Class Contribution Feature Weighting Based K-Nearest Neighbor for Fault Detection and Diagnosis in Vibration Signal Processing

The Fault Detection and Diagnosis (FFD) are the necessary in the growth of condition monitoring and vibration analysis is utilized for machine rotating in different industries. The vibration signal analysis is performed for fault detection which is suitable for classifying to carry out fault detection and diagnosis through vibrations in bearings. However, the fault detection and diagnosis are challenging due to complexity in vibration patterns which makes difficult to identify and classify abnormalities. To overcome these challenges the Machine Learning (ML) techniques is utilized for detect and diagnosis in fault detection. In this research the Class Contribution Feature Weighting based K-Nearest Neighbor (CCFW-KNN) is proposed for FFD. In the feature extraction techniques, the Time Domain Analysis (TDA) and Frequency Domain Analysis (FDA) like Root Mean Square (RMS), kurtosis and Four Fourier Transform (FFT) are performed to extract the charateristics. A proposed CCFW-KNN achieves better accuracy of 99.48% through using Case Western Reserve University (CWRU) dataset.