This report presents a technology for constructing symbolic turbulent stress and turbulent heat flux models using Gene Expression Programming, a type of genetic algorithm, with the aim of achieving high-accuracy and low-cost analysis of complex thermal-fluid fields. The report then demonstrates an application example of the technology where a turbulence model was constructed based on high-accuracy analysis results to achieve high-accuracy thermal stratification phenomena prediction at analysis costs comparable to conventional models. The report also outlines a model construction method currently under development that utilizes test results as training data. These technologies are expected as fundamental technologies that achieve both a reduction in analysis load and an improvement in predictive accuracy in product development.
