Monograph:

In industries, bearing plays a crucial part in rotating machineries. Any fault in the bearing causes major breakdown of rotating machineries. Therefore, it becomes essential to develop a system that can detect and classify the faults in bearings in order to reduce downtime. The research involves an experimental investigation to diagnose and classify the faults with the help of frequency spectrum analysis and advanced signal processing techniques. The experiment involves the vibrational signal analysis on various faults in taper roller bearing. The classifications are characterized corresponding to four load conditions i.e. no load, 100gms, 200gms, 300g, and faulty bearing i.e. outer raceway defect, inner raceway defect and roller defect. Time domain features are reformed to frequency domain features followed by bi-orthogonal wavelet transform. A comparison study is shown between a Butterworth filter and bi-orthogonal wavelet transform to show the effectiveness of the latter method. Also, the efficiency of two neural network is compared using Confusion Matrix, where Artificial neural network is 95% efficient and deep neural network is 99% efficient. The research also shows the same experimentation for acoustic emissions. Where the fault classification using ANN got a efficiency of 96% and using DNN, the efficiency was 98%.

To read further you can click on Download link.