In-Situ Weld Inspection Using 3D Defect Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional welding quality control methods are error-prone and cost-intensive, relying on postmortem analysis that lacks context for root-cause identification and are limited to specific weld processes, failing to effectively predict weld quality during the process.
Innovation Solution
An in-situ inspection system using computer vision and machine/deep learning algorithms processes sequenced imagery from cameras to provide real-time weld quality assessment, utilizing digital models for predictive insights and defect characterization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If postmortem analysis of welding failures is used, then root-cause analysis can be performed, but context information is absent for proper root-cause analysis
Solution Approach 1:
The system performs preliminary inspection during the welding process itself, capturing images of the weld pool and surrounding area in real-time. This allows defect detection to occur before the welding is complete, preserving all contextual information about the welding conditions, parameters, and environment that caused the defect. The digital model is built incrementally during welding, maintaining the temporal and spatial context of each observed feature.
2Reliability
If conventional NDE post-weld inspection is used, then weld quality can be verified, but the process is costly and time-consuming
Solution Approach 1:
The system replaces conventional mechanical NDE methods (such as ultrasonic testing, radiographic testing, or manual visual inspection) with an automated optical imaging system coupled with machine learning algorithms. Cameras capture images of the weld pool and solidified weld metal, which are then processed by a digital model to detect defects. This substitution eliminates the need for costly and time-consuming post-weld inspection while maintaining high reliability in weld quality verification.
Solution Approach 2:
The welding system performs its own inspection function through the integrated imaging system and digital model. The same system that monitors welding parameters also detects defects, eliminating the need for separate inspection equipment and personnel. The digital model automatically analyzes the captured images and provides real-time feedback on weld quality, enabling the system to self-verify its output without external intervention.
3Manufacturing precision
If process method qualification is used, then welding quality can be controlled, but the techniques are error-prone and do not regress well to desired quality features
Solution Approach 1:
The system replaces error-prone manual process qualification methods with an automated digital model that objectively analyzes weld characteristics. Instead of relying on subjective interpretation of welding parameters or experience-based quality control, the machine learning model quantitatively assesses weld pool morphology, solidification patterns, and other visual features to predict quality outcomes. This digital approach provides consistent, reproducible, and accurate quality predictions that do not suffer from human error or variability.
Data Source
AI summary
A method inspects weld quality in-situ. The method obtains a plurality of sequenced images of an in-progress welding process and generates a multi-dimensional data input based on the plurality of sequenced images and/or one or more weld process control parameters. The parameters may include: (i) shield gas flow rate, temperature, and pressure; (ii) voltage, amperage, wire feed rate and temperature (if applicable); (iii) part preheat/inter-pass temperature; and (iv) part and weld torch relative velocity). The method generates defect probability and analytics information by applying one or more computer vision techniques on the multi-dimensional data input. The analytics information includes predictive insights on quality features of the in-progress welding process. The method then generates a 3-D visualization of one or more as-welded regions, based on the analytics information, and the plurality of sequenced images. The 3-D visualization displays the quality features for virtual inspection and/or for determining weld quality.