Classification and redundancy quantitative evaluation algorithm for highly efficient fault diagnosis of rotary machines in roll-to-roll system

The roll-to-roll continuous process is a manufacturing technique widely used in several industrial appliances, including electronic devices. Real-time monitoring and fault diagnosis are crucial in this process because it involves several rotary machines, and their failure can cause speed disturbance that can result in tension disturbance and easily degrade the geometrical quality of the coated layer. Therefore, establishing an effective diagnostic system can prevent unexpected damage and reduce maintenance costs by predicting failures in advance. Herein, we propose a diagnostic model based on tension and acceleration data. Further, we establish window size calculation criteria for feature extraction and propose classification and redundancy quantitative evaluation algorithms based on the density between feature classes and the Mahalanobis distance. The results indicate that by using the proposed method, the average accuracy is improved by 6.07% compared to that of the existing method, and the learning time is reduced by 29.44%.