Due to expensive, highly complex and time-consuming processes, semiconductor-manufacturing systems have been given a special attention. In our previous work (Tang et al., 2003), a heuristic algorithm to design the reconfigurable automated production system was proposed. However, machine breakdown and planned/unplanned maintenance were not considered. This paper extends that work and addresses the related design issues in reconfigurable back-end semiconductor manufacturing systems considering failures and maintenance effects. Queuing network approaches are used to model the variance of production lines. The work enhances the proposed virtual production line (VPL) design methodology in (Tang et al., 2003) where the deterministic machine processing time is used. A priority is also introduced to each idle machine and updated based on its past performance. The algorithms for Adaptive VPL configuration/reconfiguration are then proposed. The benefit of the proposed methodology and algorithms is illustrated through a simplified back-end semiconductor manufacturing example.