Coating is the first critical step in lithium-battery manufacturing, and coating uniformity directly determines cell capacity and safety downstream. DaoAI's 3D AI-AOI inline full-inspection solution intercepts thickness anomalies and surface defects at the source.
Electrode coating is the opening step of lithium-battery manufacturing, where slurry is uniformly applied to copper or aluminum foil at micron-level thickness. Once dark spots, scratches, exposed foil or localized material drop-off appear in the coating, they are amplified through winding and formation, ultimately surfacing as capacity fade or even internal short circuits. The leading power-battery manufacturer had relied on manual lamp inspection and conventional 2D vision, which could only catch obvious surface anomalies and was nearly blind to subtle thickness undulations and low-contrast dark spots.
Coating lines run at high speed and wide web width, and the copper foil surface produces metallic glare. Conventional AOI in this scenario suffers both high false alarms and missed detections, with inspection cadence struggling to keep pace with the line. The customer needed an inline solution that could both see the surface clearly and quantify thickness.
DaoAI Solution
DaoAI deployed a 3D AI-AOI inline full-inspection system. A proprietary 3D camera captures point clouds of the coating surface and fuses them with high-resolution 2D imaging: the 2D channel handles surface defects such as dark spots, scratches and exposed foil, while the 3D channel reconstructs coating thickness undulation and edge drop-off from the point cloud. The system uses APDT positive-sample learning, building criteria from qualified-coating samples alone, going live without massive defect datasets and adapting well to color variation and glare across new slurry batches.
- Proprietary 3D camera reconstructs coating thickness, quantitatively capturing dark spots and drop-off at the 50μm level
- High-resolution 2D imaging identifies scratches and exposed foil, with point cloud and image aligned frame by frame
- APDT positive-sample learning models from qualified samples, shortening changeover and new-batch ramp-up
- 100% inline full inspection with cadence matched to high-speed coating lines
Stop the defect before winding, so no dark spot ever reaches the next process.
After deployment, the coating stage stably detects dark spots and drop-off defects better than 50μm, with composite accuracy held above 97% and false alarms effectively suppressed. Inline full inspection replaced the previous spot-check and manual lamp inspection, sharply reducing defect escape and supplying more consistent electrodes to downstream winding and formation.