Deep Learning for Anomaly Detection in CNC Machine Vibration Data: A RoughLSTM-Based Approach

Abstract

Ensuring the reliability and efficiency of computer numerical control (CNC) machines is crucial for industrial production. Traditional anomaly detection methods often struggle with uncertainty in vibration data, leading to misclassifications and ineffective predictive maintenance. This study proposes rough long short-term memory (RoughLSTM), a novel hybrid model integrating rough set theory (RST) with LSTM to enhance anomaly detection in CNC machine vibration data. RoughLSTM classifies input data into lower, upper, and boundary regions using an adaptive threshold derived from RST, improving uncertainty handling. The proposed method is evaluated on real-world vibration data from CNC milling machines, achieving a classification accuracy of 94.3%, a false positive rate of 3.7%, and a false negative rate of 2.0%, outperforming conventional LSTM models. Moreover, the comparative performance analysis highlights RoughLSTM's competitive or superior accuracy compared to CNN-LSTM and WaveletLSTMa across various operational scenarios. These findings highlight RoughLSTM's potential to improve fault diagnosis and predictive maintenance, ultimately reducing machine downtime and maintenance costs in industrial settings.