Analytical Approximation of Induction Motor Reactance Using Symbolic Machine-Derived Models

This paper predicts the induction motor magnetizing reactance using three symbolic regression methods. TuringBot, Eureqa, and HeuristicLab have been employed in this paper to develop easy and accurate mathematical models for the reactance of single and double-cage induction motors. The reactance can be instantly evaluated using these models by knowing only a few parameters from the motor nameplate data, and without doing any field measurements. The performance of the developed models has been benchmarked with various previous methods reported in the literature. The developed models have demonstrated efficient performance with a high level of predictability for the reactance. The results show that the coefficients of determination and root mean square errors for the testing phase using TuringBot, Eureqa, and HeuristicLab are 0.997946 and 0.014544, 0.998337 and 0.013088, and 0.99622 and 0.019732, respectively.