In order to investigate in detail the thermosyphon effects, the authors are studing a sim­plified thermosyphon system with a geometry that will permit both the construction of a controlled experimental set-up, and the modelling of the whole system with CFD techniques,

First studies on a bidimensional configuration have already been carried out. The two dimensional system considered in this first stage is shown in figure 6. The computational domain used, including details of the mesh, is also given in figure 6. The mesh has been concentrated through the collector and in the tank walls principally (see the solid triangles in

figure 6.b, where n represents the mesh characteristic parameter). The mesh used for the simulations here presented corresponds to n=2 (i. e. 50×90 control volumes).

The system has been simulated during 24 hours exposed to ideal outdoor conditions following a procedure similar to the one-day test proposed in the standards (ISO-9459). The initial time is 8 a. m. In figure 7 three different maps along the studied time period are shown. It can be seen how the velocity increases in the first hours of the simulation due to the received heat flow and, approximately at 6 p. m., when the heat flux stops, the inversion flow appears. Also, the benefits of the insulating material can be observed due to the time the tank remains at a high temperature.

The first point to consider is the degree of detail this method provides. From each time step all maps can be known and comparing the simulation results from experimental results becomes much easier, direct and efficient.

There are two main weak points: the time increment and the CPU time. The time increment that has been used is 0.1 s, value too low for long-term simulations [15] but needed to quickly converge the system at each time step, and to properly evaluate the transient phenomena. If greater time step is used, the number of inner iterations increases worsen the CPU costs, and errors due to temporal discretization increase. The one-day test on mesh n=2 has spent more than 8 days of CPU time. Obviously, this is not acceptable if a three-dimensional study has to be carried out so, in spite of the possibilities this method provides, it becomes necessary to study the way to shorten this CPU time. An alternative is the Multiblock method (or domain decomposition method, [1]) because it would avoid the simulation of the internal solid and improve the mesh distribution along each compound (block). Moreover, Multiblock methods would give the possibility to parallelise and use different CPUs at the same time. The critical point of the Multiblock method and which is currently focusing the attention of the authors is the transfer of information between subdomains in multiconnected grids (elliptical situations), as it occurs with thermosyphon systems, were solutions without physic sense can be obtained.

Conclusions

Computational fluid dynamics, CFD, offers a valuable tool to obtain local and extensive information of the fluid dynamic and thermal behaviour of thermosyphon systems.

Detailed CFD simulations of whole thermosyphon systems have not are not yet possible because the large computational resources (time and memory) required. However, CFD can be used to model components or parts of the components of the system with a rea­sonable CPU time. Furthermore, CFD simulations can also be used to obtain information which is required by other more simplified models like local Nusselt numbers or skin friction coefficients, with no need to construct expensive experimental units.

Due to the constant and fast improvement of the CFD techniques and increase of compu­tational resources, the authors expect to be able to perform 3-dimensional CFD simulations of complete termosyphon systems very soon. Main work currently carried out by the authors in this line focus on the development of efficient multiblock techniques.

Acknowledgments

This work has been funded in part by The European Commission under the “Energy, Environment and Sustainable Development” Programme, Framework Programme V, 1998­2002, project contract number CRAFT-1999-72476.

[1] etical and computational approaches. CRC Press, 1998. [6]