Synergism of neural networks and expert systems for system identification

Abstract An integrated identification methodology is presented that appropriately embodies (1) the nonparametric identification using neural networks, and (2) the qualitative description by means of expert systems, along with (3) the conventional identification via mathematical models. In this article, the emphasis is on the issue of neural network identification. Particularly, an improved back-propagation (IBP) algorithm supported by expert systems is proposed that proves to be much faster and better than standard back propagation. Several examples, insights, and discussions are also provided. A synergism of neural networks and expert systems is expected to develop a new generation identification technology.