Quality control is facing new challenges as lot sizes become smaller. Inspection of small lot sizes requires and automatic transfer of quality control know-how to a new product variant.
Statistical approaches that are based on large numbers are not usably any more. Inspection processes have to be set up in relation to a process and not a particular product. This will enable to close the loop from quality control to the process also for small lot sizes.
At a technological level Machine learning methods are needed that fully represent the relation between process parameters and quality data in a highly volatile production process.
Additionally, new methods are required that can robustly distinguish between defects and the normal appearance of the product. Photometric stereo provides exactly this information..
Dazu sind auch Methoden nötig, die bei der Beurteilung treffsicher und robust zwischen Defekten und Nicht-Defekten unterscheiden können. Das gelingt unter anderem mit der photometric stereo Technologie.