Neural Network Material Model for Simulation Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing numerical simulations of mechanical bodies face challenges in accurately describing the behavior of complex materials, particularly near the failure point, due to limitations in representing material behavior for lightweight construction materials.

Innovation Solution

The method employs an example-based model, such as a neural network, to simulate the behavior of complex materials by focusing on global stresses and deformations of a test body, eliminating the need to consider internal structures, and integrates numerical simulations for training, allowing for precise and reliable simulation results without requiring analytical approaches.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional measurement methods are used to determine local stresses and local deformations directly on complex materials, then measurement precision is improved, but device complexity and measurement difficulty increase significantly

Engineering Contradiction:
Improvemeasurement precisionVSAvoiddevice complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts the measurement task from the complex material itself and transfers it to a simplified test body. Instead of measuring local stresses in the complex material directly, the method measures global stresses in a test body that has been simplified to have uniform material properties, thereby reducing measurement complexity while maintaining precision through the example-based model.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces an intermediary - the test body with uniform material properties - that mediates between the complex material and the measurement system. This intermediary test body can be easily measured using conventional methods, and its measurement results are then used to train example-based models that represent the complex material behavior, thus avoiding direct complex measurements.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Ease of operation

If analytical approaches are specified for material behavior, then ease of operation is improved, but adaptability deteriorates because analytical approaches cannot handle any material type

Engineering Contradiction:
Improveease of operationVSAvoidadaptability
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent changes the approach from specifying analytical formulas to using example-based models that are trained on measurement data. This parameter change allows the model to adapt to any material type by learning from actual measurements rather than relying on pre-defined analytical forms, thereby improving adaptability while maintaining ease of operation through automated training processes.

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent creates a universal example-based model that can represent any material behavior by training on measurement data from test bodies. This universal model replaces material-specific analytical approaches, allowing the same framework to handle diverse material types (metals, polymers, composites) without requiring separate analytical formulations for each material.

Inventive Principle:
Principle #6Universality (Multi-functionality)

3Reliability

If physical testing of complex materials is performed to obtain measurement data, then reliability of material behavior description is improved, but loss of time and loss of substance increase

Engineering Contradiction:
ImprovereliabilityVSAvoidloss of time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent creates a simplified copy - the test body with uniform material properties - that replicates the essential behavior of the complex material. By measuring this simplified copy and using the results to train example-based models, the method obtains reliable material behavior descriptions without requiring extensive physical testing of the actual complex materials, thus reducing time and substance loss.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent performs preliminary measurements on simplified test bodies before attempting to characterize complex materials. These preliminary measurements on the test bodies provide the training data needed to develop example-based models that can then be applied to complex materials, avoiding the need for time-consuming direct testing of the complex materials themselves.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP2212814B1Method for making a description of complex materials by means of neural networks or related techniques
Publication Date: 2018.03.14 ANDATA ENTWICKLUNGSTECH
  • EP2212814B1 patent drawingFigure 1~2

AI summary

Disclosed is a method for making a description of complex materials for use in numeric simulations of mechanical elements, an example-based model (1) being used for the description. In order to increase the accuracy of the simulation results, the complex material is integrated into a test block (9), predefined values for the global stress and the global movement of the test block (9) are determined on the test block (9), a numeric model (2) of the test block (2) is generated for the complex material by taking into account the example-based model (1), and the example-based model (1) is trained, based on the results of numeric simulations of the test block (9), to determine the correlation between the global movements and the global stress.