EVALUATION RELIABILITY OF MULTI STATE REPAIRABLE SYSTEMS BY CONSIDERING COMMON CAUSE FAILURE AND CASCADING FAILURE USING DYNAMIC FAULT TREE AND

Accurate evaluation of system reliability has always occupied the attention of system engineers and designers. In this connection, taking account of real system features in reliability evaluation models leads to an accurate estimation of reliability. These features include multi state components, repair, dependent failures, and time. The relationship among components in a system causes the failure of one component to affect others; therefore, components fail due not only to independent failure, but also to dependent failure. There are various dependent failures, the most important of which include common cause failure and cascading failure. Common cause failure has been investigated by a number of studies but a few attention has been paid to cascading failures. Cascading failure, severely influences system reliability, such that the defective functioning of one component in some systems is propagated to other components and vice versa. The present paper develops a new method using dynamic bayesian networks to take into consideration two types of dependent failures in the reliability evaluation of systems with repairable multistate components. In so doing, a dynamic fault tree is used and converted into a dynamic Bayesian network, in such a way that repair is incorporated into reliability evaluation by drawing the conditional probability table for a multistate component. Afterwards, by adding a new gate (cascading gate) to the dynamic fault tree and converting it to the dynamic Bayesian network, the impact of cascading failure on multistate systems is examined. Finally, using the common cause failure gate as well as cascading gate, the procedure of evaluation of reliability of a system with two types of dependent failures as well as repairable multistate components is investigated. Else, use of dynamic fault tree and its conversion into a dynamic Bayesian network minimizes the explosion state in the Bayesian network. The sensitivity of the proposed method is analyzed by evaluating the reliability of the high pressure water pumping system. This pumping system uses two water pumps in parallel, which are subject to common cause failure as well as independent failure. The defective functioning of water pumps in turn impinges upon electrical panel (cascading failure), thus creating two types of dependent failures in the system. Using the proposed method, the reliability of such systems is evaluated.