Independent Component Analysis (ICA) and Statistical Tool-based Classification and Detection of Induction Motor Faults

Induction motors perform an important role in ensuring a smooth advancement of industries, but some harsh operational condition leads to premature failure in the stator winding or/and rotor, which may create a shutdown condition to many industrial applications and even creates a loss in the production process and also in money. This paper presents a technique for early detection of rotor faults and stator faults by using Park’s transformation in stator line current and get d-q current component on which Independent Component Analysis (ICA) is used to decompose in various component by MCSA (Motor Current Signature Analysis) an effective condition monitoring technique used for diagnosis of induction motor faults. The fault detection algorithm of ICA, an effective Blind Signal Separation (BSS) procedure, is implemented using MATLAB programming. Using kurtosis, margin, and peak factor as statistical quantifying tools helps create separate fault ranges to classify the fault types. Hardware experiments have been executed with same-rated induction motors with different types of fault creation set up by changing the degree of load level.