3D Shape-Based Defect Prediction for Casting Quality Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect prediction systems in industries like the automobile sector cannot accurately predict defects in products based on the shape of the product, limiting their ability to identify issues such as seizure, shrinkage, and galling in castings.
Innovation Solution
A prediction system utilizing pre-trained models that integrate defect characteristic values, three-dimensional shape features, and manufacturing conditions to predict defects in target products, including castings, by training on existing product data and applying it to new products, with the capability to display defect locations and degrees.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If defect prediction is based only on manufacturing conditions, then the prediction system is simple to implement, but it cannot predict defects due to product shape
Solution Approach 1:
The patent combines multiple input factors (manufacturing conditions, 3D shape features, and material properties) into a unified prediction model. The neural network integrates these diverse inputs to comprehensively predict defects, thereby improving prediction reliability while managing system complexity through structured data fusion.
Solution Approach 2:
The patent introduces three-dimensional shape features as an additional dimension to the traditional manufacturing condition-based prediction. By extracting and incorporating 3D geometric characteristics (volume, surface area, thickness distribution), the system transitions from 2D manufacturing parameters to multi-dimensional analysis, enabling shape-related defect prediction.
2Reliability
If 3D shape features are extracted and integrated into prediction, then shape-related defects can be predicted, but the processing complexity increases
Solution Approach 1:
The patent performs preliminary extraction of 3D shape features (volume, surface area, thickness) before the actual defect prediction process. By pre-processing and structuring the geometric data in advance, the system reduces the complexity of real-time processing while maintaining comprehensive shape analysis capability for predicting defects.
3Adaptability or versatility
If multiple defect types are predicted simultaneously, then comprehensive quality control is achieved, but the measurement precision requirement increases
Solution Approach 1:
The patent applies local quality by predicting different defect types (shrinkage, porosity, cracks, warpage) at specific locations and conditions. The neural network outputs probability distributions for each defect type based on local 3D shape features and manufacturing conditions, allowing precise characterization of specific defects while maintaining comprehensive coverage.
Data Source
AI summary
A prediction system configured to predict a defect of a target product includes a first pre-trained model trained based on a defect characteristic value indicating a defect associated with a location in an existing product, a feature of a three-dimensional shape of the existing product, and conditional information indicating a manufacturing condition of the existing product. The first pre-trained model is configured to, when a feature of a three-dimensional shape of the target product is input, output a defect characteristic value indicating a defect associated with a location in the target product.