Effect of the perturbation of the parameters of production and manufacturing models on their performance measures

Production and manufacturing systems can be very complicated and many of their parameters such as demand, processing times, failures and repairs, are stochastic and/or unknown. Therefore, approximations are employed to make the mathematical model tractable and statistical estimation methods are used to estimate the model's components and parameters. The estimated probability distributions and parameter values are not exact and errors can accumulate and affect the reliability of the mathematical model, i.e., to which extent the model remains a reliable representation of the real system. Additionally, there are no available closed form formulae for the performance measures of queueing models with general arrival and service distributions. We investigate, the effect of individual parameters and their interactions on the performance measures such as leadtime. A design of experiments is prepared in order to take account of the effect of multiple parameters as well as their interactions. We test two different models of queues (G/G/1, with and without failures) and explore the impact of the various distributions parameters on the estimated waiting time. The results provide guidance on the estimation of probability distributions and unknown model parameters of manufacturing and production systems for better performance measurements and scenarios comparison.