A continuous-flow model for unreliable production networks of the finite queue type

Production networks with finite buffers, random rates, and breakdowns are analyzed using an analytic/simulation model. Random rates are approximated by piecewise deterministic variables and segments of the network are collapsed into single servers between events. Occurrence of events represents the slow nonlinear system dynamics handled by simulation. The fast piecewise linear dynamics between events are handled algebraically. An additional approximation is used for discrete-part traffic which is represented by a continuous flow. Extensive experimentation with a multitude of networks and comparison with a conventional simulator show that the model is quite fast and accurate, thus becoming a promising candidate for design purposes.