Automated Weld Defect Detection via Machine Vision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automated welding processes, such as those for turbine rotors, face challenges in real-time defect detection due to the difficulty for human operators to monitor the welding process for extended periods and the lack of experience in identifying subtle weld defects, which leads to costly and time-consuming post-weld testing.
Innovation Solution
A system utilizing digital cameras and sensors for real-time image capture and data monitoring, correlating features with defects during a learning phase to compute an aggregate probability of defect presence during production welding, allowing for immediate detection and potential rework.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If human operators manually monitor the welding process, then they can identify potential problems, but they cannot maintain close attention for extended periods and lack experience to detect subtle defects
Solution Approach 1:
The patent replaces the mechanical human monitoring system with an automated machine vision system equipped with cameras and image processing algorithms. The system captures images of the weld pool, ripple pattern, and bead geometry, then uses automated defect detection algorithms to identify defects such as pores, inclusions, and lack of fusion. This substitution eliminates human fatigue and inexperience while maintaining continuous monitoring capability throughout the welding process.
Solution Approach 2:
The system enables self-service by allowing the welding process to monitor and detect its own defects autonomously. The machine vision system continuously captures images and the automated algorithms analyze weld quality in real-time without requiring human intervention. The system serves itself by automatically identifying defects and providing feedback, making the monitoring function independent of human operators.
2Reliability
If post-weld testing is conducted to detect defects, then defects can be identified, but significant costs and cycle time are incurred
Solution Approach 1:
The system performs preliminary defect detection during the welding process itself rather than after completion. By monitoring weld pool characteristics, ripple patterns, and bead geometry in real-time, the system identifies defects as they form. This preliminary detection allows for immediate corrective action before the weld is completed, eliminating the need for time-consuming post-weld testing and reducing overall cycle time.
Solution Approach 2:
The system skips the traditional sequential process of welding completion followed by post-weld testing. Instead, it rushes through the detection phase by performing real-time monitoring during welding, thereby identifying defects immediately and eliminating the separate post-weld testing step. This approach significantly reduces the total time required from welding initiation to quality confirmation.
3Productivity
If automated welding processes are used to improve productivity, then welding speed increases, but real-time defect detection becomes more difficult
Solution Approach 1:
The patent replaces manual visual inspection with an automated machine vision system that uses cameras and digital image processing. This substitution enables real-time defect detection during high-speed automated welding by capturing and analyzing images of the weld pool, ripple pattern, and bead geometry at welding speeds, overcoming the limitations of human visual acuity and reaction time.
Solution Approach 2:
The system implements real-time feedback by continuously monitoring weld characteristics during the welding process and providing immediate information about defect formation. The machine vision system captures images and analyzes them in real-time, providing feedback signals that can trigger alerts or automatic adjustments to welding parameters, enabling quality control to keep pace with high-speed automated welding operations.
Data Source
AI summary
A system and method detect weld defects in real time. Cameras capture images of a weld pool as well as ripple shape and fillet geometry. A processor receives the images and communicates with a database that stores correlated potential weld defects with images of a mock weld molten pool and images of a mock weld ripple shape and fillet geometry. The processor computes an aggregate probability that a weld position corresponding to the images captured by the cameras contains a defect based on the potential defects correlated in the database.