Manufacturing Quality Management Based on TimeGAN and Seq2Seq Models With Magnetic Press Machine Data

초록

Despite significant advancements in artificial intelligence (AI), the practical application of AI to realworld industrial data remains limited. Historically, research has concentrated on theoretical developments and algorithm improvements. However, there is now a growing need to apply AI collaboratively in actual production settings. This paper presents an AI-based quality management framework integrating TimeGAN and Seq2Seq models, specifically employing time-series data collected from a magnetic press machine used in permanent magnet manufacturing. The proposed Seq2Seq LSTM combined with TimeGAN achieved outstanding performance, demonstrating the lowest mean absolute error (MAE) of 0.88 and the highest R² value of 0.94 among tested models. Additional anomaly detection experiments confirmed the model’s effectiveness, exhibiting competitive recall (0.97) and F1-Score (0.93) results. These findings illustrate the significant potential of AI integration for quality control and process enhancement, suggesting broad applicability across various manufacturing industries.

키워드