A Novel Visualization Methodology for Characterizing Magnetic Flux Sensors in Induction Motors: Discrimination Properties Insights

Magnetic flux sensors used to diagnose induction motors provide a reliable and easy way to detect electrical and mechanical failures. However, those sensors are sensitive to environmental noise, positioning, and structure. A visualization technique based on statistical indices and dimensionality reduction producing a 3D plot highlighting sensor properties to characterize the sensor behavior is presented in this research. The proposed methodology uses a linear discriminant analysis (LDA) model to characterize the healthy induction motor in steady-state operation using three types of sensors: one current clamp, one triaxial magnetic Hall-effect sensor, and two handmade coil sensors. The results obtained show characteristic shapes for each type of sensor, which could be used to identify the presence of faults with the changes in the shapes produced.