Predictive Maintenance System with Secure Data Transmission using Machine Learning and Signal Processing Techniques for Vibration Analysis

This work addresses the challenge of predictive maintenance in industrial machinery, focusing on early fault detection in helical rack CNC cutting machines. The proposed solution integrates secure data acquisition and transmission, advanced signal processing techniques, and machine learning algorithms for real-time, continuous vibration analysis and fault prediction. The system incorporates token-based socket programming for enhanced security and efficient communication between system components. Key novelties include the seamless integration of cybersecurity measures with vibration analysis, a hybrid feature extraction approach combining time and frequency domain analysis, and an adaptive Support Vector Data Description (SVDD) model for evolving fault classification. The system offers significant advantages in reducing unplanned downtime, decreasing the dependency on preventive maintenance. Thereby reducing the maintenance costs, and improving overall effectiveness. Performance metrics demonstrate 92% fault detection accuracy, reduction in maintenance costs, and decrease in unplanned downtime. This solution holds significant application in the manufacturing industry, enabling tactical maintenance strategies and enhancing operational efficiency in the industrial settings.