Nuclear Engineering and Design 237 (2007) 575–590 Numerical simulation of turbulent flow in a 37-rod bundle D. Chang, S. Tavoularis ∗ Department of Mechanical Engineering, University of Ottawa, 770 King Edward Avenue, Ottawa, Ontario, Canada K1N 6N5 Received 14 December 2005; received in revised form 31 July 2006; accepted 1 August 2006 Abstract The unsteady Reynolds averaged Navier–Stokes equations, combined with a Reynolds stress model, were solved numerically to determine fully developed isothermal turbulent flow in a 60 ◦ sector of a 37-rod bundle. It was found that this flow contained large-scale coherent structures, which affected strongly the local velocity fluctuations, especially near the gaps between rods or between rods and the surrounding wall. The time-averaged mean velocity and Reynolds stresses were in good agreement with experimental results in a similar channel. Coherent velocity fluctuations at different locations throughout the entire rod bundle were strongly correlated with each other. © 2006 Elsevier B.V. All rights reserved. 1. Introduction Rod bundles are essential elements of pressurized water nuclear reactors. They consist of tightly packed arrays of rods, which contain the nuclear fuel and are surrounded by flowing liquid coolant. Flow phenomena in the subchannels bounded by adjacent rods or outer rods and the containing pressure tube walls are quite complex and exhibit patterns not present in pipe flows. In particular, rod–rod and rod–wall gap regions are char- acterized by strong transverse flow pulsations, which are largely responsible for momentum and heat transfer across the gaps. It is widely accepted that these flow pulsations are associated with large-scale vortices, which form quasi-periodically in pairs on either side of the gap (Hooper and Rehme, 1984; Rehme, 1992; Krauss and Meyer, 1998; Guellouz and Tavoularis, 2000a) and can be classified as coherent structures (Hussain, 1983). In general, previous simulations of turbulent flows in rod bun- dles have assumed that the flow field is stationary and solved the Reynolds averaged Navier–Stokes equations (RANS) using conventional gradient-transport-type turbulence models (e.g., Rapley and Gosman, 1986; Baglietto and Ninokata, 2005). Such simulations have achieved a limited agreement with experi- ments only by the direct use of empirical information applicable to the specific geometry (Wu, 1994; Baglietto and Ninokata, 2005). The use of RANS with general-purpose turbulence mod- ∗ Corresponding author. Tel.: +1 613 562 5800x6271; fax: +1 613 562 5177. E-mail address: tav@eng.uottawa.ca (S. Tavoularis). els without specific adjustments has provided poor predictions. For example, Lee and Jang (1997) performed numerical sim- ulations for a rod bundle flow using a non-linear k–ε model without any adjustments to conclude that this approach strongly underestimated the strong azimuthal turbulence intensity mea- sured by Hooper (1980). Yadigaroglu et al. (2003) conducted an in-depth review of rod bundle numerical simulations, iden- tifying the needs of light water reactor safety analyses. They summarized the numerical simulations of Tzanos (2001) in a square-lattice rod bundle and concluded that gradient-transport models, like the k–ε model, are not adequate to predict turbulent flow in the narrow gap regions. Most previous simulations of rod bundles have considered periodic arrays of rods, only solving the flow in elementary sections of such arrays and assuming symmetry across the boundaries (Rapley and Gosman, 1986; Lee and Jang, 1997; Baglietto and Ninokata, 2005). This approach greatly restricts the solution, because, although the geometric configuration may be symmetric, the flow itself may not be so, particularly as it is known to oscillate across gaps. Moreover, most numeri- cal studies have been conducted for either triangular or square subchannels, which may not be representative of entire rod bun- dles. For example, the 37-rod bundle of the CANDU reactor has three different kinds of subchannels, triangular, square and wall subchannels. Full bundle simulations are required to cap- ture possible interactions among different types of subchannels. In a recent article (Chang and Tavoularis, 2005), we have pre- sented detailed numerical simulations of isothermal turbulent flow in a rectangular duct containing a single rod, a configu- 0029-5493/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.nucengdes.2006.08.001