A multimodel approach for a nonlinear system based on neural network validity

Abstract

Purpose

The purpose of this paper is to present a new concept based on a neural network validity approach in the area of multimodel for complex systems.

Design/methodology/approach

The multimodel approach was recently developed in order to solve the modeling problems and the control of complex systems. The strategy of this approach coincides with the usual approach of the engineer which consists in subdividing a complex problem to a set of simple, manageable sub‐problems that can be solved separately. However, this approach still faces some problems in design, especially in determining models and in finding the appropriate method of calculating validities.

Findings

A novel approach based on neural network validity shows very remarkable performances in multimodel for complex systems.

Research limitations/implications

The validity of each model is based on the convergence of each neural network. For a fast convergence the proposed approach can be online to give a good performance in multimodel representation for system with rapid dynamics.

Practical implications

The proposed concept discussed in the paper has the potential to be applied to complex systems.

Originality/value

The suggested approach is implemented and reviewed with a complex dynamic and fast process compared to the residue approach commonly used in the calculation of validities. The results prove to be satisfactory and show a good accuracy.

Citation

Ben Mohamed, R., Ben Nasr, H. and M'Sahli, F. (2011), "A multimodel approach for a nonlinear system based on neural network validity", International Journal of Intelligent Computing and Cybernetics, Vol. 4 No. 3, pp. 331-352. https://doi.org/10.1108/17563781111160011