Professional Documents
Culture Documents
Safety and Reliability Modeling and Its Applications Advances in Reliability Science 1St Edition Mangey Ram Editor Full Chapter PDF
Safety and Reliability Modeling and Its Applications Advances in Reliability Science 1St Edition Mangey Ram Editor Full Chapter PDF
https://ebookmass.com/product/total-manufacturing-assurance-
controlling-product-quality-reliability-and-safety-2nd-edition-
john-cesarone/
https://ebookmass.com/product/system-reliability-assessment-and-
optimization-methods-and-applications-1st-edition-yan-fu-li-
enrico-zio/
https://ebookmass.com/product/advances-in-system-reliability-
engineering-davim/
https://ebookmass.com/product/photovoltaic-module-reliability-
wohlgemuth/
Engineering Reliability and Risk Assessment Harish Garg
https://ebookmass.com/product/engineering-reliability-and-risk-
assessment-harish-garg/
https://ebookmass.com/product/food-safety-and-human-health-ram-
lakhan-singh/
https://ebookmass.com/product/reliability-investigation-of-led-
devices-for-public-light-applications-1st-edition-edition-
raphael-baillot-and-yannick-deshayes-auth/
https://ebookmass.com/product/lead-free-soldering-process-
development-and-reliability-jasbir-bath/
Series Editor: Mangey Ram, Professor at Graphic Era University, Dehradun, India
Edited by
Mangey Ram
Graphic Era (Deemed to be University), Dehradun, India
Hoang Pham
Rutgers University, New Jersey, United States
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
Preface xi
Acknowledgement xiii
About the Editors xv
List of Contributors xvii
Appendix 277
References 280
Index 409
Preface
The editors acknowledge Elsevier and the editorial team for their adequate
and professional support during the preparation of this book. Also, we would like
to acknowledge all the chapter authors and the reviewers for their availability to
work on this book project.
Mangey Ram
Graphic Era (Deemed to be University), India
Hoang Pham
Rutgers University, USA
About the Editors
Prof. Dr. Mangey Ram received the Ph.D. degree major in Mathematics and
minor in Computer Science from G. B. Pant University of Agriculture and Tech-
nology, Pantnagar, Uttarakhand, India. He has been a faculty member for around
twelve years and has taught several core courses in pure and applied mathe-
matics at undergraduate, postgraduate, and doctorate levels. He is currently the
Research Professor at Graphic Era (Deemed to be University), Dehradun, India.
Before joining the Graphic Era, he was a deputy manager (probationary officer)
with Syndicate Bank for a short period. He is the editor-in-chief of International
Journal of Mathematical, Engineering and Management Sciences, Journal of
Reliability and Statistical Studies; the editor-in-chief of six Book Series with
Elsevier, CRC Press-A Taylor and Francis Group, Walter De Gruyter Publisher
Germany, River Publisher; and the guest editor and member of the editorial
board of various journals. He has published more than 250 research publications
(journal articles/books/book chapters/conference articles) in IEEE, Taylor &
Francis, Springer, Elsevier, Emerald, World Scientific, and many other national
and international journals and conferences. Also, he has authored/edited more
than 50 books for international publishers such as Elsevier, Springer Nature,
CRC Press-A Taylor and Francis Group, Walter De Gruyter Publisher Germany,
and River Publisher. His fields of research are reliability theory and applied
mathematics. Dr. Ram is a Senior Member of the IEEE, Senior Life Member
of Operational Research Society of India; Society for Reliability Engineering,
Quality and Operations Management in India; Indian Society of Industrial and
Applied Mathematics. He has been a member of the organizing committee of
a number of international and national conferences, seminars, and workshops.
He has been conferred with “Young Scientist Award” by the Uttarakhand State
Council for Science and Technology, Dehradun, in 2009. He has been awarded
the “Best Faculty Award” in 2011, “Research Excellence Award” in 2015, and
“Outstanding Researcher Award” in 2018 for his significant contributions in
academics and research at Graphic Era Deemed to be University, Dehradun,
India.
Specialist with the Idaho National Engineering Laboratory and Boeing Com-
pany. He received his Ph.D. in Industrial Engineering from the State University
of New York at Buffalo. His research areas include reliability modeling of
systems with competing risks and random environments, software reliability,
and statistical inference. He is the editor-in-chief of the International Journal of
Reliability, Quality and Safety Engineering and an associate editor and editorial
board member of several journals, and the editor of Springer Series in Reliability
Engineering. His numerous awards include the 2009 IEEE Reliability Society
Engineer of the Year Award. Dr. Pham is the author/coauthor of 7 books and
has published his work in over 190 journal articles, 100 conference papers,
and edited 18 books including Springer Handbook in Engineering Statistics and
Handbook in Reliability Engineering. He has delivered over 40 invited keynote
and plenary speeches at many international conferences and institutions. He is a
Fellow of the IEEE and IIE.
List of Contributors
1.1 Preamble
The development of infrastructure, in particular the transportation sector, plays
a significant role in the economic growth of any country. The economic growth
demands a good road network with good connectivity all over the country.
With the reduced availability of funds, the highway agencies are placing more
emphasis on the design and construction of pavements that require minimum
maintenance during the service life. For this, it is necessary that pavements
should be designed such that a minimum design reliability (as specified in the
country’s national specifications) is achieved and the pavement construction
should be done with the latest machinery and under stringent quality control
requirements.
In India, majority of the roads (more than 90 %) are asphalt pavements,
popularly known as flexible pavements. This is due to their low construction
cost (in comparison to cement concrete/rigid pavements), ease of maintenance,
and relatively easier construction procedure. To ensure that the pavement has
adequate strength to cater to the expected traffic, it has to be designed properly
in accordance with the national specifications. For example, IRC:37 (2018) is
followed for the design of flexible pavements in India. A flexible pavement
is a multilayer structure consisting of many layers of materials starting from
subgrade at the bottom to bituminous wearing course at the top. A typical three
layer pavement structure is presented in Fig. 1.1.
The structural design of pavements deals with determining the thicknesses
of the various component layers keeping in consideration the material prop-
erties and the amount of traffic which is expected during the design life. The
current India pavement design procedure (IRC:37, 2018) is a deterministic one
wherein the various input variables like layer thicknesses, Poisson’s ratio, elastic
Safety and Reliability Modeling and Its Applications. DOI: 10.1016/B978-0-12-823323-8.00009-X
Copyright © 2021 Elsevier Inc. All rights reserved. 1
2 Safety and reliability modeling and its applications
FIG. 1.1 A typical three layer pavement structure (Dilip et al., 2013)
modulus, and design traffic are all considered as fixed. However, in reality none
of them are deterministic; they all are stochastic (probabilistic). So, in order
to develop reliable pavement designs, the uncertainty/variability of the input
variables need to be considered in the design process. This can be addressed
through the use of reliability concepts within the pavement design process. The
details of the reliability concepts and its applications to the pavement design
process, specifically in the context of flexible pavements, are discussed in the
following sections.
Where,
t = the designated period of time or cycles for the system’s operation
T = time to failure or cycle to failure
R = reliability of the system
c1 , c2 ,…. = designated conditions, such as environmental conditions
Often, in practice, the designated operating conditions for a system c1 , c2 ,
….. are implicitly considered in the probabilistic reliability analysis and thus (1)
reduces to
R = P(T ≥ t ) (1.2)
Reliability analysis of asphalt pavements: concepts and applications Chapter | 1 3
So, R = 1 − Pf (1.3)
Where, X is the vector of basic random variables, g(X) is the limit state (or
failure) function for the failure mode considered and fX (x) is the joint probability
density function of the vector X. The expression g(X) < 0 indicates a failure
domain; g(X) > 0 indicates safe domain and g(X) = 0 denotes a failure surface.
Therefore, estimation of reliability requires the solution of a multidimen-
sional integral that can rarely be solved analytically. For this reason, other
methods such as numerical integration become essential. This might not be
practically feasible in probabilistic analysis because of the multi - dimensional
nature of the problem wherein a dimension is associated for each basic variable,
and the area of interest is usually in the tails of the distributions (Cronvall, 2011).
The uncertainty from all the sources which may affect that failure of the
component (or system) should be considered for a rigorous structural reliability
assessment.
This clearly involves taking into account all fundamental quantities entering
the problem, and also the uncertainties that arise from lack of knowledge
and idealized modeling. The structural reliability procedure is outlined by the
following steps (Cronvall, 2011):
(a) Identify all significant modes of failure of the structure or operation under
consideration, and define failure events.
(b) Formulate a failure criterion or failure function for each failure event.
(c) Identify the sources of uncertainty influencing the failure of the events,
model the basic variables and parameters in the failure functions and
specify their probability distributions.
(d) Calculate the probability of failure or reliability for each failure event,
and combine these probabilities where necessary to evaluate the failure
probability or reliability of the structural system.
(e) Consider the sensitivity of the reliability results to the input, such as basic
variables and parameters.
(f ) Assess whether the evaluated reliability is sufficient in comparison with a
target.
4 Safety and reliability modeling and its applications
(a) Level I methods : Reliability methods that employ only one ‘characteris-
tic’ value of each uncertain parameter are called level I methods. Examples
include load and resistance factor formats, including the allowable stress
formats.
(b) Level II methods : Reliability methods that employ two values of each
uncertain parameter (commonly mean and variance), supplemented with
a measure of the correlation between the parameters (usually covariance),
are called level II methods. Reliability index methods are examples of level
II methods.
(c) Level III methods : Reliability methods that employ probability of failure
as a measure, and which therefore require a knowledge of the joint distri-
bution of all uncertain parameters, are called level III methods.
(d) Level IV methods : Reliability method that compares a structural prospect
with a reference prospect according to the principles of engineering eco-
nomic analysis under uncertainty, considering costs and benefits, of con-
struction, maintenance, repair, consequences of failure, interest on capital,
etc., is called a level IV method. Such design methods are still in the process
of development.
Lemer and Moavenzadeh (1971) developed one of the first models dealing
with reliability of pavements. They pointed out that the factors affecting the
degree of variation in pavement system parameters have a significant effect on
system reliability.
The limit state function for the pavement reliability problem can be written
as:
D = log NF − log NA (1.5)
Where, NF = allowable number of axle load applications to failure
NA = number of actual axle load applications
The condition of the pavement is considered to have deteriorated below
acceptable limits when NA exceeds NF , or equivalently, D < 0. Assuming
lognormal distributions for NF and NA , the probability of failure is obtainable as
PF = φ(–β C ), where φ(.) is the cumulative distribution function of the standard
normal random variable and β C = E(D)/σ (D) is the reliability index, in which
E(D) and σ (D) are the mean and standard deviation of D (Darter and Hudson,
1973).
In the simulation model proposed by Alsherri and George (1988) for relia-
bility evaluation of pavements, the following equation based on present service-
ability index was used:
R = P p f ≥ pt (1.6)
where, pf = present serviceability index at time t, and pt = limiting (terminal)
serviceability index, generally set at 2.5 for AASHTO’s design and 3.0 for
premium design.
The following expression was used to estimate reliability under the assump-
tion that both pf and pt are normally distributed:
⎡ ⎤
⎢ μ p f − μ pt ⎥
R = ⎣ 1/2 ⎦
= (z0 ) (1.7)
σ p f + σ pt
2 2
where,
φ = standard normal distribution
μpf = mean value of pf
μpt = mean value of pt
σ pf = standard deviations of pf
σ pt = standard deviations of pt
z0 = standard normal deviate
In the AASHTO (1993) guide for the design of pavement structures, the
overall standard deviation of variation was considered by including the errors
in traffic predictions and in pavement performance prediction to analyze risk
6 Safety and reliability modeling and its applications
and reliability in the design and reliability design factor was determined. The
reliability of design was defined as:
The following equation for the reliability design factor (FR ) was derived:
(COV. ADT∗ Dd )2 + (COV.P)2 + (COV.Ld )2 + (COV.TF)2
2
SW =
5.3
(1.11)
Where ADT∗ Dd represents average daily traffic in a heavier direction; P is the
percentage of trucks in the traffic mix; Ld is the lane distribution; TF is the truck
factor (number of ESALs per truck). The growth factor and the design period
were assumed to be constants.
Using AASHTO’s flexible pavement performance prediction model, the
variance of the pavement performance prediction S2N may be obtained as
(Noureldin et al., 1994) :
2 2 2
SN2 = COV(MR) + P2 SN .COV(SN) (1.12)
P2 = variance component of SN
To determine the COV(SN), the variance of SN was estimated in the following
way :
Var(SN) ∼
= ā21 Var(D1 ) + D̄21 Var(a1 ) + ā22 m̄22 Var(D2 ) + ā22 Var(m2 )D̄22
+ Var(a2 )m̄22 D̄22 + ā23 m̄23 Var(D3 ) + ā23 Var(m3 )D̄23 + Var(a3 )m̄23 D̄23
(1.13)
Kulkarni (1994) chose traffic as a design element for evaluating the reliability
of alternate pavement designs with different types of pavements. The reliability
Reliability analysis of asphalt pavements: concepts and applications Chapter | 1 7
design procedures. The reliability indices for the pavement sections determined
by both methods were computed using the first order reliability method (FORM).
It was indicated that the RBD procedure does successfully yield cross-sections
whose reliability indices are close to the target reliability index, while the
AASHTO method does not generally produce designs of uniform reliability for
actual mechanistic failure criterion.
To incorporate reliability in pavement design, Austroads used the laboratory
fatigue relationship published by Shell Petroleum (Shell, 1978), which was
further modified to include a reliability factor (RF) corresponding to the desired
project reliability.
5
6918(0.856VB + 1.08
N = RF (1.17)
mix με
S0.36
till the fatigue cracking of bituminous surface or rutting in the pavement reaches
its terminal values, whichever happens earlier.
The general form of these fatigue and rutting models is given below:
Fatigue Model
k2
1 1 k3
N f = k1 × C × × (26)
εt MR
C = 10M (27)
Vbe
M = 4.84 × − 0.69 (28)
Vbe + Va
Rutting Model
k5
1
Nr = k4 × (29)
εv
Where,
Nf = fatigue life of bituminous layer in terms of cumulative repetitions of
equivalent 80 kN standard axle load
Nr = subgrade rutting life in terms of cumulative repetitions of equivalent 80
kN standard axle load
εt = maximum horizontal tensile strain at the bottom of bituminous layer
εz = maximum vertical compressive strain at the top of the subgrade
MR = resilient modulus of the bituminous layer (MPa)
Vbe = percent volume of effective bitumen in the mix used in the bituminous
layer (varies between 3.5 to 4.5 percent)
Va = percent volume of air voids in the mix used in the bituminous layer
(varies between 10.5 to 11.5 percent)
ki = regression coefficients (i = 1 to 5)
The values of these regression coefficients are given as : k1 = 1.6064 × 10−4
and 0.5161 × 10−4 for design traffic ≥ 20 msa and < 20 msa respectively;
k2 = 3.89; k3 = 0.854; k4 = 4.1656 × 10−8 and 1.41 × 10−8 for design traffic
≥ 20 msa and < 20 msa respectively; and k5 = 4.5337.
These fatigue and rutting transfer functions were developed and calibrated
during the database collected through R-6 and R-19 research studies sponsored
by MORTH. A total of around 120 numbers of bituminous concrete (BC) and 160
numbers of bituminous macadam (BM) road sections from R-6 and R-19 studies
were considered for development of fatigue criterion and 86 number of BC road
sections from R-6 study were analyzed for the development of rutting criterion.
These pavement sections consisted of bituminous surfacing with granular bases
and subbases and they were assumed as three layered structure. The average
annual pavement temperature (AAPT) of all the sections were around 35°C
and the bitumen of 80/100 penetration grade was used for both BC and BM
surfacing. The maximum repetitions of equivalent single-axle load for the road
sections were 50 msa only. The thickness of BC layer was 40 mm on most of
the sections and BM was used as the bituminous binder course, just below the
12 Safety and reliability modeling and its applications
BC layer. In the collected performance data, the scatter of data points was quite
large. This was attributed to the fact that the test pavements were located in
different parts of the country and they have been constructed in different climatic
conditions and probably quality control during the construction was not identical
resulting in wide variations. Because of the variabilities involved, wide scatter
was considered to be quite in order.
However, the present situation is totally different from what it was when these
transfer functions were developed and calibrated. The traffic on the pavements
has increased tremendously, both in terms of loading and number of repetitions.
The specifications of bituminous materials have changed. Flexible pavements
with thick bituminous layers are quite common these days. So, to what extent
the originally developed fatigue and rutting transfer functions are valid, is a
matter of concern and debate. There is a need for recalibration of the fatigue and
rutting transfer functions. Any changes to these transfer functions due to this
recalibration may affect the overall pavement thickness and ultimately affects
the reliability of the pavement. So, there is a need for recalibration of the fatigue
and rutting transfer functions. Different researchers have proposed different
coefficients for the fatigue and rutting equations.
The development of rutting and fatigue transfer functions is through field
calibration by ordinary least-square estimation (OLSE) technique/modelling of
the field data. The approach of OLSE in developing these equations itself violates
the basic assumptions of OLSE, due to the presence of measurement errors.
A better statistical method of functional linear-measurement error (FLME) has
been presented (Shukla and Das, 2008), which may be used for development of
fatigue and rutting equations.
1.5 Conclusions
The available methods for reliability analysis can be applied for the reliability
analysis of asphalt pavements. However, the issues indicated in the previous
section need to be taken care of during the application of such reliability analysis
for asphalt pavements.
Acknowledgements
The author would like to thank Prof. (Dr.) Satish Chandra, Director, CSIR-
Central Road Research Institute (CSIR-CRRI), New Delhi for his kind permis-
sion to publish this paper.
Disclosure statement
No potential conflict of interest was reported by the authors.
References
AASHTO, 1993. AASHTO Guide for Design of Pavement Structures. American Association of State
Highway and Transportation Officials, Washington, D.C.
Aguiar-Moya, J.P., Prozzi, J., 2011. Development of reliable pavement models. Texas Transportation
Institute, Texas A&M UNiversity System, College Station, Texas, USA.
Alsherri, A., George, K.P., 1988. Reliability model for pavement performance. J. Trans. Eng. 114
(2), 294–306.
Austroads, 2012. Guide to Pavement Technology Part 2 : Pavement Structural Design. AGPT 02-12.
Austroads Ltd., Sydney, Australia.
Cronvall, O., 2011. Structural Lifetime, Reliability and Risk Analysis Approaches for Power Plant
Components and Systems. In: VTT Publications, 775. VTT Technical Research Centre of
Finland, Vuorimiehentie.
14 Safety and reliability modeling and its applications
Darter, M.I., Hudson, W.R., 1973. Probabilistic Design Concepts Applied to Flexible Pavement
System Design. Centre for Highway Research, University of Texas at, Austin, Texas, USA.
Dilip, D.M., Ravi, P., Babu, G.L.S., 2013. System reliability analysis of flexible pavements. J.
Transport. Eng. 139 (10), 1001–1009.
Gogoi, R., Das, A., Chakroborty, P., 2013. Are fatigue and rutting distress modes related? Int. J.
Pavement Res. Technol. 6 (4), 269–273.
Hudson, W.R., 1975. State of the Art in pedicting pavement reliability from input variability.
Department of Transportation, Federal Aviation Administration, Washington, D.C.
IRC:37, 2018. Guidelines for the design of flexible pavements. Indian Roads Congress, New Delhi,
India.
Kim, H.B., Buch, N., 2003. Reliability based pavement design model accounting for inherent
variation of design parameters. 82nd Transport. Res. Board Annu. Meet., Washington, D. C.
Kulkarni, R.B., 1994. Rational approach in applying reliability theory to pavement structural design.
Transportation Research Record, Transportation Research Board. National Research Council,
Washington D.C., USA, pp. 13–17 1449.
Lemer, A.C., Moavenzadeh, F., 1971. Reliability of highway pavements. Highway Res. Rec. 36, 1–8.
Liu, H., Xu, X., 2014. Reliability analysis of asphalt pavement considering two failure modes. In:
Mohammadian, K., Goulias, K.G., Cicek, E., Jieh-JiuhWang, Maraveas, C. (Eds.), Proceedings
of 3rd International Conference on Civil Engineering and Urban Planning III. CRC Press,
Balkema, The Netherlands, pp. 291–295.
Madsen, H.O., Krenk, S., Lind, N.C., 1986. Methods of Structural Safety. Prentice-Hall Inc.,
Englewood Cliffs, New Jersey, USA.
Modarres, M., Kaminskiy, M., Krivtsov, V., 1999. Reliability Engineering and Risk Analysis - A
Practical Guide. Marcel Dekker Inc., New York, USA.
NCHRP, 2004. Guide for mechanistic-empirical design of new and rehabilitated pavement struc-
tures. NCHRP Research Report 1-37A, Transportation Research Board, National Research
Council, Washington, D.C.
Noureldin, A.S., Sharaf, E., Arafah, A., Al-Sugair, F., 1996. Rational Selection of Factors of Safety
in Reliability-Based Design of Flexible Pavements in Saudi Arabia. Transport. Res. Rec.: J.
Transport. Res. Board, SAGE Publications 1540 (1), 39–47.
Noureldin, A.S., Sharaf, E., Arafah, A., Faisal, A.-S., 1994. Estimation of Standard Deviation of
Predicted Performance of Flexible Pavements Using AASHTO Model. Transport. Res. Rec.
1449, 46–56.
Peddinti, P.R.T., Munwar Basha, B., Saride, S., 2020. System Reliability Framework for Design
of Flexible Pavements. J.Transport. Eng. Part B: Pavements 146 (3) American Society of Civil
Engineers (ASCE).
Shell, 1978. Shell Pavement Design Manual - Asphalt Pavements and Overlays for Road Traffic.
Shell International Petroleum Limited, London, U.K.
Shukla, P.K., Das, A., 2008. A re-visit to the development of fatigue and rutting equations used for
asphalt pavement design. Int. J. Pavement Eng. 9 (5), 355–364.
Timm, D.H., Birgisson, B., Newcomb, D.E., Galambos, T.V., 1999. Incorporation of reliability
into the Minnesota mechanistic-empirical pavement design method. Final Report, Minnesota
Department of Transportation, Minnesota.
Timm, D., Newcomb, D., Galambos, T., 2000. Incorporation of reliability into mechanistic-empirical
pavement design. Transport. Res. Rec. 1730 (1), 73–80.
Non-Print Items
Abstract
The variability of the inputs parameters is not considered during the pavement
design process. As the input parameters are uncertain, this needs to be accounted
for during the design process through the reliability analysis approach. However,
the application of reliability concepts to pavements is not very straight forward and
many aspects need to looked into. The present paper attempts to summarize such
concepts and applications of reliability concepts for asphalt pavement design.
Keywords
Asphalt; Material strength degradation; Mechanical behavior; Pavement design;
Reliability; Flexible Pavement
Chapter 2
2.1 Introduction
The concept of component importance is well-known among reliability en-
gineers and is mostly used in the context of two-state systems with two-
state components. There exist several different importance measures, e.g. Birn-
baum, Fussel-Vesely, Risk Achievement Worth, Risk Reduction Worth, etc. (see
Hoyland and Rausand [2009] for definitions and explanations). A comprehensive
synopsis of the topic is given in the recent survey [Kalpesh and Kirtee, 2017].
Several types of component importance are also discussed in the monograph
[Kuo and Zhu, 2012]. Finding the importance’s values may not be a simple task
for complex systems, but the effort can be worthwhile. For example, high values
can indicate critical locations in the system structure, where highly reliable
components should be placed in order to reduce or minimize the risk of system
failure. Also, as shown in this chapter, they can be essential in computing useful
reliability characteristics such as interstate transition intensities for multistate
systems with repairable components.
In this chapter, importance is attributed to a group of components rather than
to a single component alone. It is defined as the probability that simultaneous
failure or repair of all components in a set results in a system transition from
state a to state b, provided that the components are in “up” or “down” state
and a>b or a<b respectively. It will be denoted as Ia→b () and referred to as
the importance of the component set to a transition from state a to state b.
Let us stress that the considered system is assumed to be multistate and its
Safety and Reliability Modeling and Its Applications. DOI: 10.1016/B978-0-12-823323-8.00018-0
Copyright © 2021 Elsevier Inc. All rights reserved. 15
16 Safety and reliability modeling and its applications
i.e. Ia→b (,t) is the (conditional) probability that simultaneous failure of all
components in causes a transition from a to b, given that these components
are operable
Ia→b (,t) – importance of to a transition from a to b, given that a < b,
defined as follows:
Ia→b (, t ) = Pr Xt ∈ crit
a→b () | Xi (t ) = 0, i ∈ (2.6)
Markov modeling of multi-state systems with simultaneous Chapter | 2 19
2 3
i.e. Ia→b (,t) is the (conditional) probability that simultaneous repair of all
components in causes a transition from a to b, given that these components
have failed.
If is a binary function (i.e. S={0,1}) and is a one-element set (i.e. ={i}),
then the quantities defined by (5) and (6) reduce to the Birnbaum importance of
component i.
For better understanding of the introduced concepts, let us consider a power
supply system for a critical device. It consists of mains (1), a standby generator
(2), and a UPS (3) that, in case of mains outage, supplies power from batteries
for the time necessary to start the generator. The system’s RBD is shown in
Fig. 2.1.
Let S={0,1,2}, i.e. the system is in state 0 if power is not supplied to the
device, in state 1 if supplied from UPS batteries or standby generator, and in
state 2 if supplied from mains. The RBD in Fig. 2.1 implies that
(0,0,0) = (0,0,1) = (0,1,0) = 0,
(0,1,1) = 1,
(1,0,0) = (1,0,1) = (1,1,0) = (1,1,1) = 2.
If we assume that the system is fully/partly operational in state 2/1, then
the inequalities 0 < 1 < 2 define the ensuing partial order in S (note that this
order is linear, because each two elements of S are comparable). It is easily
checked that is an order-preserving function from {0,1}3 to S. We also assume
for simplicity that the system components fail independently (no simultaneous
failures) and are nonrepairable. The transition diagrams for Xt and Zt are shown
in Fig. 2.2 and 2.3. An arrow between two states indicates that they are in the
transition relation. Analyzing these figures, we conclude that (1,1,1)→(0,1,1)
yields 2→1; (1,0,0)→(0,0,0), (1,0,1)→(0,0,1), or (1,1,0)→(0,1,0) yields 2→0;
and (0,1,1)→(0,0,1) or (0,1,1)→(0,1,0) yields 1→0.
Note that it is possible to define a nonorder-preserving function from {0,1}n
to S, but it may be difficult to interpret physically. For example, if n and S are as
in the above example, then defined below is such a function.
(0,0,0) = (0,0,1) = (0,1,0) = 0,
(0,1,1) = (1,0,1) = 1,
(1,0,0) = (1,1,0) = (1,1,1) = 2.
Indeed, (1,0,0) < (1,0,1), but (1,0,0) > (1,0,1), hence is not order-
preserving.
20 Safety and reliability modeling and its applications
(1,1,1)
(0,0,0)
(1,0,0), (1,0,1)
(1,1,0), (1,1,1)
φ=2
(0,1,1)
φ=1
(0,0,0), (0,0,1)
(0,1,0)
φ=0
following formulas:
1
a→b (t ) = λ Ia→b (, t ), a > b (2.7)
Pr [Zt = a] ⊆{1,...,n}
1
a→b (t ) = μ Ia→b (, t ), a < b (2.8)
Pr [Zt = a] ⊆{1,...,n}
Ia→b (, t) = Ia→b (, t) · Pr[Xk (t ) = 1, k] = Pr Xt ∈ crit
a→b () , a > b
(2.9)
Ia→b (, t) = Ia→b (, t) · Pr[Xk (t ) = 0, k] = Pr Xt ∈ b→a () , a < b
crit
(2.10)
It is more convenient to use I a→b () instead of Ia→b () in (7) and (8), because
I a→b () is usually easier to compute than Ia→b (). It should be noted that the
above formulas are generalizations of similar ones for a two-state system with
independent components, to be found in [Korczak, 2007].
Proof: We only consider the cases a>b and a<b, because Lemma 1 states that
if a→b then either a>b or a<b. Let us assume that a>b in order to prove (7).
The law of total probability yields:
Pr [Zt+t = b, Zt = a] =
= Pr [Xt+t = y, Xt = x]
x,y: (x)=a,
(y)=b
= Pr [Xt+t = y, Xt = x]+ (2.11)
x,y: x>y, (x)=a,
(y)=b
+ Pr [Xt+t = y, Xt = x]
x,y: x>y, (x)=a, (y)=b
1
lim Pr [Xt+t = y | Xt = x] = 0 (2.12)
t→0 t
x,y: x>y, (x)=a, (y)=b
22 Safety and reliability modeling and its applications
i.e. the second sum on the right hand side of (11) is equal to zero. Expanding the
first sum we get:
Pr [Xt+t = y, Xt = x]
x,y: x>y, (x)=a,
(y)=b
= Pr [Xt+t = y, Xt = x]
⊆{1,...,n}, x>y, D(x,y)=, (x)=a, (y)=b
=∅ x,y:
= Pr [Xt+t = (x, 0 ), Xt = x]
⊆{1,...,n}, =∅ x∈crit
a→b ()
= Pr [Xt+t = (x, 0 ) | Xt = x] Pr [Xt = x]
⊆{1,...,n}, =∅ x∈crit
a→b ()
(2.13)
Let us note that for x∈a→b crit () it holds that
Pr [Xt+t = (x, 0 ) | Xt = x]
(2.14)
= Pr [Xk (t + t ) = 0, k ∈ | Xk (t ) = 1, k ∈ ]
According to the definition of λ , the right hand side of (14), divided by t,
converges to λ as t→0. Thus, for x∈a→b crit () we have:
1
limPr [Xt+t = (x, 0 ) | Xt = x] = λ (15)
t→0 t
where λ (x) and μ (x) depend on xi , i∈. However, (7) or (8) still holds if λ (x)
or μ (x) are equal for x∈crit a→b () or x∈crit a→b (), because λ or μ can
then be written without the variable x. For better explanation, this issue will also
be addressed in the next two sections.
An important conclusion can be drawn from Theorem 1. If is an order-
preserving function from {0,1} to S, then Zt is a (nonhomogenous) Markov
process with the transition intensities given by (7) and (8). Indeed, Pr(Zt = a)
Markov modeling of multi-state systems with simultaneous Chapter | 2 23
μS
111 110
λB(1− πS)
λBπS μE
μE λB
μB μB
λE λB λB
λE
μB μB
001 000
and Ia→b (,t) are functions of the state probabilities Pr(Xt =x), x∈{0,1}n , which,
as solutions of the Kolmogorov equations, only depend on t and constant
failure/repair rates λ ,/μ , ⊆{1,…,n}. In view of (7) and (8), the same property
holds for transition intensities a→b (t), i.e. they do not depend on the history
of Zt =(Xt ) before time t. Moreover, a→b (t) converge to constant values as
t→∞, because the probabilities Pr(Xt =x) also do. Zt is thus asymptotically
homogenous.
hence λ{2} (0,1,1) = λ{2} (0,1,0) = λE , but λ{2} (1,1,1) = λ{2} (1,1,0) = 0. In turn,
the repair policy yields that μ{1} =μB does not depend on the states of E and S,
μ{2} (1,0,1) = μ{2} (1,0,0) = μE, μ{2} (0,0,1) = μ{2} (0,0,0) = 0, μ{3} (1,1,0) = μS ,
and μ{3} (1,0,0) = μ{3} (0,1,0) = μ{3} (0,0,0)=0.
The transition diagram of Xt is helpful in obtaining the Kolmogorov equations
from which the state probabilities for Xt can be computed. As shown further,
these probabilities are used to find the state probabilities Pr(Zt =a), a∈S, and
transition intensities a→b (t) for the process Zt , which, in turn, are necessary
to compute useful reliability characteristics of a considered system. Analyzing
Fig. 2.3, we obtain the following Kolmogorov equations for Xt :
dP1,1,1 (t )/dt = P0,1,1 (t )μB + P1,0,1 (t )μE + P1,1,0 (t )μS +
− P1,1,1 (t )[λB (1 − πS ) + λB πS ]
dP1,1,0 (t )/dt = P0,1,0 (t )μB + P1,0,0 (t )μE − P1,1,0 (t )(λB + μS )
dP0,1,1 (t )/dt = P1,1,1 (t )λB (1 − πS ) − P0,1,1 (t )(λE + μB )
dP1,0,1 (t )/dt = P0,0,1 (t )μB − P1,0,1 (t )(λB + μE )
dP0,1,0 (t )/dt = P1,1,1 (t )λB πS + P1,1,0 (t )λB − P0,1,0 (t )(λE + μB )
dP1,0,0 (t )/dt = P0,0,0 (t )μB − P1,0,0 (t )(λB + μE )
dP0,0,1 (t )/dt = P0,1,1 (t )λE + P1,0,1 (t )λB − P0,0,1 (t )μB
dP0,0,0 (t )/dt = P0,1,0 (t )λE + P1,0,0 (t )λB − P0,0,0 (t )μB (2.19)
For simplicity, we will compute only the asymptotic values of the system
parameters, i.e. the values of Px (t), Ia→b (t) and a→b (t) for t→∞. This is
sufficient for most practical purposes. Thus, from now on these parameters will
be written without the variable t. Equating dPx (t)/dt to 0 in (19) we obtain
the following equations for the steady state probabilities Px = lim t→∞ Px (t),
x∈{0,1}n :
P0,1,1 μB + P1,0,1 μE + P1,1,0 μS = P1,1,1 λB
P0,1,0 μB + P1,0,0 μE = P1,1,0 (λB + μS )
P1,1,1 λB (1 − πS ) = P0,1,1 (λE + μB )
P0,0,1 μB = P1,0,1 (λB + μE )
P1,1,1 λB πS + P1,1,0 λB = P0,1,0 (λE + μB )
P0,0,0 μB = P1,0,0 (λB + μE )
P0,1,1 λE + P1,0,1 λB = P0,0,1 μB
P0,1,0 λE + P1,0,0 λB = P0,0,0 μB
P1,1,1 + P1,1,0 + P0,1,1 + P1,0,1 + P0,1,0 + P1,0,0 + P0,0,1 + P0,0,0 = 1 (2.20)
Markov modeling of multi-state systems with simultaneous Chapter | 2 25
The last equation expresses the obvious fact that the total probability of all
possible outcomes is equal to 1. However, we cannot solve Eq. (2.20) without it,
because the first eight equations are not algebraically independent. The solution
of Eq. (2.20) is given below.
μB μE μS (λE + μB )
P1,1,1 =
(λB πS + μS )[(λB + μB )(λB λE + λE μE + μB μE )]
λB πS
P1,1,0 = P1,1,1
μS
λB λE (1 − πS )
P1,0,1 = P1,1,1
μE (λE + μB )
λE (λB + μS )
P1,0,0 = P1,1,0
μE (λE + μB )
μE
P0,1,1 = P1,0,1
λE
(λB + μS )
P0,1,0 = P1,1,0
(λE + μB )
λB + μE
P0,0,1 = P1,0,1
μB
λB + μE
P0,0,0 = P1,0,0 (2.21)
μB
Let us note that P1,1,1 is computed first, and each subsequent state probability
of Xt is obtained from a previously computed one. Such approach allows to avoid
unnecessarily complicated formulas. Clearly, closed formulas for Px can only be
derived for a small system, in case of a more complex one a numerical method
of solving linear equations would have to be used.
An average user (electric power consumer) usually pays no attention to
the details of the system’s operation. For him/her it is essential that power be
supplied in sufficient quantity. Thus, a user perceives the considered system as a
three-state one with the state space {B,E,F}, where power is sufficiently supplied
from the mains in state B, supplied from emergency source in insufficient
quantity in state E, and not supplied in state F. Let Zt = (Xt ) be a power supply
process as perceived by the user, i.e.
F (1, 1, 1) = F (1, 1, 0) = F (1, 0, 1) = F (1, 0, 0) = B,
F (0, 1, 1) = F (0, 1, 0) = E, (2.22)
F (0, 0, 1) = F (0, 0, 0) = F.
As is usually the case, S has far less elements than {0,1}n , thus the states of
Zt =(Xt ) are obtained by merging the states of Xt . The transition diagram of Zt
is shown in Fig. 2.5.
26 Safety and reliability modeling and its applications
B
ΛB→E
ΛE→B
ΛB→0 E
Λ0→B
ΛE→0
B→E ({1}) = (1, 1, 1), (1, 1, 0); B→E ({1, 3}) = (1, 1, 1)
crit crit
crit
B→F ({1}) = (1, 0, 1), (1, 0, 0)
crit
E→F ({2}) = (0, 1, 1), (0, 1, 0)
crit
E→B ({1}) = (0, 1, 1), (0, 1, 0)
crit
F→B ({1}) = (0, 0, 1), (0, 0, 0) (2.23)
which, in view of (9) and (10), yields:
IB→E ({1}) = P1,1,1 + P1,1,0 ; I B→E ({1, 3}) = P1,1,1
IE→B ({1}) = P0,1,1 + P0,1,0
IF→B ({1}) = P0,0,1 + P0,0,0 (2.24)
Let us note that the remark to Theorem 2.1 only pertains to ={1} and
crit B→E {1}, because λ (x) or μ (x) are constant within every other set of
Another random document with
no related content on Scribd:
he should proceed from Shanghai to Tientsin in a Russian vessel
which “he might borrow for the purpose.” Li Hung-chang’s reply,
telegraphed to Yüan Shih-k’ai for transmission to the Throne, while
outwardly respectful, clearly implies that Her Majesty has been to
blame for the disasters then occurring. “I am sincerely grateful,” he
says, “for Your Majesty’s gratifying confidence in me, but cannot help
recalling to mind the folly which has now suddenly destroyed that
structure of reformed administration which, during my twenty years’
term of office as Viceroy of Chihli, I was able to build up not
unsuccessfully. I fear it will not be possible for me to resume the
duties of this difficult post at a time of crisis like the present, destitute
as I am of all proper and material resources.” He proceeds even to
criticise Her Majesty’s suggestion as to his journey, observing that
“Russia possesses no vessel at Shanghai, and would certainly
refuse to lend if she had one, in view of the state of war now
existing.” Finally, he excuses himself for deferring his departure, on
the ground that the British Minister had requested him not to leave
until the foreign Ministers had been safely escorted from Peking to
Tientsin. “I do not know,” says he, “if any such arrangements for
safely escorting them can be made,” and therefore concludes by
asking Yüan to inform the Throne that he will start northwards,
journeying by land, “as soon as his health permits it.” To this plain-
spoken message from the great Viceroy, Tzŭ Hsi replied in two lines
of equally characteristic directness:—“Li Hung-chang is to obey our
earlier Decree, and to make all haste northwards. The crisis is
serious. Let him make no further excuses for delay.”
In spite of these peremptory orders, Li Hung-chang, who had a
very definite conception of his own predicament, remained at
Shanghai, ostensibly negotiating, but in reality waiting, to see what
would be the outcome of the siege of the Legations. He was
interviewed by The Times correspondent at Shanghai on the 23rd of
July, and then stated that he would not proceed to his post in the
north until convinced by clear proofs that the Empress Dowager had
seen the folly of her ways, and was prepared to adopt a conciliatory
policy towards the outraged foreign Powers. At the end of July, when
it became clear to him that the Court had determined on flight, he
forwarded by special courier a very remarkable Memorial, in which
he called the Throne to task in the plainest possible terms, and
urged an immediate change of policy. This Memorial reached the
Empress before her departure from Peking; certain extracts from it
are well worth reproduction, as showing Li Hung-chang at his best,
and displaying that quality of courageous intelligence which made
him for twenty years the foremost official in China and a world-wide
celebrity:—
“It is true that, in times past, our Capital has been shifted on
more than one occasion of national danger, but in those days
our enemies were not able to push their armies far into the
interior of our country for indefinite periods, and were
compelled to withdraw after brief expeditions. The position of
affairs to-day, however, is very different, so that we can obtain
no reliable guidance from precedents of history. As regards
the province of Shensi, it has always been a centre of wars
and rebellions; its people are poverty stricken, and there is no
trade there. Seven centuries ago, Hsi-an was an Imperial city,
but is now anything but prosperous. Its vicinity to Kansu and
the New Dominion territories, infested with Mahomedan
rebels and adjoining the Russian Empire, renders it most
unsuitable as a site for your Majesties’ Capital. Supposing
that the Allies, flushed with success, should determine on an
advance westwards, what is there to prevent them from doing
so? If ten thousand miles of ocean have not stopped them,
are they likely to be turned back from a shorter expedition by
land?”
After referring to the fact that the cradle of the Dynasty and the
tombs of its ancestors are situated near Peking, and that it is
geographically best fitted to be the centre of Government, the
Memorialists remind the Throne that the foreign Powers have
promised to vacate Peking, and to refrain from annexing any territory
if the Court will return. These ends, they say, will not be attained
should the Court persist in its intention to proceed further westwards,
since it is now the desire of the foreign Ministers that China’s rulers
should return to Peking. In the event of a permanent occupation of
Peking by the Allies, the loss of Manchuria would be inevitable. The
Memorialists predict partition and many other disasters, including
financial distress, and the impossibility of furnishing the Throne with
supplies at Hsi-an or any other remote corner of the Empire. If the
Court’s decision to proceed to Hsi-an is irrevocable, at least a
Decree should now be issued, stating that its sojourn there will be a
brief one, and that the Court will return to Peking upon the complete
restoration of peaceful conditions. “The continued existence of the
Empire must depend upon the Throne’s decision upon this matter.”
The Memorial concludes by imploring their Majesties to authorise
Prince Ch’ing to inform the foreign Ministers that the withdrawal of
the allied armies will be followed by a definite announcement as to
the Court’s return.
In a further Memorial from the Viceroys and Governors, it is stated
that the Russian Minister for Foreign Affairs had suggested to the
Chinese Minister in St. Petersburg, that the location of the Capital at
Hsi-an would certainly prove undesirable, in view of the poverty-
stricken condition of the province, and that their Majesties would no
doubt, therefore, proceed to Lan-chou fu, in Kansu. Referring to this
interesting fact, the Memorialists observe:—