My doctoral dissertation was on the topic of “Integrated Simulation.” Together with Dr. Ingo Meents, I co-authored the first joint dissertation at TU Clausthal.
The dissertation was awarded the Förderpreis of the Freunde der TU Clausthal. The laudation was delivered by Ekkehard Schulz (at the time CEO of ThyssenKrupp AG).
Below you will find further information about EPOS and the dissertation:
- Dissertation “Integrated Simulation – Optimization Strategies and Logistical Control for Production Planning Based on Calloboratively Maintained Queueing Models“
- www.integrierte-simulation.de (links to the Stochastics pages of TU Clausthal)
- EPOS (Enterprise Production and Optimization System)
- Report on the award of the TU Clausthal Förderpreis
Summary / Abstract
Analyzing queuing models to answer questions of tactical production planning for large and complex manufacturing systems is not possible without supporting IT systems. The established mathematical models provide approximations of performance metrics such as utilization, inventory levels, and throughput times. Creating and maintaining the required queuing models for hundreds of workstations and thousands of operation sequences — connected through complex process flows — presents a significant challenge. The Integrated Simulation model is the theoretical foundation for building software systems that enable the efficient creation and maintenance of large models through distributed responsibilities. Integrated Simulation describes the import of data from line control systems and the automatic generation of queuing models from databases containing all data relevant to planning. The automation of data acquisition, the analysis of queuing models, and automated reporting enable continuous support of planning processes — even in constantly changing production environments.
Simulation and analytical performance evaluation allow the investigation of a priori defined scenarios. To find improved or even optimal parameter sets, optimization methods that can use automatically generated models are presented. Three distinct problems and their corresponding solution approaches are examined. First, the optimal product mix problem is extended to include the requirements of Integrated Simulation. A quadratic program for determining the optimal routing probabilities — which are not fixed during model generation — is then formally derived. Finally, evolutionary algorithms that use the performance metrics of queuing models as constraints or objective functions are developed.
When simulation is used as an ongoing process in production planning, integration with existing line control systems allows for the creation of operational plans, problem detection systems, and systematic validation of the simulation models used. Applications arise from comparing measured performance metrics with the production targets derived from simulation. Statistical tests, visualized through quality control charts, must be applied. Challenges arise from the autocorrelation in the logistical processes under consideration and from the number of possible charts. As a further application of integration with line control systems, a forward propagation of current inventory based on approximated remaining throughput times is presented.
To investigate the practical applicability of the proposed architecture and algorithms, a system was developed that is now used for the simulation of five production lines of IBM Deutschland Speichersysteme GmbH at three international locations. It has been shown that only the proposed holistic approach — that is, the integration of all required processes — enables the successful application of analytical performance evaluation to real-world problems.