Estimating the manufacturing cost of a metal part from textual and geometric features

Official URL: https://dx.doi.org/10.1109/UBMK63289.2024.10773581

This work presents a comprehensive approach to estimating labor hours for machining complex industrial metal parts, leveraging a rich dataset with diverse features like geometric dimensions, material properties, and operational parameters. Advanced ML models, including XGBoost, LightGBM, CNN, GBM, Random Forest, Linear Regression, Ridge, SVR, and MLP, are employed to predict labor hours. These predictions are a foundation for accurate cost estimation and are critical for efficient production planning and competitive pricing in the manufacturing sector. Our experimental study performed on a real-life metal-part dataset obtained from ERMETAL shows that the proposed LightGBM model estimates the manufacturing cost of a metal part with only 10.3% mean absolute percentage error.