An interactive decision support system using simulation-based optimization and knowledge extraction
Abstract: The use of simulation to improve existing manufacturing systems is not new, but simulation can also be used increase the understanding of production systems that have not yet been built. The power of simulation models can be further enhanced by using simulation-based optimization, in which an optimization algorithm tries to find optimal solutions, given certain objectives. However, extracting knowledge from the data resulting from simulation experiments and simulation-based optimization is a complex task. Therefore, tools are needed to assist users in this task. These tools can be visual, like diagrams, or can be generated by data mining. The process of running a study using simulation-based optimization to extract knowledge is a manual task that can in part be automated using existing tools, but to the author’s knowledge there is no software that implements the complete process. This work aims to develop a novel decision support system to support the generic decision process when using simulation and simulation-based optimization. The first step in setting up such a system is to understand how industry currently uses simulation and simulation-based optimization in manufacturing operations. Thus a questionnaire was distributed to manufacturing companies and organizations. The results showed that these techniques are being used, but that companies want more help with the analysis of the results as well as an automated guide in the decision process. This work proposes a system that supports a generic decision process by providing a tool with which a user can define a workflow in their organization, using simulation-based optimization as one component. The decision support system then provides tools for extracting knowledge in the form of diagrams and performs data mining for automated analysis. Data mining is part of the workflow as a tool for extracting knowledge after an optimization, as well as a tool for guiding optimization to suit the users’ preferences. The decision support system also provides for visualization of simulation models and optimization results using augmented reality. A head-mounted display helps users to see the results and model behaviors in 3D. This technology also makes it possible for users to collaborate, both in the same location and remotely. These visual and automatic analysis tools are shown to be effective in several application studies of real-world production scenarios in which data mining has been used to extract important knowledge that would be hard to obtain manually. Together with the automated workflow and efficient visualization of simulation and optimization results in augmented reality, the decision support system is believed to be an effective tool for extracting knowledge for general production systems design and analysis.
CLICK HERE TO DOWNLOAD THE WHOLE DISSERTATION. (in PDF format)