AI Weld Quality Prediction From Melt Pool and Keyhole Imaging

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for assessing weld quality, such as laser welding in battery modules, are inadequate and can lead to poor welds that break under vibration, and non-destructive testing methods are not applicable to all products.

Innovation Solution

An apparatus and method using a learned artificial intelligence model and a camera to predict weld quality by acquiring images of the welding process, employing shape, depth, and strength prediction models to analyze melt pools and keyholes, leveraging convolutional and deep neural networks for high-accuracy predictions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional cutting methods are used to check weld quality, then measurement precision can be achieved, but productivity decreases and the method cannot be applied to all products

Engineering Contradiction:
Improveweld quality assessment accuracyVSAvoidproduction efficiency
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent replaces the mechanical cutting method with an optical imaging system combined with AI analysis. The camera captures images of the welding area, and deep learning models automatically analyze the images to assess weld quality, eliminating the need for physical cutting and manual inspection while maintaining high measurement precision.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The patent creates a digital copy (image) of the weld area instead of physically cutting the workpiece. The camera captures visual information of the welding pool and keyhole, and AI models analyze this digital representation to determine weld quality, enabling non-contact inspection that preserves productivity.

Inventive Principle:
Principle #26Copying

2Strength

If laser welding is performed to ensure strong welds, then strength is improved, but weld quality may still be poor leading to breaks under vibration

Engineering Contradiction:
Improveweld strengthVSAvoidweld reliability under vibration
Core Design Contradiction:
StrengthVSReliability

Solution Approach 1:

The patent implements real-time feedback during the welding process by continuously monitoring the welding pool and keyhole characteristics through the camera and AI models. The system provides immediate quality assessment, allowing for real-time adjustments to welding parameters to ensure both strength and reliability, preventing defects that would cause vibration-related failures.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent performs preliminary quality assessment during the welding process itself by analyzing the welding pool and keyhole formation. This allows for early detection of potential quality issues before the weld is complete, enabling corrective actions to ensure the weld will have both the required strength and reliability under vibration conditions.

Inventive Principle:
Principle #10Preliminary action

3Measurement precision

If multiple learning datasets are used to train AI models, then measurement precision improves, but device complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent divides the quality assessment task into three separate AI models, each trained on specific learning datasets for different aspects: shape detection model for geometric features, depth prediction model for penetration depth, and strength prediction model for mechanical properties. This segmentation allows each model to specialize in one function with targeted training data, improving overall precision while managing complexity through modular architecture.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP4398187B1Apparatus and method for predicting weld quality
Publication Date: 2025.07.23 SK ON CO LTD
  • EP4398187B1 patent drawingFigure 1
  • EP4398187B1 patent drawingFigure 2
  • EP4398187B1 patent drawingFigure 3

AI summary

An apparatus for predicting weld quality includes an image sensor acquiring an image of a welding portion of a workpiece, a storage storing a shape detection model, a depth prediction model and a strength prediction model, and a controller configured to acquire shapes of a melt pool and a keyhole by receiving the image and inputting the image to the shape detection model, acquire a penetration depth of the keyhole by inputting the shapes of the melt pool and the keyhole to the depth prediction model, and acquire a tensile strength by inputting one of the shape of the melt pool, the shape of the keyhole and the shapes of the melt pool and the keyhole and the penetration depth of the keyhole to the strength prediction model. The weld quality of all products may be predicted during a welding process.