Defect Detection on Rotational Components via Trajectory Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional defect analysis methods for rotating components, such as blades in aircraft engines or wind turbines, face challenges in accurately counting and localizing defects due to varying appearances and lack of distinct features, as well as issues with object tracking under illumination changes.
Innovation Solution
A method and system that utilize trajectory fitting and clustering of defects based on fitted ellipses and rotation axes to identify distinct defects, recover missing defects, and generate modified images with accurate defect counts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional image fusion methods are used to integrate images from multiple views, then a more enhanced representation is provided, but defects may not be distinguishable and the same defect may present different appearances in different views
Solution Approach 1:
The patent segments the defect detection process into distinct stages: trajectory extraction, clustering based on rotation axes, and defect identification. By dividing the complex image fusion task into manageable segments, the system can handle the variability of defect appearances across different views more effectively, resolving the contradiction between detection accuracy and processing complexity.
Solution Approach 2:
The patent introduces a temporal dimension by tracking defect trajectories across multiple video frames. Instead of merely fusing spatial images from different views, the system adds the time dimension to observe defect movement patterns, which helps distinguish between the same defect viewed from different angles and actually different defects, thereby improving measurement precision without excessive complexity increase.
2Reliability
If object tracking methods are used to continuously track defects, then defect trajectories can be obtained, but defects easily lose track due to illumination change
Solution Approach 1:
The patent implements feedback mechanisms by continuously monitoring defect trajectories and using rotation axis information to correct tracking drift. When illumination changes cause tracking errors, the system uses the geometric constraints from rotation axes to realign and correct the trajectories, thereby maintaining reliable tracking despite harmful illumination variations.
Solution Approach 2:
The patent changes the tracking parameters from simple pixel-based tracking to trajectory-based tracking constrained by rotation axis geometry. By transforming the tracking problem into a geometric parameter space defined by rotation axes, the system becomes more robust to illumination changes, as the geometric constraints remain valid regardless of lighting conditions.
3Productivity
If conventional defect analysis methods are used, then defect detection can be performed, but counting distinct defects and finding defect locations becomes challenging
Solution Approach 1:
The patent performs preliminary action by extracting trajectories and identifying rotation axes before final defect counting. This preliminary processing organizes the data in advance, grouping potential defects by their trajectory patterns and rotation axis assignments, which simplifies the subsequent counting process and improves both speed and accuracy of distinct defect identification.
Data Source
AI summary
This disclosure relates to method and system for inspection of rotational components based on video frames of the rotational components in motion. Based on the property of projective single axis motion, motion trajectories of defects on the rotational components are modeled in a conic pattern. The defects are clustered according to their spatial information around a rotation axis of the rotational components to ascertain distinct defects. Undetected defects may be recovered by rotating their previous locations along their trajectory conics by a predetermined rotation angle. Based on the distinct defects, including the recovered defect, a modified video frame which includes an identification of the distinct defects, including the recovered defect, may be generated.