A Fuzzy Membership Function Design Methodology Based on Histogram and ROC Curve Analyses for Broken Rotor Bar Detection

The innovation of this work is the introduction of a novel fuzzy membership function design methodology, which incorporates the issue of the trade-off between the hit and false alarm rates into the design process. The method is applied to two-class broken rotor bar detection problem in squirrel-cage induction motors. The methodology is called H-ROC since it uses a weighted cost function based on ROC (Receiver Operating Characteristics) curve analysis and histogram analysis.