Integrating machine learning and feature analysis for predicting and managing thermal deformation in machine tools

• Published in: Case studies in thermal engineering 2024-05, Vol.57, p.104343, Article 104343
• Main Author: Chu, Wen-Lin
• Format: Article
• Language: English
• Subjects: Temperature sensitive point; Thermal error compensation; Time series neural network
• ISSN: 2214-157X; 2214-157X
• DOI: 10.1016/j.csite.2024.104343

This study develops a method to predict and compensate for thermal deformation in machine tools, focusing on selecting temperature-sensitive points, establishing a predictive model, and detecting temperature anomalies. Optimal sensing points were identified using feature ranking algorithms and Partial Dependence Plots, improving Z-axis deformation identification. The research also evaluated various machine learning models for Z-axis deformation prediction, notably optimizing Gaussian Process Regression for superior accuracy. Additionally, anomaly detection in temperature sensors was addressed using One-Dimensional Convolutional Neural Networks and Long Short-Term Memory Networks, enhancing system reliability and robustness. This approach offers a comprehensive solution for thermal deformation in machine tools, contributing to the field's precision and efficiency.