Data-Driven Fault Detection and Risk Assessment for Machine Tools Using Sensor Data

This paper proposes a data-driven methodology for machine tool fault detection and risk assessment, enhancing diagnostic accuracy and failure mode prioritization. The approach integrates machine learning and statistical analysis within a data-driven Failure Modes and Effects Analysis (FMEA) framework, employing a semi-supervised strategy that combines K-means clustering with similarity-based classification. To validate the method, four datasets are constructed using operational sensor data from baseline conditions and simulated failure modes under real-world scenarios. Experimental results show that the proposed framework effectively identifies high-risk failures with minimal data, outperforming traditional supervised and unsupervised methods. The findings demonstrate that this approach improves fault detection, optimizes Risk Priority Number (RPN) computation, and strengthens FMEA objectivity, contributing to predictive maintenance and machine performance optimization.