Documente Academic
Documente Profesional
Documente Cultură
15
16 A. Lisnianski & G. Levitin
The performance rates of the elements can range from perfect functioning
up to complete failure. The failures that lead to the decrease in the element
performance are called partial failures. After partial failure, elements continue to
operate at reduced performance rates, and after complete failure the elements
are totally unable to perform their tasks.
Consider the following two examples of multi-state systems:
In the power supply system consisting of generating and transmitting
facilities, each generating unit can function at different levels of capacity.
Generating units are complex assemblies of many parts. The failures of different
parts may lead to situations in which the generating unit continues to operate,
but at a reduced capacity. This can occur during the outages of several
auxiliaries such as pulverizers, water pumps, fans, etc. For example, R. Billinton
and R. Allan (1996) describe a three state 50 MW generating unit. The
performance rates (generating capacity) corresponding to these states and
probabilities of the states are presented in Table 1.1.
Pr|G>r}
Element with total failure
Element with five different
performance levels
Gj(t), j=\,2,...,n
for each system element j , and the system structure function that produces the
stochastic process corresponding to the output performance of the entire MSS
G(t) = <f>(GI(t) G„(?)).
In many practical cases, a simpler MSS model can be used, which is based
on probability distribution of performances for all of the system elements at any
instant time t during the operation period [0,T] and system structure function:
gj,P]{f),\<j<n, (1.3)
<t>{G,{t),...,Gn(t)). (1.4)
It should be noted that this simple MSS model, while being satisfactory for
most applications, fails to describe some important characteristics of MSS, such
as mean time to failure, mean number of failures during the operation period etc.
It does not matter how the structure function is defined. It can be
represented in a table, in analytical form, or be described as an algorithm for
unambiguously determining the system performance G(t) for any given set
{GM, ..., G„(0}.
Basic Concepts 19
Example 1.1.
Consider a 2-out-of-3 MSS. This system consists of 3 binary elements with
the performance rates G,(t)s {gih ga}={0, 1}, for z'=l,2,3, where
Example 1.2.
Consider a flow transmission system (Fig. 1.2A) consisting of three pipes
(elements). The oil flow is transmitted from point C to point E. The pipes
performance is measured by their transmission capacity (ton per minute).
Elements 1 and 2 are binary. A state of total failure for both elements
corresponds to a transmission capacity of 0 and the operational state
corresponds to the capacities of the elements 1.5 and 2 ton per minute
respectively so that Gi(t)e {0,1.5}, G2{i)& {0,2}. Element 3 can be in one of
three states: a state of total failure corresponding to a capacity of 0, a state of
partial failure corresponding to a capacity of 1.8 tons per minute and a fully
operational state with a capacity of 4 tons per minute so that G3(t)s {0,1.8,4}.
The system output performance rate is defined as the maximum flow that can be
transmitted from C to E.
The total flow between points C and D through the parallel pipes 1 and 2 is
equal to the sum of the flows through each of these pipes. The flow from point
D to point E is limited by the transmitting capacity of element 3. On the other
20 A. Lisnianski & G. Leviiin
hand, this flow cannot be greater than the flow between points C and D.
Therefore, the flow between points C and E (the system performance) Is
G(0=^G/(aGX0,GK0)=min{Gy(/)+GKaC?i(0}-
The values of the system structure function G(t)=j(Gi(t),G&)9G$)) for
all the possible system states are presented in Table 1.3.
A B
Fig. 1.2. Two different multi-state systems with identical structure functions.
Example 1.3.
Consider a data transmission system (Fig. 1.2B) consisting of three fully
reliable network servers and three data transmission channels (elements). The
data can be transmitted from server C to server E through server D or directly.
The time of data transmission between the servers depends on the state of the
corresponding channel and is considered to be the channel performance rate.
This time is measured in seconds.
Elements 1 and 2 are binary. They may be in a state of total failure when
data transmission is impossible. In this case data transmission time is formally
defined as oo. They may also be in a folly operational state- when they provide
Basic Concepts 21
MGl(t),...,GJ_l(t),l,GJ+l(t),...,G„(ty) =\
and (1.5)
^(G1(O,...,Gy_1(0,0,Gy+i(O,...,G„(O) = 0.
(Note that for the binary systems G,(t)& {0,1} for 1 <j<n).
When the MSS is considered, the element is relevant if some changes in its
state without changes in the states of the remainder of the elements, causes
change in the entire system state. In terms of the system structure function, the
relevancy of element7 means that there exist such G,{t),...,G„(t) that for some
22 A. Lisnianski & G. Levitin
t(Gi(t),...,Gj_i(.t),gjk,GJ+i(t\...,Gn(ty)*
t(Gl(t),...,Gj_l(t),gjm,GJ+l(t),...,Gn(t)) (Lb)
Table 1.4. Possible delays of switches and entire circuit disconnection times
Possible switch delays (sec) Possible circuit
Element Element Element disconnection
1 2 3 times
0.3,0.7 0.9,1.2 0.3,0.5,0.8 0.3,0.5,0.7
.2.2. Coherency
In the binary system context coherency means that:
All the system elements are relevant
The fault of all the elements causes the fault of the entire system
The operation of all the elements results in the entire system operation
Once the system has failed, no additional failure can make the system
function again and
When the system is operating, no repair or addition of elements can cause
the system failure.
These requirements are met in systems with monotonic structure functions:
if there is no j for which Gj > Gj (for the binary system, this can be
reformulated as follows: there is no such j that Gj = 1 and G;=0).
In a multi-state case, the system is coherent if and only if its structure
function is non-decreasing in each argument and all of the system elements are
relevant. Note that from this structure function property it follows that the
greatest system performance is achieved when the performance rates of all of
the elements are greatest and the lowest system performance is achieved when
the performance rates of all of the elements are the lowest.
1.1.2.3. Homogeneity
The MSS is homogenous if all of its elements and the entire system itself
have the same number of distinguished states. One can easily see that all of the
binary systems are homogenous.
For example, consider a system of switches connected in series (Fig. 1.3.).
Assume that all the switches are identical and have the same number of states.
The total failure of the switch corresponds to infinite delay. Since the time of
circuit closing is equal to the closing time of its fastest element and since the
elements are identical, the entire system delay can be equal only to the delay of
one of its elements. The possible system delays are the same as the delays of a
single element. This means that the system is homogenous.
Despite the fact that the homogenous MSS are intensively studied, in real
applications most of the systems do not possess this property. Indeed, even when
considering the same MSS of series switches and allowing for different switches
to have different operational delays, one obtains an MSS in which the number of
system states is not equal to the number of states of the elements (see examples
in Table 1.5.).
Table 1.5. Possible delays of switches and entire circuit disconnection times
Note that this condition is tougher than condition (1.6). Indeed, a relevant
element according to definition (1.6) can be irrelevant according to (1.8).
For example, consider a system of switches connected in a series (Fig.
1.3) and assume that the switches are binary elements with switching delays,
presented in the last row of Table 1.5. Assume that the system disconnection
time should not be greater than W: (F(G(t)=W-G(t)). Observe that for W>0.6,
the second switch is relevant since when the first and third switches do not
work, the system's success depends on the state of the second switch. For
W<0.6, the second switch is irrelevant since when the first and third switches do
not work, the system fails to meet the demand independently of the state of the
second switch. (According to definition 1.6 the second switch is always
relevant).
Using the acceptability function, one can also give the definition of system
coherency that is more closely related to the one given for the binary systems.
Indeed, the definition of coherency for binary systems operates with notions of
fault and normal operation while when applied to MSS all that is required is the
monotonic behavior of the structure function. In the context of reliability, the
MSS coherency means that the improvement in the performance of the system
elements cannot cause the entire system transition from an acceptable state to
unacceptable one:
Performance rate
Cumulative Performance
G(t)
u
Deficiency
Time
Tf T
When the system is considered in the given time instant or in a steady state
(when its output PD does not depend on time) its behavior is determined by its
performance represented as a random variable (Fig. 1.5 demonstrates the
cumulative performance curve for a MSS in a steady state). Note that in a steady
state the distribution of the variable demand can be represented (in analogy with
the distribution of the MSS performance) by two vectors (w,q), where
w={wh...,wM} is the vector of possible demand levels wn j=\,...,M and
q={q\,...,qM) is the vector of steady state probabilities of corresponding demand
levels qJ=Y>r{W=wl}, j=\,...,M- When one considers a MSS evolution in the
space of states during the system operation period T, the following random
variables can be of interest:
Time to failure, 7) is the time from the beginning of the system life up to
the instant when the system enters the subset of unacceptable states the first time.
Time between failures, Th is the time between two consecutive transitions
from the subset of acceptable states to the subset of unacceptable states.
Number of failures, NT is the number of times the system enters the subset
of unacceptable states during the time interval [0, 7].
Basic Concepts 27
PrfG>r} Performance
Deficiency*
Measures (1.14) and (1.15) are often important when logistic problems related to
MSS operation are considered (for example, in order to determine the required
number of spare parts).
MSS instantaneous (point) availability A(t) is the probability that the MSS
at instant f>0 is in one of the acceptable states:
A(t)=Pr{F(G(t),W(t))>0}. (1.16)
MSS availability in the time interval [0, T\ is defined as:
AT=±-\\(F(G(t),W{t))>0)dt, (1.17)
1
0
The random variable AT represents the portion of time when the MSS output
performance rate is in an acceptable area. For example, in Fig. 1.4 AT={T-Tr
T2)IT. This index characterizes the portion of time when the MSS output
performance rate is not less than the demand.
The expected value of AT is often used and is named demand availability
[Aven and Jensen, (1999)]:
AD=E(A7). (1.18)
For large /, the system initial state has practically no influence on its
availability. Therefore, the stationary MSS availability A for the constant
demand level W(t)=w can be determined on the base of the system PD:
A(w)=Zpk\(F(gk,w)^0), (1.19)
k=\
where pk= lim pk (t) is the steady state probability of the MSS state k with the
corresponding output performance rate gk.
In the case where F(G(t), W(t))=G(t)-W(t), we have F(gh w)=gk-w and:
k=\ gk>w
In Fig. 1.5 the stationary availability corresponds to the point where the
cumulative performance curve crosses the value of w.
When the demand is variable, the MSS operation period T is often
partitioned to M intervals T„, (l<m<M) and a constant demand level w„, is
assigned to each interval m. In this case the availability index may be written
[Billinton and Allan, (1996)]:
M M K
A{w,q)= ZA(wm)qm= £<7m UPkWigk^m) ^ °) » 0-21)
m=\ m=\ k=\
Basic Concepts 29
E*>=Y.Ph&k- (1-24)
The average MSS output performance for a fixed time interval [0,7] is
defined as:
ET=±JEtdt. (1.25)
1
0
Observe that the mean MSS performance does not depend on demand.
In some cases a conditional expected performance is used. This index
represents the mean performance of MSS on condition that it is in an acceptable
state. In the steady-state it takes the form:
£ » = TgkPkKF(gk,W)>0)/ZPkKF(gk,W)>0). (1.26)
k=\ k=\
determines the total power of consumers that must be immediately switched off
from the system.)
Since D(t) is a random variable at time /, it can be characterized by the
following measures:
- The probability that at instant t D{t) does not exceed some specified level d:
Pr{D(t)<d}, (1.28)
- The mean value of the MSS performance deviation at instant V.
D,=E{D{t)). (1.29)
When the MSS is in a steady-state and demand is constant W(t)=w,
performance deficiency is not a function of time and can be obtained from the
system steady-state PD (Fig. 1.5 dashed area) as:
K
Dx=Y1Pkmax(w-glc,0). (1.30)
k=\
The Average MSS expected performance deficiency for a fixed time interval
[0, T] is defined as follows:
DT=\;\Dtdt (1.32)
1
o
and cumulative performance deficiency for this interval is:
T
LT=DTT = \Dtdt. (1.33)
0
T T T
LT = \{W(t)-G{t))dt = \W(t)dt-\G(t)dt. (1.34)
o 0 0
As LT is a random variable, one can define the following characteristics.
- The Probability that random LT does not exceed some specified level I:
Pr{Lr<n, (1-35)
which means 66% of the nominal generating capacity for the first generator and
4
£
2oo=ZP g =0.05x0 + 0.25x0.4 + 0.3x0.8 + 0.4x1.0 = 0.74,
k=\
which means 74% of the nominal generating capacity for the second generator.
The steady-state performance deficiency (1.30) is:
A»(0.5)= I > u ( ^ - g u ) = 0.1x (0.5 -0.0) = 0.05
g\k~W<0
D2x (0.5) = X Plk (w ~g2k) = ° 0 5 x (0-5 - 0.0) + 0.25 x (0.5 - 0.4) = 0.05.
g2k-w<Q
According to the generic model, one can define different types of MSS by
describing the stochastic behavior of its elements and defining the system's
structure function. It is possible to invent an infinite number of different
structure functions in order to obtain different models of MSS. The question is
whether or not the MSS model can be applied to real technical systems. In this
Basic Concepts 33
±
Fig. 1.8. Parallel MSS with work sharing
For MSS with work sharing (Fig. 1.8), the entire system performance rate
is usually equal to the sum of the performance rates of the parallel elements for
both flow transmission and task processing systems. Indeed the total flow
through the former type of system is equal to the sum of flows through its
36 A. Lisnianski &, G. Levitin
parallel elements. In the latter type of MSS, the system processing speed
depends on the rales of the work sharing. The most effective rale providing the
minimal possible time of work completion, shares the work among the elements
in proportion to their processing speed. In this case, the processing speed of the
parallel system is equal to the sum of the processing speeds of all of the
elements.
For example, consider a system of several parallel coal conveyors
supplying the same system of boilers (Fig. 1.9.A) or a multi-processor control
unit (Fig. 1.9.B), assuming that the performance rates of the elements in both
systems can vary. In the first case the amount of coal supplied is equal to the
sum of the amounts supplied by each one of the conveyors. In the second case
the unit processing speed is equal to the sum of the processing speed of all of its
processors.
For MSS without work sharing (Fig. 1.10) the system performance rate
depends on the discipline of the elements activation. Unlike binary systems
where all the elements have the same performance rate, the choice of an active
element from the set of different ones affects the MSS performance. The most
common policy in both flow transmission and task processing MSS is to use an
available element with the greatest possible performance rate. In this case, the
system performance rate is- equal to the maximal performance rate of the
available parallel elements.
H
Fig. 1.10. Parallel MSS without work sharing.
Basic Concepts 37
The system reliability for the series, parallel and series-parallel MSS, can
be defined as the probability that the overall system performance rate meets a
specified demand.
38 A. Lisnianski & G. Levitin
Note that the parallel MSS is not only a multi-state extension of the binary
parallel structure, but it is also an extension of the binary A>out-of-« system.
Indeed, the k-out-oi-n system reliability is defined as a probability that at least k
elements out of n are in operable condition (note that k=n corresponds to the
binary series system and k=\ corresponds to the binary parallel one). The
reliability of the parallel MSS with work sharing is defined as the probability
that the sum of the elements performance rates is not less than the demand.
Assuming that the parallel MSS consists of n identical two-state elements
having a capacity of 0 in a failure state and a capacity of g in an operational
state and that the system demand is equal to kg, one obtains the binary A>out-of-
n system.
1 3
2 4
Load
Connecting
cables
I J
Generation block 2 Transformation block 2
Fig. 1.15. Bridge shaped network of roads with different speed limitations.
Note that the first example belongs to the flow transmission MSS. The
overall power supplied to the load is equal to the total power flow through the
bridge structure. The second example belongs to the task processing MSS,
where the task of a vehicle is to go from point A to point B using one of four
possible routes.
Determining the bridge performance rate based on its elements
performance rates is a more complicated problem than in the case of series-
parallel systems. It will be addressed in the coming chapters.
40 A. Lisnianski & G. Levitin
As for the series-parallel systems, the system reliability of the MSS bridge
is defined as the probability that the overall system performance rate meets a
specified demand.
/%7 .\
•'*}
s/
associated with a certain delay. Since the delay is equal to the time needed for
the digital retransmitter to processes the signal, it can be exactly evaluated and
treated as a constant value for any given type of signal. When this is so, the
total time of the signal propagation from transmitter to receiver can vary
depending only on a combination of states of multi-state retransmitters. The
entire system is considered to be in working condition if the time is not greater
than a certain specified level. Otherwise, the system fails.
In the more complex model, the retransmission delay of each multistate
element can also vary (depending on the load and the availability of processors).
In this case each state of the multi-state element is characterized by a different
delay and by a different set of following elements belonging to the range of the
element.
The system's reliability for the multi-state consecutively connected
systems can be defined as the probability that the system is connected or as the
probability that the system's signal propagation time meets the demand. The
expected system delay is also an important characteristic of its functioning.
fall into H zones (the greater the zone number the greater the parameter
deviation). The alarm search should be initiated if the total sum of the numbers
of zones the parameter falls in during r consecutive tests is greater than the
specified value W.
Service system
Consider a conveyor-type service system that can process r incoming tasks
simultaneously according to first-in-first-out rule and share a common limited
resource. Each incoming task can have different states and the amount of the
resource needed to process the task is different for each state of each task. The
total resource needed to process r consecutive tasks should not exceed the
available amount of the resource. If there is no available resource to process r
tasks simultaneously, the system fails.
The simplest example of such a model is a column of n randomly loaded
vehicles crossing a bridge. The maximum bridge load is W, the number of
vehicles crossing the bridge simultaneously is r (this number is limited by the
length of the bridge). The bridge collapses if the total load of any r consecutive
vehicles is greater than W.
Manufacturing
Consider a heating system that should provide a certain temperature along
a line with moving parts (Fig. 1.19). The temperature at each point of the line is
determined by a cumulative effect of r closest heaters. Each heater consists of
several electrical heating elements. The heating effect of each heater depends on
the availability of its heating elements and therefore can vary discretely (if the
heaters are different, the number of different levels of heat radiation and the
intensity of the radiation at each level are specific to each heater). In order to
provide the temperature, which is not less than some specified value at each
point of the line, any r adjacent heaters should be in states where the sum of
their radiation intensity is greater than the minimum allowed W.
/ \/ \/ \/ \/ v \/ v v \/ \/ \/ \
/ /\ /\ /\ /\ / \ /\ /\ /\ /\ /\ /\ \
/ w/ X X X A Nl.-Mii-p;ui- A X X \ \
in the undirected network the edges merely connect the vertices without any
consideration for direction, in the directed network, the edges are ordered pairs
of vertices. That is, each edge can be followed from one vertex to the next.
An acyclic network is a network in which no path (a list of vertices where
each vertex has an edge from it to the next one) starts and ends at the same
vertex. In reliability engineering, the directed acyclic networks are usually
considered.
Each network has a single root node (source) and one or several terminal
nodes (sinks). Examples of directed acyclic networks are presented in Fig. 1.20.
The aim of the networks is the transmission of information or material flow
from the source to the sinks. The transmission is possible only along the edges
that are associated with the transmission media (lines, pipes, channels etc.). The
nodes are associated with communication centers (retransmitters, commutation,
or processing stations etc.)
A B
Fig. 1.20. Examples of networks with single (A) and several (B) terminal nodes
Exercises
1.1. A linear consecutively connected system consists of four radio relay
stations and is aimed at providing the signal propagation from station 1 to
station 4. The transmitters allocated in the stations 1, 2 and 3 are two-state
elements with performance Gie{2, 0}, G 3 e{l, 0}, G 3 e{l, 0}. Determine the
irrelevant retransmitter.
1.2. A continuous production system consists of two units connected in
parallel. The productivity of the system is equal to the sum of the unit
productivities. The first unit has random productivity G,e{5, 4, 0}, the second
unit has random productivity G 2 e {2, 1,0}. The system should meet the demand
w. Determine the irrelevant element (if any) when w=l .5 and when w=3.
1.3. Two idenical pumps are connected in parallel. Each pump has random
productivity Ge{5, 2, 0}. The system of pumps can function with and without
work sharing. In which case is the system homogeneous?
1.4. Build the cumulative performance distribution curve for a system with
the performance distribution #={10, 8, 5, 0},/>={0.2, 0.45, 0.2, 0.15}.
1.5. The system performance distribution is the same as in exercise 1.4.
a). Find the average system performance.
Basic Concepts 49
b). Find the system availability and mean performance deficiency for the
acceptability function F(G,w)=G-w, where w=l.
1.6. The system performance distribution is g={\, 3, 5, oo}, p={0.5, 0.15,
0.3, 0.05}. The acceptability function is F(G,w)=w-G.
a). Find the system availability for w=2 and w=l.
b). Find the conditional mean performance of the system for w=4, w=20.
1.7. Two systems have performance distributions #={6, 5, 2, 0},p={0.3,
0.3, 0.25, 0.15} and g={4, 1, 0}, p={0.7, 0.1, 0.2} respectively. The
acceptability function is F(G,w)=G-w. Compare the mean performance and the
reliability of the systems, when the demand is w=3.
1.8. The system performance distribution is g={\2, 10, 7, 0},p={0.3, 0.2,
0.4, 0.1}. The demand distribution is w={10, 8, 3}, q={0.4, 0.5, 0.1}. Find the
system availability and the mean performance deficiency for the acceptability
function F(G,w)=G-w.
1.9. The flow transmitting capacity of a flow valve has the distribution
gc={30, 22, 15, 0},/>,={0.8, 0.05, 0.1, 0.05} in the closed mode and distribution
ga={5, 2, 0},/?„={0.7, 0.15, 0.15} in the open mode. In the closed mode the flow
should not be less than wc=20, in the open node it should not exceed w„=l.
Determine the valve reliability.
1.10. Which one of the electronic swiches is more reliable: one with
switching time distributions &={0.1, 0.15, 0.3, oo}, p,={0.8, 0.05, 0.1, 0.05},
g„={0.\, oo},/j„={0.85, 0.15} or one with switching time distribution gc={0.07,
0.1, 0.18, oo},pc={0.7, 0.15, 0.05, 0.1}, £„={0.18, 0.35, oo},/>,={0.7, 0.25, 0.05}
if the switching time should not exceed 0.2 in both modes.
References
Alexopoulos, C. (1995), "A note on state-space decomposition methods for analyzing stochastic
flow networks", IEEE Trans, on Reliability 44, 354-357.
Aven, T. and Jensen, U. (1999), Stochastic models in reliability, Springer-Verlag, New York.
Aven, T. (1993), "On performance measures for multistate monotone systems", Reliability Eng. and
System Safety 41, 259-266.
Barlow, R. E. and Wu, A. S. (1978), "Coherent systems with multi-state components", Mathematics
of Operations Research 3, 275-281.
Billinton, R. and Allan, R. (1996), Reliability evaluation of power systems, Plenum Press, New
York.
Block, H. and Savits, T. (1982), "A Decomposition of Multistate monotone system", J. Applied
Probability 19, 391-402.
Boedigheimer, R. and Kapur, K. (1994), "Customer-Driven Reliability Models for Multistate
Coherent Systems", IEEE Trans, on Reliability 43, 46-50.
Brunelle, R. D. and Kapur, K. C. (1999), "Review and classification of reliability measures for
multi-state and continuum models", HE Trans. 31, 1171-1180.
Doulliez, P. and Jamoulle, E. (1972), "Transportation networks with random arc capacities", RAIRO
3, 45-60.
Ebrahimi N. "Multistate reliability models", (1984), Naval Res. Logistics 31, 671-680.
El-Neweihi, E. and Proschan, F. (1984), "Degradable systems: a survey of multistate system
theory", Communication in Statistics. Theory and Methods 13, 405-432.
Evans, J. (1976), "Maximum flow in probabilistic graphs: The discrete case", Networks 6, 161-183.
50 A. Lisnianski & G. Levitin
Gnedenko, B. V. and Ushakov, I. A. (1995), Probabilistic Reliability Engineering, John Wiley and
Sons, New York.
Griffith, A. and Govindarjulu Z. (1985), "Consecutive K-out-of-N failure systems: reliability,
availability, component importance, and multistate extensions", American J. of Mathematical
and Management Sciences 5, 125-160.
Hudson J. C. and Kapur K. C. (1982), "Reliability Theory for Multistate Systems with Multistate
Components", Microelectronics and Reliability 22, 1-7.
Hwang, F.K. and Yao, Y.C. (1989), "Multistate consecutively connected systems", IEEE Trans, on
Reliability 38, 472-474.
Kossow, A. and Preuss, W. (1995), "Reliability of linear consecutively connected systems with
multistate components", IEEE Trans, on Reliability 44, 518-522.
Levitin, G. and Lisnianski, A. (2001), "Structure Optimization of Multi-state System with Two
Failure Modes", Reliability Eng. & System Safety 72, 75-89.
Levitin, G. (2002), "Optimal allocation of elements in linear multi-state sliding window system",
Reliability Eng. & System Safety, 76, 247-255.
Lin, J.S., Jane, C.C. and Yuan, J. (1995), "On reliability evaluation of a capacitated-flow network in
terms of minimal pathsets", Networks 25, 131-138.
Malinowski, J. and Preuss, W. (1995), "Reliability of circular consecutively connected systems with
multistate components", IEEE Trans, on Reliability 44, 532-534.
Natvig B. (1984), "Multi-state coherent systems". In Encyclopedia of Statistical Sciences, vol. 5, ed.
N. Jonson and S. Kotz, Wiley&Sons, New York.
Nordmann, L. and Pham, H. (1999), "Weighted voting systems", IEEE Trans, on Reliability, 48, 42-
49.
Ohi F. and Nishida T. (1984), "Multistate systems in reliability theory", in: Stochastic Models in
Reliability Theory . Edited by S.Osaki & Y.Hatoyama. Springer Verlag, Berlin, 12-22.
Reinshke K. and Ushakov 1. (1988), Application of Graph Theory for Reliability Analysis, (in
Russian), Radio i Sviaz, Moscow.
Ross S. (1993), Introduction to probability models, Academic Press, Inc, Boston. Ross S. M. (1979)
Multivalued state component systems, Annals of Probability 7, 379-383.
Shinmori S. and Ohi F. (1994), "On structural relations for component sets of multi-state systems".
Mathematica Japonica 40, No. 1, 135-142.
Ushakov I. (ed.) (1994), Handbook of Reliability Engineering, Wiley&Sons, New York.
Yang, C. and Kubat, P. (1989), "Efficient computation of of most probable states for communication
networks with multimode components", IEEE Trans, on Communications 37, 535-538.
Zuo. M. and Liang, M. (1994), "Reliability of multistate consecutively connected systems",
Reliability Engineering & System Safety 44, 173-176.