PhD Thesis - staffweb - University of Greenwich

PhD Thesis by John Ewer.research based code because it drastically reduces the time taken to assess and monitor thebehaviour of newly developed research algorithms.1.6 Background to this researchThe origins of this current work go back to 1987 when Knight, Cowell and Edwards[KNIGHT87] investigated the benefits (to CFD) of using strict Software Engineering design anddevelopment techniques for the development of reliable and extensible framework of CFDresearch. This theoretical consideration highlighted some of the problems that are discussed inthis thesis but was not researched in practice due to resource limitations. The investigation wasextended into practical research by Petridis in 1995, for his PhD research [PETRIDIS95]. Thislater research primarily investigated the potential benefits of an integrated Knowledge BasedSystem (KBS) for the dynamic control of a Heat transfer code during the numerical simulation.One problem which was noted during the research was the lack of expertise that was requiredto control the CFD code in terms of appropriate decision making. This was largely due to thenature of most CFD codes which use batch mode of processing that is pre-configured to solvesome flow scenario without any form of user intervention. Further problems, facing the research,were the limited time constraints for development of a prototype system and the lack of accessto comprehensive and reliable CFD software which meant that the prototype system only hadquite limited capabilities when compared with the fully coupled flow, heat, turbulence andradiation algorithms that are commonly found in commercial CFD codes or more recent in-houseresearch codes. The lack of flow and turbulence handling was particularly restrictive becausethese sub-models constitute the majority of the complexity of any general purpose CFD system.The previous research demonstrated that there were important improvements to be made toCFD software if appropriate control expertise could be determined and encapsulated in a KBS.Whilst some expertise had been gained during this prior research, it was by no means completebecause of the limited scope and capability of the prototype system. Another problem that wasidentified was the question of how sufficient information about a particular solution status couldbe represented for initiating any KBS reasoning [EWER93-1].1-11