A Decomposition Approach for the Approximate Evaluation of the Output Variability in Multi-Stage Production Lines

The evaluation of the average performance of manufacturing systems has been widely investigated in the manufacturing system engineering literature. However, there is industrial evidence that production variability due to random disturbances causes the observed production rate to be different from its average value. In this paper, a fast and accurate approximate analytical method for the evaluation of the output variability in capacitated multi-stage production lines where machines are prone to random failures is proposed. The method decomposes the production line into two-machine one-buffer subsystems and propagates the first two asymptotic moments of the output throughout the production line. Numerical results show that the proposed method has good accuracy if compared with discrete event simulation and outperforms existing methods for the output variance estimation. The industrial benefits derived by the use of this method are shown through application to real manufacturing contexts.