Wing mesh module

The wing mesh module has also gone through many evolutions in order to improve the correlation with the experimental results. The simulations by Mahon [12] have been used as a starting point, however Mahon already stated that his 3D study was still at an early stage. Initial improvements involved a larger computational domain (upstream and downstream) and the use of more structured mesh blocks, especially in the critical region underneath the wing towards the ground. Since the correlation did not improve sufficiently, other aspects had to be checked as well. Accurate measurements of the wind tunnel model revealed that the profile geometry differed slightly from the CAD design drawings. The trailing edges of both elements were thicker for the actual models than in the CAD design (1.6mminstead of 0.9mm). This was a result of the production process, during which the wing was constructed from an upper and lower carbon fibre skin, which were bonded together. The combined thickness of the skins at the trailing edge was larger than intended and the actual profile was therefore deformed as if it was stretched open at the back and rotated relative to the joining point at the nose.

One simulation based on a grid that incorporated these reversed engineering insights by adjusting the geometry in a similar way showed improved results. This grid - as well as several others during the development stage - made use of infinitely thin endplates for which the endplate inside and outside coincided in the same plane. These promising results prompted a further investigation into the existence of geometrical anomalies and the complete pressure tapped wing was 3D laser scanned with a 30,000 points cloud to obtain a CAD representation of the actual physical model. CFD simulations using these new data showed that global dimension and orientation differences, extracted from the scanned data, had much more influence than local geometrical discrepancies. The final wing mesh module has thus been based on a selection of the scanned data, capturing the essential differences compared to the design CAD data, without using all the data points. Figure 2.16 shows the final wing mesh module in the form which has been used for the presented simulations. Seven streamwise stations were created for which the scanned data was used to determine the continuous profiles at each location (the blue connectors in streamwise direction in figure 2.16); the wing surface between these stations was obtained through linear interpolation. This practice limited the complexity of the geometrical data, while discrepancies in dimensions and orientation were still captured at regular intervals. The grid was constructed in such a way that the boundary layer would be resolved on both elements as well as on all sides of the endplate. The area-weighted average of y+ _is

Research description

slightly below one on the elements and around three for the endplate. The boundary layer blocks were wrapped around the elements (see figure 2.17) and endplates (see figure 2.18) in order to save cells in the far field. The connector dimensions have been chosen in such a way that all opposite outer domains of the module feature the same amount of nodes, which allows for straightforward creation of a fully structured grid from the module.

The final wing mesh module is an accurate representation of the pressure tapped wing elements combined with the less complicated endplates of the force wing. Endplate thickness and fillet radii on the upstream, downstream and top edge of the endplate are the same as for the real force wing model. The endplate of the final wing mesh module evolved from an infinitely thin flat endplate, via a 5mmthick rectangular endplate with sharp corners into the realistic endplate with 2.5mm fillet radius. Since these changes have been made one at a time, while keeping the rest of the grid the same, it is possible to derive the influence of each step. The infinitely thin endplate baseline case underpredicted the downforce by 8.1%, adding thickness to the endplate, reduced this underprediction to 7.2% and finally using fillets on the edges brought the correlation to within 6.3% of the experimental value. The respective suction peak on the main element went from 8.0% underpredicted, to 7.9% and finally to 6.9% compared to the experimental results. The wing drag was overpredicted by 1.8% for the first case, 3.7% for the thick endplate and 2.1% for the final version. From this it can be concluded that the realistic geometry of the endplate has a noticeable positive influence on the correlation.

Finally, it needs to be mentioned that the wing mesh module incorporates laminar zones17_{for the main element and for the flap, just like in Mahon’s simulations [12]. Mahon}

performed a correlation study between oil flow results and a numerical simulation based on the Orr-Sommerfield equation to show that the transition position can be derived from the experimental oil flow data. This same reasoning and method has been used within the current research.

The laminar zones are included to model the laminar flow along the elements near the leading edges, for better correlation. The locations of these zones have been derived from oil flow visualizations by looking at the transition position. Transition in the simulations is enforced instantaneously at the boundary between the laminar and turbulent zones, in contrast to in the experiments where transition takes place over a certain interval. The

17_{Ideally a method predicting transition would be used like described by Czerwiec et al. [98], however for}

the applied solver such a function is not available and therefore laminar flow has to be modeled in zones with an instant transition into turbulent flow at the zone boundary.

Research description

laminar zones were kept constant in spanwise direction, as in Mahon’s simulations, and the length was based on that occurring in the symmetry plane. The transition regions in the oil flow experiments were found to be curving downstream towards the wing tips for the higher ride heights, but it proved difficult to include this in the grids. The length of the laminar zones were varied in the simulations depending on the wing ride height, in line with the experimental oil flow data. The computational laminar zones extended all the way upstream to the velocity inlet, which gave slightly better results than Mahon’s localized laminar zones, because the current solution meant that no upstream turbulence quantities were transported through the laminar zone, giving the turbulent boundary layer a fresh start at the end of the laminar zones.