Prediction of broken rotor bar in induction motor using spectral entropy features and TLBO optimized SVM

The information of the fault frequency characteristics is great importance for all associated diag nostics. This requires a high-resolution spectrum analysis to achieve efficient monitoring machinery faults, especially while diagnosing rotor bar breakage under light load conditions, because frequencies almost overlap with fundamental. In this context, rather than looking several bands are observed separately in terms entropy contained within these bands. First, motor current signal has been divided into using continuous wavelet transform (CWT), and spectral calculated from each band as features describe condition. Principal component (PCA) used feature reduction tool, projected onto first two principal components have fed SVM inference. supervised learning method classification regression analysis. To improve performance, radial basis function (RBF) kernel employed, find optimal value parameters, metaheuristic approach, namely teaching learning-based optimization (TLBO), utilized. ANSYS 2D workbench simulate finite element model (FEM) an induction broken bars, efficacy proposed then tested simulated data. investigate robustness white Gaussian noise also added data, performance features.