NLP group

Using adaptive neural networks for optimising discrete event simulation

Josef Baloun and Ladislav Lenc
International Journal of Simulation Modelling (IJSIMM) (2024)

PDF

Abstract

The paper presents the use of adaptive neural networks for carrying out simulation optimisation using digital models (discrete event simulation models) created in accordance with the Industry 4.0 concept. The digital models reflect different problems in industrial engineering. The simulation optimisers use an adaptive neural network to find the best settings of the digital models according to defined objective functions for each model. We compared the effectiveness (using different evaluation criteria) of the adaptive neural network (ANN) optimisation method used on 6 different discrete event simulation models. We compared adaptive neural networks with 11 optimisation methods – pseudo gradient, metaheuristic, evolutionary and swarm optimisation methods (and their combinations). The ANN method demonstrated the ability to efficiently find the global optimum of the objective function in different cases of the objective function – the ANN method is in the top 5 best tested methods from the 12 optimisation methods.