1. Advances in Fault Detection and Diagnosis in Industry
May 20, 2017 Bangalore India
2. • Introduction to Condition Monitoring in Industry
• Fault Detection, Diagnosis and Prognosis
• Fault Management
• Condition Monitoring System
• Fault Detection – Process Model
• Fault Diagnosis - Plausibility Checks
• Data Analysis
• On-line Enterprise Asset Management
•
Overview
3. Run to Break - Unplanned Maintenance
Maintenance as Cost Centre
Maintenance cost greater than 20% of overall cost
Preventive Maintenance – Planned Maintenance
Too Much Maintenance
Predictive Maintenance - “Condition Based Maintenance
Profit Centre
Condition Monitoring in Industry
5. o Fault is a deviation / abnormal condition – a fault may initiate
failure.(State)
o Failure is a permanent interruption of system (Event)
o Malfunction is temporary interruption
Fault Detection: Early detection of faults (small faults) with abrupt /
incipient time behaviour. E.g. Limit Checking / Threshold checking,Trend
checking, Signal analysis….
Fault Diagnosis : Diagnosis of faults in the processes / parts/ devices.
E.g. Analytical / Heuristic symptoms and their relations to faults, Fault
evaluation (Hazard class)
Fault Prognosis: Predicting remaining useful life
E- MAINTENANCE
Fault Detection, Diagnosis & Prognosis
6. Avoid shutdowns by early detection and actions like condition based
maintenance / repair
Fault tolerant systems for sudden faults / failures / malfunctions –
Reconfiguration / Redundant components
Maintenance on demand
Tele-monitoring
100% Quality Control
Effective asset management
Improving total life cycles of products and processes
Fault Management
8. Fault Detection -Process Model
RTU
RealTime DataTransfer
OLE
Monitoring
System
Microsoft Excel /VB/
Python
P
P
P P
P
If ΔP High –> Change Suction Filter
Atm. Pressure
Compressor
Suction Pressure
9. First step towards Model Based Fault Detection
- By checking the plausibility of its indicated values (Rough Process Model)
- Using Microsoft Excel / VB / Python
- For above example:
IF [ΔP > ΔPmax] THEN [CHANGE SUCTION FILTER]
- More examples: Oil Pressure of engine Poil with speed N and C.W Tw
- IF[N < 2000 rpm] AND [Tw < 40 C]THEN [2.5 bar < Poil > 4 bar}
- Trend Analysis using “LOOKUP” function in Excel
- Use VBA macros to “Diagnose Faults”
Fault Diagnosis – Plausibility Checks
10. Process Model
Mass of Air in, mai
Mass of water in, mwi
Temperature
Tai
Two
Twi
Tao
ma
mw
Tai
Ta0
Twi
Two
Overall Fault Detection of H.E – Static Model
Heat loss, Ql = ma cpa (Tai-Tao) – mw cpw (Two-Twi)
Overall Heat transfer coeff, U = Q/(AΔTm)
ΔTm = (Tai-Two)-(Tao-Twi)/ ln (Tai-Two)/(Tao-Twi)
Residual , r(k) = Ql(k)/cp, k = t/To, r(k) will change in
case of faults in sensors, leaks, insulation …
Evaluating U(operation) with U(design )
can determine contamination / fouling in
H.E.
12. Data Analysis
Name Cell Sim# Graph Min Mean Max 5% 95%Errors
Range: RISK (Loss Hours)
New Reliable Plant with Backup /
RISK (Loss Hours)
F5 1 13.04665 46.77504 113.4808 20.81343 79.74958 0
New Reliable Plant with Backup /
RISK (Loss Hours)
F5 2 13.04665 46.77504 113.4808 20.81343 79.74958 0
New Reliable Plant with Backup /
RISK (Loss Hours)
F5 3 13.04665 46.77504 113.4808 20.81343 79.74958 0
New Reliable Plant with Backup /
RISK (Loss Hours)
F5 4 13.04665 46.77504 113.4808 20.81343 79.74958 0
Low customer demand / RISK (Loss
Hours)
F6 1 32.93708 125.7655 312.3814 53.29559 217.4649 0
Low customer demand / RISK (Loss
Hours)
F6 2 32.93708 125.7655 312.3814 53.29559 217.4649 0
Inputs in
Scenario For
F5 >75%
Cell Name Description Risk Assessment!F5
New Reliable Plant
with Backup / RISK
(Loss Hours)
(Sim#1)
Percentile
Risk Assessment!F5
New Reliable Plant
with Backup / RISK
(Loss Hours)
(Sim#1)
Percentile
Risk Assessment!F5
New Reliable Plant
with Backup / RISK
(Loss Hours)
(Sim#1)
Percentile
Risk Assessment!F5
New Reliable Plant
with Backup / RISK
(Loss Hours)
(Sim#2)
Percentile
Risk Assessment!F5
New Reliable Plant
with Backup / RISK
(Loss Hours)
(Sim#2)
Percentile
Risk Assessment!F5
New Reliable Plant
with Backup / RISK
(Loss Hours)
(Sim#2)
Percentile
Risk Assessment!F5
New Reliable Plant
with Backup / RISK
(Loss Hours)
(Sim#3)
Percentile
>75% <25% >90% >75% <25% >90% >75%
#1 D5
New Reliable Plant with Backup /
Probability of occurance(Failure)
RiskWeibull(2,0.327,RiskShift(0.1),Ris
kStatic(0.1))
0.874 0.126 0.95 0.874 0.126 0.95 0.874
- D8
Unreliable Backup/
Maintenance/Equipment Service /
Probability of occurance(Failure)
RiskWeibull(5,0.5,RiskShift(0.3),RiskS
tatic(0.6))
- - - - - - -
- D7
Unreliable Maintenance / Equipment
Service / Probability of
occurance(Failure)
RiskWeibull(5,0.696,RiskShift(0.1),Ris
kStatic(0.6))
- - - - - - -
- D6
Low customer demand / Probability
of occurance(Failure)
RiskWeibull(2,0.38,RiskShift(0.1),Risk
Static(0.5))
- - - - - - -
13. On-line Enterprise Asset Management
• Uptime / Downtime – Plant Availability
• Total Production – Revenue Generation
• Total cost of Maintenance …
• ROCE – Return on Capital Employed
RTU
14. • Sarathsri E-Technologies, Bangalore
• Institutions of Mechanical Engineers, South Asia, India
• Air Liquide , France
• BOSCH , India
• Indian Institute of Technology
References
Fault Diagnosis Systems by Rolf Isermann, Springer 2006
Fault Diagnosis Applications by Rolf Isermann, Springer
Vibration based Condition Monitoring by Robert Bond Randall, WILEY 2011
Artificial IntelligenceTools: Decision support Systems in Condition Monitoring
and Diagnosis by Diego Galar Pascual, CRC Press 2015
https://opcfoundation.org/
Acknowledgment