Annular Disk Frame Inspection Using Auxiliary-Line AI References
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing defect inspection methods for frame members on annular disks are prone to erroneous determinations due to noise in image data and the need for specified values, which can be inappropriate for different types of friction plates, leading to inaccurate positional deviation assessments.
Innovation Solution
An inspection method using machine learning-based determination models that utilize training data with auxiliary lines to detect positional deviations in frame members, eliminating the need for specified values and reducing noise-related errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image processing is performed on image data to calculate values such as area, average gray value, and arc distance, then positional deviation of frame members can be detected, but specified values must be set in advance which may not align with the inspection target causing erroneous determination
Solution Approach 1:
The patent applies preliminary action by generating training data in advance that includes auxiliary lines drawn at predetermined positions. These auxiliary lines are incorporated into training images before machine learning processing, allowing the determination model to learn the correct reference positions beforehand. This eliminates the need to set specified values for each inspection target, as the model has already learned the appropriate references during training.
Solution Approach 2:
The patent uses copying by creating training images that replicate the inspection scenario including auxiliary lines. The training data copies the structure and reference elements (auxiliary lines) that will be present during actual inspection, allowing the machine learning model to learn from these copied examples and generalize to new inspection targets without requiring manual specification of comparison values.
2Measurement precision
If image processing is performed on image data to determine positional deviation, then detection capability is improved, but noise in image data adversely affects the comparison between calculation values and specified values causing erroneous determination
Solution Approach 1:
The patent introduces an intermediary element - the auxiliary lines - that serves as a stable reference framework between the image data and the determination model. These auxiliary lines are drawn at predetermined positions and incorporated into both training and inspection images, providing a consistent reference that is independent of noise in the original image data. The machine learning model learns to use these auxiliary lines as intermediaries for accurate positional deviation measurement.
Solution Approach 2:
The patent replaces the mechanical system of manual specified value setting and direct calculation comparison with a machine learning-based system. Instead of calculating values and comparing them against pre-set specified values, the system uses a determination model trained with auxiliary lines that automatically performs the comparison, substituting the manual mechanical process with an intelligent system that is more robust to noise.
3Ease of operation
If specified values are set in advance for comparison with calculation values, then positional deviation can be determined, but the specified values may not correspond appropriately to the inspection target leading to erroneous determination
Solution Approach 1:
The patent applies self-service by enabling the determination model to automatically adapt to different inspection targets through machine learning. Instead of requiring manual setting of specified values for each target, the system uses training data with auxiliary lines to create a model that self-adjusts to the appropriate references. The model serves itself by learning the correct comparison criteria from training examples, eliminating the need for operator intervention in setting specified values.
Data Source
AI summary
An inspection image generation unit 12 configured to perform image processing on a captured image of an annular disk, and draws an auxiliary line in the vicinity of a position where a plurality of frame members exist to generate an inspection image equipped with auxiliary lines, and a defect determination unit 13 configured to apply the generated inspection image equipped with auxiliary lines to a trained determination model to detect a positional deviation of the frame members are provided. According to this, it is not necessary to set a specified value for comparison with various calculation values calculated for the frame members through image processing, and even in a case where a noise is included in the captured image, it is possible to reduce the possibility of occurrence of erroneous determination by an adverse effect on the comparison between the calculation values and the specified value due to the noise. In addition, it is easy to perform learning and determination of the positional deviation occurring in the frame members by using the auxiliary line as a reference.