Wavelet in power system application-Fault

Power System

1. Bearing Faults Classification by SVM & SOM Using Complex Gaussian Wavelet Kalyan M Bhavaraju P.K.Kankar S.C.Sharma S.P.Harsha Vibration & Noise Control Laboratory Mechanical & Industrial Engineering Department Indian Institute of Technology Roorkee India

2. What is the need for Fault Diagnosis of Bearings in Electric Motors??????

3. Fault Diagnosis Strategy Proposed for Effective Diagnosis of Bearing Faults

4. Cases Considered for the Present Study Bearing with no fault “BNF” Spall on outer race “SOR” Spall on Inner race “SIR” Spall on Ball “SOB” Combination of bearing component faults “CFB” Parameter Value Outer race diameter 28.262 mm Inner race diameter 18.738 mm Ball diameter 4.762 mm Ball number 8 Contact angle 0° Radial Clearance 10 µm

5. Wavelet Complex Gaussian Wavelet 7 th level of decomposition Maximum Energy Criterion (MEC) Scale no.24 has max energy

6. Statistical Features Skewness kurtosis Skewness Kurtosis Standard Deviation

7. Artificial Intelligence Techniques Input Bundle of synaptic connections Winning neuron Two dimensional array of postsynaptic neurons Supervised Technique SVM Un-Supervised Technique SOM

8. Results SOB SIR CFB BNF SOR Classified as 14 1 SOB 15 SIR 15 CFB 15 BNF 1 14 SOR SOB SIR CFB BNF SOR Classified as 13 1 1 SOB 11 2 2 SIR 3 12 1 1 CFB 4 2 9 BNF 2 1 1 1 10 SOR Confusion Matrix for SVM Confusion Matrix for SOM Parameters SVM SOM Correctly Classified Instances 73(97.3333%) 55(73.333%) Incorrectly Classified Instances 2(2.6667%) 20(26.6667%) Total No. of Instances 75 75

9. Conclusions MEC could be effectively applied for fault features detection Effective application of AI techniques for bearing faults classification Efficiency of SVM is more than SOM

10. Thank U Vibration & Noise Control Laboratory Mechanical & Industrial Engineering Department Indian Institute of Technology Roorkee India