ANALYSIS CONDITIONS IN COMPUTATIONAL HOMOGENIZATION OF HETEROGENEOUS MATERIALS WITH RANDOM MICROSTRUCTURE USING A

Computational homogenization is a powerful technique used to analyse the behaviour of heterogeneous materials with random microstructures. However, traditional approaches for determining the effective properties of such materials can be computationally expensive and time-consuming. In recent years, artificial intelligence (AI) has emerged as a promising tool to accelerate the homogenization process by learning the complex relationships between the microstructure and macroscopic properties. We investigate the application of AI techniques in the analysis of conditions for computational homogenization of heterogeneous materials with random microstructures. We explore the use of machine learning algorithms, such as deep neural networks, to predict the effective properties of materials based on their microstructural characteristics. By training the AI models on a dataset of representative microstructures and their corresponding effective properties, we aim to develop accurate and efficient prediction models. To ensure the reliability of the AI-based homogenization process, we carefully analyse the conditions required for effective learning and prediction. This includes studying the influence of training dataset size, data representation, and model architecture on the accuracy and generalization capabilities of the AI models. We also investigate the effect of different microstructural descriptors and feature extraction techniques on the prediction performance. we explore the transferability of the trained AI models across different material systems and microstructural variations. We analyse the robustness of the models by testing their performance on unseen microstructures and evaluating their ability to capture the underlying physics governing the effective behaviour of the materials. we aim to establish a framework for the efficient and accurate computational homogenization of heterogeneous materials with random microstructures using AI. The findings can significantly impact the design and optimization of materials in various engineering applications, where the understanding of effective material properties is crucial for performance predictions and reliability assessments.