Application of a system for the automatic generation of fuzzy neural networks

To facilitate the transfer of technology emerging from theoretical research of fuzzy neural networks into industrial applications, a fuzzy neural networks system for the automatic generation (FNNAGS) is proposed in this paper. In FNNAGS, the fuzzy model constructed by the system can be expressed as either a Mamdani model or a Takagi-Sugeno model, according to the preference of the user. Off-line design and on-line applications are incorporated into an interactive software system. In the stage of off-line design, only the training data need to be provided in order to construct a process model. Users do not need to give the initial fuzzy partitions, membership functions or fuzzy logic rules. These initial parameters will be set up automatically by the FNNAGS, in accordance with the properties of the training data. After off-line design has been completed, the model can be expressed as a fuzzy rule base, which can be used to control, estimate, identify or predict a process or plant through an application interface between FNNAGS and the external world. (C) 2000 Elsevier Science Ltd. All rights reserved.