EFRC Training Workshop

Condition Monitoring for

Reciprocating Compressors

Block II – Standards/Performance/Big Data

Mr. Guillaume Christin – Baker Hughes, a GE Company
Mr. Jan Löken – PROGNOST Systems GmbH

Expert session condition monitoring September 13/14 2017 1

 Topics
• Industry standards
• Performance monitoring
• Industry 4.0
• Big Data

Expert session condition monitoring September 13/14 2017 2

 CM Standards: Recip
compressors
• Limited information from standards for Condition Monitoring

• API 618: List of Machine Protection Parameters: Frame

vibration, Crosshead vibration
• API 670: Protection: Frame, Crosshead vibration, Rod
Drop
• EFRC: Evaluation: Frame, Cylinder (rod & lateral),
Foundation
• ISO: standards for acceptable vibration levels for
reciprocating machinery

Expert session condition monitoring September 13/14 2017 3

 Standards: API 618/API 670
API-618 5th Edition (Table 5) API-670 5th Edition Table P1
Condition
Alarm Shutdown Alarm Shutdown
High Gas discharge Temperature X X X
Low Frame Lube-oil Pressure X X X X
Low Frame Lube-oil Level X - X -
Cylinder Lubricator System Failure X - X -
High Oil-Filter differential Pressure X - X -
High Frame Vibration X X X X
High level in Separator X X X X
Jacket Water System Failure X - X -
High Crosshead Pin, Main Bearing,
1 1 - -
Vent Packing Temperature
High Crosshead Vibration - - X X
Rod Drop - - X X
1 –API 618 5th Edition 7.6.6.7 – if specified and is present

Expert session condition monitoring September 13/14 2017 4

 Standards: Experience
Crosshead vibration
Machine
Frame vibration
Protection
Piston rod
Condition displacement
Monitoring

Dyn. pressure
Cylinder vibration
Process data
Technology developments and practical experience over recent years
lead to additional recommended measurements for monitoring and
diagnostics.

Expert session condition monitoring September 13/14 2017 5

 Best practice / Sensors
Packing Case Rod Position
Temperature Frame Vibration

Packing Vent Line

Temperature Crosshead
Acceleration

Cylinder Pressure

Cylinder Valve
Temperature
Cylinder
Acceleration
Suction/Discharge
Connecting Rod Gas
Bearing Temperature
Temperature

Keyphasor System
Main Bearing
Temperature
Keyphasor Backup
System

Cylinder pressure sampling enables processing of performance parameters: Degrees Rod Reversal,
Discharge & Suction Power Loss, Gas Rod Load, Indicated Horsepower, Peak Rod Loads, etc.
Expert session condition monitoring September 13/14 2017 6
 Best practice / Crosshead
Accelerometers can detect the typical malfunctions as loose running gear
components, liquid ingestion, that generate impact-type vibration response
better than a velocity sensor due to the high frequencies generated by
impact events.

– Measurement should be acceleration units / Do not integrate the signal to velocity

– Possible Analyses: Peak or RMS

– Which band pass filter to apply for protection ?

Expert session condition monitoring September 13/14 2017 7

 Best practice / Crosshead
Valve Cover Acceleration Crosshead Acceleration

Mechanical Valve Noise + Mechanical

Mechanical Valve Noise + Mechanical

Expert session condition monitoring September 13/14 2017 8

Best practice / Crosshead
Filtered and unfiltered crosshead during development of failure
(loose piston rod connection)

Expert session condition monitoring September 13/14 2017 9

 Best practice / Crosshead
Case with Loose Piston
• Crosshead acceleration normal conditions 0.5 g pk
• Diagnostic alarm “Alert - Piston Looseness” 0.8 g pk
• Diagnostic alarm “Danger - Piston Looseness” 1.2 g pk
• First Alert was four weeks before the protection alarm (2 g pk)
• Inspection: broken roll pin, loose nut

Expert session condition monitoring September 13/14 2017 10

Best practice / Crosshead
Case with lubrication issue
Raw data Good signal

Segmented view (RMS 36 Seg.)

Protection limits

Expert session condition monitoring September 13/14 2017 11

Best practice / Crosshead
Case with lubrication issue
Raw data Bad signal

Segmented view (RMS 36 Seg.)

Protection limits

Expert session condition monitoring September 13/14 2017 12

 Best practice / Crosshead

• Alert & Shutdown based on reference

period
• Crank Angle bands alarming (fixed or
custom width)
• 1 alarm threshold per segment

Use crosshead vibration for shutdown

Expert session condition monitoring September 13/14 2017 13

 Best practice / Rod position
Rod Position continuously monitors the crosshead looseness, piston
rod vibration and rider band condition of each cylinder in a reciprocating
compressor. Rod position shall be used as shutdown parameter

Broken Rod case. Early

detection with Pk-pk rod
displacement

Expert session condition monitoring September 13/14 2017 14

 Best practice / Rod position
Case with broken rod
Normal signals- 8 minutes before failure
Raw data

Segmented view (Peak to Peak 8 Segments)

Protection limits

Expert session condition monitoring September 13/14 2017 15

 Best practice / Rod position
Case with broken rod
First indications – 90 sec to failure
Raw data

Segmented view (Peak to Peak 8 Segments)

Protection limits

Expert session condition monitoring September 13/14 2017 16

 Best practice / Rod position
Case with broken rod
Bigger increase – 41 sec. to failure
Raw data

Segmented view (Peak to Peak 8 Segments)

Protection limits

Expert session condition monitoring September 13/14 2017 17

 Best practice / Rod position
Case with broken rod
Shutdown released – 32 sec. to failure
Raw data

Segmented view (Peak to Peak 8 Segments)

Protection limits

Expert session condition monitoring September 13/14 2017 18

Best practice / Rod position
Case with broken rod
Piston rod broke

Automatic trip was no active!

Expert session condition monitoring September 13/14 2017 19

 Emissions
• Gas emission are a critical issue for recips
• Solution: monitoring packing case status, adopting better sealing
solutions.
• Condition monitoring can help identifying such leakages and provide
corrective actions.
• Pressure packing leaks detection with temperature, vent
temperature, cylinder pressure

Expert session condition monitoring 20

 Performance Indicators
• Volumetric Efficiency
– Fraction of stroke used by suction or discharge process
• Capacity
– Amount of gas processed per specified unit of time
• Flow Balance
– Ratio of suction capacity to discharge capacity
• Indicated Horsepower
– Power required for the cycle (area inside P-V curve)
• Power Losses
– Power lost to flow restrictions in suction and discharge path

Expert session condition monitoring September 13/14 2017 21

 Volumetric Efficiency – Defined
Performance Indicators
Discharge Volumetric Efficiency Suction Volumetric Efficiency
𝑉𝐷 𝑉2 − 𝑉3 𝑉𝑠 𝑉1 − 𝑉4
𝑉𝐸𝐷 = = 𝑉𝐸𝑆 = =
𝑉𝑃𝐷 𝑉𝑃𝐷 𝑉𝑃𝐷 𝑉𝑃𝐷
Clearance Volume, VCL

Piston Displacement, VPD

Cylinder Pressure

Discharge Volume, VD

3 2

4 1
Suction Volume, VS

0% Displaced Volume (%) 100%

Link to page 16 Expert session condition monitoring September 13/14 2017 22

Volumetric Efficiency – Measured
Performance Indicators

We will use these

measurements in the
following pages to
Discharge VE = 48% calculate capacity
(48% - 0%) and flow balance
values.

Suction VE = 86%
(100% - 14%)

Expert session condition monitoring September 13/14 2017 23

 Capacity
Performance Indicators

• Capacity is the amount of gas (mass or

volume) that is processed by a compressor per
specified unit of time.
• Mass Units (kg/hr, lbmol/stroke, etc.)
• Volume Units (mmscfd, CFM, etc.)
• Capacity can be measured by flow instruments
or calculated from measured values of
volumetric efficiency, pressure, cylinder
volume, and compressor speed.

Expert session condition monitoring September 13/14 2017 24

 Suction Capacity Calculation (SI)
Performance Indicators
•Information for • Calculation
cylinder on page 42:
𝑃𝑉 = 𝑍𝑛𝑅𝑇
– Ps = 310 psig = 2137.4 kPa =
(2137.4+ 101.3) kPa abs = 𝑃𝑉
𝑛𝑠 =
2239 kPa abs 𝑍𝑅𝑇

– Vs = (86%)(2.83m3) = 2.44 m3 2 239 000 𝑁

2 2.44 𝑚3
𝑛𝑠 = 𝑚
– Z = 1.0136 (dimensionless) 3784 𝑁 ∙ 𝑚
1.0136 302.43 𝐾
𝑚𝑜𝑙 ∙ 𝐾
– R = 3784 J/mol K
– Ts = 84.7º F = 29.28º C = 𝑛𝑠 = 4.71 𝑚𝑜𝑙 (𝑝𝑒𝑟 𝑠𝑡𝑟𝑜𝑘𝑒)
(29.28 + 273.15) K = 302.43º
K

Expert session condition monitoring September 13/14 2017 25

 Flow Balance
Performance Indicators

• Flow Balance is defined as Suction Capacity

divided by Discharge Capacity.
𝑛𝑠
𝐹𝑙𝑜𝑤 𝐵𝑎𝑙𝑎𝑛𝑐𝑒 =
𝑛𝑑
• If the cylinder is in perfect condition, with no
leakage past the valves, packing or rings, Flow
Balance is exactly 1.

Expert session condition September 13/14 2017 26

monitoring
 Suction Valve Leak Example
Performance Indicators
– Some of the gas leaks out of
the chamber through the
suction valve during the
compression process.

– Fewer moles of gas are 0.1 mole

(leak)

discharged than are brought 0.9

in during the suction mole

process. Discharge Capacity = 0.9 mole

– nS > nD, so Flow Balance >1

Expert session condition monitoring September 13/14 2017 27

 Discharge Valve Leak Example
Performance Indicators
– Additional gas leaks into the
chamber through the 1.0 mole

discharge valve during the 0.1 mole (leak)

suction process.
Suction Capacity = 1.0 mole

– This “extra” gas is discharged

along with the gas that was 1.1
mole
admitted through the suction
valve. Discharge Capacity = 1.1 mole

Flow Balance = 1.0 / 1.1 = 0.9

– nS < nD, so Flow Balance < 1

Expert session condition monitoring September 13/14 2017 28

 Discharge Valve Leak Example
Performance Indicators
P SUC ST1, P SUC ST1, Suction pressure
1390
psia TDC BDC

1300

1200

1100

1000

900
pV-diagram 1st & 2nd stage shortly before issue
800

700

600

500

400

0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0 8,0 9,0 10,0 Piston stroke [inch] 12,0

Machine Measuring Point Data name TAG-Name Value Unit

P CYL1 ST2 HE p-V-Diagram 302 PT 7015 14,50 - 2335 psia
P CYL1 ST2 CE p-V-Diagram 302 PT 7016 18,02 - 2339 psia
P CYL2 ST1 HE p-V-Diagram 302 PT 7013 12,60 - 882,8 psia
P CYL2 ST1 CE p-V-Diagram 302 PT 7014 15,42 - 885,6 psia
P SUC ST1 Suction pressure 302 PT 7001A 0 - 580,2 psia
P DIS ST1 Discharge pressure 302 PT 7001B 0 - 870,2 psia
P SUC ST2 Suction pressure 302 PT 7004A 0 - 870,2 psia
P DIS ST2 Discharge pressure 302 PT 7004B 0 - 2321 psia

Expert session condition monitoring September 13/14 2017 29

 Discharge Valve Leak Example
Performance Indicators
P CYL2 ST1 HE, p-V-Diagram
1400
psia TDC BDC

1300

1200

1100
pV-diagram of 1st stage CE side changed:
- Steeper compression line
1000
- Flatter re-expansion line
900
-> Clear hints for discharge valve leakage
800

700

600

500

400

300
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0 8,0 9,0 10,0 Piston stroke [inch] 12,0

Machine Measuring Point Data name TAG-Name Value Unit

P CYL1 ST2 HE p-V-Diagram 302 PT 7015 14,50 - 2335 psia
P CYL1 ST2 CE p-V-Diagram 302 PT 7016 18,01 - 2339 psia
P CYL2 ST1 HE p-V-Diagram 302 PT 7013 12,58 - 882,8 psia
P CYL2 ST1 CE p-V-Diagram 302 PT 7014 14,28 - 884,5 psia
P SUC ST1 Suction pressure 302 PT 7001A 0 - 580,2 psia
P DIS ST1 Discharge pressure 302 PT 7001B 0 - 870,2 psia
P SUC ST2 Suction pressure 302 PT 7004A 0 - 870,2 psia
P DIS ST2 Discharge pressure 302 PT 7004B 0 - 2321 psia

Expert session condition monitoring September 13/14 2017 30

 Discharge Valve Leak Example
Performance Indicators
P CYL1 ST2 HE, p-V-Diagram
1400
psia TDC BDC

1300

1200

1100
pV-diagram of 1st stage CE side got worse:
1000 - Again steeper compression line
900 - And flatter re-expansion line
800
-> Clear hints for a bigger DV leakage
700

600

500

400

300
0,0 1,0 2,0 3,0 4,0 5,0 6,0 7,0 8,0 9,0 10,0 Piston stroke [inch] 12,0

Machine Measuring Point Data name TAG-Name Value Unit

P CYL1 ST2 HE p-V-Diagram 302 PT 7015 14,50 - 2335 psia
P CYL1 ST2 CE p-V-Diagram 302 PT 7016 19,38 - 2340 psia
P CYL2 ST1 HE p-V-Diagram 302 PT 7013 13,35 - 883,6 psia
P CYL2 ST1 CE p-V-Diagram 302 PT 7014 8,977 - 879,2 psia
P SUC ST1 Suction pressure 302 PT 7001A 0 - 580,2 psia
P DIS ST1 Discharge pressure 302 PT 7001B 0 - 870,2 psia
P SUC ST2 Suction pressure 302 PT 7004A 0 - 870,2 psia
P DIS ST2 Discharge pressure 302 PT 7004B 0 - 2321 psia

Expert session condition monitoring September 13/14 2017 31

 Discharge Valve Leak Example
Performance Indicators
50,00
%
Clear change of characteristic pV values
46,50
45,00
can be seen after one poppet failed
43,50
42,00
40,50
39,00
37,50
36,00
34,50
33,00
31,50
30,00
28,50
27,00
25,50
24,00
22,50
21,00
19,50
18,00
16,50
15,00 Again change of characteristic pV
13,50
12,00
values due to 2nd damaged poppet
10,50

10.09.2014 17.09.2014 23.09.2014 28.09.2014 03.10.2014 09.10.2014 12.10.2014

14:00:00 18:00:00 02:00:00 10:00:00 18:00:00 02:00:00 14:00:00

Machine Measuring Point Data name TAG-Name Segment Value Unit Status Min and Max Values
P CYL2 ST1 CE Break through suction pressure 302 PT 7014 1 0 - 150,0 ° CA 0 - 69,10
P CYL2 ST1 CE Break through discharge pressure 302 PT 7014 1 200,0 - 350,0 ° CA 200,0 - 261,2
P CYL2 ST1 CE Polytropic exponent 1, Expansion line 302 PT 7014 1 0 - 5,000 0 - 1,207
P CYL2 ST1 CE Polytropic exponent 2, Expansion line 302 PT 7014 1 0 - 5,000 0 - 1,247
P CYL2 ST1 CE Polytropic exponent 1, Compression line 302 PT 7014 1 0 - 5,000 0 - 2,581
P CYL2 ST1 CE Polytropic exponent 2, Compression line 302 PT 7014 1 0 - 5,000 0 - 1,791

Expert session condition monitoring September 13/14 2017 32

Suction Valve Leak Example
Performance Indicators

Expert session condition monitoring September 13/14 2017 33

Suction Valve Leak Example
Performance Indicators

Expert session condition monitoring September 13/14 2017 34

Suction Valve Leak Example
Performance Indicators

Expert session condition monitoring September 13/14 2017 35

Suction Valve Leak Example
Performance Indicators

After Valve Replacement

Cylinder1 HE:
Suction and Discharge Valve
Temperatures increase

Expert session condition monitoring September 13/14 2017 36

 Indicated Horsepower
Performance Indicators

The WORK done on the gas for one

complete cycle is shown by the cross-
sectional area inside the measured
cycle curve on a P-V plot.

Power is WORK per unit

TIME, so it takes machine
speed into consideration.

Expert session condition monitoring September 13/14 2017 37

 Power Losses
Performance Indicators

The work that is done to overcome

flow restrictions in the suction and
Discharge Work Loss discharge paths is “lost” to the
compression process.

Power is WORK per unit

TIME, so it takes machine
speed into consideration.

Suction Work Loss

Expert session condition monitoring September 13/14 2017 38

Diagnostic Software Calculations
Performance Indicators

Expert session condition monitoring September 13/14 2017 39

 Industry 4.0

Source: http://www.allaboutlean.com/

Expert session condition monitoring September 13/14 2017 40

 What’s So Different?
Data Technology
• Volumes • Smart devices &
• Variety of sources machines
• Ubiquitous access & • Connectivity
speed • Compute power
• Low cost

Expert session condition monitoring September 13/14 2017 41

 Big data – Example I
• Twin-engine civil aircraft
• 6 h flight NY  LA
• Produces 240TB data
• Analysed on-board with
Engine-indicating and
crew-alerting system
(EICAS)

Expert session condition monitoring September 13/14 2017 42

Big data – Example II
• 4-throw compressor
– 21 probes

• 45 MB per minute
= 64,8 TB per day

• Monitoring use
software analytics
methods

Expert session condition monitoring September 13/14 2017 43

 Characteristics of the Digital
Enterprise
• Information-driven
– Leverages data, advanced analytics, and modern platforms throughout
enterprise
• Core and contract
– Optimize around core strategic capabilities, outsource balance
• Smart, connected ecosystem
– Dynamic, real-time information throughout value chain; adaptive
business processes
• Value-based products/services, smart environments
– Data–rich, outcome oriented, remote monitoring, as-a-service enabled
• Service savvy
– Services as/more important than physical equipment; connected
business models
• Quickly adaptive to customer
– Solve customer problems through new services

Expert session condition monitoring September 13/14 2017 44

 Tools of the Digital Enterprise

Visual Analytics Machine Learning Streaming Analytics Smart Machines

Cognitive Computing/AI Predictive Analytics Augmented Reality Enhanced Plant Apps

Expert session condition monitoring September 13/14 2017 45

 Big Data development
• Big Data is known as „Data Warehousing“
and „Data Mining“ since the late 1990´s
• First application was ERP/CRM
• Data analyses was performed with
economically-oriented statistical methods
• What happened? (Description)
• Why did it happen? (Diagnostics)

Expert session condition monitoring September 13/14 2017 46

 Big Data development
• Future of Big Data is „Prediction“ of e.g.
demographical developments or customer
churn or traffic jam predictions
• Amazon & Co. are the drivers
• Engineering/mechanical statistical
methods are in their early days
• What will happen? (Prediction)
• What to do? (Prescription)
Expert session condition monitoring September 13/14 2017 47
 Big Data development
What to do?
Prescriptive Analysis

What will happen?

Predictive Analysis

Machine Learning
Why did it happen?
Diagnostic Analysis

What happened?
Describing Analysis

Data Mining

Expert session condition monitoring September 13/14 2017 48

 Machine Learning
• „Machine Learning“ = Knowledge from
Experiences
• Different scientific methods are used to
enable machines to make experiences, i.e.
to learn
• Supervised Learning
• Regression cluster
• Decision Trees
• Support Vector Machines
• Fuzzy Systems
• Neuronal Networks

Expert session condition monitoring September 13/14 2017 49

 Big Data in CMS
• Diagnostics of machines, the reliable
identification of failure pattern and the
output of clear information is key for
meaningful condition monitoring
• Today, some systems are in the beginning
of the „Prediction“ phase
• Mostly linear developments, e.g. rider
band wear

Expert session condition monitoring September 13/14 2017 50

Big Data – Example of usage in
Condition Diagnostics
Pressure
Pattern recognition
compares signals based on
known cases and results in
Vibration automatic message:
„Blocked discharge Valve
issue Cyl. XY”

Temperatures

Expert session condition monitoring September 13/14 2017 51

Big data – Example of usage in
Condition Diagnostics
Multivariate
predictive empirical
modeling engine
that uses the
operating and
environmental
profile for each
piece of equipment,
combined with first
principles rules and
FMEA patterns to
provide diagnostic
advisories:
„Valve Leak
Cylinder X”
Expert session condition monitoring September 13/14 2017 52
 Industry Goals
Need to make operation safer and more
reliable while helping to ensure optimal
performance at a lower sustainable cost.
An intelligent asset strategy for each asset is
needed to collect and manage all data and
information to make the best decisions that
help maximize overall asset and operational
performance.

Expert session condition monitoring September 13/14 2017 53

Elements of Asset Performance
Management Platform
• Collect, analyze, and visualize the data that is crucial to assets’ health
• Prioritize work by determining which assets are most critical to the
overall health of operation
• Understand the true status of each asset
• Predict equipment and process issues with greater accuracy before
they occur
• Assure mechanical integrity of assets and compliance with regulations
• Learn from continuous, real-time analysis and visibility to understand
the options for mitigating potential problems while balancing associated
costs, risks, and benefits
• Define and manage asset strategies with a risk-based approach for
improved planning and efficiency

Expert session condition monitoring September 13/14 2017 54

 Summary
By leveraging cutting-edge data science, domain
experience and analytics technology, today’s
solutions for the Industrial Internet are able to
balance traditionally competing priorities –
reducing costs, improving availability, and
managing risk—to provide intelligent asset
strategies that help maximize asset performance
and productivity.

Expert session condition monitoring September 13/14 2017 55