Recently we have presented nonparametric predictive inference (NPI) for system reliability [1, 2], with specific attention to redundancy allocation. Series systems were considered in which each subsystem i is a k_i-out-of-m_i system. The different subsystems were assumed to consist of different types of components, each type having undergone prior success-failure testing. This work uses NPI for Bernoulli variables [3], which enables prediction for m future variables based on n observations, without the need of a prior distribution. In this paper, we present a generalization of these results by considering multiple subsystems which all consist of one type of component, which provides an important step to wider applicability of this approach.