EBSD Microstructure Analysis Using Neural Networks for Faster Mechanics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional micromechanical simulations of material microstructures are complex and time-consuming, making it inefficient to evaluate mechanical properties of workpieces.
Innovation Solution
A neural network is trained to process Electron Backscatter Diffraction (EBSD) images to generate micromechanical response arrays, providing quicker and more efficient evaluation of mechanical properties by replacing traditional simulations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional Finite-Element-Method simulation is used to analyze micromechanical response, then measurement precision of mechanical properties is improved, but processing time and computational complexity increase significantly
Solution Approach 1:
The patent creates a digital copy of the physical microstructure through EBSD imaging, then uses this digital representation as input for both conventional FEM simulation and neural network analysis. This copying enables parallel processing approaches where the neural network can quickly predict mechanical properties from the microstructure image without requiring time-consuming physical simulation for each analysis case.
Solution Approach 2:
The patent replaces the conventional mechanical FEM simulation system with a neural network-based prediction system. The neural network learns the relationship between microstructure morphology and mechanical response through training on FEM data, then uses this learned model to predict mechanical properties directly from EBSD images, substituting the computationally intensive mechanical simulation with a faster data-driven approach.
2Productivity
If neural network is used to replace conventional simulation, then processing time is reduced, but reliability of prediction may be compromised
Solution Approach 1:
The patent implements a feedback mechanism where the neural network's prediction uncertainty is evaluated and fed back into the analysis process. When uncertainty exceeds a threshold, the system automatically triggers conventional FEM simulation to provide ground truth data for retraining or validation, ensuring that reliability requirements are met while maintaining high productivity for confident predictions.
Solution Approach 2:
The patent changes the operational parameters of the analysis system by introducing uncertainty quantification as a key output parameter from the neural network. This allows the system to adapt its behavior based on prediction confidence levels, using faster neural network analysis for certain predictions while falling back to more reliable FEM simulation for uncertain cases, thus balancing productivity and reliability.
Data Source
Figure 1
Figure 2
AI summary
The present invention relates to a computer-implemented method for evaluating mechanical properties of a material particularly of a workpiece, comprising the steps of: - Providing an artificial neural network (3) being trained to associate an EBSD image with one or more micromechanical response arrays; - Providing a sample EBSD image of a material to be probed for a micromechanical response; - Applying the EBSD image to the neural network (3) to obtain as network output one or more corresponding micromechanical response arrays and/or a prediction uncertainty of the neural network (3); - Determining mechanical properties of the material depending on the network output.