Jacqueline Schmitt from project B3 successfully defended her dissertation titled "Methodology for process-integrated inspection of product quality by using predictive data mining techniques" on February 04, 2021. The verbal examination took place in digital form. The results of the dissertation were presented in a public 45-minute lecture on Zoom. The examination committee was formed by Prof. Dr.-Ing. Andreas Menzel (examination chairman), Prof. Dr.-Ing. Jochen Deuse (rapporteur), Dr.-Ing. Ralph Richter (co-rapporteur) and Prof. Dr. Claus Weihs (co-examiner).

We cordially congratulate her on completing her doctorate!

Abstract of the thesis -- In the conflicting areas of productivity and customer satisfaction, product quality is becoming increasingly important as a competitive factor for long-term market success. In order to simultaneously counter the steadily increasing cost pressure on the market, this means the consistent concentration on internal company processes that influence quality, in particular to reduce technology-related output losses as well as defect and inspection costs. An essential prerequisite for this, in addition to defect prevention and avoidance, is the early detection of deviations as the basis for process-integrated quality control. Increasingly, growing demands for safety, accuracy and robustness are counteracting the speed and flexibility required in the production process, so that a process-integrated inspection of quality-relevant characteristics can only be carried out to a limited extent using conventional methods of production measurement technology. This implies that quality deviations are not immediately recognised and considerable productivity losses can occur. 

In the present work, a holistic methodology for process-integrated inspection of product quality by using predictive data mining algorithms is developed. The core of the methodology is a new, data-based procedure for the conformity assessment of product characteristics by predictive data mining models. In order to integrate this procedure into the existing quality assurance and to guarantee a reliability of the inspection equivalent to conventional measuring and inspection procedures, a holistic methodology for the planning and design of the process-integrated inspection is also developed. While analytical modelling approaches are used at the core of the methodology, the structure is decisively characterised by the integration of expert knowledge. This combination of data- and expert-based modelling enables the functional and plausible mapping of causal, quality-related relationships, so that a contribution is made to reliable quality assurance in industrial production.

The developed method was empirically validated using selected industrial case studies. The results of the validation show that the developed method can generate shortened quality control loops and identify savings and optimisation potentials of the inspection and production processes. The use of predictive quality inspection thus leads to an increase in productivity and a reduction in quality costs.