KOMPASS AI brings dent detection to formed, machined and finished parts. Cameras and lighting reveal changes in the surface, while the model checks them against your acceptance criteria, so damaged parts can be separated before coating, assembly or packing adds more cost.
Dent detection gets difficult when a shallow depression reflects the light almost like the surrounding surface. A part can pass a quick visual check, reach the next operation and reveal the damage only when someone turns it.
Our AI Product Inspection system, KOMPASS™, combines trained visual models with modular vision hardware to identify dents on the surfaces your line needs to inspect. The setup brings the defect into view, applies your quality criteria and passes the result to the line controls.
Reflections make polished rods a demanding inspection task. Jidoka's demonstration shows automated surface inspection of these parts, with dents among the defect types described. It brings together controlled imaging, visual classification and automated separation.
Choose camera and lighting angles that reveal the depression on the actual finish. Curved parts need a view of the relevant circumference, with handling that presents it repeatably.
KOMPASS checks the captured image for the trained defect categories. Good parts and borderline samples establish which surface changes your process accepts and which it rejects.
The inspection decision reaches the configured line response. A defect image gives quality teams a starting point for checking whether similar marks recur after a particular operation.
Send us a dented part, an acceptable part and the criteria that separate them. We will assess the imaging and handling your inspection needs.
Get In TouchEach method depends on a suitable view of the surface. The difference is how it handles normal variation and applies the acceptance decision during production.
Jidoka's bearing inspection system uses rotation and flipping to expose the top, bottom, inner and outer surfaces. Its published defect categories include dents, scratches, tool marks and unwashed areas across 18 bearing variants.
These are results for the complete bearing inspection application, rather than a separate dent-only benchmark.
It checks the captured surface for depressions or deformation that fail your visual acceptance standard. Dent inspection starts with the surfaces that matter to the product, the point in production where they are accessible and representative examples of acceptable and damaged parts. KOMPASS then classifies the visible defects and passes the result to the configured response. Functional acceptance still follows your engineering requirements, especially where a mark could affect sealing, fit or another critical feature.
Shallow dent detection depends on making the change in shape visible in the image. Reflective surface inspection may need several lighting directions, different camera views or rotation of the part. A faint depression that disappears under one arrangement may become clear under another. The right test uses your actual finish, including the hardest borderline examples, at the intended production speed. That establishes which defects the station can reliably distinguish before automatic rejection is enabled.
Machine vision lighting establishes a repeatable view, then the model is trained with both acceptable surface variation and real defects. Include normal contours, tool marks, oil conditions and finish differences in the sample set. For surface deformation detection, test the difficult examples that sit close to the acceptance boundary. Reviewing missed defects and good parts incorrectly rejected together helps avoid improving one result at the expense of the other during validation.
Dent measurement requires a setup that can measure geometry against a defined reference. A visible dark patch or distorted reflection does not, by itself, establish depth. If your decision depends on a depth tolerance, specify that requirement at the start so the team can assess suitable 3D sensing or profiling. Where the requirement is cosmetic defect detection, a validated visual classification may be sufficient. The inspection method should follow the acceptance criterion on your drawing or quality specification.
Yes, where imaging, handling, processing and rejection can all fit the available cycle. A dent detection camera needs enough light and a suitable exposure to capture the relevant detail without movement blur. Rotating or flipping a part also takes time. Feasibility should therefore be checked against the complete station cycle, part spacing and reject position. A software inference figure alone does not establish the production rate that the installed inspection can support.
Include each material, finish and geometry that changes the appearance of the inspected area. Surface topography, gloss, curvature and normal manufacturing texture all affect the image. For metal surface inspection, that can mean samples from different machining, polishing or forming conditions. Add the smallest rejectable defect and acceptable marks that look similar. If the line has several variants, record their differences so validation represents the intended mix instead of a single easy sample.
Measure defect escapes and the false reject rate separately on a labelled set that represents production. Review results by defect category, severity, part variant and inspected region. Include fresh samples that were not used for training, then repeat checks at the intended speed and presentation conditions. Jidoka's NSK case reports performance for a combined bearing application. Your acceptance decision should use the measured results for your own parts, with clear criteria for moving into automatic operation.
Potentially, if the imaging setup reveals each defect that matters. Surface defect detection can bring dents, scratches and other visible anomalies into one inspection, but they may need different views or lighting. Define the complete defect list early and validate each category individually. Otherwise, a station that performs well on an obvious scratch could still miss a shallow depression. The review should also identify any regions hidden by a fixture or inaccessible at that production stage.
Share your part, finish, smallest rejectable dent and target line speed. We will assess what it takes to bring that defect into view and make the decision at your chosen checkpoint.
