3-Chapter Three Inspection Decisions 2010 (Compatibility Mode)

Chapter 3
Inspection Decisions
© A.K.S. JARDINE
Reliability Improvement
through Inspection Decisions
© A.K.S. JARDINE
Maintenance Optimization
Optimizing Equipment Maintenance and Replacement Decisions
Component Inspection Capital Equipment Resource

Replacement Procedures Replacement Requirements
Maintenance Management System (CMMS/EAM/ERP)
© A.K.S. JARDINE 3
Inspection Problems
1. Inspection frequencies for equipment which is in

continuous operation and subject to breakdown.
2. Inspection intervals for equipment used only in

emergency conditions. (Failure finding intervals)
3. Condition monitoring of equipment: Optimizing

condition-based maintenance decisions.
© A.K.S. JARDINE 4
Inspection Problems
System Failures
Decreasing
system
failures
Component 2
Component 1
Component 3
Component 5
Component 4 Increasing
inspection
frequency
Inspections &
Minor
Maintenance
© A.K.S. JARDINE 5
Examples of equipment subject to regular
inspections – PAM participants
•?
•?
•?
© A.K.S. JARDINE 6
Examples of equipment subject to regular
inspections
• Haul trucks : A,B,C, D inspections at

250 hours – manufacturer’s
recommendation
• Compressor: Daily check list – present
practice. RCM suggests weekly
• Conveyor belt: Daily check list – RCM
basis
• Screening station: Weekly check –
present practice
© A.K.S. JARDINE 7
Optimal Inspection Frequency: D(n) Model
(Section 3.3, page 106)
Total Downtime versus Inspection Frequency

λ( n ) n
D(n) = +
µ i
Total Downtime, D(n)
Downtime due to Inspections

and Minor Maintenance
Total Downtime (D)

Inspection Frequency (n) Downtime due to
System Failures
Optimal inspection frequency minimizes total downtime, D(n)
© A.K.S. JARDINE 8
Inspection Frequency to Minimize the Total Downtime per Unit
Time of Buses at an Urban Transit Authority
(Section 3.3.5, page 108)
© A.K.S. JARDINE 9
Montreal’s Transit Commission’s Bus
Inspection Policy
Inspection Type
Km (1000) “A” “B” “C” “D”
5 X
10 X
15 X Where:
20 X Ri = No. of type i
25 X inspections / Total No. of
30 X inspections i = A, B, D, or
35 X D
40 X
45 X
50 X
55 X
60 X
65 X
70 X
75 X
80 X
Total 8 4 3 1 Σ = 16
R 0.5 0.25 0.1875 0.0625 Σ = 1.0
© A.K.S. JARDINE 10
Mean Distance to Failure versus
Inspection Interval
6000
Mean time 5000

between Actual
failures
4000
(km) Predicted
3000
2000
4500 5500 6500 7500 8500
Inspection interval (km)
Downtime As a Function of the Inspection
Frequency
3.5 repair
3 inspection
2.5 total
Downtime
2
(% of total
bus-hours/
year) 1.5
0.5
0
4000 5000 6000 7000 8000 9000
Interval Between Inspections (km)
Optimal Inspection Interval

Failure Finding Intervals (FFIs) for Single
Protective Devices
Your
Protective
Device
From: Moubray. J.M., Reliability - centred Maintenance II,
1999, page 172
“Failure-finding applies only to hidden or unrevealed failures. Hidden failures in

turn only affect protective devices.
If RCM is correctly applied to almost any modern,complex industrial system, it is
not unusual to find that up to 40% of failure modes fall into the hidden category.
Furthermore, up to 80% of these failure modes require failure finding, so up to
one third of the tasks generated by comprehensive, correctly applied maintenance
strategy development programs are failure –finding tasks.
A more troubling finding is that at the time of writing, many existing maintenance
programs provide for fewer than one third of protective devices to receive
attention at all (and then at inappropriate intervals)”………………….
.
.
“This lack of awareness and attention means that most of the protective devices
in industry-our last line of protection when things go wrong- are maintained
poorly or not at all.
This situation is completely untenable”
Examples of Protective Devices: PAM
program participants
1. Fire extinguisher (monthly)

2. Smoke detector ( quarterly)
3. Public address system (quarterly)
4. Exit sign ( monthly)
Examples of Protective Devices: PAM
program participants
1. Pressure relief valve

2. Circuit breakers
3. Power off switch
4. Ozone detector
Maximizing Availability (Section 3.4, page 110)
• The problem is to determine the best interval between

inspections to maximize the proportion of time that the
equipment is in the available state
• Two possible cycles of operation are:
Inspect Inspect
Ti Ti Tr
ti ti
0 0
Cycle 1 Cycle 2
Good Failed
Cycle Cycle
Where: Ti = time required to effect an inspection, Tr = time required to

effect a repair or replacement, and ti = interval between inspections
Numerical Example
Problem: A protective device with a normally distributed MTBF = 5
months (and standard deviation = 1 month) has:
• Ti = 0.25 month
• Tr = 0.5 month
Find the optimal inspection interval, ti, to maximize availability.
Result:
ti 1 2 3 4 5 6
A(ti) 0.8000 0.8905 0.9173 0.9047 0.8366 0.7371
Therefore, the optimal inspection interval is 3 months
Inspection Interval vs. Availability
ti
100 ti × R(ti ) + ∫ t × f (t )dt

90 −∞
80
A (ti ) =
ti + Ti + Tr × [1- R(ti )]
70
60 Where:
50 A(ti) = availability per unit time
40 R(ti) = reliability at an inspection interval of
30 ti
% Availability
20 f(t) = density function of the time to failure
Optimal
10 of the equipment
Interval
0
0 1 2 3 4 5 6 Ti , Tr , and ti have been defined in the
previous slide
Inspection Interval (Months) In this case, Ti = 0.25 month and Tr = 0.5
month
F.F. Interval vs. Availability:
Moubray (Horton) Model
I = 2(1-A)M
Where:
A = Availability
M = MTBF
I = Failure Finding Interval
Source: Moubray, RCM II, pp. 177

An Example Using the Moubray Model
Problem: Brake lights on motorcycles
• MTBF = 10 years
• FFI = 10% of MTBF of 10 years = 1 year
Result: Availability = 95.0%
Problem source : Moubray, RCM II., 1999, pp. 175-176

Example Using Moubray (Horton)
Model
Problem: A protective device has a MTBF = 10

years. Determine the FFI which will yield
an Availability = 99%.
Answer: FFI = 10 weeks (roughly)
Example for FFI (Failure Finding Interval) for a
Protective Device (Section 3.4.6, Page 115)
Pressure safety valve

§ 1000 valves in service
§ Current inspection interval = 12 months
§ Time to inspect = 1 hour
§ Time to replace a defective valve = 1 hours
§ 10% of valves found defective at inspection, therefore
MTTF = (1000 x 1) /100 = 10 years
What is valve availability for the current inspection interval?
Answer
Failure Finding Pressure Valve Pressure Valve Availability Chart
102
Interval (Weeks) Availability (%) 100
98
1 99.9 96
94
92
Availability (%)
5 99.5 90
88
0 20 40 60 80 100 120
10 99.0
Inspection Interval (Weeks)
15 98.5
21 98.0
54 95.0
104 90.0
When to Apply FFI?
• Situation to apply FFI

– Failure is hidden (or not self-announced),
and
– Hidden failure may (with a significant
possibility) cause severe consequences, or
major disaster
C-More Consortium © A.K.S.25

JARDINE
Meeting, April 29, 2008
Scenarios When FFI Is Applied
• Two different scenarios when FFI is applied

– Protective device, its function is not the
main function, but just to protect the main
function
– Operating device, its function is the main
function

JARDINE
Different Consequences
• Hidden failure of a protective device does not have any consequence until a
failure of the protected function (system) occurs
No consequence Disaster
Protected system X X
Protective device X X
• Hidden failure of an operating device has immediate consequence
Immediate and continuous Immediate and continuous

consequence consequence
Operating device X X
C-More Consortium Meeting, April 29, 2008 27
Structure of FFI Optimizer
© A.K.S. JARDINE
The FFI Software

JARDINE
Condition Monitoring Decisions
© A.K.S. JARDINE
Optimising Condition Based Maintenance
(CBM) (Section 3.5, page 115)
Objective is to obtain the

maximum useful life from
each physical asset before
taking it out of service for
preventive maintenance...
RCM Methodology Logic
Condition-Based Maintenance
Condition Monitoring: An Analogy
HEART FAILURE
EQUIPMENT FAILURE
Hazard or Risk = f (Age) + f (Risk factors)
Risk
Riskfactors:
factors:
••Oil
Cholesterol
Analysis (Fe,
levelCu, Al,
• Cr,
Blood
Pb…..etc.)
pressure
••Vibration
Smoking (Velocity and
• Acceleration)
Lifestyle
••Thermography
Levels of protein
• Visual
Constituent
Inspection
…………………
Homocysteine
…………………
Real world research
Managing Risk: A CBM Optimization Tool
Condition Monitoring via Warning Limits
• Simple to
understand
• Limitations: Alarm > 300ppm
– Which
Warning > 200ppm
measurements?
– Optimal limits?
Normal < 200ppm
– Effect of Age?
– Predictions?
• CBM optimization
extends and
WorkingAge
enhances the Control
Chart technique
EXAKT Optimal Decision –
A New “Control Chart”
Irving Pulp and Paper
Analysis of Goulds 3175L Pumps

Bearings Vibration Data
– 56 vibration measurements
provided by accelerometer
Using <EXAKT>:
– 2 measurements significant
A Check:
– Had <EXAKT> model been
applied to previous histories
– Savings obtained = 33 %
EXAKT status today @ Irving Pulp & Paper
• Irving P&P have installed the software

for everyday use, and have developed a
tool to link the relevant databases.
• Steps have also been taken to make
small modifications to the pumps to
improve reliability.
• No pump failures in 22 months and
counting, after 15 failures in 5 years for
12 pumps.
Early work with PHM
Estimated Hazard Rate at Failure
Number Flight Hours Fe Cr Hazard Rate

1 11770 5 6 0.043
0.043
2 11660 2 6 0.012
3 8460 12 2.4 0.0071
4* 12630 8 1 0.0014
5 7710 8 0 0.00094
6* 9240 2 3 0.00029
7* 5660 10 1 0.00020
8* 7190 2 2.5 0.000073
*Doubtful Removal
The hazard rate equation is:
3.47
h(t,z) = 4.47 t (.41z1 +.98z )
) e 2
24100 ( 24100
where z1 is Fe concentration and z2 is Cr concentration
Data to Hazard
DATA PLOT
Data
Age
Hazard PLOT
Hazard PHM
Age
© A.K.S. JARDINE
Optimal Policy - Optimal Hazard Level
Hazard PLOT Ignore

Hazard Hazard
Optimal
Hazard
level
Age
COST PLOT Replace at
failure only
Cost/unit time
minimal cost
optimal
hazard Hazard
Optimizing CBM Decisions: EXAKT
Three Keys:
• Hazard
• Transition Probabilities
• Economics
From Data to Risk Model
• The probability of failure can change:

– As equipment ages
– As conditions change
• Risk Model uses data to determine:
– The likely contributions of age and
condition monitoring indicators to the
chance of failure
– How the condition monitoring indicators
evolve in time
Step 1: Hazard Model
β −1
β t  γ 1 z1 ( t ) + ...+ γ n z n ( t )
HAZARD (t ) =   e
η η 
Failures/op.hour Contribution of
Failure/flying hour condition information
Contribution of
Failures/km. to hazard
age to hazard
Failures/tonne
Failures/cycle
0.483
Etc… 1.483  t 
=   e 0.0518 Fe +1.867 Al + 0.01183 Mg
148790  148790 
Constants estimated from data
How does data model hazard?
Sample History History consists of:
•Beginning time
•Inspection results
•Ending time and reason
Lead and Silicon are

the significant
covariates.
Age Effect
EQUIPMENT FAILURE
47
Combine Data From All Histories
β −1
β  t 
exp γ z ( t ) + ... + γ z ( t )
HAZARD ( t ) = η  η 
 
{11 n n }
1.523
2.523  t 
=   exp {0.2293* Pb +0.4151* Si}
3402  3402 
Step 2: Model for Covariate Evolution
Sample
Lead History
(Pb)
Moderate
Probability
Small
Covariate Bands
Probability
Moderate
Probability
High
Probability
49
Step 2 (Cont.): Transition Probabilities
• Transition probabilities for all significant

covariates are calculated from all histories
• Probabilities can change over time
– e.g. probability of worsening wear can
increase with age.
Transition Probabilities
Inspection Interval = 30 days
VEL #1A
Age 0 0 0.1 0.15 0.22 Above
to 180 to to to to
(Days) 0.1 0.15 0.22 0.37 0.37
Very Smooth
0 to 0.1 0.575373 0.224181 0.145161 0.040554 0.014731
Smooth
0.1 to 0.15 0.205868 0.249818 0.330898 0.137414 0.076002
Rough
0.15 to 0.22 0.055426 0.137583 0.37788 0.229398 0.199714
Very Rough
0.22 to 0.37 0.012852 0.047422 0.190398 0.24338 0.50699
Failure
Above 0.37 0.00048 0.002696 0.017039 0.052114 0.927672
7/27/2010 © A.K.S. JARDINE

Data to Model: Summary
1.523
2.523 t  
  exp {0.2293* Pb +0.4151* Si} Hazard
3402 3402 
PLUS
Covariate
Evolution
AND
Cf = Total cost of failure

replacement
Optimal Hazard Level
Cp = Total cost of
preventive replacement
52
Optimal Decision Chart
Condition-Based Maintenance
See: www.omdec.com
EXAKT Procedures Window
Oil Analysis Data
Oil Analysis Events Data
Summary of PHM Parameters
Summary of PHM Parameters
Oil Analysis Decision
Decision Model
Vibration Monitoring Data
Vibration Analysis Events Data
Transition Probability Matrix
Inspection Interval = 30 days
VEL #1A
Age 0 0 0.1 0.15 0.22 Above
to 180 to to to to
(Days) 0.1 0.15 0.22 0.37 0.37
Very Smooth
0 to 0.1 0.575373 0.224181 0.145161 0.040554 0.014731
Smooth
0.1 to 0.15 0.205868 0.249818 0.330898 0.137414 0.076002
Rough
0.15 to 0.22 0.055426 0.137583 0.37788 0.229398 0.199714
Very Rough
0.22 to 0.37 0.012852 0.047422 0.190398 0.24338 0.50699
Failure
Above 0.37 0.00048 0.002696 0.017039 0.052114 0.927672
7/27/2010 © A.K.S. JARDINE

Vibration Monitoring Decision
Warning Limits in ‘ppm’
Normal Warning Alarm
Al < 20 20 - 40 > 40
Cr <10 10 - 20 >20
Cu <50 50 - 100 >100
Fe <200 200 - 300 >300
Si <15 15 - 25 >25
Measurements & Decision
In Operation
Cardinal River Coals
Oil Analysis data from 50 Wheel Motors
– Twelve covariates
measured
– Covariates used: Iron
and Sediment
– Estimated Saving in
Maintenance Costs:
22% for cost ratio 3:1
Sensitivity of Optimal Policy to Cost Ratio
Recent Developments in CBM Optimization
• Marginal Analysis: addressing

individual system failure modes
• Remaining Useful Life (RUL):
Reporting RUL and standard deviation
for selected state of covariates and
working age
• Probability of failure in a short
interval as a part of decision report
• Availability Optimization
• Availability and Cost Optimization
#1. Marginal Analysis: Modeling of Diesel
Engines employed on T23 Frigates
Diesel Engines: Failure Modes
Type of Failure Count

0 Not Known 3
1 Cooling System 13 Unrelated
2 Fuel System 6 to oil
readings
3 Generator 9
4 Accessories 9
5 Cylinder Liners and Rings 24
6 Valves and Running Gear 20 Possibly
7 Pistons, Articulations and Bearings 15
related to
oil
8 Cylinder Heads 14 readings
9 Misc, including cylinder block failures 12
Simultaneous decisions for each failure mode of a
repairable system
#2. Conditional Density Function &
Remaining Useful Life
• Shows the shape of the

distribution of the time
to failure given current
conditions
• Expected time to failure
(Remaining Useful Life,
or RUL)
• Standard deviation
#3. Probability of failure between
inspections
#4. New Criterion: Availability
• Preventive downtime: tp
• Failure downtime: tf
# 5. Cost and Availability Combined
• New dialog prompts

for:
– Fixed costs: Cp, Cf
– Downtimes: tp, tf
– Hourly costs: ap, af
Summary: Principle of CBM Optimization
Age Data
Maintenance
Condition- Decision
Monitoring Data
Hazard Model Cost and

Transition Model Availability Model
Supplied by user
Software engine
Intermediate result
Remaining Engineering
Final result Useful Life Judgment
CBM OPTIMIZATION
Executive Summaries
© A.K.S. JARDINE
Campbell Soup Company
Analysis of Shear Pump

– 21 vibration measurements
provided by accelerometer
Using <EXAKT>:
– 3 measurements significant
A Check:
– Had <EXAKT> model been
applied to previous histories
– Savings obtained = 35 %
Source: Jardine, AKS, Joseph, T and Banjevic, D, “Optimizing condition-based maintenance decisions for equipment
subject to vibration monitoring” , Journal of Quality in Maintenance Engineering, Vol. 5. No. 3, pp 192-202, 1999
Had we replaced at 175 days…..!!!
Failed at WorkingAge = 182 days
Inspection at
WorkingAge = 175 days
Hong Kong Mass Transit Railway
Corporation
– Had excessive traction – Reduced total costs by

motor ball bearing 55%
failures
– CBM to monitor bearing The reality:
grease colour Failures reduced to 2/yr.
– Changed inspections
from every 3.5 years to
annual
– Reduced failures/yr..
From 9 to 1
Sasol Plant
Analysis of Warman-Pump
– Total 8 pumps each with two bearings
(16 bearings) analyzed
– 12 vibration covariates identified
Using <EXAKT>:
– 2 covariates significant
– Annual replacement cost savings= 42 %
Feedback:
– Model results found realistic by Sasol
plant
– Significant vibration covariates identified Vlok, et al, "Optimal Component Replacement
by <EXAKT> are agreed as a major Decisions using Vibration Monitoring and the PHM",
JORS, 2002.
problem
Open Pit Mining Operation
CAT 793B Transmissions Oil data analyzed
– Covariates used:
Iron, Aluminum,
Magnesium
– Saving in
Maintenance Costs:
25%
– Average replacement
time increase: 13%
– Warranty limit could
be increased
Nuclear Generating Station
Hydrodyne seals
prevent leakage of
heavy water during
fuelling operation:
Seal Leak Rate data
from 4 reactors
Jardine, A.K.S., Kahn, K., Banjevic, D., Wiseman, M. and Lin, D., “An Optimized Policy for the Interpretation of Inspection Data
from a CBM Program at a Nuclear Reactor Station”, Proceedings, COMADEM, Sweden August 27-29, 2003
Maintenance and Diagnostic Data
EXAKT Modeling with MWM Diesel Engine

employed on Halifax Class ships.
Condition-based Optimal
Replacement Policy
CBM Optimization Studies: A summary
Additional CBM References
1. www.mie.utoronto.ca/cmore (For
information about the CBM research
activities at the University of Toronto)
2. www.omdec.com (For information

about the EXAKT: CBM Optimization
software)
EXAKT Tutorials
Learn the fundamentals of EXAKT by going through the EXAKT
tutorial for single items, i.e.,components or systems as a whole:
www.omdec.com/articles/p_exaktTutorial.html
Link to advanced EXAKT tutorial for complex items,

e.g., systems consisting of components with different
failure modes:
www.omdec.com/articles/p_ExaktTutorialComplexItems.html
Note: Just follow the instructions on the web pages.
Maintenance Management
Inspection Decisions
Problems: Pages 130-134
Answers: Page 311

3-Chapter Three Inspection Decisions 2010 (Compatibility Mode)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

3-Chapter Three Inspection Decisions 2010 (Compatibility Mode)

Uploaded by

Copyright:

Available Formats

Chapter 3

Optimizing Equipment Maintenance and Replacement Decisions

Component Inspection Capital Equipment Resource

Maintenance Management System (CMMS/EAM/ERP)

1. Inspection frequencies for equipment which is in

2. Inspection intervals for equipment used only in

3. Condition monitoring of equipment: Optimizing

• Haul trucks : A,B,C, D inspections at

Total Downtime versus Inspection Frequency

Downtime due to Inspections

Total Downtime (D)

Optimal inspection frequency minimizes total downtime, D(n)

Mean time 5000

4500 5500 6500 7500 8500

Inspection interval (km)

Optimal Inspection Interval

“Failure-finding applies only to hidden or unrevealed failures. Hidden failures in

1. Fire extinguisher (monthly)

1. Pressure relief valve

• The problem is to determine the best interval between

Where: Ti = time required to effect an inspection, Tr = time required to

Therefore, the optimal inspection interval is 3 months

100 ti × R(ti ) + ∫ t × f (t )dt

Source: Moubray, RCM II, pp. 177

Problem: Brake lights on motorcycles

Result: Availability = 95.0%

Problem source : Moubray, RCM II., 1999, pp. 175-176

Problem: A protective device has a MTBF = 10

Answer: FFI = 10 weeks (roughly)

Pressure safety valve

Failure Finding Pressure Valve Pressure Valve Availability Chart

• Situation to apply FFI

C-More Consortium © A.K.S.25

• Two different scenarios when FFI is applied

C-More Consortium © A.K.S.26

Immediate and continuous Immediate and continuous

C-More Consortium © A.K.S.29

Objective is to obtain the

Analysis of Goulds 3175L Pumps

• Irving P&P have installed the software

Number Flight Hours Fe Cr Hazard Rate

Hazard PLOT Ignore

• The probability of failure can change:

Constants estimated from data

Lead and Silicon are

• Transition probabilities for all significant

7/27/2010 © A.K.S. JARDINE

Cf = Total cost of failure

7/27/2010 © A.K.S. JARDINE

Normal Warning Alarm

Cu <50 50 - 100 >100

Fe <200 200 - 300 >300

• Marginal Analysis: addressing

Type of Failure Count

• Shows the shape of the

• New dialog prompts

Hazard Model Cost and

Analysis of Shear Pump

Failed at WorkingAge = 182 days

– Had excessive traction – Reduced total costs by

CAT 793B Transmissions Oil data analyzed

EXAKT Modeling with MWM Diesel Engine

2. www.omdec.com (For information