Deep Material Network for Multi-Scale Material Property Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a need for an improved prediction method for macroscopical physical properties of multi-scale materials, as existing methods are inefficient and lack accuracy in handling varying morphologies.
Innovation Solution
A computer-implemented method is provided for training a Deep Material Network (DMN)-based neural network to predict macroscopical physical properties of multi-scale materials. The method includes obtaining a dataset with tensors describing physical properties at both macroscopical and microscopical levels, along with morphological parameters. The neural network is trained to predict macroscopical physical properties based on these inputs, using a DMN architecture that integrates morphological parameters into the training stage.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional mechanical tests are used to determine macroscopical physical properties, then measurement accuracy is maintained, but time consumption and cost increase significantly
Solution Approach 1:
The patent creates a virtual copy of the physical material testing process through neural network simulation. Instead of performing actual mechanical tests on physical specimens, the system uses trained neural networks to predict macroscopical physical properties from microscopical tensors, thereby eliminating time-consuming physical experimentation while maintaining prediction accuracy.
Solution Approach 2:
The patent replaces the mechanical testing system with a computational prediction system. Traditional mechanical tests that require physical equipment, samples, and manual operation are substituted by a digital neural network model that processes microscopical material tensors to directly output macroscopical property predictions, dramatically reducing time and resource requirements.
2Productivity
If existing prediction methods are used for multi-scale materials, then computational speed is maintained, but prediction accuracy deteriorates due to inability to handle varying morphologies
Solution Approach 1:
The patent implements a dynamic prediction system where the neural network adapts to varying material morphologies through morphological parameters. Instead of using fixed prediction models, the system dynamically adjusts its prediction based on input morphological characteristics, allowing accurate predictions across different material configurations while maintaining computational efficiency.
Solution Approach 2:
The patent introduces morphological parameters as additional inputs to the neural network, enabling the model to account for variations in material structure and morphology. By changing the parameter space to include these descriptive features, the system achieves high prediction accuracy for multi-scale materials with diverse morphologies without sacrificing computational speed.
3Measurement precision
If detailed microscopical analysis is performed to improve prediction accuracy, then measurement precision improves, but computational cost increases
Solution Approach 1:
The patent performs preliminary analysis by pre-training neural networks on comprehensive microscopical data sets. The heavy computational work of learning complex microscopical-to-macroscopical relationships is done once during the training phase, storing the knowledge in the network weights. During actual prediction, only lightweight inference is required, dramatically reducing computational cost while maintaining high accuracy.
Solution Approach 2:
The patent introduces neural networks as an intermediary between microscopical material tensors and macroscopical property predictions. This intermediary model learns the complex relationships during training and then efficiently translates microscopical inputs to macroscopical outputs without requiring direct, computationally expensive physical measurements or simulations.
Data Source
AI summary
A method for training a Deep Material Network-based neural network configured to predict a macroscopical physical property of a multi-scale material. The multi-scale material comprises one or more components. The method includes obtaining a dataset, each entry of the dataset corresponding to a respective multi-scale material object. The entry includes a tensor describing the physical property of the object at a macroscopical level, one or more tensors each describing the physical property of a component of the object at a microscopical level, and one or more morphological parameters each describing a morphology of the object. The method further includes training, based on the dataset, the neural network to predict a tensor describing the physical property of a multi-scale material object at a macroscopical level based on the one or more tensors for the object and based on the one or more morphological parameters for the object.