Model representation methods in simulation of manufacturing systems

A survey of simulation practitioners revealed that they hardly use any formal model representation techniques prior to construction of computer models. The data and operating logic of the system under investigation is directly represented in the computer model. Although this may shorten the model building time in the short term, a poorly detailed and documented model can cause problems in the long term. For example, subsequent changes to the model become difficult as the modeller needs to interpret the logic by reverse engineering the computer model. Therefore, it is beneficial to document the knowledge gathered during the first analysis of the system and its subsequent model changes. This paper presents alternative model representation methods suitable for manufacturing system modelling and an approach to identify an appropriate representation method for a given class of system features. Reported model representation methods were studied to identify a set of suitable methods for manufacturing applications. A comprehensive list of criteria was developed to benchmark these methods. For example, factors such as: abstraction level; reusability; convertibility into other methods; and level of detail have been included in this process. By analysing a variety of manufacturing systems, core sets of features have been identified to represent system behaviour. By combining these features a series of reference models were built to represent systems with varying operating complexities. The representation methods identified above have been evaluated using reference models. The paper includes the results of this analysis.