Surrogate modelling of thermal and residual stress fields in cold-spray additive manufacturing using machine learning

Abstract

In cold spray additive manufacturing, temperature and residual stress evolution significantly affect product quality and service life. Conventional finite element (FE) simulations provide predictions but are computationally expensive for large-scale applications. To address this, we developed machine learning (ML) surrogate models with physics-guided features to predict temperature and residual stress fields efficiently. FE simulations generated datasets for training and testing, with extracted physical features serving as model inputs. Several ML methods, including XGBoost, Neural Networks, and Extra Trees, were implemented. Model performance was validated, and generalisation was tested under varying process parameters such as path pattern, deposition speed, and initial temperature. Across these tests, all models achieved R² values around or above 0.9, demonstrating strong predictive capability. Neural Networks provided the highest accuracy for temperature prediction, while Extra Trees excelled for residual stress. These results indicate that ML surrogate models can efficiently and accurately predict thermal and mechanical behaviour in cold spray additive manufacturing, offering a promising alternative to conventional FE simulations for process optimisation.