Development and Implementation of a Simplified Self-Tuned Neuro–Fuzzy-Based IM Drive

Abstract

A novel simplified self-tuned neuro-fuzzy controller (NFC) for speed control of an induction motor (IM) drive is presented in this paper. The proposed NFC combines fuzzy logic and a four-layer neural network structure. Only speed error is employed as input to the NFC so that the computational burden of the NFC is reduced and it becomes suitable for real-time industrial drive applications. Based on the knowledge of back-propagation algorithm, an unsupervised self-tuning method is developed to adjust membership functions and weights of the proposed NFC so that the performance will be similar to that of the conventional two-input NFC. The complete drive incorporating the proposed self-tuned NFC is experimentally implemented using a digitalsignal-processor board DS-1104 for a laboratory 1/3-hp motor. The effectiveness of the proposed NFC-based vector control of IM drive is tested in both simulation and experiment at different operating conditions. Comparative results show that the simplification of the proposed NFC does not decrease the system performance as compared to the conventional NFC. In order to prove the superiority of the proposed simplified NFC, the performances of the proposed NFC are also compared to those obtained by a conventional proportional-integral controller.