Morphed Surface Repair Modeling for Defect-Deformation Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for defect removal in manufactured objects, such as those produced by additive or subtractive manufacturing, often require time-consuming, part-variant-specific processes and struggle to distinguish between manufacturing deformations and actual defects, leading to inefficient and costly defect removal.
Innovation Solution
A computer-aided manufacturing program employs an image-to-image translation based machine learning algorithm, trained using pairs of input images representing nominal and deformed surfaces with defects, to generate a 3D model that differentiates between deformations and defects, allowing for efficient removal of defects without excess material loss.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing smoothing methods are used to remove defects from manufactured objects, then defect removal can be achieved, but the process is time-consuming and requires part-variant-specific procedures
Solution Approach 1:
The patent uses a learned model (copy) trained on training data representing various part variants to automatically identify and remove defects. Instead of requiring manual scanning and processing for each new part variant, the system creates a universal defect removal model that can be applied across multiple variants, significantly reducing the time and effort required for defect removal while maintaining high productivity
Solution Approach 2:
The system performs preliminary training using training images and training masks that represent various part variants and their defects. This preliminary action creates a pre-trained model that can immediately be applied to new parts without requiring time-consuming re-scanning and re-processing for each variant, thereby improving defect removal efficiency while minimizing time loss
2Manufacturing precision
If existing smoothing methods are used, then defects can be removed, but it is difficult to distinguish between manufacturing deformations and actual defects
Solution Approach 1:
The patent introduces training masks as an intermediary element that explicitly marks the location and extent of actual defects separate from manufacturing deformations. The learned model uses these masks during training to understand the distinction between normal variations and true defects, enabling accurate identification and removal of only actual defects while preserving manufacturing deformations
Solution Approach 2:
The system uses training data consisting of pairs of images and corresponding masks that provide feedback during the learning process. This feedback mechanism allows the model to continuously improve its ability to distinguish between deformations and defects by learning from labeled examples, thereby enhancing both defect identification accuracy and measurement precision
3Reliability
If customized defect removal solutions are used for each part variant, then specific defects can be addressed, but the process becomes complex and costly
Solution Approach 1:
The patent creates a universal learned model that can handle multiple part variants simultaneously. The training data includes images and masks from various part variants, enabling the model to generalize across different geometries and manufacturing processes. This universal approach maintains reliable defect removal effectiveness while eliminating the need for complex customized solutions for each variant, thereby reducing overall process complexity
Solution Approach 2:
The system changes the parameters of the defect removal process by using a data-driven learned model instead of traditional geometry-based methods. The model learns optimal defect identification and removal parameters from training data, allowing it to adapt to different part variants automatically. This parameter transformation simplifies the process while maintaining reliability across diverse manufacturing scenarios
Data Source
AI summary
Methods, systems, and apparatus, including medium-encoded computer program products, for computer aided repair of physical structures include: generating a two dimensional difference image from a first three dimensional model of at least one actual three dimensional surface of a manufactured object, and a second three dimensional model of at least one source three dimensional surface used as input to a manufacturing process that generated the manufactured object; obtaining from an image-to-image translation based machine learning algorithm, trained using pairs of input images representing deformed and deformed plus surface defected added versions of a nominal three dimensional surface, a translated version of the two dimensional image; generating from the translated version of the two dimensional image a third three dimensional model of at least one morphed three dimensional surface corresponding to the at least one source three dimensional surface. Further, defects can be removed based on the third three dimensional model.