h1

%0 Thesis
%A Kraus, Andreas
%T Anwendungsorientierte Methodik zur frühzeitigen Integration datenbasierter Ansätze in die Anlaufphase disruptiver Produkte
%V 26
%I RWTH Aachen University
%V Dissertation
%C Aachen
%M RWTH-2023-10476
%@ 978-3-98555-179-8
%B Ergebnisse aus der Elektromobilproduktion
%P 1 Online-Ressource : Illustrationen, Diagramme
%D 2023
%Z Druckausgabe: 2023. - Auch veröffentlicht auf dem Publikationsserver der RWTH Aachen University. - Weitere Reihe: Edition Wissenschaft Apprimus. - Weitere Reihe: Elektromobilproduktion
%Z Dissertation, RWTH Aachen University, 2023
%X Manufacturing companies are progressively confronted with the challenge of an increasing number of production ramp-up phases, which are centrally characterized by the instability of production, especially in case of disruptive products. In recent years, data-based approaches from the area of machine learning have increasingly been established in the production environment as an effective tool for improved process stability and thus have the potential to reduce the loss in added value during ramp-up phases. Especially at the beginning of ramp-up phases, the essential prerequisite for an effective use of data-based approaches (i.e., a sufficient data basis for the required training process) is not given. The generation of a sufficient data basis prior to the start of production is in most cases accompanied by a high expenditure of time and costs and furthermore increases strongly with the complexity of the underlying use case. The aim of this dissertation is to develop an application-oriented methodology that supports manufacturing companies in the early identification of potential failures in production as well as in the development and integration of suitable data-based countermeasures. By applying the methodology in early stages of the product development process, companies have the opportunity to integrate the developed measures into their production environment prior to the beginning of the ramp-up phase of new products, thus preventing the loss in added value during the production ramp-up phase itself. The methodology developed within the scope of this dissertation consists of three successive modules that systematically guide users from the identification of potential failures to the development of data-based countermeasures as well as their integration into production. In the first module, all relevant process variables and product parameters along the process chain under consideration are identified. Based on this, the impacts and sensitivities of the individual process steps are quantified, followed by the derivation of an interdependence graph along the process chain. Correspondingly, a failure cost graph along the process chain is determined on the basis of the process costs. The results of the first module serve as a basis for analysis at the beginning of the second module, in order to detect quality and cost critical process steps. As a countermeasure for potential failures within the critical process steps, data-based approaches from the area of machine learning are derived and subsequently complemented by the identification of suitable data-based and knowledge-based sources for the required model training. In the third module of the methodology, the modeling and corresponding model training of the data-based approach is performed and thus yields the final trained model. Lastly, a complete catalog of measures is derived in order to complement the integration in production.
%F PUB:(DE-HGF)11 ; PUB:(DE-HGF)3
%9 Dissertation / PhD ThesisBook
%R 10.18154/RWTH-2023-10476
%U https://publications.rwth-aachen.de/record/972935