Abstract
In this paper, we study reliability properties of k-out-of-n system consisting of l \((1\le l\le n)\) different types of components with discrete, independent lifetimes. We obtain some conditional survival functions of lifetime of a used system. Next, we use them to calculate two conditional failure probabilities of k-out-of-n systems and show that they are equal to unconditional failure probability of a k-out-of-\((n-r)\) system, \(r<n-k+1\). These results are extended versions of the respective ones existing in the literature.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
A technical system has a k-out-of-n structure if it works when at least k of the n components operate. It fails if \(n-k+1\) or more components fail. Two important particular cases of k-out-of-n systems are parallel and series ones corresponding to \(k=1\) and \(k=n\). In the literature, many authors paid attention to the reliability and aging properties of k-out-of-n systems and their variants and extensions, see, for example Eryilmaz (2011, 2012, 2013), Navarro and Duarte (2017), Navarro et al. (2017), Misra and Francis (2018), Zhang et al. (2018), Balakrishnan et al. (2018) and Salehi et al. (2019). Most of these results have been restricted to the case when the component lifetimes are independent and identically distributed. However, in some practical situations systems might be composed of independent and nonidentical components. The most recent results in this direction are in Li and Chen (2004), Xu (2008), Sadegh (2008), Zhao et al. (2008), Gurler and Bairamov (2009), Kochar and Xu (2010), Salehi and Asadi (2010), Salehi et al. (2011) and Sutar and Naik-Nimbalkar (2019), under the assumption that the component lifetimes have absolutely continuous distributions.
The situation becomes more complicated in the case in which the parent distribution of the component lifetimes is discrete. This is so due to the presence of ties between components failures. This assumption might be more adequate for example when the component lifetimes represent the numbers of turn-on and switch-off up to the failure or when the system’s elements operate in discrete cycles, or are exposed to shocks occurring in discrete times. Reliability properties of k-out-of-n systems composed of components which have discrete operation times have been considered by Weiss (1962), Young (1970), Tank and Eryilmaz (2015), Dembińska and Goroncy (2020) and Dembińska et al. (2021).
Dembińska (2018) established explicit expressions for unconditional and some conditional probabilities of a failure of a k-out-of-n system whose component lifetimes \(X_{1},\ldots ,X_{n},\) are not necessarily independent nor identically distributed discrete variates. Let \(X_{1:n}\le \ldots \le X_{n:n}\) stand for the order statistics corresponding to \(X_{1},\ldots ,X_{n}\) and \(T_{k,n}\) denote the lifetime of the k-out-of-n system. In particular, she obtained the formula describing the conditional probability that this system will break down at time \(t_{j}\) given the times of failures of its components which occurred up to time \(t_{i}\):
\(r\!<\!n\!-\!k\!+\!1\) and \(t_{i_{1}}\!\le \!t_{i_{2}}\!\le \!\ldots \!\le \!t_{i_{r}}\!\le \!t_{i}\!<\!t_{j}\), where \(t_{i_{1}}\!\le \!t_{i_{2}}\!\le \!\ldots \!\le \!t_{i_{r}}\!\le \! t_{i}\) are such that \({\text {P}}\left( X_{1:n}\!=t_{i_{1}},X_{2:n}\!=t_{i_{2}},\ldots ,X_{r:n}\!=t_{i_{r}},X_{r+1:n}\!>t_{i}\right) >0\). She also considered the situation when at time \(t_{i}\) we registered a failure of a component of the k-out-of-n system and we observed that at this time exactly r components were broken, \(r<n-k+1\). Then, repeating similar arguments as for (1) she computed the conditional probability that the system will fail to function at time \(t_{j}>t_{i}\):
where \(t_{i}\) is such that \(P(X_{r:n}=t_{i},X_{r+1:n}>t_{i})>0\). Next, the probabilities (1) and (2) were applied to obtain the corresponding residual lifetimes of a used system. Under the assumption that \(X_{1},\ldots ,X_{n}\) are identically distributed with common cumulative distribution function (cdf) F, she observed that the probability (1) does not depend on \(t_{i_{1}}\le t_{i_{2}}\le \ldots \le t_{i_{r}}\) and as well as the probability (2) is equal to unconditional probability that a k-out-of-\((n-r)\) system, consisting of homogeneous elements with lifetimes \(Y_{1},\ldots ,Y_{n-r}\) having cdf given by
will brake down at time \(t_{j}\) \((t_{j}>t_{i})\). Our aim is to extend these results by considering k-out-of-n systems with independent component lifetimes that are of l \((1\le l\le n)\) different types and adding some extra information in the conditions of the probabilities (1) and (2) which concerns failures of these components. This is done in Sect. 2.
Throughout the paper we write \({\text {I}}(\cdot )\) for the indicator function, that is \({\text {I}}(x \in A) = 1\) if \(x \in A\) and \({\text {I}}(x \in A) = 0\) otherwise.
2 Main result
Consider a k-out-of-n system which is composed of n independently operating components. We assume that the lifetimes of the components, \(X_{1},X_{2},\ldots ,X_{n}\), are discrete random variables (rvs) of l \((1\le l\le n)\) different types. There are exactly \(n_{w}\) rvs of type w having cdf \(F_{w}\), \(w=1,\ldots ,l\) (the cdfs \(F_{w},\) \(w=1,\ldots ,l,\) are pairwise different and \(n_{1}+n_{2}+\ldots +n_{l}=n\)). Without loss of generality we can assume that
The k-out-of-n system functions as long as at least k of its n components function. It fails when the \((n-k+1)\)-th component failure occurs. Thus the lifetime of the k-out-of-n system is \(T_{k,n}=X_{n-k+1:n}\). The discrete case becomes more complicated than the continuous one due to possible ties between component failures with non-zero probability. In this case at the moment of the system failure the number of inoperative elements can be larger than \(n-k+1\).
Let \({\mathcal {T}}\) \(=\{t_{1},\ldots ,t_{N}\}\), where \(N\le \infty \), be the union of the supports of \(F_{w}\), \(w=1,\ldots ,l\), and assume that \(t_{1}<t_{2}<\ldots <t_{N}\). Next, if \(X_{i}\sim F_{w},\) \(w=1,\ldots ,l,\) then \(p_{w}(t)={\text {P}}_{w}(X_{i}=t)\), i.e. \(p_{w}\) is the probability mass function (pmf) corresponding to \(F_{w}\), \(F_{w}(t^{-})={\text {P}}_{w}(X_{i}<t)\) and \({\overline{F}}_{w}(t)=1-F_{w}(t)\).
We need the following definition.
Definition 1
For a fixed \(\omega \in \Omega \) we write \(X_{j:n}(\omega )\rightsquigarrow F\) instead of \(X_{h_{j}(\omega )}\sim F\), where the function \(h_{j}\) is defined as follows:
-
(i)
if there is exactly one p such that \(X_{j:n}(\omega )=X_{p}(\omega )\), then \(h_{j}(\omega )=p\),
-
(ii)
otherwise, if \(X_{1:n}(\omega )=X_{j:n}(\omega )\) and \(X_{j:n}(\omega )=X_{p_{1}}(\omega )=X_{p_{2}}(\omega )=\ldots =X_{p_{m}}(\omega )\), where \(j\le m\) and \(1\le p_{1}<p_{2}<\ldots <p_{m}\le n\), then \(h_{j}(\omega )=p_{j}\); and if \(j_{1}\) is the largest integer satisfying \(X_{j_{1}:n}(\omega )<X_{j:n}(\omega )\) and \(X_{j:n}(\omega )=X_{p_{1}}(\omega )=X_{p_{2}}(\omega )=\ldots =X_{p_{m}}(\omega )\), where \(1\le p_{1}<p_{2}<\ldots <p_{m}\le n\), then \(h_{j}(\omega )=p_{j-j_{1}}\).
Now we are able to define
where
Observe that \(S_{i}^{(w)}\) informs us how many of \(X_{j}\)’s of type w are not greater than \(X_{i:n}\) and \(G_{i}^{(w)}\) limits to such of them which are equal to \(X_{i:n}\). Moreover \(S_{i}^{(l)}\!=i-\sum _{w=1}^{l-1}S_{i}^{(w)},\) \(i\!=1,\ldots ,n.\) Definition 1 with the example of its application as well as the constructions of vectors \(S_{i}\) and \(G_{i}\), \(i\!=1,\ldots ,n\) were proposed by Jasiński (2020).
Firstly, we assume that at the moment \(X_{r:n}=t_{i}\) we registered \(g_{r}^{(w)}\) failures of components of type w, \(w=1,\ldots ,l\) of a used k-out-of-n system. We also assume that at this time exactly \(s_{r}^{(w)}\) elements of type w were broken, where \(\sum _{w=1}^{l}s_{r}^{(w)}\!=r\) and \(r<n-k+1\). Then it is of interest to obtain the following conditional survival function of \(T_{k,n}\)
where \(t_{i}\), \(g_{r}=\left( g_{r}^{(1)},\ldots ,g_{r}^{(l)}\right) \) and \(s_{r}=\left( s_{r}^{(1)},\ldots ,s_{r}^{(l)}\right) \), are chosen so that the probability \(p^{*}_{k;r}(t_{i},g_{r},s_{r})\) of \(\{X_{r:n}=t_{i},X_{r+1:n}>t_{i},G_{r}=g_{r},S_{r}=s_{r}\}\) is not equal to 0.
We begin with an observation that
where
It follows immediately that
Now we determine the probability
where
For abbreviation, let
Then
where
and \(B_{g_{r}^{(l)},s_{r}^{(l)},v-{\tilde{v}}_{l-1}}^{t_{i},t_{j}}\) is given by (8) with \(w=l\) and \(v_{w}\) replaced by \(v-{\tilde{v}}_{l-1}\). Denoting
we have
Similarly, we obtain \({\text {P}}\left( B_{g_{r}^{(l)},s_{r}^{(l)},v-{\tilde{v}}_{l-1}}^{t_{i},t_{j}}\right) \). Now combining (3) with (4) and (7) with (10), after simple algebra, we derive the desired conditional probability as follows
Since
applying (11), we immediately get the conditional probability that the system will fail to function at time \(t_{j}\).
Notice that
where \(Y_{1},\ldots ,Y_{n-r}\) are independent rvs of possibly l different types having cdfs \(F_{w}^{Y},\) \(w=1,\ldots ,l,\) given by
More precisely, there are exactly \(n_{w}-s_{r}^{(w)}\) rvs of type w having cdf \(F_{w}^{Y}\), \(w=1,\ldots ,l\), where \(\sum _{w=1}^{l}n_{w}-s_{r}^{(w)}=n-r\).
Because the survival function uniquely determines the distribution, this proves the following theorem.
Theorem 1
Under the above assumptions and notation, for \(r<n-k+1\) the conditional distribution of \(T_{k,n}\) given \(X_{r:n}=t_{i},X_{r+1:n}>t_{i},G_{r}=g_{r},S_{r}=s_{r}\)
-
(i)
does not depend on \(g_{r}^{(1)},\ldots ,g_{r}^{(l)}\),
-
(ii)
is just equal to the unconditional distribution of \(T^{Y}_{k,n-r}\) the lifetime of k-out-of-\((n-r)\) system consisting of components with independent lifetimes \(Y_{1},\ldots ,Y_{n-r}\), where there are exactly \(n_{w}-s_{r}^{(w)}\) of \(Y_{i}\)’s of type w having cdf \(F_{w}^{Y},\) \(w=1,\ldots ,l,\) defined in (12).
Now for \(r<n-k+1\) we are interested in finding the following conditional survival function of \(T_{k,n}\)
where \(t_{i_{1}}\le \ldots \le t_{i_{r}}\le t_{i}\) and \(s_{j}\!=\left( s_{j}^{(1)},\ldots ,s_{j}^{(l)}\right) \), \(j=1,\ldots ,r,\) are chosen so that the probability of \(\{X_{1:n}\!=t_{i_{1}},\ldots ,X_{r:n}\!=t_{i_{r}},X_{r+1:n}\!>t_{i},S_{1}\!=s_{1},\ldots ,S_{r}\!=s_{r}\}\) is not equal to 0. Here \(t_{i_{1}},\ldots ,t_{i_{r}}\) are the ordered failures times of the components of the system which occurred up to time \(t_{i}\). Moreover, there are exactly \(s_{j}^{(w)}\) elements of type w, \(w=1,\ldots ,l,\) that were broken at the time \(t_{i_{j}}\), \(j=1,\ldots ,r\).
Using the concept of tie-runs proposed by Gan and Bain (1995), let us assume that \(t_{i_{1}}\le t_{i_{2}}\le \ldots \le t_{i_{r}}\) have m tie-runs with lengths \(z_{1},z_{2},\ldots , z_{m}\) \((z_{1}+\ldots +z_{m}=r)\), i.e.
We begin with the probability
Notice that
where
Thus
with \(s_{z_{0}}^{(w)}=0\).
Now we will obtain the probability \(p^{**}_{k;1,\ldots ,r}(t_{i_{1}},\!\ldots \!,t_{i_{r}},t_{i},t_{i}+x,s_{1},\!\ldots \!,s_{r})\) of the event \(\{X_{1:n}\!=\!t_{i_{1}},\ldots ,X_{r:n}\!=\!t_{i_{r}},X_{r+1:n}\!>\!t_{i},X_{n-k+1:n}\!>\!t_{i}+x,S_{1}\!=\!s_{1},\ldots ,S_{r}\!=\!s_{r}\}\). Since \(z_{1}+\ldots +z_{m}=r\), with the notation (6), we get
where \(t_{j}\) is defined in (5) and
Observe that \(B_{s_{z_{1}}^{(l)},\ldots ,s_{z_{1}+\ldots +z_{m}}^{(l)},v-{\tilde{v}}_{l-1}}^{t_{i_{z_{1}}},\ldots ,t_{i_{z_{1}+\ldots \!+z_{m}}},t_{i},t_{j}}\) is given by (16) with \(w=l\) and \(v_{w}\) replaced by \(v-{\tilde{v}}_{l-1}\). Using (9) we have, for \(w=1,\ldots ,l\),
where \(v_{l}=v-{\tilde{v}}_{l-1}\). Now combining (13) with (14) and (15) with (17), we are able to determine the conditional probability
Applying (18), we obtain the conditional probability that the system break down at \(t_{j}\) because
Notice that the formula in (18) is the same as in (11). Then we have the following result.
Theorem 2
Under the above assumptions and notation, for \(r<n-k+1\) and \(t_{i_{1}}\le t_{i_{2}}\le \ldots \le t_{i_{r}}\le t_{i}\), the conditional distribution of \(T_{k,n}\) given \(X_{1:n}=t_{i_{1}},X_{2:n}=t_{i_{2}},\ldots ,X_{r:n}=t_{i_{r}},X_{r+1:n}>t_{i},S_{1}=s_{1},\ldots ,S_{r}=s_{r}\),
-
(i)
does not depend on \(t_{i_{1}},t_{i_{2}},\ldots ,t_{i_{r}}\) nor on \(s_{1},s_{2},\ldots ,s_{r-1}\),
-
(ii)
is the same as the unconditional distribution of \(T^{Y}_{k,n-r}\) the lifetime of k-out-of-\((n-r)\) system consisting of components with independent lifetimes \(Y_{1},\ldots ,Y_{n-r}\), where there are exactly \(n_{w}-s_{r}^{(w)}\) of \(Y_{i}\)’s of type w having cdf \(F_{w}^{Y},\) \(w=1,\ldots ,l,\) defined in (12).
Remark 1
Theorems (1) and (2) corresponds to Theorem III.4 of Dembińska (2018).
3 Summary and conclusions
A k-out-of-n system is a technical device which plays an important role in the reliability theory. It has various applications in engineering. For example, it is used as the design of servers in internet service or the design of engines in the automotive industries. In this paper, we have considered k-out-of-n systems that consist of multiple types of components. Although such systems are more common in real life situations, their reliability analysis is more difficult and complicated. The operators of the systems are interested in getting inference about the reliability or other specifications of the system but they usually have only some partial information about the lifetime of the system e.g. they registered the times of failures of the components up to and including the fixed time. The presented results would allow the operators for greater planning and more efficient use of resources to reduce unexpected costs of utilization.
Data availability
Not applicable
References
Balakrishnan N, Barmalzan G, Haidari A (2018) On stochastic comparisons of \(k\)-out-of-\(n\) systems with Weibull components. J Appl Probab 55:216–232
Dembińska A (2018) On reliability analysis of \(k\)-out-of-\(n\) systems consisting of heterogeneous components with discrete lifetimes. IEEE Trans Reliab 67:1071–1083
Dembińska A, Goroncy A (2020) Moments of order statistics from DNID discrete random variables with application in reliability. J Comput Appl Math 371:112703
Dembińska A, Nikolov NI, Stoimenova E (2021) Reliability properties of \(k\)-out-of-\(n\) systems with one cold standby unit. J Comput Appl Math 388:113289
Eryilmaz S (2011) Dynamic behavior of \(k\)-out-of-\(n\): G systems. Oper Res Lett 39:155–159
Eryilmaz S (2012) On the mean residual life of a \(k\)-out-of-\(n\): G system with a single cold standby component. Eur J Oper Res 222:273–277
Eryilmaz S (2013) On reliability of a \(k\)-out-of-\(n\) system equipped with a single warm standby component. IEEE Trans Reliab 62:499–503
Gan G, Bain LJ (1995) Distribution of order statistics for discrete parents with applications to censored sampling. J Stat Plan Inference 44:37–46
Gurler S, Bairamov I (2009) Parallel \(k\)-out-of-\(n\): G systems with nonidentical components and their mean residual life functions. Appl Math Model 33:1116–1125
Jasiński K (2020) Some concept of Markov property of discrete order statistics arising from independent and non-identically distributed variables. Statist Probab Lett 160:108718
Kochar S, Xu M (2010) On results lifetimes of \(k\)-out-of-\(n\) systems with nonidentical components. Probab Eng Inf Sci 24:109–127
Li X, Chen J (2004) Aging properties of the residual life length of \(k\)-out-of-\(n\) systems with independent but nonidentical components. Appl Stoch Models Bus Ind 20:143–153
Misra N, Francis J (2018) Relative aging of \((n-k+1)\)-out-of-\(n\)-systems based on cumulative hazard and cumulative reversed hazard functions. Naval Res Logist 65:566–575
Navarro J, Durante F (2017) Copula-based representations for the reliability of the residual lifetimes of coherent systems with dependent components. J Multivar Anal 158:87–102
Navarro J, Longobardi M, Pellerey F (2017) Comparison results for inactivity times of \(k\)-out-of-\(n\) and general coherent systems with dependent components. TEST 26:822–846
Sadegh MK (2008) Mean past and mean residual life functions of a parallel system with nonidentical components. Commun Stat Theory Methods 37:1134–1145
Salehi ET, Asadi M (2010) Results on the past lifetime of \((n-k+1)\)-out-of-\(n\) structures with nonidentical components. Metrika 75:439–454
Salehi ET, Asadi M, Eryılmaz S (2011) Reliability analysis of consecutive \(k\)-out-of-\(n\) systems with nonidentical components lifetimes. J Stat Plan Inference 141:2920–2932
Salehi M, Shishebor Z, Asadi M (2019) On the reliability modeling of weighted \(k\)-out-of-\(n\) systems with randomly chosen components. Metrika 82:589–605
Sutar S, Naik-Nimbalkar UV (2019) A load share model for nonidentical components of a \(k\)-out-of-\(m\) system. Appl Math Model 72:486–498
Tank F, Eryilmaz S (2015) The distributions of sum, minima and maxima of generalized geometric random variables. Stat Pap 56:1191–1203
Weiss G (1962) On certain redundant systems which operate at discrete times. Technometrics 4:169–174
Xu M (2008) Ordering residual lifetimes of \(k\)-out-of-\(n\) systems with nonidentical components. Portland State University, Tech Rep, Department of Mathematics and Statistics
Young D (1970) The order statistics of the negative binomial distribution. Biometrika 57:181–186
Zhang T, Zhang Y, Du X (2018) Reliability analysis for \(k\)-out-of-\(n\) systems with shared load and dependent components. Struct Multidiscip Optim 57:913–923
Zhao P, Li X, Balakrishnan N (2008) Conditional orderings of \(k\)-out-of-\(n\) systems with independent but nonidentical components. J Appl Prob 45:1113–1125
Funding
Not applicable
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author states that there is no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Jasiński, K. Some conditional reliability properties of k-out-of-n system composed of different types of components with discrete independent lifetimes. Metrika 84, 1241–1251 (2021). https://doi.org/10.1007/s00184-021-00826-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00184-021-00826-1