Deep Learning Welding Spot Detection for Full Quality Coverage
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Manual sampling for welding spot quality detection in resistance welding is labor-intensive, time-consuming, and cannot cover all welding spots, leading to higher detection costs.
Innovation Solution
A method using deep learning to detect welding spot abnormalities by acquiring dynamic welding parameters, inputting them into a pre-trained simulation model, and determining deviations from a threshold to identify abnormal spots.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual sampling by tearing down is used for welding spot quality detection, then detection can be performed, but it cannot cover all welding spots and requires labor which is time-consuming and leads to higher detection costs
Solution Approach 1:
The patent replaces the manual mechanical sampling method with an automated detection system that uses sensors to collect welding parameters and a deep learning model to analyze them. This substitution eliminates the need for manual tearing down and sampling, enabling automated real-time detection of all welding spots without labor time consumption.
Solution Approach 2:
Instead of physically tearing down and sampling actual welding spots, the system creates a virtual model by collecting welding parameters (current, voltage, resistance) and using a deep learning model to predict quality outcomes. This copying approach allows comprehensive detection without physical intervention.
2Measurement precision
If manual sampling by tearing down is used for welding spot quality detection, then detection can be performed, but it requires labor which is time-consuming and leads to higher detection costs
Solution Approach 1:
The system replaces manual inspection with an automated deep learning-based detection system that processes welding parameters in real-time. This substitution maintains high detection accuracy through sophisticated algorithm analysis while dramatically improving productivity by eliminating manual labor and enabling parallel processing of multiple welding spots.
Solution Approach 2:
The detection system performs quality assessment automatically using the welding parameters already collected during the welding process. The deep learning model self-evaluates the quality by analyzing patterns in the parameter data, eliminating the need for external manual inspection and enabling continuous autonomous detection.
3Quantity of substance
If manual sampling by tearing down is used for welding spot quality detection, then detection can be performed, but it cannot cover all welding spots
Solution Approach 1:
The detection system is designed to be universally applicable to all welding spots by using sensors that collect standard welding parameters (current, voltage, resistance) that can be measured for any welding spot. The deep learning model is trained on diverse data to handle various welding conditions, enabling comprehensive coverage of all welding spots with a single standardized detection approach.
Solution Approach 2:
By replacing manual sampling with automated sensor-based parameter collection and algorithmic analysis, the system can detect all welding spots without the physical constraints and costs associated with manual tearing down and inspection. The automated system processes multiple spots efficiently, reducing per-unit detection cost while increasing total coverage.
Data Source
AI summary
The present application relates to a method, device, and system for detecting welding spot quality abnormalities based on deep learning. The method includes: acquiring a dynamic welding parameter in a welding process corresponding to any target welding spot; inputting the dynamic welding parameter into a pre-trained dynamic welding parameter simulation model for simulation, and acquiring a welding simulation parameter output by the dynamic welding parameter simulation model; determining a deviation of the dynamic welding parameter from the welding simulation parameter, and determining that the target welding spot is an abnormal welding spot when the deviation is greater than a preset threshold. The solution of the present application can reduce the frequency of manual tearing down and batches for abnormality detection, which has a faster abnormality detection speed and may cover all welding spots.