Quantification of Feature Effects on the Accuracy of Bearing Fault Diagnosis in Inverter-Fed Induction Motors

Induction motors have an important role in industrial plant. They can suffer from different kinds of faults including electrical and mechanical faults. Machine Learning (ML) algorithms are effectively used to detect these faults. However, the application of inverter fed induction motors affects fault diagnosis accuracy significantly. The basic requirement for accurate performance of ML techniques is based on appropriate feature engineering. This paper evaluates and quantifies the effect of each feature on accuracy of fault classifier. Feature engineering steps progressively move from raw motor phase current to multi-sensor fusion. Results show significant improvement in fault diagnosis accuracy especially due to integration of voltage and current feature fusion.

کلیدواژه ها:

Inverter fed motor ، Machine Learning ، Bearing faults ، Feature engineering