Performance Analysis of Unreliable Manufacturing Systems with Uncertain Parameter Estimates

Traditional systems engineering methods for the performance evaluation of manufacturing systems assume that machine reliability parameters (Mean Time to Failure and Mean Time to Repair) are precisely known. However, in practical situations, these parameters are either estimated from real life data or based on experts’ knowledge. In both cases, they are subject to uncertainty. This paper proposes for the first time an approach for the performance evaluation of unreliable manufacturing systems that considers uncertain machine parameter estimates. The proposed method is based on the combined use of Bayesian estimation, probability density function discretization and existing decomposition-based techniques for analyzing manufacturing lines composed of unreliable machines and capacitated buffers. Numerical results show that neglecting uncertainty in the input parameter estimates generates consistent errors in the output performance measure estimates, thus making the consequent system design and operation decisions sub-performing. An industrial case is proposed to show the benefits of this method in real production settings.