Numerical simulation-based optimization of riser design and defect control for ductile iron castings

Abstract

This study proposes a riser optimization framework combining numerical simulation and Gaussian Process Regression (GPR) to reduce shrinkage defects in ductile iron brake calipers. ProCAST analysis identified severe shrinkage at the piston cavity bottom. Orthogonal experiments showed that the riser diameter ratio (DR/T) and modulus (S) are the dominant factors affecting feeding performance. To address the limitations of small-sample simulations, a Bayesian-based GPR surrogate model was established, achieving high predictive accuracy (R² = 0.9954) and outperforming conventional BP neural networks. Using the optimized parameters, the shrinkage volume was reduced from 7.8418 cm³ to 1.2549 cm³, corresponding to an 84% reduction. Final simulations confirmed that shrinkage defects were effectively transferred into the riser, improving densification in critical regions. The proposed framework offers an efficient digital optimization strategy for complex ductile iron castings.