Modeling and optimization in complex systems based on computational intelligence

Abstract

Purpose

For many optimization problems such as optimal techniques, compositions, producing process, the optimizing objectives in systems are complex relations with respect to a great deal of parameters. Generally, the objective function is hardly obtained, even the searching objective is unquantifiable. So it is difficult to model and optimize the complex systems to some extent.

Design/methodology/approach

To the above purpose, a novel approach is presented in this paper. It firstly utilizes the excellent fitting performance of neural network (NN) combined with expert knowledge (EK) to obtain the objective function, and secondly searches the optimal influential parameters with genetic algorithm (GA).

Findings

Peaks function inside Matlab and the acural application of comprehensive performance optimization in analog PID control system are studied, respectively. The results of simulation and the actual experiment both show that the modeling and optimizing method presented in the paper are effective.

Research limitations/implications

Expert knowledge is needed for the fuzzy/unquantifiable objective.

Practical implications

The paper presents a useful way to model and optimize complex systems.

Originality/value

The combined approach based on NN, EK and GA is firstly presented and is effectively used in modeling and optimizing complex systems.

Citation

Su, Y., Li, T., Wang, D. and Liu, X. (2012), "Modeling and optimization in complex systems based on computational intelligence", Kybernetes, Vol. 41 No. 9, pp. 1235-1243. https://doi.org/10.1108/03684921211275252