Data-driven uncertainty process modeling and robust design of superalloy investment casting

Uncertainty in process parameters is a universal challenge in materials manufacturing, leading to deviations between designed and actual outcomes. This issue is exacerbated in investment casting due to its lengthy process chain and multiple manufacturing steps, where traditional design methods struggle to maintain robustness against fluctuations. This study presents a data-driven reliability-based approach for gating system design, integrating uncertainty modeling and robust optimization. By sampling real-world production data and quantifying uncertainties from various manufacturing processes, the impact of key process parameters was evaluated, and a predictive model linking design and processing variables to defect formation was developed. By incorporating the uncertainty model into the data-driven predictive model via reliability-based algorithms, a robust gating system design was achieved. Compared to conventional deterministic design and empirical thermal modulus methods, the proposed approach achieves 24.4% improvement of production rate and 1.6 0.2% reduced failure probability. Simulation and experimental results consistently validate its robustness against process variations. The established data-driven workflow provides a generalizable template for improving investment casting reliability across other structural geometries and process conditions.