OVERVIEW OF METHODS FOR APPLYING ARTIFICIAL INTELLIGENCE IN THE DESIGN OF FOUNDRY TECHNOLOGIES AND METAMATERIALS

Abstract

Artificial Intelligence (AI) is emerging as a transformative catalyst, unlocking new frontiers for optimizing and enhancing traditional industries like foundry production, while also fostering the development of innovative fields such as metamaterial design. This review analyzes key AI methods and applications in these domains, highlighting its potential to elevate product quality, reduce costs, and accelerate innovation cycles. In foundry production, AI ensures significant improvements across all stages of the product lifecycle. Machine learning and deep learning systems enable the prediction of casting defects (such as porosity, shrinkage, cracks) based on the analysis of historical data and process parameters. This facilitates prompt correction of manufacturing conditions and prevention of product rejection. AI holds substantial potential for optimizing casting process parameters and improving casting quality. The application of computer vision and deep learning is transforming automated quality control, enabling rapid, real-time defect detection. Integrating AI with 3D printing allows for optimized design of casting technologies, molds, and patterns, creating digital twins that accelerate development and testing phases. The article provides an industrial application example from Sarginsons Industries, which leveraged AI to almost halve the weight of automotive castings. AI also serves as a tool for optimizing the Life Cycle Assessment (LCA) of castings, contributing to resource conservation and circular economy principles. In the realm of metamaterial design, AI is revolutionizing the traditional approach by shifting towards inverse design. Generative models, such as GANs and VAEs, enable the creation of entirely new topologies with unique properties that fulfill desired functional characteristics. The methodology for designing metamaterials based on graph neural networks is presented, allowing for rapid generation of structures with specified properties and their optimization through iterative learning. Specifically, the potential of spherene structures for developing functionally optimized casting patterns for the Lost Foam Casting (LFC) process is explored, which will facilitate controlled thermal decomposition and efficient gas removal, thereby improving the quality of metal castings. Convolutional Neural Networks (CNNs) are effectively utilized for predicting metamaterial properties, and the integration of AI with topological optimization algorithms allows for finding their optimal structures. A separate aspect addresses the role of AI in invention. AI systems are already capable of generating novel and original solutions that may be deemed patentable, as demonstrated by the precedent with the DABUS system. This opens new stages in the development of intellectual property, where AI acts not merely as an assistant but as a full participant in the process. Despite significant advantages, the implementation of AI faces challenges such as the need for large volumes of high-quality data, interpretation of “black box” models, and substantial computational resources. Nevertheless, the prospects are  extremely broad, including the development of hybrid models (physics-informed AI), the creation of digital twins, and the full integration of AI into the concept of “smart manufacturing” (Industry 4.0). Thus, AI is not merely an automation tool but a driver of scientific and technological transformation, enhancing human intelligence, allowing focus on creativity, strategic decisions, the creation of more advanced products, and sustainable development.