Development and repeatability assessment of a machine-vision system for receptacle alignment in ICT needle-bed manufacturing

This paper presents a machine-vision measurement workflow to assess receptacle alignment in in-circuit test (ICT) needle-bed manufacturing. Its methodological contribution is an application-specific geometric model that maps the image-plane displacement between the detected receptacle center and its nominal reference position to the projected displacement at the printed circuit board (PCB) contact plane, and subsequently to an equivalent inclination angle. The model makes the link between image measurement and the engineering acceptance criterion explicit, while using deterministic processing implemented in an industrial vision controller. The system was assessed on a filled needle-bed sample comprising 42 receptacles, with five repeated measurements per receptacle, and a separate 25-reading test on one receptacle. The maximum within-receptacle sample standard deviation across the 42 receptacles was 0.0111 mm; the single-receptacle test yielded a standard deviation of 0.00110 mm and a range of 0.004 mm. Average image-processing time was approximately 0.5 s per inspection. Measurements differed between the two expected assembly-condition regions, but their distributions overlapped and they were not calibrated OK/NOK classes. The results support short-term repeatability under the tested fixed setup and preliminary functional integration feasibility. They do not establish absolute measurement accuracy, a complete measurement-uncertainty budget, robustness to changed operating conditions, or validated industrial classification performance.