A Novel Approach to a Few Shot Learning Techniques Based on Thermal Error Modeling for Slant Bed CNC Lathe Machine

Thermal errors in precision machining can influence machine tool accuracy and effectiveness. This research addresses these challenges by emphasizing temperature sensitivity. An enhanced approach is presented that combines the fuzzy C-means clustering algorithm with the Grey Wolf Optimizer, followed by the Grey Relation Coefficient, to select temperature sensitivity points (TPS). To address the issue of limited datasets in machine learning, this study introduces a new method based on Few-shot learning models, including Model-Agnostic Meta-Learning (MAML), Matching Networks, and Siamese Networks. The models are evaluated for their ability to predict thermal errors using a limited dataset. The MAML model demonstrates notable predictive performance, with an R² value of 0.956, RMSE of 0.208, and MAE of 0.160 in the X-direction; and an R² value of 0.996, RMSE of 0.657, and MAE of 0.495 in the Y-direction. Furthermore, MAML effectively predicts thermal errors, contributing to the enhancement of tool machine accuracy during operation. This research fills a gap in current techniques and serves as a foundation for future studies on machine tool thermal error correction.