Search

This study uses a multilayer perceptron deep learning model to predict creep strain vs time curves in composite materials based on their microstructural features. Finite element simulations generate ground truth data for model training and validation. The multilayer perceptron model, trained on this comprehensive dataset, effectively captures the complex relationships between the microstructure and creep behavior, achieving high accuracy. Comparative analysis with traditional models shows the multilayer perceptron model’s superior performance. This demonstrates the model’s potential for reliable application in various engineering fields, offering improved predictions of composite material behavior under creep conditions.

Predicting the microstructure of composite plates based on deformation data under static loads is crucial for advanced materials design and optimization. This study utilizes finite element simulations to generate deformation data, capturing the complex mechanical behavior of composite materials under static loading conditions. We developed a deep learning model based on a multi-layer perceptron (MLP) architecture to predict the microstructure of these composite plates from the simulated deformation data. The model is trained on a dataset comprising diverse microstructural patterns and their corresponding deformation responses. Our results demonstrate the MLP’s capability to accurately infer microstructural details, highlighting the potential of deep learning in materials science. This approach not only enhances the understanding of the relationship between deformation and microstructure but also provides a robust framework for designing composite materials with desired properties through computational methods.

This paper presents an innovative approach for multiple particle tracking within complex systems, utilizing convolutional neural networks in conjunction with multi-output models. Accurate particle tracking is a critical prerequisite for unraveling the dynamic behaviors of particles in a myriad of research domains, encompassing colloidal particles, biological cells, and molecular dynamics. Different from conventional methodologies, our approach combines data preprocessing, multilayer perceptron model training, and multi-output model integration to yield precise and efficient particle tracking results. The significance of this research lies in the adaptability and versatility of the trained models, which are designed to surmount challenges, including crowded and noisy environments. This work represents a substantial step forward in particle tracking methodologies, providing a robust and efficient alternative to conventional methods, promising more profound investigations into particle dynamics within complex systems, and contributing to a deeper understanding of the microscale world.

Performing time-dependent finite element simulations for wave propagation in composites is a particularly complex task that consumes a lot of computational energy as it involves modeling the interactions between waves and various constituents that make up the composite material. In this study, we have developed a surrogate model of elastic wave propagation in composites based on three-dimensional conventional neural networks. The input to the model consists of a three-dimensional matrix representing the architecture of the composites and a vector representing the input waves, while the output is a vector representing the output elastic waves. After training the model using 60 000 randomly generated samples, it has shown high accuracy and efficiency in predicting the output elastic waves. This significantly reduces computational resources required to conduct simulation using commercial software, making it a more practical solution for real-world applications, such as composite optimization, nondestructive testing, and material characterization.