3D Printing Inspection Using Correlated Image Subareas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing quality inspection methods for additive manufacturing, such as 3D printing, struggle to accurately detect defects, particularly in thin or delicate structures, due to limited spatial resolution of powder bed cameras and reliance on single image analysis, leading to false error detections and missed flaws.
Innovation Solution
A method involving virtual splitting of digital snapshots into subareas, followed by correlation and aggregation of these subareas using machine-learning techniques, to identify defects by comparing pre- and post-coating images, enhancing detection accuracy by accounting for relationships between adjacent areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If single image analysis is used for defect detection, then the inspection process is simple and fast, but the detection accuracy is insufficient leading to false error detections and missed flaws
Solution Approach 1:
The patent divides the inspection process into multiple stages: capturing images at different process stages (before coating, after coating, after sintering), splitting each image into subareas for detailed analysis, and performing correlation analysis across multiple images. This segmentation approach enables comprehensive defect detection while maintaining operational efficiency through automated multi-stage processing.
2Measurement precision
If powder bed camera spatial resolution is increased to detect thin structures, then small defects become visible, but the device complexity and cost increase
Solution Approach 1:
Instead of improving spatial resolution in a single dimension, the patent adds the time dimension by capturing images at multiple process stages. This temporal dimension allows detection of defects that may not be visible in single images, such as powder dispersion issues before coating or sintering defects after heating, effectively compensating for limited spatial resolution without increasing camera complexity.
3Adaptability or versatility
If rendering is used to provide abstract representation for defect detection, then comparison with CAD model is enabled, but information about grayscale patterns in real images is lost
Solution Approach 1:
The patent merges multiple image types and analysis methods: combining real images with rendering images for comparison, merging correlation analysis results from multiple image pairs, and integrating subarea-based analysis with overall image assessment. This merging approach preserves grayscale pattern information from real images while incorporating the structural comparison capabilities of rendering, achieving both versatility and information retention.
Data Source
AI summary
A quality inspection method and a quality inspection arrangement for 3D printing is provided.