Fault Diagnosis in Rotating Machines using Vibration Signal Analysis and Machine Learning

Timely diagnosis of bearing fault is vital for various industries, where reliability and efficiency of rotatory machines are of paramount importance. In rotatory machines, bearing failures not only lead to catastrophic breakdowns but also pose a significant safety risk to human beings. Therefore, early detection of bearing faults can prevent accidents, ensuring a safer working environment, and maintain the desired production level. Here, we proposed a machine learning algorithm to detect and diagnose 4 types of rotating machine faults operating under different speed conditions. First, we obtained the vibration signals from the bearing under test. Then, we segmented the data into a total of 56 segments of equal time duration. Next, we extracted Mel Frequency Cepstral Coefficients Delta (MFCC-Delta) from the segmented data. In the end, we classified the faults by providing these delta cepstral features as input to different classifiers. Quadratic Discriminant (QD) outperforms all other classifiers with an accuracy of 98.7%. We expect that our proposed approach will be beneficial for various industries in predicting various bearing faults, reducing maintenance and downtime expenses.