Rotating Machine Fault Detection Using Support Vector Machine (SVM) Classifier

Abstract

Automatic machine fault detection is a critical task in the industrial sector as it allows for the early identification of problems and prevents equipment breakdowns, which can reduce costly downtime and repairs. Artificial intelligence (AI) technology will be an essential tool for diagnosing rotating machine health. The key challenge in modern industry is to detect rotary machine faults in time and location of the fault. Rotating machines generally suffer from unwanted breakdowns because of defective bearings, miss-alignment, unbalance, or looseness. An efficient and accurate method for detecting defects in rotary machines using a machine-learning algorithm is proposed in this research. A data acquisition system is designed to collect necessary information from different parts of the rotating machines, and an AI-based fault detection model determines the machine's health. Furthermore, it validated the result of the actual industrial data collected from one cement industry in Chittagong. The proposed method uses Support Vector Machines (SVM) with a Gaussian kernel function and Principal Component Analysis (PCA) technique for feature extraction and dimensionality reduction. Bayesian optimization is also employed to optimize the hyperparameters of the model. The support vector machine (SVM) classifier achieved 98.2% accuracy, and implementing the model using the medium class neural network with the Rectified Linear Unit (ReLU) activation function resulted in 95.6% accuracy. The proposed method could detect faults in rotary machines accurately, which has the potential to enhance the overall efficiency and reliability of industrial operations. Using the PCA technique and Bayesian optimization in this work provides a promising approach for researchers and practitioners in the field of machine fault detection.