Nuclear Thermal-Hydraulic (TH) experiments provide valuable insight into the underlying physics of heat and mass transfer and qualified data that can be utilized for code development and validation. However, these measurements are normally sparsely collected which provides limited coverage over the entire domain of interest. Determination of the spatial configuration of these sensors is crucial and challenging during the pre-test design.

This paper aims to investigate the application of data-driven techniques for sparse sensor placement in nuclear TH experiments. Two examples of Thermocouples (TCs) placement in (i) an integral test facility (PANDA) with steam injection through a multi-hole sparger into a subcooled pool [1] and (ii) a 3D test section of a Lead-bismuth eutectic (LBE) loop (TALL-3D) [2] are presented. Both cases share a common feature that the utilization of optical techniques, such as Particle Image Velocimetry (PIV), is difficult or impractical to implement. Therefore, the quantification of transportation of momentum and energy relies heavily on readings from TCs. The ideally positioned TCs are expected to provide sufficient measurements for model calibration and validation to distinguish the major uncertainties from the model inputs.

The datasets are generated through Computational Fluid Dynamics (CFD) simulations, where input parameters with high uncertainty are systematically varied. QR decomposition is employed on the solution matrix to determine the most influential sensor positions. Several configurations have been proposed and compared.
 

Data-driven,sensor placement,sparse measurement,CFD