Methoden zur datengetriebenen Lokalisierung des Verbesserungspotentials in Produktionsabläufen

Abstract

The aim of this scientific treatise is to support manufacturing companies in increas-ing the productivity of their production processes, one of their classic challenges. The methods presented in this thesis are developed to localize the potential for im-provement in production processes. On this basis, appropriate measures can be tak-en to improve the production process, for example through targeted financial in-vestments. The work presented is at the cutting edge of current developments: On the one hand, today's production is a complex and dynamic system that is comprehensively influenced by its environment. Such a production contains production processes, which are characterized by the demand for an increased number of product variants to be produced at the same time in combination with strongly fluctuating demand loads. This increases both, the dependencies and the fluctuations, within the production processes, which ultimately leads to increased dynamics. On the other hand, the changes that accompany the digitization and interconnection of production processes show new possibilities to counteract this complexity with the help of data-based methods and algorithms. By digitizing the relevant production data, it is possible to interconnect production processes internally and thus capture the production process as a whole. This enables the localization of improvement potentials in dynamic pro-duction processes. Three data-driven methods are developed to localize the potential for improvement in production processes that master the increased complexity and utilize the potential of the production data. The first method localizes the potential for productivity improvement by detecting bottlenecks with a data-driven algorithm. For the first time, productvariant-specific and time-variant bottlenecks can be localized. This is of particular importance for modern production lines, where a large number of product variants are produced simultaneously and in a changing mix. In addition, the method has the advantage over existing procedures that no model assumptions and no simulation models are necessary. The presented method for the localization of bottlenecks answers the question "where" the problem occurs in a production process. However, it makes no statement about "what" the problem is. The second method - the information stream mapping - localizes the improvement potential in information streams and thus provides the methodical tool for visualization, analysis and evaluation of information streams. Similar to value stream map-ping, an information stream map is created, which is then evaluated using five newly developed parameters for information streams (level of automation, media disruption rate, real-time capability, centrality index and first pass yield for information). Based on this evaluation, information streams in production processes can be ana-lysed and optimized. The method opens the possibility of increasing productivity in an area of production to which little attention has been paid so far. Information stream mapping deals with the question whether the right information is in the right place at the right time. It therefore answers the "what" related to information in pro-duction processes. The third method externalizes the questions of "where" and "what" in order to in-crease productivity in production processes through data-driven supply chain collaboration. An external partner improves – data-driven – the production by analysing selected parameters. Using the example of the cooperation between a machine user and a machine supplier, a five-stage procedure is developed that follows the classic approaches to business model innovation. First, the current business model is identified. Based on this, possible partners of the cooperation are evaluated. In a third step, there follows the development and evaluation of data products from the perspective of the data producer according to the criteria "acquisition effort" and "criticality for the existing business model". Based on the data products, a new business model is developed in the fourth step, which is implemented in step five. In order to ensure the applicability and universality of the developed methods, all three presented methods were validated in close cooperation with an industrial partner. The results of the validations are promising: bottlenecks in dynamic production processes can for the first time be localized for product-variant specific and time-variant production. The underlying problems solved with the new methods are real and generally valid and can be transferred to other manufacturing companies, especially mass manufacturers with a line production.