Laser Welding State Detection for Molten Shape Abnormalities
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining the processing state in laser welding, such as PTL 1, struggle to accurately determine the position, number, and size of molten shape abnormalities, particularly when foreign substances are present, as they primarily rely on threshold comparisons rather than detailed signal analysis.
Innovation Solution
A determination method and device that utilize an optical sensor to detect heat radiation, visible light, and reflected light during laser processing, generating signals that are used to calculate feature quantities, which are then input into a trained determination model to precisely identify the position, number, and size of molten shape abnormalities, even when foreign substances are involved.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If threshold comparison method is used to determine welding state, then the determination process is simple, but the measurement precision of molten shape abnormality is insufficient
Solution Approach 1:
The patent transforms the determination approach from simple threshold comparison to multi-parameter analysis by extracting multiple feature quantities (average intensity, standard deviation, maximum value, minimum value, gradient) from the light emission signal. This parameter transformation enables precise identification of molten shape abnormalities while maintaining systematic processing through the determination model.
Solution Approach 2:
The patent adds temporal dimension analysis by examining the time-series characteristics of light emission signals during laser processing. By analyzing how signal parameters change over time rather than using single-point threshold checks, the system achieves higher precision in detecting molten shape abnormalities without excessive complexity.
2Measurement precision
If detailed signal analysis is performed to identify molten shape abnormalities, then the measurement precision improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary action by pre-establishing a determination model through machine learning training before actual welding detection. The model is trained offline with labeled data containing various welding states, so during real-time processing, only feature extraction and model inference are needed, significantly reducing online processing time while maintaining high detection accuracy.
Solution Approach 2:
The patent segments the light emission signal into distinct time sections corresponding to different welding stages (heating phase, melting phase, cooling phase). By analyzing each segment separately and extracting specific features from relevant phases, the system achieves detailed defect detection without needing to process the entire signal at maximum complexity.
3Measurement precision
If machine learning model is used for determination, then the determination accuracy of processing state improves, but the device complexity increases
Solution Approach 1:
The patent uses copying by creating a virtual training dataset that simulates various welding conditions and defect scenarios. Instead of requiring extensive physical experiments to gather training data, the system generates synthetic data representing different molten shape abnormalities and welding states, which is then used to train the determination model. This reduces the complexity of data collection while maintaining model accuracy.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables detailed determination of the processing state by accurately identifying the position, number, and size of molten shape abnormalities, improving the assessment of welding quality and defect detection in lap welding processes.
Implementation Method 1
detecting, using an optical sensor, at least one of heat radiation light, visible light, and reflected light generated at a welded portion
Implementation Method 2
detecting, using an optical sensor, at least one of heat radiation light, visible light, and reflected light generated at a welded portion
Data Source
AI summary
A determination method for determining a processing state includes detecting, using an optical sensor, at least one of heat radiation light, visible light, and reflected light generated at a welded portion formed at a surface of a workpiece by emission of a laser beam on the workpiece, obtaining, from the optical sensor, a signal indicating a change in one of heat radiation, visible light, and reflected light in a time section corresponding to a welding time of each workpiece, determining, as the processing state, the position and number of molten shape abnormality in a welded region having a molten length and a molten width by inputting a feature quantity to a determination model that determines the processing state, the feature quantity including signal intensity of the signal, the molten shape abnormality occurring when a foreign substance exists at an overlapping surface of the workpiece, and outputting a determination result.