Deep Learning Model Predicts Stress Fields from Microstructure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional manufacturing methods and AI methods face challenges in designing composite materials with superior mechanical properties due to difficulties in conjunction of disparate materials and predicting physical fields like strain or stress tensors, limiting the optimization of composite designs.
Innovation Solution
A deep learning-based approach using a game-theoretic Generative Adversarial Network (GAN) that translates material geometry and microstructure into strain or stress fields, enabling direct prediction of physical fields and improving the efficiency of evaluating physical properties of hierarchical materials.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional manufacturing methods are used to create composite materials, then material combinations can be achieved, but the ability to optimize spatial distributions and manipulate complex microstructures is limited
Solution Approach 1:
The patent changes the manufacturing approach from traditional physical assembly to additive manufacturing with digital control, enabling precise manipulation of spatial distributions and microstructures through software parameters rather than physical constraints
2Measurement precision
If multiscale modeling approaches like FEM or MD simulations are used to calculate physical properties, then detailed physical field information can be obtained, but the computational cost and time required increase significantly
Solution Approach 1:
The patent creates a digital twin or surrogate model through neural network training that copies the input-output relationships of expensive FEM simulations, enabling rapid prediction of physical fields without repeating full computational simulations
Solution Approach 2:
The patent performs preliminary training using a limited dataset of FEM simulations to build the neural network model, which then handles all subsequent predictions without requiring additional FEM computations
3Productivity
If AI methods optimize composite designs based on FEM calculations, then design optimization can be achieved, but direct prediction of physical fields and connection to microstructure is lost
Solution Approach 1:
The patent creates a multi-functional neural network that simultaneously performs design optimization and physical field prediction by incorporating multiple output branches, eliminating the need to choose between optimization efficiency and physical insight
4Reliability
If the design space of composites is explored exhaustively, then optimal designs can be found, but the intractable number of combinations makes this approach impossible
Solution Approach 1:
The patent focuses the search on locally optimal microstructural configurations rather than exhaustively searching all possible global combinations, using the neural network to evaluate only promising designs generated through targeted optimization algorithms
Data Source
AI summary
Materials-by-design is a new paradigm to develop novel high-performance materials. However, finding materials with superior properties is often computationally or experimentally intractable because of the astronomical number of combinations in design spaces. The disclosure is a novel AI-based approach, implemented in a game-theory based generative adversarial neural network (GAN), to bridge the gap between the physical performance and design space. A end-to-end deep learning model predicts physical fields like stress or strain directly from the material geometry and microstructure. The model reaches an astonishing accuracy not only for predicted field data but also for secondary predictions, such as average residual stress at R2˜0.96). Furthermore, the proposed approach offers extensibility by predicting complex materials behavior regardless of shapes, boundary conditions and geometrical hierarchy. The deep learning model demonstrates not only the robustness of predicting multi-physical fields, scalability, and extensibility. The disclosure may alter physical modeling and simulations by incorporating material geometry and boundary conditions into a graphical representation, and vastly improves the efficiency of evaluating physical properties of hierarchical materials directly from the geometry of its structural makeup.