A Multiagent design methodology for the manufacturing execution system domain

Abstract

While technological inventions and progress are driving the linkage of cyber-physical-production systems (CPPS), there is still room for a suitable control structure. Multiagent system (MAS) are proper candidates since they establish optimization through competition and flexibility. Although MAS in manufacturing are not new, they still lack momentum. One reason is the lack of specified design methodologies meeting the demands of the manufacturing domain. The manufacturing domain is mainly driven by mechanical or industrial engineers as stakeholders, with no prior MAS knowledge. Although numerous MAS design methodologies exist, none address mechanical or industrial engineers as target designers. Among others, this doctoral thesis addresses this issue, as it is built on two pillars. The first pillar of this doctoral thesis is the design of a MAS as a process control for the integrated tool lifecycle of the TU Wien Pilotfabrik. The TU Wien Pilotfabrik is a demonstrator factory offering students and scientists a state-of-the-art production environment. The tool cycle combines the tool disposition, the tool supply, and the tool use. The tool cycle was chosen as demonstration process since it manages physical and digital entities and encompasses different production layers. Its biggest challenge is the highly heterogeneous hard- and software landscape, causing missing data exchanges and a lack of interfaces. Although it can be assumed, that the tools have a high impact on the overall production cost and thus a high saving potential, the true tooling cost can only be estimated. According to Sharit and Elhence cutting tools cause 25% of all operation costs [SE89]. As groundwork for this thesis, all linkage issues of the tool cycle in the TU Wien Pilotfabrik were resolved and published as the “integrated tool lifecycle” [FTP19]. Besides providing a process control, the MAS maps the true tooling cost for each order and aims to minimize it if possible, by negotiating agents finding a beneficial order sequence. During the first attempts of designing such a MAS, the lack of feasible design methodologies for the manufacturing domain was discovered. The second pillar of this doctoral thesis addresses this issue by developing a multiagent Design Methodology for the Manufacturing Execution System Domain (DM-MESD) especially addresses mechanical and industrial engineers. The DM-MESD offers the designer the possibility to evaluate if the design methodology is suitable for the desired application in advance. This will be achieved by a defined investigation area, which serves as gauge for other applications. Comparing the application with the investigation area resolves, if the DM-MESD can be used for a specific application. The presented design methodology consists of six phases. The DM-MESD focuses on building services based on existing functions and tasks. It uses textual and structural analysis methods to elaborate those tasks and functions to describe existing services and elaborate new services if possible. It is a pure design methodology with no implementation tools or suggestions. This thesis examines the DM-MESD on the integrated tool cycle. By doing so the desired MAS is being d