Maintenance cost analysis of a two-component parallel system with failure interaction

We consider a two-component standby system which can successfully operate upon a demand if at least one component is not failed. Component failures can occur randomly but are detected only by periodic testing or inspections. We suppose that tests or inspections are staggered. If a component is found failed a corrective maintenance action is made. Besides periodic inspections, preventive maintenance actions are scheduled for the system at a fixed time after the last preventive or corrective maintenance action. We show that when there is failure interaction and both components are not identically distributed the choice of the first component to inspect to obtain the minimum cost rate depends on the degree of s-dependence between components lifetimes. Numerical experiments presented show that not to consider the failure interaction when it exists might lead to important underestimations of the cost rate. Besides, this underestimations seem to be more important as the expected cost per unit time of unavailability increases.