3D Deep Learning for Ultrasonic Defect Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional ultrasonic testing methods using one-dimensional signal data can only determine the position of defects with coordinate values, while two-dimensional image data can assess the shape but not the depth of defects, making it difficult to analyze three-dimensional defects effectively.
Innovation Solution
A device and method utilizing a three-dimensional deep learning model to analyze defects by collecting and stacking two-dimensional inspection images from ultrasonic testing, generating input data through data augmentation, and deriving representation data to predict the type of defect as a probability using a generation and detecting network within the deep learning model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If one-dimensional signal data is used for defect analysis, then the position of defects can be determined with coordinate values, but the shape of defects cannot be checked
Solution Approach 1:
The patent transitions from one-dimensional signal data to three-dimensional image data by stacking multiple two-dimensional inspection images along the depth direction. This dimensional transformation enables simultaneous acquisition of position, shape, and depth information, resolving the contradiction between position detection precision and shape information completeness.
2Loss of information
If two-dimensional image data is used for defect analysis, then the shape of defects can be checked, but the depth of defects cannot be determined
Solution Approach 1:
The patent adds the depth dimension by stacking two-dimensional inspection images along the Z-axis to create three-dimensional data. This enables depth measurement capability while preserving shape information from the two-dimensional images, thus resolving the contradiction between shape analysis and depth measurement.
3Loss of information
If three-dimensional deep learning model is used to analyze defects, then three-dimensional defect analysis is enabled, but computational resources and processing time increase
Solution Approach 1:
The patent performs preliminary actions by pre-processing and stacking two-dimensional inspection images into three-dimensional data structures before feeding them to the deep learning model. This preparation step organizes the data in advance, allowing the model to process three-dimensional defect information more efficiently and reducing overall processing time.
4Loss of information
If three-dimensional raw data is processed through deep learning, then comprehensive defect analysis is achieved, but data processing complexity increases
Solution Approach 1:
The patent segments the complex three-dimensional data processing into manageable steps: collecting two-dimensional images, stacking them into three-dimensional data, performing data augmentation, and feeding to the deep learning model. This segmentation reduces the perceived complexity by breaking down the overall process into discrete, manageable operations.
Data Source
AI summary
A device and method for analyzing a defect in ultrasonic testing using a three-dimensional deep learning model is proposed. The method includes preparing three-dimensional raw data by collecting a plurality of two-dimensional inspection images obtained by ultrasonic testing of an inspection object and stacking the plurality of two-dimensional inspection images; generating input data for a deep learning model by processing the three-dimensional raw data; deriving representation data, which is an inferenced three-dimensional image representing the defect of the inspection object, through a first feature transformation by applying first weights in a trained state to the input data within a generation network of the deep learning model; and determining a defect type of the inspection object by inputting the representation data to a detecting network of the deep learning model.