Bearing Industry

Therefore, enterprises urgently require dedicated automated visual inspection equipment to resolve the dual pain points of accuracy and efficiency in manual testing while enabling data traceability. Addressing these challenges, we custom-developed an AI Intelligent Roller Inspection Machine, integrating a full workflow of automatic feeding, 360° appearance inspection, automatic sorting, and data traceability. The feeding system utilizes an imported membrane-coated vibration feeder that seamlessly interfaces with the inspection equipment, ensuring orderly roller presentation to match the high inspection speed of 180 pcs/minute. The vision system employs a high-precision 8-camera line-scan system​ paired with an AI algorithm module, capturing defects as small as 0.01 mm​ (scratches, dents, burrs) at full speed.

Crucially, to solve misjudgment issues caused by air bubbles and surface contaminants adhering during online inspection, the system incorporates an AI Secondary Verification Module​ trained on big data, which automatically differentiates between bubbles/contaminants and genuine defects. Actual validation shows this effectively corrects misjudgments common in traditional vision systems, increasing the product qualification rate by over 5%, perfectly aligning with the development trends of industrial automation and intelligent quality control.

The project faced three core challenges: First, due to the rollers' heavy weight, traditional fixtures and manual handling caused secondary surface damage during transfer. Second, larger rollers presented significant difficulty for outer-surface inspection, where traditional area-scan cameras struggled to balance full coverage, precision, and speed. Finally, machining oil often remained on the roller surface, visually mimicking real defects and causing high false rejection rates.

To resolve this, we customized a full-process solution encompassing Cleaning – Robotic Automatic Loading – Online Inspection – Automatic Sorting: A robotic handling system was introduced to eliminate secondary damage during transfer; high-precision line-scan cameras with multi-angle exposure​ were employed to solve the imaging challenges of large outer surfaces, balancing accuracy and speed; and an AI model​ was utilized to accurately distinguish machining oil from genuine defects, effectively reducing misjudgments. This solution successfully meets the stringent quality standards of the wind power industry and aligns with the national strategy for domestic substitution of heavy equipment.

To address this, we launched a comprehensive solution based on a flexible vision platform and 3D inspection technology. By constructing a configurable software platform adaptable to multiple types and integrating 3D vision inspection technology capable of capturing minute deformations (dents and collapses), we effectively resolved the precision bottlenecks of traditional equipment in detecting slow-motion deformation.

This achieves high-precision, rapid identification of subtle surface defects across various bearing types, thereby significantly enhancing inspection efficiency and accuracy.