Abstract

As the term implies, simulation data management (SDM) is essentially the managed recording of simulation data, specifically the metadata that describes a simulation so that its purpose can be immediately understood without having to load it up in its associated simulation software. Ignoring for now the business aspects of a simulations study (for example, who is the client and its representative), since this is not specifically simulation data, there are two aspects to SDM: 1) planning what to do before the start of the analysis, and 2) recording what was done during the course of the analysis. It is important to clearly set out what the aim of the analysis is at the start through appropriate planning and then subsequently to check whether the analysis has addressed the initial aim. For routine analyses, the intended approach and actual approach may be very similar or indeed identical, and this may lend itself to a template driven study that, once developed, can be used by less experienced users – so called democratization. For novel analyses, however, the planned analysis will often differ from the actual analysis as limitations in the simulation software become apparent, or the solver parameters need to be adjusted in order to achieve a solution. A successful SDM system should be capable of easily enabling the planning and recording of simulation data for both routine and novel analyses, and for the spectrum in between – for example, capturing modifications to previously routine workflows such that processes can be improved for future use. In its simplest form, an SDM system can (and often does) store metadata in the form of a spreadsheet. This is at least a start and simulation metadata is recorded. A more powerful approach, however, is to use a database system since this allows the simulation data to be stored more robustly and searched more easily, and it enables further opportunities such as the automated creation and execution of simulation activities which can significantly reduce effort and improve quality. With traditional database thinking, however, the difficulty is: how to design a data structure that is sufficiently flexible to allow different metadata to be stored depending upon the type of study, and even evolve during the course of a novel study where the precise metadata to store may not be clear even to the most experienced user a priori. This presentation discusses precisely how to set up the flexible database structure required for successful SDM, and how useful information can be retrieved in an easily understood format. Specifically, the database structure is based upon the principal of inheritance and captures only the changes in metadata from one simulation to another. Everything presented is programmed within Microsoft Office, and is therefore easily accessible to many simulation users.