Data-driven modeling of the flow field between two PWR surrogate bundles under seismic conditions using Bagging-Optimized Dynamic Mode Decomposition (BOP-DMD)

The safe operation of nuclear reactors under severe operating conditions is a mandatory requirement for any plant design. One of the most challenging conditions is represented by earthquakes: however, works of literature discussing the response of reactor core bundles and their interaction with the moderator under seismic conditions are limited. Direct measuring techniques of flow field parameters are highly complicated and extremely expensive, hence fluid-structure interaction models are proposed to solve such problems. However, these models will still have some limitations, for example, high computational costs and systematic uncertainty. Data-driven modeling as a data analysis technique can be used to highly reduce computational costs whilst obtaining results of high enough accuracy. In this work, we use Bagging-Optimized Dynamic Mode Decomposition (BOP-DMD) to provide stable forecasting of some flow field parameters with spatial and temporal uncertainty quantification. This is an equation-free Model Order Reduction technique (MOR) built using MATLAB, it is suitable for future data prediction with large accuracy limits and low computational time. The benchmark of this work is the ICARE experimental facility. Data used in this work for the training and validation of the model are obtained from 2D experimental measurements of flow velocity fields around PWR surrogate bundlesunder seismic conditions using the Particle Image Velocimetry technique (PIV).