Speed Control of an IM with Adaptive Neuro-Fuzzy Inference Strategy

Abstract

Abstract. The paper proposes development of a robust control scheme for speed control of an Induction  Motor (IM) by making use of Neural Networks (NN) and Fuzzy Logic techniques. IM is an integral part of modern industries. Unlike DC motors, controlling of an IM is complex due to its highly nonlinear characteristics. The conventional controllers are not robust and accurate. The Artificial Intelligent controllers are gaining popularity due to the adaptability they bring into the system. Fuzzy logic is capable of gaining the knowledge of system involved and NN is capable of training the system to be adaptable. The strategy developed with such a combination of technique is referred as ANFIS. The developed ANFIS speed controller for an IM consists of five main blocks which are Fuzzification process, Rule base, NN training algorithm, De-fuzzification process and Power modulator. The crisp data inputs, speed error and change in speed error are converted into system understandable linguistic values. The knowledge block is made up of pre-computed input and output sets called as rule base. A network is trained with Back propagation algorithm to select a proper rule. The selected rule is triggered to obtain a proper control signal which is converted to the crisp form before being fed to the power modulator. The power modulator provides the controlled step input voltage to an IM. The results are obtained by simulating the system model in Matlab Simulink environment. The results obtained with developed ANFIS controller when compared to other existing models, shows the improved transient response and considerable decrease in the Torque and Current waveforms. The settling time of speed of a rotor has decreased slightly.