Metal Sheet Reverse Loading Prediction from Tensile Test Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the model constants of material models for predicting tension-compression reverse loading behavior in metal sheets require specialized testing machines and jigs, making them costly and time-consuming, especially for high-tensile steel sheets where buckling deformation is a significant issue.
Innovation Solution
A method using a neural network to predict model constants based on uniaxial tensile test data, including point sequence data from stress-strain curves, which allows for the determination of model constants without the need for a tension-compression test, thereby simplifying the process and reducing costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a tension-compression test is performed to determine model constants for predicting reverse loading behavior, then the accuracy of press forming FEM analysis is improved, but the test requires specialized equipment and jigs which increases cost and time
Solution Approach 1:
The invention creates a virtual copy of the tension-compression test through machine learning. A neural network is trained using data from standard uniaxial tensile tests to predict the model constants that would otherwise require actual tension-compression testing. This virtual replication eliminates the need for specialized equipment while maintaining prediction accuracy
Solution Approach 2:
The invention replaces the mechanical testing system with a computational system. Instead of physically applying tension and compression loads to measure material behavior, the system uses machine learning algorithms to calculate the reverse loading behavior from uniaxial tensile test data alone, substituting mechanical experimentation with computational prediction
2Measurement precision
If a tension-compression test is conducted on high-tensile steel sheets to evaluate bauschinger effect, then material model constants are accurately determined, but buckling deformation occurs during compression making the test difficult to conduct
Solution Approach 1:
The invention extracts only the necessary information from a simplified test. Instead of performing the complete tension-compression cycle that causes buckling, the method extracts model constants by training a neural network on data from uniaxial tensile tests only, removing the problematic compression phase entirely while still obtaining the needed material parameters
Solution Approach 2:
The invention performs preliminary training with comprehensive data from multiple high-tensile steel sheets before actual prediction. The neural network is pre-trained on a dataset containing uniaxial tensile test results and corresponding tension-compression test outcomes from various steel sheets, enabling it to predict model constants for new materials without performing actual compression tests
3Measurement precision
If thin sheet tension-compression tests are performed using buckling prevention jigs, then test accuracy is improved, but financial and time costs increase due to specialized equipment requirements
Solution Approach 1:
The invention makes the uniaxial tensile testing machine universal by enabling it to provide data for determining model constants that traditionally required separate tension-compression testing equipment. The neural network processes uniaxial tensile test data to predict reverse loading behavior, allowing one piece of equipment to serve multiple material characterization functions
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables accurate prediction of tension-compression reverse loading behavior without the need for specialized testing equipment, improving analysis accuracy and reducing financial and temporal costs, while maintaining sufficient regression accuracy for press forming FEM analysis.
Implementation Method 1
a value of the model constant of a prediction metal sheet is acquired by causing a neural network to perform machine learning
Implementation Method 2
plastic deformation has been performed
Implementation Method 3
The early yielding phenomenon during reverse loading as described above is called a bauschinger effect
Implementation Method 4
a material may undergo buckling deformation during compression
Data Source
AI summary
A method of predicting a tension-compression reverse loading behavior predicted by determining a model constant of a material model expressing the tension-compression reverse loading behavior of a metal sheet includes acquiring a value of the model constant of a prediction metal sheet by inputting metal materials test data, including a factor related to a uniaxial tension behavior, of the prediction metal sheet to a learned model that has been caused to perform machine learning using, as an input variable, metal materials test data of a learning metal sheet and using, as an output variable, a value of the model constant, which has been determined based on a tension-compression test of the learning metal sheet. The factor related to a uniaxial tension behavior includes point sequence data obtained by discretizing a stress-strain curve of uniaxial tensile test obtained from a uniaxial tensile test.