7.1 Whole computational domains: sym­metrical configuration.- For these configu­rations, it was observed that spatial periodicity in each cell of the slats structure for A’ < 1.053 range exists, and that the behaviour of the variables «*, u* and T* was similar to cases when parallel slats were in contact with cold isothermal wall only. But in case of pressure, the pe­riodic behaviour was not obvious. Due to that fact two additional cases considering values of A = 30.5 and 60 and preserving values of A’ in 0.9528 were computated. In these new cases, it was demonstrated that distributions have had a completely periodic behaviour. Therefore, for symmetrical configurations the periodicity behaviour exists for all variables only when the overall aspect ratio A is higher than 20. In Fig. 4 the streamlines are pre­sented for cases studied.

7.2 Whole computational domains: asymmet­rical configurations close to hot isothermal wall.- It was demonstrated that spatial periodic­ity for each cell of the slats structure for all cases proposed in Table 2 exist.

7.3 Reduced computational domains: asym­metrical configurations close to hot isother­mal wall.- In Table 9 numerical results of the Nus — selt numbers obtained on whole computational domains are compared with the Nusselt numbers calculated on reduced domains. It was observed that the percentage differences |<й/|% increases with the decrease in A’.

An additional study was performed progres­sively increasing the value of from 20 to 86.696

and maintaining the values of the other parame­ters constant. Numerical results are presented in Table 10.

8.2 Verification.- For the finest meshes the

observed order of accuracy p approaches the theoretical values of the differential scheme used (between 1 and 3). The percentage of Richardson nodes obtained was high for Th = 30 and 70°С cases, but decreases in the case Th = 12й0С, where the physical phenomena was more complex. For all cases, the GCI values decrease as n increases. These results indicate that the estimator GCI is reliable only for Tft = 30 and 70°С cases.

8.3 Experimental setup.- A schematic of the setup is shown in Fig. 6. The internal dimen­sions of the cavity are = 300mm high, L = 70mm wide and F = 500mm deep. The slats consist of 25 rectangular glass sheets (dimensions are 1.5mm thick, 40mm high and 500mm deep). The rectangular sheets are distributed in a uniform manner, and are separated from the front and back plates by air gaps of lh = lc = 15mm.

Over the cold isothermal wall, an aluminum plate; type K thermocouples; another aluminum plate; a cooling system formed by a single-pass heat exchanger; and insulation mate­rial of armaflex were placed consecutively.

The hot isothermal wall structure consisted of: an aluminum plate; type K thermocouples; glued electrical heaters; and rock wool insulation material. The cavity was closed at its sides by lids covered with insulation material on the outside. Two glass windows were mounted on one side wall (window A and B, see Figs. 6a and 7), in order to allow the visualization of the air flow in the cavity.

Control, regulation and data acquisition were carried out with a data acquisition unit managed by a software applica­tion programmed in HPVEE language. The air is seed by aerosols of olive oil generated by a Laskin nozzle. The images of the flow were captured using a Digital Particle Image Ve- locimetry (DPIV). Errors of the measured velocities due to DPIV device, the data acquisition and the post-processing were expected to be below ± 0.0014 m/s.