Handbook of Reliability, Availability, Maintainability and Safety in Engineering Design - Part 31 docx

3.4 Application Modelling of Reliability and Performance in Engineering Design 283
Fig. 3.83 Design specification FMECA—final absorption tower
Total failures + suspensions = 72
Mean time to failure (MTTF) = 2.35 (days)
The Kolmogorov–Smirnovgoodness-of-fittest The Kolmogorov–Smirnov(K–S)
test is used to decide if a sample comes from a population with a specific distribu-
tion. The K–S test is based o n the empirical distribution function (e.c.d.f.) whereby,
given N ordered data points Y
1
,Y
2
, Y
N
, the e.c.d.f. is defined as:
EN = n(i)/N , (3.212)
where n(i) is the number of points less than Y
i
,andtheY
i
are ordered from smallest
to largest value. This is a step function that increases by 1/N at the value of each
ordered data point. An attractive feature of this test is that the distribution of the
K–S test statistic itself does not depend on the cumulativedistribution function being
tested. Another advantage is that it is an exact test; however, the goodness-of-fit test
depends on an adequate sample size for the approximations to be valid. The K–S
test h as several important limitations, specifically:
• It applies only to continuous distributions.
• It tends to be more sensitive near the centre o f the distribution than at the tails.
284 3 Reliability and Performance in Engineering Design
Table 3.26 Acid plant failure data (repair time RT and time before failure TBF)
Failure time RT

(min)
TBF
(day)
Failure time RT
(min)
TBF
(day)
Failure time RT
(min)
TBF
(day)
7/28/01 0:00 38 0 9/25/01 0:00 31 5 11/9/01 0:00 360 1
7/30/01 0:00 35 2 9/27/01 0:00 79 2 11/10/01 0:00 430 1
7/31/01 0:00 148 1 9/29/01 0:00 346 2 11/20/01 0:00 336 10
8/1/01 0:00 20 1 9/30/01 0:00 80 1 11/26/01 0:00 175 6
8/5/01 0:00 27 4 10/1/01 0:00 220 1 11/28/01 0:00 118 2
8/7/01 0:00 15 2 10/4/01 0:00 63 3 12/1/01 0:00 35 3
8/11/01 0:00 5 4 10/7/01 0:00 176 3 12/2/01 0:00 556 1
8/12/01 0:00 62 1 10/8/01 0:00 45 1 12/5/01 0:00 998 3
8/13/01 0:00 580 1 10/10/01 0:00 52 2 12/6/01 0:00 124 1
8/14/01 0:00 897 1 10/10/01 0:00 39 0 12/11/01 0:00 25 5
8/15/01 0:00 895 1 10/11/01 0:00 55 1 12/12/01 0:00 120 1
8/16/01 0:00 498 1 10/12/01 0:00 36 1 12/17/01 0:00 35 5
8/17/01 0:00 308 1 10/14/01 0:00 10 2 12/26/01 0:00 10 9
8/19/01 0:00 21 2 10/18/01 0:00 1,440 4 1/2/02 0:00 42 7
8/21/01 0:00 207 2 10/19/01 0:00 590 1 1/18/02 0:00 196 16
8/22/01 0:00 346 1 10/22/01 0:00 43 3 1/29/02 0:00 22 11
8/23/01 0:00 110 1 10/24/01 0:00 107 2 2/9/02 0:00 455 11
8/25/01 0:00 26 2 10/29/01 0:00 495 5 2/10/02 0:00 435 1
8/28/01 0:00 15 3 10/30/01 0:00 392 1 2/13/02 0:00 60 3

9/4/01 0:00 41 7 10/31/01 0:00 115 1 2/13/02 0:00 30 0
9/9/01 0:00 73 5 11/1/01 0:00 63 1 2/17/02 0:00 34 4
9/12/01 0:00 134 3 11/2/01 0:00 245 1 2/24/02 0:00 71 7
9/19/01 0:00 175 7 11/4/01 0:00 40 2 3/4/02 0:00 18 8
9/20/01 0:00 273 1 11/8/01 0:00 50 4 3/9/02 0:00 23 5
• The distribution must be fully specified—that is, if location, scale, and sha pe
parameters are estimated from the data, the critical region of the K–S test is no
longer valid, and must be determined by Monte Carlo (MC) simulation.
Goodness-of-fit results The K–S test r esult of the acid plant data given in
Table 3.26 is the following:
Kolmogorov-Smirnov(D) statistic = 347
Modified D statistic = 2.514
Critical value of modified D = 1.094
Confidence levels = 90% 95% 97.5% 99%
Tabled values of K–S statistic = 0.113 0.122 0.132 0.141
Observed K–S statistic = 325
Mean absolute prob. error = 0.1058
Model accuracy = 89.42% (poor)
The hypothesis that the data fit the two-Weibull distribution is rejected with 99%
confidence.
3.4 Application Modelling of Reliability and Performance in Engineering Design 285
Fig. 3.84 Weibull distribution chart for failure data
Three-parameter Weibull fit—ungrouped data (Fig. 3.84):
Minimum life = 0.47 (days)
Shape parameter BETA = 1.63
Scale parameter ETA = 1.74 (days)
Mean life = 2.03 (days)
Characteristic life = 2.21 (days)
Standard deviation = 0.98 (days)
Test for random failures The hypothesis that failures are random is rejected at 5%

level.
3.4.3 Application Modelling Outcome
The acid plant failure data do not suitably fit the Weibull distribution, with 89%
model accuracy. However, the failures are not random (i.e. the failure rate is not
constant), and it is essential to determine whether failures are in the early phase or
in the wear-out phase of the plant’s life cycle—especially so soon af ter its installa-
286 3 Reliability and Performance in Engineering Design
tion (less than 24 mon ths). The distribution must be fully specified—that is, the K–S
test is no longer valid, and must be determined by Monte Carlo (MC) simulation.
However, prior to simulation, a closer definition of the source of most of the failures
of the critical systems (determined through the case study FMECA) is necessary.
Table 3.27 shows the total downtime of the acid plant’s critical systems. The down-
time failure data grouping indicates that the highest downtime is due to the hot gas
feed induced draft fan, then the reverse jet scrubber, the drying tower blowers, and
final absorption.
Engineered Installation Downtime
Table 3.27 Total downtime of the environmental plant critical systems
Downtime reason description Total hours Direct hours Indirect hours
Hot gas feed, hot gas fan total 1,514 1,388 126
Gas cleaning, RJS total 680 581 99
Drying to wer, SO2 blowers total 496 248 248
Gas absorption, final absorption total 195 100 95
Total 2,885 2,317 568
Monte Carlo simulation With the K–S test, the distribution of the f ailure data must
be fully specified—that is, if location,scale and shapeparameters are estimated from
the data, the critical regionof the K–S test is no longer valid,and must be determined
by Monte Carlo (MC) simulation.
MC simulation emulates the chance variations in the critical systems’ time be-
fore failure (TBF) by generating random numbers that form a uniform distribution
that is used to select values from the samp le TBF data, and for which various TBF

values are established to develop a large population of representative sample data.
The model then determines if the representative sample data come from a popula-
tion with a specific distribution (i.e. exponential, Weibull or gamma distributions).
The outcome of the M C simulation gives the following distribution parameters
(Tables 3.28 and 3.29):
Time Between Failure Distribution
Table 3.28 Values of distribution models for time between failure
Distribution model Parameter Parameter value
1. Exponential model Gamma 4.409E-03
2. Weibull model Gamma 1.548E+00
Theta 3.069E+02
3. Gamma model Gamma 7.181E-01
Theta 3.276E+02
3.4 Application Modelling of Reliability and Performance in Engineering Design 287
Repair Time Distribution
Table 3.29 Values of distribution models for repair time
Distribution model Parameter Parameter value
1. Exponential model Gamma 2.583E-01
2. Weibull model Gamma 8.324E-01
Theta 3.623E+00
3. Gamma model Gamma 4.579E-01
Theta 8.720E+00
The results of the MC simulation are depicted in Fig. 3.85. The representative sam-
ple data come from a population with a gamma distribution, as illustrated. The me-
dian (MTTF) of the representative data is given as approximately 2.3, which does
not differ greatly from the MTTF for the three-parameter Weibull distribution for
ungrouped data, which equals 2.35 (days). This Weibull distribution has a shape pa-
rameter, BETA, o f 1.63, which is greater than 1 , indicating a wear-out condition in
the plant’s life cycle.
Fig. 3.85 Monte Carlo simulation spreadsheet results for a gamma distribution best fit of TBF data

288 3 Reliability and Performance in Engineering Design
Conclusion From the case stu dy data, the assumption can be made that the critical
systems’ specific high-ranking critical components are inadequately designed from
a design integrity point of view, as they indicate wear-out too early in the plant’s
life cycle. This is with reference to the items listed in Table 3.25, particularly the
drying tower blowers’ shafts, bearings (PLF) and scroll housings (TLF), the hot gas
feed induced draft fan (PFC), the reverse jet scrubber’s acid spray nozzles (TLF),
the final absorption tower vessel and cooling fan guide vanes (TLF), and the IPAT
SO3 cooler’s cooling fan control vanes (TLF).
Figure 3.85 shows a typical Monte Carlo simulation spreadsheet of the critical
systems’ time before failure and MC results for a gamma distribution best fit of TBF
data.
3.5 Review Exercises and References
Review Exercises
1. Discuss total cost models for design reliability with regard to risk cost estimation
and project cost estimation.
2. Give a brief account of interference theory and reliability modelling.
3. Discuss system reliability modelling based on system per formance.
4. Compare functional failure and functional performance.
5. Consider the significance of functional failure and reliability.
6. Describe the benefits of a system breakdown structure (SBS).
7. Give reasons for the application and benefit of Markov modelling (continuous-
time an d discrete states) in designing for reliability.
8. Discuss the binomial method with regard to series networks and parallel net-
works.
9. Give a brief account of the principal steps in failure modes and effects analysis
(FMEA).
10. Discuss the different types of FMEA and their associated benefits.
11. Discuss the advantages and disadvantages of FMEA.
12. Compare the significant differences between failure modes and effects analysis

(FMEA) and failure modes and effects criticality analysis (FMECA) .
13. Compare the advantages and disadvantages of the RPN technique with those of
the military standard technique.
14. Discuss the relevance of FMECA data sources and users.
15. Consider the significance of fault-tree analysis (FTA) in reliability, safety and
risk assessment.
16. Describe the fundamental fault-tree analysis steps.
17. Explain the basic properties of the hazard rate function and give a brief descrip-
tion of the main elements of the hazard rate curve.
18. Discuss component reliability and failure distributions.
3.5 Review Exercises and References 289
19. Define the application of the exponential failure distribution in reliability anal-
ysis and discuss the distribution’s statistical properties.
20. Define the application of the Weibull failure distribution in reliability analysis
and discuss the distribution’s statistical properties.
21. Explain the Weibull shape parameter and its use.
22. Discuss the significance of the Weibull distribution function in hazards analysis.
23. Describe the principal properties and use of the Weibull graph chart.
24. Consider the a pplication o f reliability evaluation of two-state device networks.
25. Describe the fundamental differences between two-state device series networks,
parallel networks, and k-out-of-m unit networks.
26. Consider the application of reliability evaluation of three-state device networks.
27. Briefly describe three-state device parallel networksand three-state device series
networks.
28. Discuss system performance measures in designing for r eliability.
29. Consider pertinent approaches to determination of the most reliable design in
conceptual design.
30. Discuss conceptual design optimisation.
31. Describe the basic comparisons of conceptual designs.
32. Define labelled interval calculus (LIC) with regard to constraint labels, set la-

bels, and labelled interval inferences.
33. Consider the application of labelled interval calculus in designing for reliability.
34. Give a brief description with supporting examples of the methods for:
a. Determination of a data point: two sets of limit intervals.
b. Determination of a data point: one upper limit interval.
c. Determination of a data point: one lower limit interval.
d. Analysis of the interval matrix.
35. Give reasons for the application of FMEA and FMECA in engineering design
analysis.
36. Define reliability-critical items.
37. Describe algorithmic modelling in failure mod es and effects analysis with re-
gard to numerical analysis, order of magnitude, qualitative simulation, and fuzzy
techniques.
38. Discuss qualitative reasoning in failure modes and effects analysis.
39. Give a brief accountof the conceptof uncertaintyin engineeringdesign analysis.
40. Discuss uncertainty and incompleteness in knowledge.
41. Give a brief overview of fuzziness in engineering design analysis.
42. Describe fuzzy logic and fuzzy reasoning in engineering design.
43. Define the theory of approximate reasoning.
44. Consider uncertainty and incompleteness in design analysis.
45. Give a brief account of modelling uncertainty in FMEA and FMECA.
46. In the development o f the qualitative FMECA, describ e the concepts of lo gical
expression and expression of uncertainty in FMECA.
47. Give an example of uncertainty in the extended FMECA.
48. Describe the typical results expected of a qualitative FMECA.
290 3 Reliability and Performance in Engineering Design
49. Define the proportionalhazards model with regard to non-parametric model for-
mulation and parametric model formulation.
50. Define the maximum likelihood estimation parameter.
51. Briefly describe the characteristics of the one-parameter exponential distribu-

tion.
52. Explain the process of estimating the parameter of the exponential distribution.
53. Consider the approach to determining the maximum likelihood estimation
(MLE) parameter.
54. Compare the characteristics of the two-parameter Weibull distribution with
those of the three-parameter Weibull model.
55. Give a brief account of the procedures to calculate the Weibull parameters
β
,
μ
and
γ
.
56. Describe the procedure to derive the mean time between failures (MTBF)
μ
from the Weibull distribution model.
57. Describe the procedure to obtain the standard deviation
σ
from the Weibull
distribution model.
58. Give a brief account of the method of qualitative analysis of the Weibull distri-
bution model.
59. Consider expert judgment as data.
60. Discuss uncertainty, probability theory and fuzzy logic in design ing f or reliabil-
ity.
61. Describe the application of fu zzy logic in reliability evaluation.
62. Describe the application of fuzzy judgment in reliability evaluation.
63. Give a brief account of elicitation and analysis of expert judgment in designing
for reliability.
64. Explain initial reliability calculation using Monte Carlo simulation.

65. Give an example of fuzzy judgment in reliability evaluation.
References
Abernethy RB (1992) New methods for Weibull and log normal analysis. ASME Pap no 92-
WA/DE-14, ASME, New York
Agarwala AS (1990) Shortcomings in MIL-STD-1629A: guidelines for criticality analysis. In: Re-
liability Maintainability Symp, pp 494–496
AMCP 706-196 (1976) Engineering design handbook: development guide for reliability. Part II.
Design for reliability. Army Material Command, Dept of the Army, Washington, DC
Andrews JD, Moss TR (1993) Reliability and risk assessment. American Society of Mechanical
Engineers
Artale A, Franconi E (1998) A temporal description logic for reasoning about actions and plans.
J Artificial Intelligenc e Res JAIR, pp 463–506
Ascher W (1978) Forecasting: an appraisal for policymakers and planners. John Hopkins Univer -
sity Press, Baltimore, MD
Aslaksen E, Belcher R (1992) Systems engineering. Prentice Hall of Australia
Barnett V (1973) Comparative statistical inference. Wiley, Ne w York
Barringer PH (1993) Reliability engineering principles. Barringer, Humble, TX
Barringer PH (1994) Management overview: reliability engineering principles. Barringer, Hum-
ble, TX
3.5 Review Exercises and References 291
Barringer PH, Weber DP (1995) Data for making reliability improvements. Hydrocarbons Process-
ing Magazine, 4th Int Reliability Conf, Houston, TX
Batill SM, Renaud JE, Xiaoyu Gu (2000) Modeling and simulation uncertainty in multidisciplinary
design optimization. In: 8th AIAA/NASA/USAF/ISSMO Symp Multidisciplinary Analysis and
Optimisation, AIAA, Long Beach, CA, AIAA-200-4803, pp 5–8
Bement TR, Booker JM, Sellers KF, Singpurwalla ND (2000a) Membership functions and proba-
bility measures of fuzzy sets. Los Alamos Nat Lab Rep LA-UR-00-3660
Bement TR, Booker JM, Keller-McNulty S, Singpurwalla ND (2000b) Testing the untestable: re-
liability in the 21st century. Los Alamos Nat Lab Rep LA-UR-00-1766
Bennett BM, Hoffman DD, Murthy P (1992) Lebesgue order on probabilities and some applica-

tions to perception. J Math Psychol
Bezdek JC (1993) Fuzzy models—what are they and why? IEEE Transactions Fuzzy Systems
vol 1, no 1
Blanchard BS, Fabrycky WJ (1990) Systems engineering and analysis. Prentice Hall, Englewood
Cliffs, NJ
Boettner DD, Ward AC (1992) Design compilers and the labeled interval calculus. In: Tong C,
Sriram D (eds) Design representation and models of routine design. Artificial Intelligence in
Engineering Design vol 1. Academic Press, San Diego, CA, pp 135–192
Booker JM, Meyer MA (1988) Sources and effects of inter-expert correlation: an empirical study.
IEEE Trans Systems Man Cybernetics 8(1):135–142
Booker JM, Smith RE, Bement TR, Parkinson WJ, Meyer MA (1999) Example of using fuzzy
control system methods in statistics. Los Alamos Natl Lab Rep LA-UR-99-1712
Booker JM, Bement TR, Meyer MA, Kerscher WJ (2000) PREDICT: a new approach to product
development and lifetime assessment using information integration technology. Los Alamos
Natl Lab Rep LA-UR-00-4737
Bowles JB, Bonnell RD (1994) Failure mode effects and criticality analysis. In: Annual Reliability
and Maintainability Symp, pp 1–34
Brännback M (1997) Strategic thinking and active decision support systems. J Decision Systems
6:9–22
BS5760 (1991) Guide to failure modes, effects and criticality analysis (FMEA and FMECA).
British Standard BS5760 Part 5
Buchanan BG, Shortliffe EH (1984) Rule-based expert systems. Addison-Wesley, Reading, MA
Buckley J, Siler W (1987) Fuzzy operators for possibility interval sets. Fuzzy Sets Systems 22:215–
227
Bull DR, Burrows CR, Crowther WJ, Edge KA, Atkinson RM, Hawkins PG, Woollons DJ (1995a)
Failure modes and effects analysis. Engineering and Physical Sciences Research Council
GR/J58251 and GR/J88155
Bull DR, Burrows CR, Cro wther WJ, Edge KA, Atkinson RM, Hawkins PG, Woollons DJ (1995b)
Approaches to automated FMEA of hydraulic systems. In: Proc ImechE Congr Aerotech 95
Seminar , Birmingham, Pap C505/9/099

Carlsson C, Walden P (1995a) Active D SS and hyperknowledge: creating strategic visions. In: Proc
EUFIT’95 Conf, Aachen, Germany, August, pp 1216–1222
Carlsson C, Walden P (1995b) On fuzzy hyperknowledge support systems. In: Proc 2nd Int Worksh
Next Generation Information Technologies and Systems, Naharia, Israel, June, pp 106–115
Carlsson C, Walden P (1995c) Re-engineering strategic management with a hyperknowledge sup-
port system. In: Christiansen JK, Mouritsen J, Neergaard P, Jepsen BH (eds) Proc 13th Nordic
Conf Business Studies, Denmark, vol II, pp 423–437
Carter ADS (1986) Mechanical reliability. Macmillan Press, London
Carter ADS (1997) Mechanical reliability and design. Macmillan Press, London
Cayrac D, Dubois D, Haziza M, Prade H (1994) Possibility theory in fault mode effects analyses—
a satellite fault diagnosis application. In: Proc 3rd IEEE Int Conf Fuzzy Systems FUZZ-
IEEE ’94, Orlando, FL, June, pp 1176–1181
292 3 Reliability and Performance in Engineering Design
Cayrac D, Dubois D, Prade H (1995) Practical model-based diagnosis with qualitative possibilistic
uncertainty. In: Besnard P, Hanks S (eds) Proc 11th Conf Uncertainty in Artificial Intelligence,
pp 68–76
Cayrol M, Farency H, Prade H (1982) Fuzzy pattern matching. Kybernetes, pp 103–106
Chiueh T (1992) Optimization of fuzzy logic inference architecture. Computer, May, pp 67–71
Coghill GM, Chantler MJ (1999a) Constructive and non-constructive asynchronous qualitative
simulation. In: Proc Int Worksh Qualitative Reasoning, Scotland
Coghill GM, Shen Q, Chantler MJ, Leitch RR (1999b) Towards the use of multiple models for
diagnoses of dynamic systems. In: Proc Int Worksh Principles of Diagnosis, Scotland
Conlon JC, Lilius WA (1982) Test and evaluation of system reliability, availability and maintain-
ability. Office of the Under Secretary of Defense for Research and Engineering, DoD 3235.1-H
Cox DR (1972) Regression models and life tables (with discussion). J R Stat Soc B 34:187–220
Davis E (1987) Constraint propagation with interval labels. Artificial Intelligen c e 32:281–331
de Kleer J, Brown JS (1984) A qualitative physics based on confluences. Artificial Intelligence
24:7–83
Dhillon BS (1983) Reliability engineering in systems design and operation. Van Nostrand Rein-
hold, Berkshire

Dhillon BS (1999a) Design reliability: fundamentals and applications. CRC Press, LLC 2000,
NW Florida
Dubois D, Prade H (1988) Possibility theory—an approach to computerized processing of uncer-
tainty. Plenum Press, New York
Dubois D, Prade H (1990) Modelling uncertain and vague knowledge in possibility and evidence
theories. Uncertainty in Artificial Intelligence vol 4. Elsevier, Amster da m, pp 303–318
Dubois D, Prade H (1992a) Upper and lower images of a fuzzy set induced by a fuzzy relation:
applications to fuzzy inference and diagnosis. Information Sci 64:203–232
Dubois D, Prade H (1992b) Fuzzy rules in knowledge-based systems modeling gradedness, un-
certainty and preference. In: Zadeh LA (ed) An introduction to fuzzy logic applications in
intelligent systems. Kluwer, Dordrecht, pp 45–68
Dubois D, Prade H (1992c) Gradual inference rules in approximate reasoning. Information Sci
61:103–122
Dubois D, Prade H (1992d) When upper probabilities are possibility measures. Fuzzy Sets Systems
49:65–74
Dubois D, Prade H (1993a) Fuzzy sets and probability: misunderstandings, bridges and gaps. Re-
port (translated), Institut de Recherche en Informatique de Toulouse (I.R.I.T.) Université Paul
Sabatier, Toulouse
Dubois D, Prade H (1993b) A fuzzy relation-based extension of Reggia’s relational model for diag-
nosis. In: Heckerman, Mamdani (eds) Proc 9th Conf Uncertainty in Artificial Intelligence, WA,
pp 106–113
Dubois D, Prade H, Yager RR (1993) Readings in fuzzy sets and intelligent systems. Morgan
Kaufmann, San Mateo, CA
Dubois D, Lang J, Prade H (1994) Automated reasoning using possibilistic logic: semantics, belief
revision and variable certainty weights. IEEE Trans Knowledge Data Eng 6:64–69
EPRI (1974) A review of equipment aging theory and technology. Nuclear Safety & Analysis
Department, Nuclear Power Division, Electricity Power Research Institute, Palo Alto, CA
Fishburn P (1986) The axioms of subjective probability. Stat Sci 1(3):335–358
Fullér R (1999) On fuzzy reasoning schemes. In: Carlsson C (ed) The State of the Art of Infor-
mation Systems in 2007. Turku Centre for Computer Science, Abo, TUCS Gen Publ no 16,

pp 85–112
Grant Ireson W, Coombs CF, Moss RY (1996) Handbook of reliability engineering and manage-
ment. McGraw-Hill, New York
ICS (2000) The RAMS plant analysis model. ICS Industrial Consulting Services, Gold Coast City,
Queensland
IEEE Std 323-1974 (1974) IEEE Standard for Qualifying Class IE Equipment for Nuclear Power
Generating Stations. Institute of Electrical and Electronics Engineers, New York