Developing an Automated Fluid Activation Residence Time CFD Database
Fluid activation is of critical importance to fusion power plant design, as it can impact dose rates to maintenance personnel, heating in sensitive components, and radionuclide inventories with implications for accident scenarios and waste. The levels of fluid activation in a system are dependent on the neutron flux spectrum and the exposure time of the fluid. Accurately modelling the fluid residence time distribution (RTD) through a pipe system component requires computational fluid dynamics (CFD). However, performing CFD on whole pipe systems can be computationally expensive which limits frequency of design iterations. To address this concern, a new code was developed: FARBASE (the Fluid Activation Residence time dataBASE). This paper details the development and functionality of FARBASE, focusing on its two key features:
- An automated CFD pipeline which accepts a parametric description of a pipe component under given flow conditions, generates the pipe geometry, runs a steady-state OpenFOAM simulation, and returns the resulting RTD.
- A Gaussian Process Regression (GPR) surrogate model which can be trained on the CFD database and queried to provide uncertainty quantified predictions of RTDs. Where the uncertainty of the GPR prediction exceeds a given threshold, FARBASE can automatically perform additional CFD to update the database, improving the accuracy of future predictions.