An evolutionary approach to optimize speed controller of DC machines

The conventional controllers used for DC machines are static and their parameters are fixed through proper design. The classical approach is to use a PID controller with constant parameters after analyzing the stability criterion. The modern approach is to use controllers based on fuzzy logic or other AI techniques. The authors have chosen a speed-tracking problem where a DC machine has to follow a time varying speed demand. The controller coefficients are fixed through an evolutionary algorithm. Representative values of steady state error, maximum overshoot and transient rise time are computed through feature extraction algorithms. Now, the fitness of each member is computed as a fuzzy value based on some predefined fuzzy functions involving the feature values. This fuzzy fitness value governs the selection of coefficients through a genetic algorithm until convergence is obtained. The performance has been studied with various fitness functions and the results are found to be satisfactory.