The mainstream of diagnostic approaches of printed circuit board assemblies is optimised for large-scale industrial production, and is less suited for repair and maintenance during later stages of the product life-cycle. Main challenges include sparse documentation, small batch sizes and high product diversity, to which existing solutions are too costly, bulky or parameter-sensitive for industrial use. The paper presents a novel approach using computer vision, visual servoing and digital twins for robotic positioning of a measuring probe head. Subject to patent application, the method has proven its feasibility and affordability in live industrial practice with 99.80% success rate.