Weld Surface Image Analysis for Spatter-Based Quality Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing welding process quality assessment methods are complex, costly, and require extensive equipment, making them impractical for industrial-scale production, especially for small-scale structures, due to the need for real-time dynamic analysis and complex software installations.
Innovation Solution
A method using an image-capturing device to acquire images of welded surfaces, applying digital processing and convolutional neural networks to classify spatters, determining parameters like area, density, and distance from the weld bead, allowing for flexible and simpler assessment of welding performance post-process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If real-time dynamic analysis of spatters is performed using high-speed cameras and complex software, then measurement precision and reliability of welding quality assessment is improved, but device complexity and cost increase significantly
Solution Approach 1:
The patent uses static images captured by simple cameras as copies of the welding spatter patterns, replacing the need for complex high-speed dynamic analysis systems. The image processing algorithms analyze these static copies to determine welding quality parameters, achieving accurate assessment without requiring expensive real-time dynamic equipment.
Solution Approach 2:
The patent replaces complex mechanical and electronic measurement systems (high-speed cameras, sensors) with a computational approach using standard cameras and image processing algorithms. The mechanical system of capturing and analyzing spatters is substituted with an optical-digital hybrid system that uses software-based image analysis instead of hardware-based dynamic measurement.
2Measurement precision
If complex equipment and software are installed on-site for real-time welding analysis, then measurement precision is improved, but ease of operation and adaptability for small-scale structures deteriorates
Solution Approach 1:
The patent captures static image copies of welding spatters using simple cameras that can be operated without specialized training. These images are then processed by algorithms that automatically extract quality metrics, making the system easy to operate while maintaining detection precision through sophisticated image analysis.
Solution Approach 2:
The image processing system performs self-service by automatically analyzing captured images and generating welding quality assessments without requiring operator intervention or interpretation. The algorithms independently detect spatters, classify them, and compute quality parameters, making the system both precise and easy to operate.
3Productivity
If high-speed cameras and complex processing systems are used for real-time spatter analysis, then productivity monitoring is improved, but loss of time for setup and operation increases
Solution Approach 1:
The patent uses static image copies captured at regular intervals during welding to monitor productivity, eliminating the need for continuous high-speed recording. This approach provides sufficient monitoring capability while dramatically reducing data processing requirements and setup time compared to real-time dynamic analysis systems.
Solution Approach 2:
The system performs periodic image capture and analysis at predetermined intervals during the welding process, rather than continuous real-time analysis. This periodic approach maintains productivity monitoring capability while reducing computational load and operation time, as only discrete moments in the welding process need to be analyzed.
Data Source
AI summary
The invention relates to a method for determining the performance of a welding method carried out on a metal workpiece, in particular an electric arc welding or laser welding method, with the following steps: introducing one or more extracts of the initial image each having at least one presumed projection, as input to at least one neural network, in particular a convolutional neural network, so as to classify the presumed projections as confirmed or unconfirmed projections, carrying out a second digital processing operation on the initial image comprising the previously classified projections so as to determine at least one parameter representative of the quantity of confirmed projections chosen from the surface of one or more projections.