Diagnosis of induction machine stator faults by parameter estimation technique based on direct search on sparse grid

Abstract

Fault diagnosis is widely applied in asset condition monitoring of induction machines to identify potential problems before serious deterioration or breakdown occurs. The confronted challenge is that the estimation problem exhibits nonlinear property and multiple local minima and the objective function surfers from measurement noise. To tackle this challenge, a novel parameter estimation technique based on a global direct search method, named Hyperbolic Cross Points (HCP) algorithm, is proposed. Accordingly, it is applied for the non-intrusive detection of stator winding short circuit fault in squirrel-cage induction machines. The simulated stator current used in testing HCP method is generated from a MATLAB/SIMULINK induction motor model. The algorithm has been validated by considering various stator short circuit levels and fault locations. After the algorithm has been carefully tested, it is used to estimate fault parameters from recorded current from a 800W squirrel-cage induction motor in the laboratory environment. It demonstrates the effectiveness of the proposed method by detecting both the fault location and short circuit level with high accuracy and reduced computational burden.