Advanced AI Techniques applied to structural integrity 

Filippo Berto and Marcello Laurenti

Artificial intelligence (AI) has become a transformative tool in materials science, offering innovative approaches to address challenges in modeling and understanding material behavior. This lesson explores the core mechanic and the application of advanced AI architectures, such as Multi-Layer Perceptrons (MLPs), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs) and Transformers, in analyzing and predicting the static and dynamic mechanical properties of additively manufactured materials. Key topics include how AI can model material anisotropy, address geometric imperfections, and predict performance under diverse testing conditions.

The lesson also covers the use of generative AI, such as AutoEncoders, to create realistic topologies for strut-like structures. These virtual models closely replicate real-world geometries, making them valuable for simulations and structural analyses.

Dimensionality reduction techniques, such as Principal Component Analysis (PCA), and ensemble learning methods, including bagging, are presented as strategies to improve model efficiency and accuracy. Additionally, the lesson examines the integration of unsupervised learning for data clustering and reinforcement learning for optimizing manufacturing processes.

Through this comprehensive overview, participants will gain insights into how AI techniques are reshaping the field of materials science, enabling more efficient analysis, better performance predictions, and innovative solutions to longstanding challenges.