Welding Learning Model for Stable Molten Pool and Arc Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing welding information learning models lack stability in extracting accurate welding information due to varying welding conditions, leading to potential erroneous processes when the learned and actual welding conditions differ.
Innovation Solution
A welding-information learning-model generation method that uses image data from a visual sensor to generate a learning model capable of extracting accurate welding information, considering multiple settings of welding conditions such as electrode extension, welding current, and arc voltage, by incorporating information about the molten pool, welding wire, and arc behavior.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a learning model is trained with welding information extracted from image data captured under specific welding conditions, then the learning model can accurately extract welding information under those specific conditions, but the accuracy deteriorates when the actual welding conditions differ from the training conditions
Solution Approach 1:
The patent applies parameter changes by systematically varying multiple welding condition parameters (electrode extension, welding current, arc voltage, welding speed, weaving conditions, welding position) during the data collection phase. This creates a comprehensive training dataset that covers the range of actual welding conditions, enabling the learning model to maintain high extraction accuracy across different operational parameters rather than being limited to a single fixed condition set
2Reliability
If image data is captured under a single welding condition setting, then the data collection process is simple and quick, but the learning model cannot generalize to different welding conditions
Solution Approach 1:
The patent implements preliminary action by pre-capturing image data under multiple welding condition settings before actual welding operations begin. This advance data collection and model generation ensures the learning model is ready to handle various welding conditions without causing delays during production, as the generalization capability is established beforehand through comprehensive training data
Solution Approach 2:
The patent achieves universality by creating a learning model that functions accurately across multiple welding conditions rather than being specialized for a single condition. The model is designed to extract welding information universally applicable to different electrode extensions, currents, voltages, and positions, making it adaptable to various welding scenarios without requiring separate models for each condition
Data Source
AI summary
This learning model generation method is for generating a learning model for learning by taking, as teaching data, image data from a visual sensor and welding information extracted from the image data, wherein: a plurality of welding conditions to be used for generating the learning model correspond to the difference in setting pertaining to at least one item; the image data corresponding to each of the plurality of welding conditions includes at least one of a molten pool, a welding wire, and an arc; the welding information includes at least one of information pertaining to the behavior of the molten pool, information pertaining to the position of the welding wire, and information pertaining to the arc; and the learning model which receives the image data as an input and outputs the welding information is generated.