Mixture Based Classifier Using Gaussian Processes for Induction Motor Diagnosis

This paper explores the use of Gaussian Processes (GPs) and Gaussian Mixture Models (GMMs) for diagnosing induction motors. GPs provide flexible, non-linear models that handle noisy data, while GMMs offer robust probabilistic classification by modeling data as mixtures of Gaussian distributions. By integrating Bayesian inference with GMMs and utilizing Stan for complex model management, the study enhances classification accuracy. Experimental data from induction motors under various conditions were analyzed, identifying patterns indicative of motor health. The approach, leveraging synthetic data generation, demonstrates effectiveness in proactive maintenance and fault detection, reducing downtime and costs.