Using neural-fuzzy in control applications

Fuzzy logic is gaining widespread acceptance in the control engineering community because of its continued success in control applications. However, certain inherent difficulties of the approach are restricting its growth. Difficulties include: developing fuzzy rules by hand for large systems, selecting appropriate membership function shapes, fine-tuning fuzzy solutions for specific levels of accuracy, and guaranteeing the reliability/robustness of solutions. A neural-fuzzy based approach to developing fuzzy logic designs eliminates some of these problems. This paper discusses using a neural-fuzzy approach, based on NeuFuz, to designing control systems. It presents the steps involved in the NeuFuz approach, as well as a real design example for DC motor control. The objective of the example was to design a controller that minimized overshoots during motor start-up and maintained the desired motor speed as the motor load varied. This example has widespread applicability because of the number of applications that involve some type of motor control. Whether a small induction motor is needed for proper autofocus in a camera or a servo motor is needed for proper positioning of the head in a high-speed hard disk drive, the algorithm for motor control presented in this paper may be used to develop a solution. This paper also includes information on the performance of the NeuFuz-designed motor controller and some comments on other types of control