Fault Diagnosis for Rotating Machine Utilizing Machine Learning with FFT-based Feature Extraction

Rotating machines have an important role in industrial production which requires a regular scheduled service and routine checks. Any unexpected failures can result in significant financial losses and personnel damage. Thus, fault diagnosis is crucial. Among the common types of faults, unbalanced and misalignment which cause the high vibration levels are rarely studied. Therefore, this paper proposes a method for fault diagnosis using the Fast Fourier Transform (FFT) for extracting features from a collected dataset of vibration signals. Several machine learning models are utilized for training after applying feature extraction such as Random Forest (RF), Multilayer Perceptron (MLP), and Xtreme Gradient Boosting (XGBoost). Based on the outcomes of various tests, Random Forest has outperformed the others in the frequency range of 30Hz-40Hz, with an accuracy of 91.14%.