Multi-constraint optimization of riser design for low-pressure casting of a gearbox housing

Abstract

Dispersed hot spots in ZL105 aluminum-alloy gearbox housings for heavy vehicles make it difficult for conventional riser design to balance feeding reliability and material utilization. A multi-constraint riser design method was therefore developed for an existing mold interface and a fixed layout of two top risers and two side risers. The objective was to minimize the total volume of the riser bodies and necks, while modulus, feeding-capacity, geometric-process, and installation-space requirements were formulated as 13 inequality constraints and one equality constraint. Analytical elimination of the equality constraint reduced the model to seven independent variables. Riser dimensions were solved using a particle swarm optimization-sequential quadratic programming (PSO-SQP) strategy and independently verified against genetic algorithm (GA), PSO, and multi-start SQP solutions under a common evaluation budget, with 30 runs per method. All four methods converged to the same feasible engineering-boundary solution, with a total volume of 176494.675 mm 3 . Feasibility was retained under a joint adverse combination of ±10% parameter variations and in 100,000 independent perturbation samples. ProCAST simulations and low-pressure casting experiments showed that the average shrinkage-porosity fraction decreased from 13.93% to 2.84%, and the defect morphology changed from connected clusters to dispersed spots. The field-averaged porosity in the top and side sampling regions decreased to 0.68% ± 0.28% and 0.42% ± 0.19%, respectively, with no evident surface sinks, hot tears, or macroscopic shrinkage cavities at the corresponding hot spots. The proposed workflow integrates hot-spot identification, constraint modeling, dimensional optimization, and numerical and experimental validation, providing an accurate and efficient engineering approach for intelligent riser design in complex thin-walled housings.