AI Crack Probability Prediction for Deep-Drawn Components
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep-drawn components often experience cracks due to the limitations of material properties and manufacturability, making it challenging to predict and prevent crack formation during the development process.
Innovation Solution
An AI system that uses two AI programs to detect cracks in deep-drawn components, relate crack data to component geometry, and determine the probability of crack formation based on empirical data, providing accurate predictions for design and development.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Shape
If developers push the boundaries of shape complexity in deep-drawn components, then the desired shape complexity is improved, but cracking occurs due to material property limitations
Solution Approach 1:
The AI system performs preliminary analysis of crack formation probability during the design phase by evaluating geometric data against trained models. This allows developers to identify potential cracking risks before manufacturing, enabling preventive design modifications to avoid cracks while achieving desired shape complexity
Solution Approach 2:
The system provides feedback on crack formation probability based on component geometry analysis. This feedback loop allows developers to adjust design parameters iteratively, optimizing the balance between shape complexity and crack prevention by viewing probability assessments for different design configurations
2Measurement precision
If traditional methods relying on experience and simulations are used, then the development process is simpler, but the accuracy of crack prediction is insufficient
Solution Approach 1:
The patent replaces traditional mechanical/simulation-based assessment methods with an AI-based system that uses image processing and machine learning models. This substitution achieves higher prediction accuracy by learning from empirical crack data, despite the increased software and computational complexity
Solution Approach 2:
The system transforms the crack prediction approach by changing from simulation-based parameters to AI-model-based probability parameters. By training models on geometric and crack data, the system generates probability assessments that are more accurate than traditional simulations, accepting the complexity of model training and data processing
3Measurement precision
If more empirical data is collected and AI programs are trained, then the crack formation probability prediction accuracy is improved, but the data processing time and computational resources increase
Solution Approach 1:
The AI models are trained in advance using historical crack and geometric data before actual production use. This preliminary training phase allows the system to provide rapid predictions during design iterations without requiring real-time data processing, thus improving accuracy while minimizing time loss during actual design work
Solution Approach 2:
The system uses a two-stage approach: first, a first AI program performs selective crack detection in images; second, a second AI program uses the detected cracks and geometric data for probability assessment. This partial processing approach balances accuracy with computational efficiency by focusing resources on relevant data analysis
Data Source
Figure 1~2
AI summary
The invention relates to an AI system (10) and a method for providing a probability of crack formation of a deep-drawn component (12).The method comprises capturing (S10) images of a plurality of deep-drawn components (12) by means of an image acquisition device (16); determining (S12) cracks in the deep-drawn components (12) by means of a first AI program from the captured images, wherein at least one position of the respective crack is determined and stored as metadata; assigning (S14) the metadata to geometric data of the deep-drawn components (12); training (S16) a second AI program with the assigned metadata, wherein the trained second AI program determines the probability of crack formation as a function of component data of the deep-drawn component (12), wherein the component data includes at least the geometric data of the deep-drawn components (12); and providing (S18) the probability of crack formation dependent on the component data by the trained second AI program.