solution

14
 Mathematical
Interpretation 3 Solution

Validation of a simulation model ces, need to be known accurately

by mirroring actual operational to ensure model predictive quality
performance in a plant/unit or and what level of validation is
pilot/laboratory scale equipment necessary.
can provide confidence of reliable
There are various possibilities to
results. It is crucial that simulation
validate a process model. The
results are not blindly accepted.
actual validation method strongly
The success of a simulation
depends on the process, on the
depends on the user’s ability to
modelling goal and on the
interpret input and output data.
possibilities of getting indep-
You need to have procedures and endent information for validation.
tests to check whether your
model has been correctly
implemented and then ask
whether it imitates the real world
to a sufficient accuracy to do the
intended job. You will need to
determine what process
parameters, inputs, or disturban-

15
The possibilities include but are not limited to:

➢ verify experimentally the simplifying assumptions,

➢ compare the model behavior with the process behavior,
➢ develop analytical models for simplified cases and
compare the behavior,
➢ compare with other models using a common problem,
➢ compare the model directly with process data.

As a result of the validation, you might find out that your model needs
to be changed, and you’ll have to return to step 2 and perform the
sequence again.

Usually, validation results indicate how to improve the model. You

can often identify the inadequate areas in our model development
and, therefore, not all modelling efforts are lost. As mentioned before:
No one gets it right the first time. Some iterations should be expected.

16
 Real World
problem

Interpretation

In the last step you are faced with These issues are just some of the
the implementation of the model many which arise as models are
results back into the real-world conceptualized, developed, solved,
problem you originally addressed. tested, and implemented. What is
Some issues that arise are: clear from the above discussion is
the fact that modelling is far more
➢ For applications where
than simply the generation of a set
speed might be essential, do
of equations.
I need to reduce the model
complexity? The result, though, is that you can
➢ How can model updating be get to your solution quicker,
done and what data are cheaper and/or more reliably. So, it
needed to do it? is well worth the effort.
➢ Who will actually use the
results and in what form
should they appear?
➢ How is the model to be
maintained?
➢ What level of document-
tation is necessary?

17
 REFERENCE CASE

Mobatec regularly does smaller projects and interesting case studies.

This case discusses the use of a dynamic simulation model during the
design phase of a steam boiler project of one of our customers.

A Chemical Company invested in two The new setup would decrease the
steam boilers, which should provide capacity (smaller holdup because of
steam to a high-pressure steam header shorter distances) and therefore the
and a connected steam network. Before company was interested in the dynamic
the company invested in the new boilers, behavior of the various steam headers in
the steam network was pressurized by case of failure of one of the steam
the high-pressure steam network of an suppliers or other non-nominal behavior
external supplier (via a pipeline of several (e.g. trips in parts of the connected
kilometers). plants).

Using the supplied documents (PFD’s, ➢ The main customers of steam

P&ID’s, equipment details, etc.) and headers were modeled as “tunable
consultation with company process customers”, such that they show
engineers, we set up a dynamic model, realistic steam consumption under
tested and tuned for the steam network. different conditions.
The model consisted of the following ➢ All controllers, control loops and
components: logics of interest were included in
the model.
➢ Simplified models of all boilers,
based on the boiler volume and the For the simulation only the water / steam
heat input to generate steam. system was important. For the process
➢ The steam network with the main conditions on the part of the steam
steam headers and the important consumers, reasonable estimates were
control valves and controls. made, as far as was relevant, in
consultation with company’s process
engineers.

18
 REFERENCE CASE

The company engineers thought of a lot of different possible scenarios and it was easy
for them to implement and run these and plot and discuss relevant results. As expected,
the dynamic model confirmed that the steam headers reacted a lot quicker to offsets
and disturbances with the new boiler setup as compared to previous setup (with the
external high-pressure steam supplier).

The new boiler setup was designed to between two headers was too small to
operate at roughly 50% of its capacity, cope with dynamic changes, causing one
such that the boilers could ramp-up (or header to blow off steam, while the other
down) in case of a disturbance (like the dropped in pressure. Therefore, the actual
failure of an external boiler). It was found valve was replaced before the new boilers
that for most disturbances the new where in place.
design and the proposed control system All in all, it was a very nice project, with
were just ok and the pressures in the very positive results for the involved
headers would not drop below critical companies. Our customer gained better
values, such that no (parts of) plants insight in the dynamic behavior of their
should be shut down. steam network and managed to
An important change had to be successfully make some necessary
implemented in the field, though, to changes to both their equipment and
make sure that certain plant parts would their control system, before the boilers
not trip after a disturbance in the steam were in service.
network. The model predicted that the
current size of a main “let down valve”

19
 Strategic Partnerships for
Modelling and Simulation
Sometimes, projects that involve modelling and simulation are straightforward,
and can be performed with inhouse engineers. Often though, projects benefit
from an interdisciplinary approach where research, laboratory work, modelling,
simulation and plant data work hand in hand. When looking for a modelling
and simulation partner, you should consider the following aspects:

Research: Your partner should have a strong background in the latest trends
and findings regarding the field of modelling and simulation methods, and
preferably also has knowledge in your field. This way the partner immediately
understands your specific needs and can cater for them optimally.

Software: A state-of-the-art modelling and simulation software is essential for

success. To save your time and resources, your partner should be able to
provide or build models for a large variety of industrial processes.

Experience: Optimally, the partner should have vast experience of prior

projects that are similar to yours. This ensures efficiency and reliability.

Versatility: Combining different methods is the key to a successful project.

Data-driven, physics-based, steady-state, dynamic – don’t settle for one
approach. Your partner should be able to combine experimental methods,
research, plant data and straightforward simulations.

20
Next Steps
There are two next steps you can take now.

Watch our
Introduction to Modelling Webinar
In this free Introduction Course, you will learn about (equation based) process
modelling. it is especially suitable if you want to be able to setup your own models,
either with or without additional help. No prior experience with mathematical
modelling or other modelling tools is required. It is a do-it-yourself course that
consists of 3 one-hour online sessions and has a total of 5 exercises which will on
average take you between 2 to 6 hours to work through.

Go to Webinar

Schedule a Strategy Call

During this conversation I will ask you a number of questions about you, your
company, and your goals and will I help you make good strategic choices. Free of
charge.

If at the end of the conversation we come to the conclusion that we can help you,
we will talk further about how that works.

If, during the conversation we conclude that you or your company are not ready
yet, or we think that this method is not suitable, we will discuss other possible
options or refer you to another party.

I promise you that you’ll always get one step ahead because of this conversation. I
have reserved several spots in my agenda. Click on the button to directly book an
appointment in my calendar.

Schedule
21
 How to get from your process
to a model of your process:

➢Do it yourself
➢Done with you
➢Done for you

Resources:
1. Process Modelling and Model Analysis - Ian Cameron, Katalin Hangos
2. Five Steps for Building Consitent Dynamic Process Models and their
Implementation in the Computer Tool Modeller - Mathieu Westerweele

www.mobatec.nl
22