Application of AI Solder Joint Inspection Equipment in the Welding Field

Application of AI Solder Joint Inspection Equipment in the Welding Field

Modern welding technologies encompass a variety of types, including resistance welding, thermocompression welding, soldering iron welding, laser solder welding, laser fusion welding, and wave soldering, among others. These techniques are widely applied in industrial production such as 3C electronics and new energy manufacturing.

Given the current state of welding technology and the influence of various factors during the welding process, manual inspection of solder joints can only be accomplished through multi-station collaboration when faced with different inspection requirements, making omissions and misjudgments highly likely. Additionally, manual inspection suffers from inconsistencies in standards among different inspectors and the fatigue caused by repetitive detection tasks for workers.

Traditional visual inspection methods are gradually being replaced by machine vision detection in automated production due to their high subjectivity, missed detection, and other shortcomings.

The machine vision AI solder joint inspection system utilizes a combination of industrial cameras, lenses, and lighting to capture clear images of any solder material under different colored light sources. Paired with detection software for automatic recognition, it significantly improves the efficiency and quality of solder joint inspection.

To address these challenges, Songsheng Optoelectronics has developed an AI solder joint inspection device capable of measuring multiple parameters—such as height, area, and volume—of solder joints on circuit boards in a single pass. This enables better data integration and comprehensively enhances the inspection efficiency and accuracy of circuit boards, thereby resolving key industrial vision challenges such as defect detection, dimensional measurement, and visual guidance positioning in production processes.

AI Welding Inspection Software Training Process

The labels on the diagram sequentially read:
Small Batch Image Collection → Model Pre-training → Online Testing → Model Iterative Optimization → Model Delivery

The solder joint inspection system utilizes AI algorithms to detect various defects. During the training process, manufacturers must provide defect standards and categorized images. By labeling different defect types, the AI model automatically learns defect characteristics.

Detectable Defect Types

• Common  solder joint defects in resistance welding, thermocompression welding, soldering iron welding, laser solder welding, laser fusion welding, and wave soldering

• Includes cold solder joints, over-welding, non-wetting, misalignment, solder balls, missing wires, missing pads, broken traces, excess solder, and more

• Compatible with semi-automatic and fully automatic production lines

AI Welding Inspection Software Features

1. Production Statistics
    Real-time data collection and analysis of production metrics

2. Deep Learning-Based Pad Auto-Location
    recise automatic positioning of solder pads using deep learning algorithms

3. Multiple Trigger Modes

• External hardware trigger for camera capture

• PLC/industrial PC communication trigger

• Manual "soft trigger" button operation

4. Intelligent Defect Visualization

• Clear bounding boxes highlight wires and defects

• Green: Compliant solder joints

• Red:  Non-compliant solder joints

• Customizable pass/fail criteria via [Settings]

5. Customizable Process Parameters

• NG judgment thresholds (e.g., cold solder joint length ratio)

• Defect everity standards adjustment

• Solder ball rejection criteria (size/position-based)

6. Automated Data Management

• Auto-sorting of OK/NG images by date and defect category

Case Studies

Enameled Wire Spot Welding Software Interface

Comparative Display of Enameled Wire Welding Defects

Laser Solder Welding Software Inspection Interface

Laser Solder Welding Defect Comparison Chart