Thermal Striping in LWR Piping Systems

Request Copy
Report Date: August 2001
Appendices: No
 
Abstract

Thermal striping characterizes the phenomenon where hot and cold flow streams join and result in random temperature fluctuations of the coolant near the piping wall. The wall temperature fluctuations can cause cyclical thermal stresses and subsequent fatigue cracking of the piping. For LWRs, the issue of thermal striping came into increased attention after recent incidents due to thermal striping at some nuclear power plants (Oskarshamn/Ringhals1 /Barsabeck2 in Sweden and Tsuruga in Japan) that raised safety concerns. The current study investigates thermal striping at tee junctions of light water coolant systems. The objectives of the current study are to a) identify the key phenomena and parameters of tee junctions that would lead to thermal striping through numerical simulations, b) examine importance of individual parameters, such as Reynolds number, temperature difference between hot and cold streams, and geometries through turbulent simulations as well as experimental data obtained from the literature or acquired from the Japanese Working Group, and c) establish guidelines that the designers should follow to minimize thermal fatigue failures in piping systems.

The FLUENT5 code was chosen for this study. FLUENT5 is a state-of-the-art Computational Fluids Dynamics (CFD) program for modeling fluid flow and heat transfer in complex geometries. In order to validate the simulation results, measurements from a series of thermal striping experiments for water flow in tee junctions that were performed by a working group consisting of several Japanese utilities in the early 1980’s were selected for the initial benchmark study. To simulate turbulent fluctuation with FLUENT, a two-stage approach was employed. First, the Reynolds stress model (RSM) was used to obtain the flow and thermal field over the entire subject system and to determine the boundary conditions for detailed large eddy simulations (LES) of the regions of interest. It is noted that unsteady simulation results using the RSM model showed either stable or oscillatory coolant temperatures depending on the type of numerical discretization scheme used in the simulations. When the first order scheme was used, the coolant temperatures were constant and no temperature fluctuations were found. However, when the second order scheme was used, converged solutions could not be obtained after more than a thousand iterations and the coolant temperature fluctuations were found to persist in unsteady calculations. The LES simulations are challenging to perform due to the extensive computer capability required. The mesh size required for coolant flow with higher Reynolds number (~ 105 to 106) is on the order of 1 to 3 mm and even finer resolution is required for the boundary layer region. To achieve that type of resolution, only a small region of interest was modeled using boundary conditions obtained from the RSM model of the entire system. For these two LES simulations performed, the calculated results showed reasonable agreement with the experimental results.

To increase confidence in the FLUENT5 results, we obtained JNC’s computer codes, AQUA and DINUS-3, for thermal striping modeling. DINUS-3 is a direct numerical simulation code and was previously benchmarked for thermal striping modeling for parallel sodium and water coolant jets. Further validation for tee junctions is currently ongoing at JNC. The DINUS-3 code will be used as a second validation source of the FLUENT5 model.

Calculations for a second benchmark case using both FLUENT5 and AQUA/DINUS-3 were initiated. This set of experimental data was obtained from Hitachi through TEPCO. Hitachi, which participates in the Japanese working group, has performed water experiments for various flow conditions for tee junctions. The FLUENT5 calculations have been used to compare with the experiment and also to study the effect of various parameters. Calculations using AQUA/DINUS-3 codes are still ongoing due to minor difficulties in running the codes.

Calculated results of the Hitachi experiment benchmark case using FLUENT5 show reasonable fluctuations near the tee junction but higher coolant temperature fluctuations downstream of the mixing junction than those reported by Hitachi. Comparisons of the normalized fluctuating temperatures show that the calculated maximum normalized temperature fluctuations remain further downstream than that of the measurements. Although the calculation showed agreement between the values of the maximum normalized fluctuating temperatures, the predicted locations of the peak fluctuations are about 2 to 3D further downstream than the experiment. The experimental value of maximum normalized fluctuating temperature of 0.24 at both the bottom and the side of the pipe (at L/D=0.5 and L/D=1, respectively) compares well to the calculated value of 0.251 at the side of the pipe at L/D=4. However, the measured normalized mean temperature seems to approach uniform mixing at L/D=4 while the calculated ones seem to be still in a stage of transition.

Sensitivity studies were performed for three parameters: time step size, the Smagorinsky constant, and coolant temperature sampling location. In the sensitivity study of time step size, it was found that the normal mean temperatures for both cases have similar trends and are about equal in magnitude. However, a smaller time step leads to a larger maximum calculated normalized fluctuating temperature for all locations; 0.251 at the side of pipe at L/D=4 if 0.01 s is used and 0.218 at the bottom of the pipe at L/D=2.5 if 0.05 s is used. The maximum measured normalized fluctuating temperature is 0.24. In the sensitivity study of the Smagorinsky constant, the normalized mean temperatures for both cases seem to track each other with some small variations (≅0.05). The effect of damping of the calculated normalized fluctuating temperature is apparent for Cs=0.23 for up to L/D=3, however, the effect is not significant downstream of L/D=4. The coolant temperature gradient was found to be significant in the near wall coolant region up to L/D=5. Calculated coolant temperature at depths of 3 mm and 5 mm were compared. The maximum normalized mean temperature difference between 3 mm and 5 mm depth is 0.2, or 6.58 °C, located at L/D=2. The maximum temperature gradient is 3.29 °C/mm. This temperature gradient is attributed to turbulent mixing since the wall is assumed to be adiabatic. Further downstream the coolant temperatures at both depths approach that of complete mixing.

Program:     NSP Nuclear Systems Enhanced Performance

Type:     TR 

RPT. No.:  7