Deep Learning For Fault Diagnosis In Rotating Machine

Abstract

Rotating machines are crucial in industries for reliable system operation, but unexpected failures can result in significant financial losses and personnel damage. Hence, fault diagnosis is important. Among the common types of faults in rotating machines, unbalanced and misalignment are important types of faults in operation but are rarely studied. This article uses data from four types of operating conditions, namely (i) normal, (ii) imbalanced, (iii) imbalanced associated with horizontal misalignment, and (iv) imbalanced associated with vertical misalignment. This study proposed a method for fault diagnosis using Continuous Wavelet Transform (CWT) for extracting vibration signals into RGB images and using Convolutional Neural Network (CNN) models for fault diagnosis. Moreover, several machine learning models are also utilized for comparison with the proposed method such as Random Forest (RF), Multilayer Perceptron (MLP), and Xtreme Gradient Boosting (XGBoost). The results show that the diagnosis method CNN achieves an overall accuracy of 93.27%, which is superior to other machine learning methods.