We developed and implemented the inSpect visual control system to provide a comprehensive quality assessment for extruded catalytic converter (CAT) substrates. Utilizing advanced industrial vision technologies, the system performs precise 2D and 3D measurements.
Servo-driven components automatically adjust to accommodate both square and round parts, ensuring precise inspection across all sizes produced. The fully automated process maintains consistent cycle times, ergonomic operation, and consistent results, with robotic manipulation preserving substrate integrity during inspection.
Our setup utilizes multiple industrial computers with GPUs, ensuring all measurements are processed within the required cycle time. We integrated two control stations along the existing conveyor line, positioned immediately after the substrates are cut to size and cleaned. At the first station, the system simultaneously inspects for anomalies of the substrate’s circumference and edges.
Laser sensors and profilers measure circumference at three points (beginning, middle and end). Substrates are rotated by servo-driven clamps until the system identifies the barcode which serves as a reference point for aligning the honeycomb structure. The precise angle of the honeycomb allows the system to ensure alignment, parallel to the inspection surface, which is critical for detecting defects, especially cracks. Optimized lighting, activated in a strategic sequence, improves crack detection. This approach is also employed for identifying chips and gouges that could affect CAT performance.
The second station conducts two inspections: it checks for surface defects and examines the honeycomb cell structure for missing walls, filled or compressed cells.
Laser sensors and profilers assess the perpendicularity of the top and bottom faces relative to the side walls, and the side walls themselves are scanned by servo-driven laser profilers positioned beneath the cameras to analyse their perpendicularity. The outcomes indicate if the substrate walls are tilted.
Both stations are equipped with AI-powered industrial cameras and LED lighting, using a combination of analytical and AI-driven inspection techniques. Traditional mathematical methods analyse precise geometric measurements such as circumference and perpendicularity. Meanwhile, AI detects potential anomalies and defects on the substrate’s side walls, top and bottom faces, and internal honeycomb structure, including side cracks, scratches, edge chips, and issues like missing cell walls or compressed cells. The system inspects the surface area by generating point clouds to identify discrepancies. Deep learning algorithms further assess the depth, area, location, and volume of defects across all surfaces, enabling accurate evaluations.
