Enhancing Fault Detection in CNC Machinery: A Deep Learning and Genetic Algorithm Approach

Abstract

This paper introduces a deep learning-based approach (DLBA) tailored for fault detection and condition monitoring in industrial machinery. The presented DLBA architecture is assessed utilizing a dataset derived from a CNC milling machine, as part of the University of Michigan's System-level Manufacturing and Automation Research Testbed (SMART). The results underscore the substantial efficacy of the LSTM model, evidenced by a precision of 94% and an F1 score of 95%. These findings serve as a robust foundation for the identification of CNC machining failures within the manufacturing industry. The DLBA architecture constitutes a comprehensive framework for efficient fault detection, incorporating a variety of network models, including MLP, CNN, CNN auto-encoder, LSTM, and ResNet. Each model leverages its unique strengths in the analysis of complex data. Genetic Algorithm (GA) optimization is employed for parameter tuning across all models, enhancing their performance. The findings from this study contribute significantly to the development of more reliable and cost-effective predictive maintenance systems, enabling manufacturers to detect faults early on, thus preventing expensive downtime and mitigating safety risks.