Validation of STAR-CCM+ with the OECD/NEA T-junction Blind Benchmark

Printer-friendly versionPDF version

Thermal fatigue is a degradation mechanism induced on the primary piping system of a nuclear power plant. Consequences of thermal fatigue are often critical, ranging from structural damage to a complete shut-down as happened with the French pressurized water reactor (PWR) Civaux in 1998. Thermal fatigue has been a very persistent problem and has also occurred in the Japanese PWR Tsuruga-2 in 1999, and the Japanese PWR Tomari-2 in 2003. Hence, it is considered to be a serious safety concern and is seen as one of the most influential parameters on the ageing and life management of nuclear power plants.
An advanced Computational Fluid Dynamics methodology such as Large Eddy Simulation (LES) has emerged to be an effective tool to study the thermal fatigue phenomena. For nuclear applications, any model of a CFD solver should be extensively validated so as to establish confidence in the code capabilities. The OECD/NEA-Vattenfall T-Junction Benchmark was initiated to test the ability of state of the art Computational Fluid Dynamics (CFD) codes to predict the important parameters affecting high-cycle thermal fatigue in mixing tees. Participants in the benchmark exercise were given the (steady) volumetric flow rates in the two inlet pipes of the T-junction. The cold and hot temperatures in the two inlet pipes were also given. In return, numerical data were requested at various measuring stations downstream of the junction for later comparison against the test data once these had been released, thus creating the conditions of a blind benchmark exercise.
Nuclear Research & Consultancy Group (NRG) was one of the 29 world-wide participants in this benchmark. NRG used STAR-CCM+ CFD solver for this benchmark. WALE model is used to account for the sub-grid-scale stresses. Bounded central scheme is chosen for spatial discretization, mainly for its proven stability in LES simulations. Second order implicit formulation is employed for temporal discretization. As an accuracy requirement, the physical time-step was chosen in such a way that the average CFL number in the domain is around 1. The submissions were ranked according to three separate metrics, based on velocity data; temperature data and Fourier transform (of velocities) data. NRG’s results were consistently good in all three metrics.
The RMS temperature predictions which are generally most hard to predict were very well captured. The combined NRG results based on the above mentioned three metrics stood first when compared to rest of the participants. In this presentation  the results obtained by NRG will be discussed in detail.

Author Company: 
Author Name: 
S. Jayaraju
E. Komen
E. Baglietto