Fusion Engineering and Design 87 (2012) 556–560 Contents lists available at SciVerse ScienceDirect Fusion Engineering and Design journal homepage: www.elsevier.com/locate/fusengdes CFD analysis of flow boiling in the ITER first wall Phani Domalapally a,∗ , Enrico Rizzo b , Laura Savoldi Richard a , Fabio Subba a , Roberto Zanino a a Dipartimento di Energetica, Politecnico, I-10129 Torino, Italy b Institut fuer Technische Physik, Karlsruhe Institute of Technology, Karlsruhe, Germany article info Article history: Available online 23 February 2012 Keywords: Boiling First wall Two phase flow Computational Fluid Dynamics abstract This paper compares two Computational Fluid Dynamic (CFD) approaches for the analysis of flow boil- ing inside the first wall (FW) of the International Thermonuclear Experimental Reactor (ITER): (1) the Rohsenow model for nucleate boiling, seamlessly switching to the Volume of Fluid (VOF) approach for film boiling, as available in the commercial CFD code STAR-CCM+, (2) the Bergles–Rohsenow (BR) model, for which we developed a User Defined Function (UDF), implemented in the commercial code FLUENT. The physics of both models is described, and the results with different inlet conditions and heating levels are compared with experimental results obtained at the Efremov Institute, Russia. The performance of both models is compared in terms of accuracy and computational cost. © 2012 Elsevier B.V. All rights reserved. 1. Introduction In ITER, the FW heat load during operations can be as high as several MW/m 2 , which should be removed by a proper cooling sys- tem to prevent damage of the component [1]. The proposed sink to handle this load consists of copper alloy (CuCrZr) hypervapotron or swirl tubes, which exploit the large heat transfer coefficient (HTC) characteristic of highly sub-cooled boiling [2]. In order to predict the system performance, a dedicated thermal hydraulic analysis has to be conducted, and several papers have been devoted during the last few years to this task on different geometries foreseen for the ITER first wall [3–7]. Here we test and compare two approaches, based respectively on an enhancement of the Rohsenow model [8] and on the classical Bergles–Rohsenow (BR) work [9]. The first model is available in the commercial STAR-CCM+ code, including a smooth transition to the VOF model for film boiling. We implemented the second in the FLUENT code, specifically for this work. We then compared our computed results with experimental data from Efremov Institute, Russia [10]. Although our ultimate goal is to develop predictive capabilities for the hypervapotron, we start by analyzing in some detail the simpler case of a rectangular (so-called flat) channel, heated on one side. This allows us concentrating on the details of the heat transfer model performances. Preliminary results for the hypervapotron are discussed at the end of the paper. The layout of this work is as follows: in the next section we discuss the system geometry and meshing issues. In Section 3 we ∗ Corresponding author. E-mail address: phani.domalapally@polito.it (P. Domalapally). describe the physical models. In Section 4 we present and discuss our results, in terms of both accuracy and numerical performance. To the best of our knowledge, this is the first time that the Rohsenow and BR models are critically compared for an ITER-relevant applica- tion. Also, we are not aware of an implementation of the BR model in FLUENT so far and, indeed, previous applications of FLUENT to ITER-relevant boiling problems met only limited success [6]. 2. Geometry and mesh Fig. 1 represents the experimental test section. It also shows the location of a set of thermocouples inserted 1.5 mm below the heated surface. Fig. 2 shows a cross-section highlighting the inner dimensions of a flat channel. In the hypervapotron case, a number of teeth are machined at the top of the fluid flow channel, along with a side channel. The hypervapotron channel dimensions are given in Fig. 3, showing cross and longitudinal sections. In both cases, the top wall is heated over a length of 100 mm, the system being otherwise thermally insulated. For the flat-channel geometry, we tested a number of hexa- hedral non-uniform meshes created with the commercial mesh generator GAMBIT, ranging from 0.3 to 1.7 Mcells. We found a rela- tive variation of the computed temperature of ∼12% (using the inlet subcooling as the reference value) when the mesh size varied from 0.3 to 1.0 Mcells. This variation was reduced to ∼2% when the mesh size went from 1 to 1.7 Mcells. Then, ∼1 Mcells are a good compro- mise between reasonable grid-independence of the solution and computational cost. This size of the mesh is also comparable to that chosen in previously published works on the same subject [6,7]. For the hypervapotron, we kept the same average cell size (∼0.5 mm), but we used tetrahedral cells because of the more complex flow 0920-3796/$ – see front matter © 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.fusengdes.2012.01.024