Devising a Vibration-Based Fault Detection System for Textile Machinery

The textile sector is the backbone of the economy of many developing countries in South Asia. Diverse machinery fault caused by intensive production schedules during operation is a major concern for industries in this sector. There exist several systems in the state-of-the-art literature for detecting textile machinery faults where faulty output is already produced before machine fault detection. In this study, we propose a vibration-based machinery fault detection system for the textile industry. We use a highly sensitive accelerometer to detect even the tiniest vibration changes. Using the accelerometer, we produce a data set by creating six artificial faults in the machine and measuring the vibration of the machine during those faults. Next, we perform Fast Fourier analysis to derive the machine frequency and statistical analysis to detect vibration variation during different faults. We find that there is a change in the machine frequency and vibration respectively during different faults. Then, we run eight different machine learning algorithms to detect the type of fault in the machine. We measure the precision, recall, and F1 score of our machine learning models through ten-fold cross-validation. We get the highest F1 score of 98.9% using the Decision Tree classifier. Finally, we construct a real device by implementing our trained machine learning model in Arduino to identify machine faults which demonstrate the utility of our proposed approach in real scenarios.