In this paper, we study the effect of repair dependence function on three repairable
parallel systems with identical components under the following configurations: (a) A twocomponent
parallel system with active components (b) A two-component parallel system
with active and warm standby components (c) A three-component parallel system with active, warm standby, and cold standby components. Switching among the components in the respective configurations is considered as perfect. The failure time of components follow Weibull distribution and their repair time follow exponential distribution. The number of failed components at time is a stochastic process. The system of differential difference equations of the non-homogenous continuous time Markov process is used for reliability and
availability evaluation of the systems, under configurations (a), (b) and (c). The repair dependence function is classified into weak, linear and strong dependence classes. The cost incurred per unit time of availability is a random variable. The cost analysis is performed by computing the expected cost of availability for the various combinations of dependence classes. The computation of expected cost helps in choosing the appropriate combination of
dependence classes in the respective systems. Finally, we present the numerical calculations through MATLAB programming.