Abstract:

The paper deals with testing the frequency of using the database records of discrete event simulation model input parameters during the simulation optimization. We developed a simulation optimizer using the Server-Client architecture used for testing the behaviour of different optimization methods on discrete event simulation models. The server simulation optimizer reads and writes records (the settings of the simulation model input parameters and its objective function value) to internal and external databases. The data sent to the external database are encrypted. If the simulation optimizer wants to perform the simulation run with the concrete settings of the simulation model it searches for these concrete settings in the local database first. If this record is not found, the optimizer searches the external database. If the external database contains these settings of the simulation model and the objective function optimizer, it downloads these encrypted data and decrypts them. We implemented modified pseudo-gradient heuristic methods - Hill Climbing, Tabu Search, Local Search - to analyse their behaviour of searching for the global optimum in the defined Search Space. We tested different settings to reduce the impact of incorrect settings.