Reflective Gear Disc Loading

Gears are essential automotive components with massive market demand. Gear discs come in hundreds of specifications and considerable weights, making manual loading and unloading slow and labor-intensive. Highly reflective surfaces and frequent product changeovers add further difficulty to automation. A gear manufacturer deployed a 3D vision-guided robotic solution built on the DexForce embodied AI architecture, achieving flexible, unmanned loading and unloading of gear discs with ±3 mm repeat positioning accuracy.

• Highly reflective surfaces.

Conventional structured light cameras struggle to image reflective gear discs, producing incomplete point clouds.

• Hundreds of disc specifications.

Traditional template matching is inefficient when parts come in over a hundred variants.

• Small batches and frequent changeovers.

High-mix, low-volume production demands fast adaptation.

• Heavy parts at high volumes.

Manual loading of heavy gear discs is inefficient and physically demanding.

An AI-powered structured light camera with a wide field of view performs fast imaging of gear discs arranged on divider trays, outputting complete, high-quality 3D point clouds. A deep learning general-purpose circle model recognizes disc contours and fits the disc center point, then outputs the optimal grasp pose. The system guides the industrial robot to precisely pick each gear disc and place the disc onto the positioning platform. Once machining is complete, finished parts are automatically unloaded and stacked in an orderly manner.

• Complete imaging of highly reflective parts
• The wide-FOV structured light camera with integrated AI algorithms intelligently reconstructs missing point cloud data on reflective disc surfaces, producing complete, high-quality 3D point clouds.
• One model for hundreds of disc specifications
• The deep learning general-purpose circle model accurately recognizes discs across hundreds of specifications without frequent creation of 3D registration templates, enabling flexible production.
• Millimeter-level repeat positioning.
• Visual repeat positioning accuracy of ±3 mm meets gear disc machining and production requirements.

The DexForce Spatial and Embodied Intelligence Suite overcame the long-standing limitation of conventional structured light on highly reflective parts, effectively answering the challenges of small-batch, high-mix production with frequent changeovers. Mixed-model production of multiple gear disc categories now runs smoothly, with significantly improved line stability and production flexibility, and measurable gains in both yield and cycle time.