A mathematical model of a flexible manufacturing system with limited in-process inventory

This paper describes a computationally simple, asymptotic model of a flexible job shop, especially designed for estimating the influence of limited in-process inventory level on the production rate. Its main features make it very similar to the one by Solberg. While Solberg's model consists of a closed queuing network, we propose an open queuing network with a limited amount of inprocess customers; a single customer class is assumed, the various actual processing routes being accounted for by routing probabilities. For such a queuing network, the product form of state probabilities is valid, and the normalization constant can be very simply obtained by a convolution algorithm, close to the one used by Solberg. Various performance indices are calculated, regarding the job shop behaviour over a long period of time. Comparison of analytical results of the model and simulation results are provided in order to estimate the amount of error introduced by assuming exponentially distributed processing times and Poisson inputs in the mathematical representation. Simulations were carried out in FORTRAN-based SLAM language.