Fault Diagnosis of Ball Bearing Using Low-Cost Edge Device and Deep Learning

The ongoing competitive environment in the industries forces them to adopt economical and efficient methods for fault diagnosis in the components of rotating machinery. Keeping this into consideration, the present work focuses on the economical and efficient diagnosis of faults in the bearing. In the present work, a low-cost edge device and an economical transducer are used for the acquisition of signal information from the rotating machines with different fault conditions. Further, the obtained signals are processed through the Hilbert transform to streamline the fault information. The processed signal is then used for the generation of time–frequency spectrograms using Continuous Wavelet Transform (CWT). The obtained spectrograms are used as input to a designed Deep Convolutional Neural Network (DCNN) for the diagnosis of various faults in different speed conditions. The results are analyzed using DCNN, and its robustness is validated at different speed conditions. The obtained results demonstrate that the proposed method has the capability to diagnose different bearing faults with substantial accuracy, having the highest value of 98.7%.