An expert model of switched reluctance motor using decision tree learning algorithms

In this paper two different decision tree learning systems for modeling of a switched reluctance motor have been developed. The design vector consists of the design parameters in the first one whereas in the second one, it is a combination of hysteresis current band in the current limiter and the switching angles. The output performance variables are efficiency and torque ripple in both systems. An accurate analysis program based on improved magnetic equivalent circuit (IMEC) method has been used to generate the input-output data. These input-output data is used to produce the decision trees for predicting the performance of switched reluctance motor (SRM). The performance prediction results for a 6/8, 4 kw, SR motor show good agreement with the results obtained from IMEC method or finite element (FE) analysis. The developed decision tree systems can be used for fast prediction of motor performance in the optimal design process or on-line control schemes of SR motor.