Input analysis in simulation: A case study based on the variability in manufacturing lines

Abstract

Simulation is a powerful tool with acknowledged capabilities that has proved to be a valuable support instrument to decision making. In order to attain a proper representation from the system, it’s necessary to perceptively observe the system, to collect the data corresponding to the model’s input, and to perform an accurate analysis of the same data. The results and recommendations subsequent to the simulation are as legitimate as its modeling and inputs. In manufacturing systems build-to-order, that generates multiple products with identical characteristics and high levels of customization, the input analysis earns a vital role. On this article is proposed a methodology to deal with the variability inherent to systems, resorting to statistical inference methods: hypotheses testing. These methods are a vigorous tool on the analysis of data collected from the system. Furthermore, a case study based on a real manufacturing line will be presented, where the impact that the information inputted onto the model has in the simulation’s results shall be analyzed, regarding the processing lead time. In addition, the presented case study provides evidence of the two foremost pitfalls, referred by Law, which ought to be avoided. The usage of the mean towards a statistical distribution misleads the system’s analyst, whereas the normal distribution doesn’t accurately represent the processing times. An adequate replication of the variability over the manufacturing processes from the real system, throughout the probability’s distribution on an “off-line” simulation, comprises as a vital element to support decision making.