Additive manufacturing enables multi-material functionally graded materials (FGMs) with expanded functionality, yet post-machining remains challenging because flow stress varies with composition. This paper presents a data-efficient, physics-guided Gaussian process (GP) model for predicting milling forces in SS316/IN718 FGMs without repeated force-coefficient identification. A physics-based milling model with mixture-law baselines is corrected using sparse force measurements, while the GP learns the residual as a smooth function of composition and parameters. The model reduces prediction error from 26.6 to 19.2% for feed force and from 19.8 to 10.6% for normal force, while adaptive feed scheduling lowers peak-force variation from ∼50 to ∼8%.
Jin et al. (Wed,) studied this question.