References of "PHYSOR 2018"
     in
Bookmark and Share    
Full Text
Peer Reviewed
See detailDevelopment of a Data-Driven Approach based on Kalman filtering for CFD Reactor Analysis
Introini, Carolina; Cammi, Antonio; Lorenzi, Stefano et al

in PHYSOR 2018 (2018)

In the last several years, computer-based simulation has become an important analysis and design tool in many engineering fields. The common practice involves the use of low-fidelity models, which in most ... [more ▼]

In the last several years, computer-based simulation has become an important analysis and design tool in many engineering fields. The common practice involves the use of low-fidelity models, which in most cases are able to provide fairly accurate results while maintaining a low computational cost. However, for complex systems such as nuclear reactors, more detailed models are required for the in-depth analysis of the problem at hand, due for example to the complex geometries of the physical domain. Nevertheless, such models are affected by potentially critical uncertainties and inaccuracies. In this context, the use of data assimilation methods such as the Kalman filter to integrate local experimental data witihin the numerical model looks very promising as a high-fidelity analysis tool. In this work, the focus is the application of such methods to the problem of fluid-dynamics analysis of the reactor. Indeed, in terms of nuclear reactor investigation, a detailed characterization of the coolant behaviour within the reactor core is of manda- tory importance in order to understand, among others, the operating conditions of the system, and the potential occurrence of accident scenarios. In this context, the use of data assimilation methods allows the extraction of information of the thermo-dynamics state of the system in a benchmarked transitory in order to increase the fidelity of the com- putational model. Conversely to the current application of control-oriented black-box in the nuclear energy community, in this work the integration of the data-driven paradigm into the numerical formulation of the CFD problem is proposed. In particular, the al- gorithm outlined embeds the Kalman filter into a segregated predictor-corrector formu- lation, commonly adopted for CFD analysis. Due to the construction of the developed method, one of the main challenges achieved is the preservation of mass-conservation for the thermo-dynamics state during each time instant. As a preliminary verification, the proposed methodology is validated on a benchmark of the lid-driven cavity. The obtained results highlight the efficiency of the proposed method with respect to the state-of-art low fidelity approach. [less ▲]

Detailed reference viewed: 107 (3 UL)