Spot Weld Inspection
Accurately classify spot weld quality while minimizing false positives
A breakthrough in complex inspection, part location, classification, and OCR
To ensure a firm electrical connection, wires must be spot welded to each other or to terminals. Welding involves melting together two different metals to form a firm connection. It is important that the weld be adequately melted, have a sufficient but not excessive volume, have a good shape, and be positioned properly. Manufacturers tend to have many lines for a wide variety of electronic parts and need to ensure that all connections are secure.
Because spot welds have a high variation and non-uniformity in appearance, including shape, position, color, reflectivity, texture, and surface markings, the inspections can have a high rate of false positives, otherwise known as overkills. Overkills lead to good parts being discarded. Good welds incorrectly marked as defective have to go to manual inspection, which is extremely slow compared to line speeds, and still often results in improperly identified weld defects.
Parts also vary in size, color, and other features from one lot to another. The wide range of variation, and the difficulty of distinguishing good connections from bad ones makes rule-based machine vision impractical.
Users train Cognex Deep Learning’s defect detection tool on a wide selection of properly spot-welded connections to learn the full variation of normal parts. As the tool scans through spot welds it analyzes and flags any that are outside of the acceptable range, while minimizing false positives.
Cognex Deep Learning’s classification tool can then train on a range of labelled welding defects and learn to categorize specific defect types, such as s poor shape, blowhole, crack, burn, or surface contamination. These categorized defects can then be used for upstream process control to minimize defects over time.