Discrete-event simulation is used by thousands of companies to design new manufacturing systems and to improve the performance of existing ones. Manufacturing systems contain numerous sources of randomness such as machine times to failure and machine repair times, which greatly impact on system performance. If each source of system randomness is not modeled by an appropriate probability distribution, then it is highly likely that the simulation model will produce erroneous performance results, resulting in costly decisions. If the system of interest exists in some form, then it will often be possible to collect date and to use statistical techniques to determine an appropriate probability distribution. However, if the system does not exist, then collecting date is impossible and an analyst is forced to use a somewhat arbitrary distribution. To address this real and important problem, we propose a Phase 1 research study to determine the technical feasibility of developing (in Phase 2) a library of probability distributions that would be appropriate for difference common sources of randomness encountered in simulation models of manufacturing systems. COMMERCIAL APPLICATIONS: The research is directly applicable to discrete-event simulation of manufacturing systems when data on system randomness (times to failure, times to repair, processing times, ect.) are not available - a commonly occurring situation. The results of the research will be implemented in a computer program in Phase 3.
