SOFTWARE ARCHITECTURE FOR PARALLEL SIMULATION OPTIMIZATION

Abstract

Manufacturing simulation issues can be solved in a variety of different COTS (Commercial off The Shelf) simulation software – for example ARENA, QUEST, SIMUL8, Siemens PlantSimulation, etc. The input and output values can be stored in databases, but the interface has to be implemented by a developer. The output values (such as manufacturing time, average time of product waiting, etc.) from one simulation run (simulator) are not the optimized values, but only the result from a particular run. The optimized values can be found after performing more simulation experiments. The number of experiments required increases with the complexity of the model and the independent input values. There is also the need to set up new input parameters correctly with consideration of optimization gradient prediction. If the model is very simple and accuracy is not important, the input settings can be made manually. For common cases this process must be automated. A software solution for the automation of the simulation optimization process in parallel logic SIMD (Single Instruction – Multiple Data Stream) was developed using Rockwell ARENA simulation software. The simulation can be speeded up rapidly with the aid of a parallel approach. This software also supports networking and also acts as the reference for further work on methodology for the optimization method and its choice of presets.