A failure interaction model for multicomponent repairable systems

Abstract

Modeling of stochastic dependency among components in a repairable system is still a challenging task when dealing with the maintenance of multicomponent systems. With the help of stochastic dependency information, failure of a component brings attention to the components having strong interactions with the failed component. With this information, one can plan the maintenance of components in a better way. Since a change in failure probability of a component (due to deterioration or failure of a component in a given time interval) influences the failure probabilities of other components in the system, therefore, in this article, we consider probability of failure to represent the state of the component to model the stochastic dependency among components. We apply the Bayesian belief network to model such scenario of dependency among the components and present two case studies to compute various probabilities. In the first study, expert elicitation is being used, whereas the time between failure of the components is used in the second case to calculate failure probabilities. To illustrate the applicability of the proposed approach, one case study for each is presented. The first case study takes the case of an army truck through expert elicitation approach whereas the second case study deals with a rolling mill gearbox whose time between failure of components was available.

Multi-component maintenance grouping optimization based on stochastic dependency

“…The prior probabilities of each node shown in Figure 5 are calculated using RC algorithm. 20 With this structure, we experimented on the resultant changes in failure probability of other components by instantiating the condition of failure for each component, one at a time, by varying the node probability. For instance, Figure 6 shows the BN on instantiated condition of turbocharger failure, that is, turbocharger is in failed state (failure probability = 1) and resultant change in failure probabilities of others.…”