Prediction Model Using LSTM-Based Double-GAN in CNC Machining
Junhae Lee, Sooeun Lee, Sangjoon Park · Information (Koganei) · 2024
This paper proposes an anomaly prediction model that can detect machining defects using the GANs and LSTM, and analyzes the data extracted from the sensors built into the CNC machine. To achieve this, we first select meaningful features from CNC facility data and apply them to the GAN, augmenting data of CNC machining. This approach helps overcome issues related to data scarcity and imbalance, ultimately generating anomaly detection factors for identifying defective products. In numerical results, through the proposed model of LSTM-based Double-GAN, their application in manufacturing processes, we contribute to substantial improvements in practical quality and cost savings. Key Words: CNC machining data, Double-GAN, Anomaly prediction, Data analysis, Quality Defects