• A novel deep learning retrieval method integrates process design features (modulus, wall thickness, symmetry, length-to-height ratio) with CLIP-based shape features, significantly improving process reuse.
• Incorporating process design features boosts retrieval accuracy by 6.09% to 97.82% and increases process similarity by 30.25% on a database of 230 production casting models.
• The retrieved process, validated via EasyCast simulation, yields minimal shrinkage in the target model, demonstrating superior process reuse capability.
• The method requires no large training dataset, making it highly applicable to casting process design and related manufacturing.