Intelligent Monitoring System for Machinery Manufacturing Process Based on Deep Learning

Abstract

This study introduces an Intelligent Monitoring System for Machinery Manufacturing Process (IMS-MMP) that leverages deep learning techniques and threshold-based anomaly detection algorithms to enhance operational efficiency and minimize downtime in manufacturing environments. The system utilizes Recurrent Neural Networks (RNNs), specifically Long Short-Term Memory (LSTM) networks, to analyze time-series sensor data collected from machinery components and predict potential equipment failures. Additionally, threshold-based anomaly detection algorithms are integrated to identify deviations from normal operating conditions and trigger timely alerts for maintenance interventions. Through comprehensive experimental validation, IMS-MMP demonstrates high predictive accuracy, with LSTM-based predictive models achieving a mean absolute error (MAE) of 3.2% in forecasting machinery failures. The anomaly detection algorithms exhibit robust performance, with a true positive rate (sensitivity) of 92%, a true negative rate (specificity) of 95%, and a low false positive rate (FPR) of 3%. Moreover, the deployment of IMS-MMP results in significant improvements in operational efficiency, including a 30% reduction in unplanned downtime, a 20% decrease in maintenance costs, a 15% increase in overall equipment effectiveness (OEE), and a 25% improvement in production throughput compared to traditional reactive maintenance approaches. This study highlights the potential of IMS-MMP as a valuable tool for proactive maintenance and anomaly detection in machinery manufacturing processes, offering manufacturers a data-driven approach to optimize operational performance and maximize productivity in the Industry 4.0 era.