Assembly Confirmation: Verifying Assembly State Before Release

Accuracy requirements should be derived backwards from the product tolerance, rather than starting from a "the more accurate the better" target. If the positioning tolerance on the drawing is ±0.5 mm, it is enough for the measurement repeatability of the inspection system to be clearly better than that order of magnitude; pushing accuracy higher than necessary brings unnecessary cost.

Yes for many structures, provided that a feature reflecting depth can be imaged (such as the exposed length of a shaft or the position of a press-fit shoulder). If that feature is occluded or not visible, consider changing the tooling, adding a side camera or switching to another sensing method (such as a displacement sensor) rather than forcing vision to do the job.

The quality of manual confirmation fluctuates with fatigue and cycle time, whereas visual confirmation applies the same set of acceptance criteria every time. In addition, the results of visual confirmation can be retained as image evidence for later traceability and process improvement, which manual confirmation can hardly achieve.

Note that the range is two-sided: an incomplete assembly is a defect, and over-pressing is equally a problem. The acceptance criteria should set both an upper and a lower limit, rather than only a "not less than". The dimension ranges in the table above are only examples of the form of the criteria; the specific tolerances must come from the product drawing or process documents and cannot be assumed from experience.

The most common technical route for seating judgement is measurement-based judgement: use the image to measure a quantity that reflects assembly depth (such as the length of shaft exposed after press-fitting, the gap width after snapping together, or the distance from the seating mark to the datum), then compare this quantity with the acceptable range. Compared with whole-image comparison that asks "does it look alike", the advantage of measurement-based judgement is that the criteria can be quantified, tolerances can be set and results can be traced. The prerequisite is that this physical quantity can be measured consistently in the image — which brings us back to the questions of optics and mounting angle.

View the object characteristics, acceptance criteria and optical considerations for each object; it covers 8 common object types including gasket presence/absence inspection, spring presence/absence inspection, seal ring presence/absence inspection and dowel pin presence/absence inspection.

What visual inspection replaces is repetitive visual judgement, not manual labor for everything. The typical division of labor is: vision performs part-by-part full inspection and judgement, while people handle re-judgement of borderline samples, changeovers and exceptions, and maintenance of the optics and tooling. How many people are actually involved depends on the degree of automation and the re-judgement strategy.

Please provide photos of the assembled in-place state (including conforming parts and the common not-fully-seated parts), together with the in-place tolerance requirement from the drawing; an engineer will judge whether that acceptance criterion can be measured stably in the image.