Using AI to Enhance Quality of the 3D Sand Printing-Metal Casting Manufacturing.

Abstract

The integration of 3D sand printing (3DSP) with conventional metal casting presents an innovative solution to address the complexities of modern manufacturing in achieving high-quality, high-volume production with reduced lead times. To achieve these benefits of 3DSP requires optimizing the sand printing process parameters to obtain ideal casting molds. This study explores a hybrid modeling framework that merges physics-based simulations with data-driven predictive analytics and artificial intelligence (AI) techniques to optimize the 3DSP-metal casting process, focusing on quality improvement, defect minimization, and determining optimal parameters.MAGMA simulation software is used to take input 3DSP model casting parameters (e.g., binder composition, sand properties, pouring temperature, and cooling rates) and provides insights into solidification, thermal profiles, and the risk of defects, such as porosity and shrinkage cavities, occurring in the casting. Building on the simulation data, machine learning (ML) models (e.g., random forest, support vector machine (SVM), and neural networks) will be introduced to provide a predictive defect analysis. These ML models were trained on 3DSP process parameters and corresponding simulation data to predict potential casting defects, enabling proactive mitigation strategies. Next, genetic algorithms (GA) will be applied to explore the relationships learned by the machine learning models to find the optimal process parameters. The integrating of physics simulations, ML and GA creates an adaptive optimization loop that continuously refines the process, leading to improved efficiency and cost-effectiveness, thus supporting sustainability goals and advancing smart manufacturing.