Authors: Li, Lei & Ma, Yongsheng & Lange, Carlos
The complexity in configuring the CFD solver imposes a barrier for users to efficiently setup the solver and obtain satisfactory results. Such kind of deficiency becomes more obvious when it comes to simulation-based design where the CFD solver is expected to respond to design changes automatically. By applying artificial intelligence, expert systems can be used to capture the knowledge involved in CFD simulation and then assist the solver configuration. This paper proposes an expert system for both dry and wet steam simulation. According to the product design, the expert system is able to select the right module to model the steam flow. Based on the derived non-dimensional numbers, appropriate physics models can be selected to run the simulation. Grid adaption, higher order schemes, and a subroutine for advanced turbulence models help to improve the accuracy of the CFD model after rounds of simulation. The output of the expert system is a robust simulation model with accurate results which are guaranteed by flow regime validation, grid independence analysis, and error estimation. The effectiveness of the proposed system is demonstrated by the analysis of a contracted pipe. In dry steam simulation scenario, the error induced by the expert system is smaller than that of the traditional ANSYS batch mode. The results obtained by the expert system also match well the empirical results when it comes to wet steam simulation.
