CFD model of fluid flow in reactor : A simulation of velocity and heat distribution in a channel

The basic problem of operating a boiling water nuclear reactor (BWR) is that of maximizing the power output while avoiding fuel rod over-heating (dry-out). For the safe operation of BWRs this entails a detailed understanding of the flow of water and steam through the reactor core. In a BWR the water functions not only as a coolant, but also as a moderator for the neutrons emitted in the fission process. To describe the thermohydraulic properties of the reactor a number of parameters are of common interest. Examples of such parameters are void, pressure, temperature, water and steam velocities, pressure, sheer forces and turbulent kinetic energy. There are a few ways of revealing these values such as experiments built up to behave like a reactors and computer simulations using models based on the laws of fluid dynamics and thermodynamics. This research concerns a computer based model which uses a continuous fluid dynamic, (CFD), calculation program called OpenFOAM (Open Field Operation and Manipulation). OpenFOAM uses Navier Stokes equations for continuity, momentum- and energy conservation to simulate flows. The method used in this research has been to first build a model which describes the adiabatic flow correctly by using an already existing solver which uses the continuity and momentum conservation laws. In order to achieve a model that can solve temperature distributions in the flow the energy equation is added to the program coding in OpenFOAM. There are totally 12 turbulence models. Some of the models have not produced results on account of that they either diverged or needed input that was not attainable. The models which were tested and used were four k -ε models, one RSTM model and two low-Re models. A question that is addressed in this report is which of the many turbulence models that describes the experimental flow most correctly. The low-Re model LienLeschzinerLowRe produces the results with best congruence to the experimental data. The k -ε models model RNGkEpsilon and the RSTM model LRR were also fairly close. It is found that there are a few turbulence models that describe the experimental flows sufficiently well. LaunderSharmaKe was the turbulence model which simulated the temperature distribution best and was almost within the error bar limit of 5 K in all of the plots. It is interesting that the two low-Re models show the best results if only one characteristic at the time is studied. One of the turbulence models describes only the velocity profile well and the other one oppositely describes the temperature distribution the best. It can thereby be stated that if the user wants a turbulence model that describes both velocity profiles and temperature distribution the RSTM model LRR is the best one. If on the other hand computer capacity is a limiting factor it might be profitable to use the simpler k -ε model RNGkEpsilon. An other conclusions of this thesis is that the LRR and RNGkepsilon models are suited for the simulations of the geometries described in this work provided that the channel is wide enough for the model to simulate a correct temperature profile. With the use of gmsh a case geometry with wider channel area could easily be created. It would of course be necessary to use experimental data to validate the assumption that more realistic results can be obtained on wider channels