Computational Fluid Dynamics Modeling and Analysis For the Giant Magellan Telescope (GMT)

Transcription

1 Computational Fluid Dynamics Modeling and Analysis For the Giant Magellan Telescope (GMT) John Ladd *a, Jeffrey Slotnick b, William Norby a, Bruce Bigelow c, William Burgett c a Boeing Research & Technology, The Boeing Company, St. Louis, MO, USA b Boeing Research & Technology, The Boeing Company, Huntington Beach, CA, USA c GMTO Corporation, 465 N. Halstead St., Pasadena, CA, USA ABSTRACT The Giant Magellan Telescope (GMT) is planned for construction at a summit of Cerro Las Campanas at the Los Campanas Observatory (LCO) in Chile. GMT will be the most powerful ground-based telescope in operation in the world. Aero-thermal interactions between the site topography, enclosure, internal systems, and optics are complex. A key parameter for optical quality is the thermal gradient between the terrain and the air entering the enclosure, and how quickly that gradient can be dissipated to equilibrium. To ensure the highest quality optical performance, careful design of the telescope enclosure building, location of the enclosure on the summit, and proper venting of the airflow within the enclosure is essential to minimize the impact of velocity and temperature gradients in the air entering the enclosure. High-fidelity Reynolds-Averaged Navier Stokes (RANS) Computational Fluid Dynamics (CFD) analysis of the GMT, enclosure, and LCO terrain is performed to study (a) the impact of either an open or closed enclosure base soffit external shape design, (b) the effect of telescope/enclosure location on the mountain summit, and (c) the effect of enclosure venting patterns. Details on the geometry modeling, grid discretization, and flow solution are first described. Then selected computational results are shown to quantify the quality of the airflow entering the GMT enclosure based on soffit, site location, and venting considerations. Based on the results, conclusions are provided on GMT soffit design, site location, and enclosure venting. The current work is not used to estimate image quality but will be addressed in future analyses as described in the conclusions. Keywords: GMT, Telescope, Enclosure, Los Campanas Observatory (LCO), Soffit, Magellan, CFD, RANS 1. CFD MODELING 1.1 Introduction Goals of this work are to utilize CFD to understand the sensitivity of the flow characteristics affecting the ingestion of low-level air from the terrain with design considerations such as site location, enclosure design type (i.e. open vs. closed soffit), and enclosure venting configurations. The air flow is considered to be a viscous ideal gas with no temperature variation as detailed in Section 1.4. The Reynolds-Averaged Navier Stokes (RANS) equations are solved via a pressure-based algorithm and all cases are computed using the same procedure so that meaningful comparisons are made. In the CFD analysis procedure, geometry for the desired configuration is assembled and processed in order to obtain a suitable computational mesh (or grid) for the numerical analysis. Once the grid is generated with the desired resolution and quality, the flow solution is obtained by executing the flow solver software using an appropriate number of computer processors, or cores. Details of each of these elements of the process are further discussed below. 1.2 Geometry For the GMT CFD analysis, the computational model is composed of a simplified geometrical representation of the telescope primary and secondary mirrors, enclosure, and surrounding terrain. Including a high level of geometric detail is desirable but must be weighted with the corresponding increase in computational time (cost) to obtain the flow solution. Boeing engineers worked closely with the Giant Magellan Telescope Organization (GMTO) to include geometric features of interest yet exclude components considered too small and insignificant to the resulting flow field for this initial study. The telescope components as designed and modeled in the CFD simulations are described in the following sections. *john.a.ladd@boeing.com phone: boeing.com Modeling, Systems Engineering, and Project Management for Astronomy VII, edited by George Z. Angeli, Philippe Dierickx, Proc. of SPIE Vol. 9911, SPIE CCC code: X/16/$18 doi: / Proc. of SPIE Vol

2 1.2.1 Telescope A side and back view of the GMT installation showing key geometric dimensions is depicted in Figure 1a and 1b. The telescopee center is defined to be 10.7 meters from the observing floor, and a total of 22.5 meters from ground level m on Axis -, 107m ng Floor m (a) Side view 1 Level- (b) Back view Figure 1. Views of the GMT installation and CFD telescope model Figure 2. CFD model of telescope and supports A simplified telescope Computer Aided Design (CAD) geometry definition in the STandard for the Exchange of Product model data (STEP) format was obtained from GMTO as shown in Figure 2. Essential features include a representation of the seven primary mirror segments and actuator housings as connected slab hexagons, the secondary mirror as a single thick hexagon, and the main truss support structure underneath the telescope. The bottom of the GMT model is positioned on the flat observing floor inside the enclosure. The telescope elevation angle in the current work was kept at a nominal value of 60 degrees Enclosure The simplifications of the enclosure models for the GMT for both the open and closed enclosure base soffit configurations are shown in Figure 3. The original models, Figure 3a and 3b, were simplified first by omitting the truss framework within the enclosures and below the open soffit enclosure as shown in Figures 3c and 3d. Next, the fully open venting pattern omitted the thinner vertical support members between the top and bottom of each horizontal vent opening, shown in Figure 4a and 4b, as they are not considered to be significant for the current flow simulations. (a) Open soffit CAD model (b) Closed soffit CAD model Proc. of SPIE Vol

3 CEEC= CC 111 (c) Open soffit simplified CAD model (d) Closed soffit simplified CAD model Figure 3. GMT enclosure soffit configuration comparison (a) Original vent openings (open soffit) (b) Simplified vent openings (open soffit) Figure 4. Nominal GMT venting patterns SUBJECT: AVAILABLE PROJECT CONTOURS DATE', 09N MODEL RESOLUTION - PROVIDE 2.0m CONTOURS WITHIN 5 DIA. OF ENCLOSURE - PROVIDE 10.0m CONTOURS FOR REMAINING AREA PREDOMINANT WIND DIRECTION ;. n,)r-! AVAILABLE SURVEYED CONTOURS CENTER OF GMT TELESCOPE PIER PREVIOUS SITE MODEL AREA (965m X 845m) PROVIDED TO BOEING NEW SITE MODEL AREA - 2,000m X 2,000m (CENTERED ON TELESCOPE PIER) TO BE PROVIDED TO BOEING SCALE IN METERS Figure 5. Topographic map of LCO and identification of GMT CFD surface domain Proc. of SPIE Vol

4 1.2.3 Terrain The terrain model used in the computational analysis is provided by M3 Engineering as shown in Figure 5. This topographic image illustrates a 2000 square meter section of the LCO, roughly centered on the GMT enclosure, which was extracted for use in the CFD model. To enable accurate terrain modeling in the CFD simulations, appropriate surface sampling resolutions are chosen to adequately represent the features of the terrain around the enclosure. On the summit, within approximately 5 enclosure diameters around the telescope, surface features of 2 meters in length or greater are captured. Everywhere else, surface features of 10 meters in length or greater are captured. The terrain geometry file was originally obtained as a Tagged Image File Format (tiff) file with color mapped to elevation. The tiff file was converted in AutoCad to a Civil 3D drawing (dwg) file, and was then made available as a STEP file as depicted in Figure 6. The blue arrow in this figure denotes the approximate location of the telescope on the summit. Due to the large number of small triangular surface entities (~152,000) present in the file, direct processing of this model for grid generation could not be completed with standard geometry manipulation tools. As an alternative, a separate software tool was written to generate a regular ordered surface definition at a similar resolution. This was accomplished by computing the intersection curve of a vertical planar surface with the terrain surface and systematically marching in 4 meter increments from one boundary to the opposite side of the domain. For each intersection line, points were then redistributed such that each line contained points that were equally spaced 4 meters apart. The end product of this processing was a regular lattice of points that contained cells of 4 square meters, as shown in Figure 7. The yellow cylinder indicates the approximate location of the baseline telescope location on the summit. To ensure that the choice of 4 meters was sufficient to accurately capture the surface features of the terrain, the surrogate lattice surface was overlaid onto the original CAD surface, as seen in Figure 8. As depicted, the surrogate model retains the key geometry features of the original terrain model, including the edges of the summit where the telescope/enclosure will be located. Figure 6. Original surface definition of the LCO terrain Figure 7. Surrogate geometry model of LCO terrain Figure 8. Overlay of LCO surface (red) with surrogate model (blue) Proc. of SPIE Vol

5 1.3 Grid Generation The Boeing-developed Modular Aerodynamic Computational Analysis Process (MADCAP) code is used to first combine the telescope, enclosure, and terrain geometry models into one single geometry model, as shown in Figure 9. Once the geometry of all the components is assembled, the resolution of the computational grid discretization required for an accurate CFD analysis is determined. The cell spacing on the surface mesh are tailored depending on their location as described below. Figure 9. Combined GMT, enclosure, and terrain models 1. Surface grid cell edge lengths on the enclosure are held to 1 meter everywhere except for 0.25 meters near the vent openings to ensure generation of a minimum of 2 cells across the thickness of the opening. 2. Cell edge lengths are specified at 0.25 meters over the entire telescope, including support truss, so that there are at least 4 to 5 cells across the thickness and width of support members. 3. Cell spacing is held to 1 meter along the perimeter of enclosure intersection with site terrain summit surface. 4. From the edges of the summit, cell spacing is allowed to grow along the terrain surface to 20 meters at the outer boundaries of the computational domain. 5. In the direction normal to the terrain surface, the first computational cell spacing is 3 millimeters, which provides a minimum of 25 cells in the velocity boundary layers. With these grid rules, MADCAP is used to generate a high-quality unstructured tetrahedron surface mesh. Selected images of the surface grid and a cutting plane near the vent slots for the closed soffit enclosure configuration are shown in Figure 10. Once the surface mesh is generated, a volume mesh to fill in the computational domain, as depicted in Figure 11, is computed using the Advancing Front Local Reconnection (AFLR) code developed by Mississippi State University 1. This technique utilizes the orthogonal properties of a prism layer near the surface which transitions to all tetrahedron cells away from the boundaries. The grid cells grow significantly in size as the distance away from the enclosure increases since flow gradients vanish and free-stream flow properties are recovered near the outer boundaries. The nominal grid size used in the current studies totals approximately 16M cells. Like many ground based telescopes, the GMT design will incorporate a variable height, semipermeable windscreen over the main enclosure opening to help shield the telescope from excessive wind velocities and flow gradients that would otherwise be present. A windscreen was modeled in the CFD simulations using both a porous-jump internal boundary condition (BC) as well as a discrete slab windscreen with circular holes. Both of these techniques are briefly described next. Proc. of SPIE Vol

6 J 2 f :» \. Ç /If!: Figure 10. Selected images of the GMT computational grid for the closed soffit configuration (a) Entire CFD domain (b) Close-up of telescope/enclosure on summit Figure 11. Volume mesh for CFD simulations Proc. of SPIE Vol

7 To model the windscreen using the porous wall BC in the flow solver, a curved grid plane is defined and embedded in the volume grid system just inside the front opening of the enclosure, as depicted in Figure 12. Grid points are clustered ahead of, and behind, the embedded surface to properly compute the effect of the windscreen on the flow entering the enclosure, as shown in yellow. The geometric relationship between the embedded grid surface and the telescope is depicted in Figure 13. The height of the embedded plane is 33 meters, which is the maximum height of a screen which will not intrude into the optical path of the telescope at an elevation of 60. For the current study, a 25% porous screen was assumed which is correlated to the momentum loss coefficient through empirical data for thin screens with round holes. As will be shown in Section 2, the primary effect of the screen is to block/deflect the incoming flow leading to reduced velocities and increased pressures on the upstream side...- `1`:..:-!:! Figure 12. Embedded grid for porous wall BC Figure 13. Relation of windscreen surface with telescope seeing path Figure 14. Relation of windscreen surface with telescope seeing path For the modeling of the windscreen using the discrete CAD definition, grid clustering near the holes is illustrated in Figure 14. This level of resolution is chosen to minimize the number of grid points and computation time for the CFD simulation while capturing the key effect of flow blockage on the air entering the enclosure. The addition of the discrete windscreen model increases the size of the computational grid by a factor of 3. However, this direct method should be used for irregular porosity concepts since the porous wall boundary condition uses empirical relationships based on small sharp-edged circular holes. 1.4 Flow Solution Boundary Conditions and Domain The Ansys Fluent flow solver software package 2 is used to perform the CFD analysis. Fluent has a variety of modeling capabilities including fluid flow, heat transfer, and reacting gas chemistry. The computational grid system generated with AFLR is first converted to a Fluent Case file (.cas). Once the grid is converted and read into the Fluent code, initial and boundary conditions are set. Because of the very low speed (< 15 m/s) of the air flow for the GMT analyses, the flow is treated as incompressible (no change in density). A pressure- based solver is used for the momentum equations. In the current analysis, the energy equation not solved and temperature is constant. The pressurevelocity coupling scheme used is the Semi-IMplicit Pressure-Linked Equations (SIMPLE) approach. The momentum equations are solved using a 2 nd -order upwind spatial integration. The flow is assumed fully turbulent, and the twoequation k-omega Shear Stress Transport (SST) model is used to provide predictions of the turbulent viscosity. Further details on the solution procedure are available 3. The computational domain including specification of boundary conditions is shown in Figure 15. The domain edges are oriented along the primary wind direction so that the boundary conditions on the lateral side planes can be set as inviscid and impermeable. The velocity at the inflow plane is set perpendicular to the boundary. The downstream boundary is set to free stream pressure (zero gauge pressure) for all cases. The choice of the velocity distribution used as the inflow boundary condition is guided by similar relationships used in wind tunnel testing 4. A power-law distribution is used to relate the local and a reference velocity (U and U ref ) to the local and reference height (Z and Z ref ): = (1) Proc. of SPIE Vol

8 For the GMT analysis, a terrestrial velocity profile with exponent n=0.16, appropriate for rough open flow, is assumed. Based on the Reference 4 wind tunnel data, a reference height (Z ref ) of 500m was chosen. Choice of the referencee velocity (U ref f) is guided by wind speed measurements at the Cerro Los Campanas summit 5. The wind speed is defined as percentiles of velocity magnitude data collected from a measurement location approximately 10 meters above grade on the north end of the LCO summit as depicted in Figure 16. For this study, the 75th percentile wind speed from night measured data is 9.8 m/s. To achieve this velocity, a value of U ref at the upstream boundary is initially assumed. Using this U ref, the Fluent CFD code is then run to obtain the velocity profile along the tower axis including the 10m elevation point location. Two iterations were required to determine that an inflow velocity profile with U ref =7.3 m/s at a Z r ref=500 meters (Figure 17) results in a velocity of approximately 10 m/s at the tower location as desired (Figure 18). For the 25 th percentile wind speed of 4.0 m/s, a value of U ref =3.0 is required. Vmag (n io ni a Figure 15. Fluent boundary conditions for CFD simulations Figure 16. Location of LCO summit velocity measurement -0-Tower Locatic U, = 7.3 7f = 500 -IMIKEILMCCE071E Velocity (m/s) = 9. Zw.=10./ O U = 9.8 1ol o oax o ( Velocity (m /: sl Figure 17. Inflow and tower velocity profiles Figure 18. Enlarged view of tower velocity profile A key physical phenomenon that is important in the CFD analysiss is the acceleration of the flow from the inflow boundary along the terrain up to the summit. A near-surface velocity deficit in the inflow profile of ~ 4 m/s (at 10m above terrain) increases to a near-surface velocity excess of ~10 m/s (at 10m above terrain) at the tower location on the Proc. of SPIE Vol

9 summit as shown in Figures 17 and 18. This occurs because the rising terrain slope contracts and accelerates the flow, reaching a maximum above the summit. 2. CFD RESULTS Results of the CFD simulations to assess the effect of the soffit configurations, location of the telescope on the summit, and the presence or absence of the windscreen on the airflow entering the enclosure are presented here. Simulation results are reviewed primarily for the 75 th percentile head wind velocity and direction (azimuth=20 ) but also the telescope looking 90 degrees counterclockwise (right cross wind, azimuth = 110 ) for both open and closed soffit enclosure configurations as depicted in Figure 19. The 75th percentile wind speed from night measured data is 9.8 m/s. The full open venting pattern (as shown in Fig. 4b) is used. The effect of the soffit and trends in flow with the telescope at alternate locations is presented followed by the effects of the windscreen and alternate venting configurations. Primary Wind Direction N Primary Wind Direction N (a) Head-wind (b) Right-cross-wind Figure 19. GMT orientations for nominal site location used in Task Effect of Soffit at Baseline Site Flow Characteristics Analysis of the CFD simulations is performed to better understand the relevant flow physics that drive the effects of soffit configuration on the quality of the airflow entering the enclosure. The CFD post-processing code Fieldview from Intelligent Light is used to process all CFD simulations. 13i 0 --NWALnQ1JWl0 0 Ñ c VIOk \ A P!-NH:4P19":-'5 P ;vv.; A ' (a) Closed soffit (b) Open soffit Figure 20. Comparison of velocity and flow streamlines for soffit configurations, 75 th percentile headwind, porous windscreen Proc. of SPIE Vol

10 The direct effect of the soffit geometric configuration on airflow approaching the enclosure (left to right) is shown in Figure 20 for the 75% headwind case. In this comparison, velocity vectors in the lateral centerplane of the enclosure are shown and colored according to the local velocity magnitude. The vertical black line on the windward side is the location of the porous jump boundary condition plane. While the velocity magnitudes within the enclosure are similar between the solutions of the two soffit types, there is a large recirculating region above the summit just in front of the aperture for the closed soffit configuration not present in the open soffit. This is the center of the so called necklace vortex resulting from the stagnation of the flow on the blunt cylindrical enclosure since the vortex hangs around the obstacle as a necklace as shown as the red streamlines in Figure 21a for the closed soffit configuration. As will be shown, it is this region that elevates more near-surface air up and into the enclosure compared to the open soffit. For the open soffit, as shown in Figure 21b, the flow still stagnates on the smaller support cylinder under the enclosure, however the flow is also contained on three sides under the enclosure which tends to break up this rotational behavior and the flow more uniformly passes around the smaller cylindrical support column. (a) Closed Soffit (b) Open Soffit Figure 21. Streamlines patterns from different soffit configurations The computed pressure contours in the same center cutting plane for the two soffit types are shown in Figure 22. The pressure is lower in the center of the necklace vortex compared to the surrounding flow as expected. While this low pressure region is not seen in the open soffit configuration, there is a flow expansion over the overhanging windward edge which drops the pressure considerably. The average pressure within and above the open soffit enclosure is seen to be approximately 5 Pa lower than for the closed soffit. The open soffit provides lower flow blockage than the closed soffit and as a result the surrounding velocities are slightly higher and the pressure slightly lower. (Pa) Ó.Ó P9 (Pa) (a) Closed soffit (b) Open soffit Figure 22. Comparison of gauge pressure on enclosure symmetry plane as a function of soffit type baseline site, 75 th percentile wind speed, head-wind, porous windscreen, vents full open Proc. of SPIE Vol

11 2.1.2 H min Analyses For the current GMT flow analysis, solution of the energy equation to compute temperature gradient effects was not included in the CFD simulations as mentioned previously. However, the heating of the air mass on the terrain that surrounds the telescope and enclosure on the LCO summit, and the flow of that air into the enclosure, is of primary concern. In lieu of a computed temperature gradient, the key metric used in this analysis is the minimum height above the terrain (H MIN ) that a flow particle entering the enclosure would experience as it traveled from the inflow boundary to the enclosure. Figure 23 graphically illustrates the definition of this parameter as well as the height above grade (H AG ) on the summit. Figure 23. Definitions of streamline heights relative to enclosure A more quantitative analysis of the CFD simulations is performed to characterize the flow entering the enclosure. As mentioned previously, minimum height above terrain, H MIN, is definedd as a proxy for temperature gradient. Tracking H MIN is used to estimate the amount of undesirable air entering the enclosure. Also, it is assumed that air closer to the ground will exhibit larger temperature gradients than air further above the terrain. To this end, the flow at points at the enclosure entrance, definedd along lines at various heights above grade H AG (as defined in Fig 23) ), is characterized by tracing streamlines upstream (backwards in time) to the domain inflow boundary and computing the H MIN along each streamline path. An illustration of the lattice of release points (shown for the enclosure with the closed soffit) is given in Figure 24. Figure 24. Streamline release points at opening of enclosure Proc. of SPIE Vol

12 The streamlines are then traced backward in time so that their trajectory can be analyzed and their minimum distance above the terrain computed. With a value H MIN determined for each release point, a contour plot of values can be constructed from the aperture array as shown in Figure 25. A point is removed from the plot (represented by white regions) if that streamline cannot be integrated upstream because of reversed flow in these locations. The flow entering the enclosure with the closed soffit originates from locations closer to the terrain (darker blue/black contours) than does flow entering the enclosure with the open soffit. Hmin (m) mi1 I 1111! r Hmin (m) I NI (a) Closed soffit (b) Open soffit Figure 25. H MIN contours at enclosure entrance baseline site, 75 th percentile wind speed, head-wind, porous windscreen, vents full open This trend illustrated in the bar chart in Figure 26. In this plot, the percentage of flow from a given H MIN value is grouped in the following buckets : 0-8m, 8-16m, 16-32m, and 32-50m. This data quantitatively shows that the effect of the closed soffit configuration results in more of the lower level air entering the enclosure as compared to the open soffit configuration, for both the enclosure with and without the windscreen modeled. Specifically, of the total amount of air entering the enclosure with the windscreen, 17% of the flow for the open soffit originates between the ground and 16m above terrain. For the closed soffit, 32% of the flow originates between the ground and 16m above terrain. Air that originates from between 16m and 32m represents 67% of the total air mass entering the enclosure with the open soffit and 53% of the total air mass entering the enclosure with the closed soffit. In contrast, of the total amount of air entering the enclosure with no windscreen, 19% of the flow for the open soffit originates between the ground and 16m above terrain. For the closed soffit, 27% of the flow originates between the ground and 16m above terrain. However, air that originates from between 16m and 32m represents only 42% of the total air mass entering the enclosure with the open soffit and 34% of the total air mass entering the enclosure with the closed soffit. This data further corroborates the lifting effect of the flow, particularly in the range of H MIN from 16m-32m, due to the presence of the windscreen. % Flow Entering Enclosure Open Soffit, Windscreen Off Closed Soffit, Windscreen Off Open Soffit, Windscreen On Closed Soffit, Windscreen On 0 to 8m 8 to 16m 16 to 32m 32 to 50m H MIN Buckets Figure 26. Amount of flow entering enclosure sorted by H MIN buckets as a function of soffit type and windscreen baseline site, 75th percentile wind speed, head wind, vents full open Proc. of SPIE Vol

13 A quantitative comparison of the magnitude of H MIN at the enclosure opening between simulations with both open and closed soffit configurations with and without the windscreen modeled is shown in Figure 27. These simulations are performed at the baseline site with the 75 th percentile winds and the enclosure positioned in the head wind orientation with vents full open. In this plot, the minimum distance above the terrain is computed for each line of release points, given as height above grade (H AG ), at the enclosure opening. The data shows that the flow for both closed soffit configurations originates from closer to the terrain as compared to the open soffit cases. Additionally, as mentioned above, the effect of the windscreen tends to lift more of the lower level air higher vertically into the enclosure opening starting around 26m height above ground. Height Above Grade (m) Open Soffit, Windscreen Off Closed Soffit, Windscreen Off Open Soffit, Windscreen On Closed Soffit, Windscreen On Minimum Distance Above Terrain H min (m) Figure 27. Comparison of the distribution of H MIN at the enclosure opening as a function of soffit type and windscreen baseline site, 75th percentile wind speed, head-wind, vents full open 2.2 Effect of Telescope/Enclosure Location on Summit Terrain-Only CFD Simulation Results To guide selection of alternate locations for the telescope on the summit, CFD analysis was conducted for the 75% wind condition without any telescope or enclosure present. Contours of computed surface pressure variations and nearsurface oilflow are shown in Figure 28. Regions of flow expansions, such as around the sharp summit edge, are seen as blue/black contours. Regions of higher relative pressure are shown as yellow/red contours. The near-surface streamlines indicate straighter approach flow on the windward and leeward side of the eastern portion of the summit. Results from this analysis were used to propose two alternate locations (with the telescope and enclosure) as shown in Figure 29. Computed velocity magnitude contours and streamlines near the top of the summit are shown in Figure 30 at each of the three examined site location with the flow direction from left to right. The outline of the enclosure is shown for reference but was not included in the CFD simulations. The computed data show how the upward sloping terrain near the windward summit edge for the baseline location causes the flow to decelerate considerably compared to similar locations at the alternate sites. The approaching flow for the Alternate #2 site is seen to be the smoothest in terms of terrain profile, streamline pattern, and velocity gradients although also having the largest peak velocity near 12 m/s as shown in Figure 30c. Proc. of SPIE Vol

14 V N Baseline Site Figure 28. Computed terrain surface pressures and oilflow lines Figure 29. Locations of baseline and alternate site locations on summit ":- (a) Baseline (b) Alternate #1 (c) Alternate #2 Figure 30. Terrain-only velocity magnitude contours at site locations Flow Characteristics at Alternate Site Locations In addition to analyzing the telescope and enclosure at the northwest side of the summit (baseline), two alternate sites were also analyzed with CFD to understand this sensitivity. The Alternate #1 was approximately one enclosure diameter from the Baseline while the Alternate #2 location was centered on the southeast end of the summit as shown in Figure 29. The 75% headwind condition was assumed for all simulations and the windscreen was not modeled. Proc. of SPIE Vol

15 Closed Soffit (a) Baseline Site (b) Alternate #1 Site (c) Alternate #2 Site Figure 31. Velocity vectors in the lateral mid-plane of the enclosure, 75 th percentile headwind, no windscreen Despite the different site locations, the flow stagnation into the enclosure base from the simulations with the open soffit at all three site locations is similar in shape and magnitude as shown in Figure 31. Inside the enclosure, however, the velocity distribution across the primary mirror shows slightly different trends. In all three cases, the flow entering the bottom of the enclosure impinges on the base of the primary mirror causing a local region of recirculating flow. At the baseline location, some of the lower air flows over the primary mirror generating a thicker boundary layer as compared to the predicted flow at the alternate locations. From the Alt1 and Alt2 simulations, the flow entering at the bottom of the enclosure turns more towards the floor of the enclosure resulting in higher velocity air flowing over the primary mirror, as well as some flow being diverted down and around the backside of the mirror. In contrast, the simulations with the closed soffit show significantly different flow characteristics ahead of and inside the enclosure. The size of the region of flow stagnation at the base of the closed soffit decreases significantly based on the simulations at the Alt1 and Alt2 locations as compared to the baseline location. As a result, there is less of a lifting effect of the flow into the enclosure at the Alt1 and Alt2 locations, which diminishes the area of flow recirculation above the primary mirror in the optical path. In fact, this data suggests that the closed soffit velocity profile along the beam axis for the Alt2 location is only slightly less favorable as compared to the velocity profile computed for the open soffit configuration at the baseline location Hmin Analyses at Alternate Sites A quantitative comparison of the magnitude of H MIN at the enclosure opening for simulations with the 75th percentile winds for both open and closed soffit configurations, with the enclosure in the head-wind orientation positioned at the three site locations, is shown in Figure 32. In this plot, the minimum distance above the terrain is computed for each line of release points, given as height above grade (H AG ), at the enclosure opening. The data shows that the flow for all closed soffit configurations originates from closer to the terrain as compared to the open soffit cases. Near the bottom of the enclosure opening, at the elevation where data for all six simulations is available (H AG =21m), the lowest value of H MIN is computed as approximately 7m for the closed soffit at the baseline location. The highest value of H MIN is computed as approximately 15m for the open soffit configuration at the Alt1 location. The difference between the highest and lowest value is approximately 8m, which is more than double the closed soffit value. This data suggests that the location of the enclosure on the summit has a larger effect on the flow entering the enclosure for the closed soffit configurations than it does for the open soffit configurations. It is also noted that there is an 11% difference in H MIN between the simulation with the open soffit configuration at the baseline site location and the closed soffit configuration at the Alt2 site location. Although the uncertainty in the computational results is not known, this difference suggests that the closed soffit configuration at the Alt2 site location may provide similar flow quality inside the enclosure compared to Proc. of SPIE Vol

16 the open soffit at the baseline site location. This conclusion is also supported by the data in Fig. 31 which shows differences in velocity magnitude over the mirror along the beam axis between these two cases appear to be small. Height Above Grade (m) Minimum Distance Above Terrain Hmin (m) Figure 32. Comparison of the distribution of H MIN at the enclosure opening at GMT site locations 75 th percentile wind speed, head-wind, no windscreen, vents full open In Figure 33, the percentage of flow from a given H MIN value is grouped in the following buckets : 0-8m, 8-16m, 16-32m, and 32-50m, consistent with the analysis presented in Fig. 24. This data quantitatively shows that the effect of the closed soffit configuration results in more of the lower level air entering the enclosure as compared to the open soffit configuration. Further, the smallest amount of low-level air entering the enclosure for the open soffit occurs at the Alt1 site location, while the smallest amount of low-level air entering the enclosure for the closed soffit occurs at the Alt2 site location. Specifically, 19%, 7%, and 10% of the total flow entering the enclosure with the open soffit configuration at the baseline, Alt1, and Alt2 site locations, respectively, originates from between 0 and 16m above the terrain. In contrast, 30%, 24%, and 18% of the total flow enters the enclosure with the closed soffit configuration at the baseline, Alt1, and Alt2 site locations, respectively. In general, there is no significant difference between simulations by soffit type or site location on the amount of flow entering the enclosure from between 16m and 32m above terrain. Also, small amounts of near-ground layer air from heights between 0 and 8m are predicted to enter the enclosure for the closed soffit configuration only at the baseline site location, but not at the Alt 1 or Alt2 site locations. Finally, more of the air originating from heights of 32m above terrain enters the enclosure for the open soffit as compared to the closed soffit. % Flow Entering Enclosure Baseline - Open Soffit Baseline - Closed Soffit Alternate #1 - Open Soffit Alternate #1 - Closed Soffit Alternate #2 - Open Soffit Alternate #2 - Closed Soffit 0 Baseline Open Baseline Closed Alternate 1 Open Alternate 1 Closed Alternate 2 Open Alternate 2 Closed Fil i EI 11 i i 0 to 8m 8 to 16m 16 to 32m 32 to 50m H MIN Buckets Figure 33. Amount of flow entering enclosure sorted by H MIN buckets as a function of soffit type and site location 75 th percentile wind speed, head-wind, no windscreen, vents full open Proc. of SPIE Vol

17 2.3 Effect of Telescope/Enclosure Orientation (a) 75% Headwind (b) 75% Tailwind... ::: ta..a t (c) 75% 90-Degree Crosswind (d) 75% 45-Degree Crosswind Figure 34. Comparison of flow vectors colored by velocity magnitude in plane parallel to flow direction through enclosure center as function of telescope pointing baseline site, closed soffit, 75th percentile winds, porous windscreen, vents full open Numerous CFD simulations were conducted for the 75th percentile wind at several pointing directions of the closedsoffit telescope with the initial ventilation scheme fully open and a porous windscreen model included. Velocity vectors colored and scaled by total velocity magnitude from the CFD simulations for each of the four enclosure orientations, in the plane through the mirror center point aligned with the primary flow direction, are presented in Figure 34. For the head-wind case, the flow deceleration approaching the windscreen (represented by thin black vertical line) is evident. Similarly, for the tail wind case, flow stagnation into the back enclosure wall is clearly seen since this vent configuration terminates toward the rear centerplane. For the right cross wind case, flow directly enters all four open vent levels, and high velocity air is shown streaming over and around the mirror with a considerable recirculation region directly over the primary mirror. For the right 45 cross wind case, the flow stagnates on the outside of the enclosure door facing the primary wind direction. For this simulation, flow enters the enclosure on the left and exits the enclosure on the right similar to the right cross wind case, albeit at a lower overall velocity magnitude. 2.4 Effect of Telescope/Enclosure Venting Configuration E É Ai t.k'9tipf'p --.R\ R a td1r (a) Vents full open (b) Video1, VRCW1 venting configuration Figure 35. Comparison of velocity vectors in a plane parallel to flow direction through enclosure center as function of venting configuration baseline site, closed soffit, 75th percentile winds, right cross wind, porous windscreen Proc. of SPIE Vol

18 For the right cross-wind telescope orientation, improved venting configurations are explored in order to improve the quality of the flow in and around the primary and secondary mirrors inside the enclosure. Velocity vectors in the plane through the mirror center point aligned with the primary flow direction, for the right cross wind orientation with two venting configurations (full open and VRCW1) are presented in Figure 35. The simulations are obtained at the baseline site location with 75th percentile winds for the closed soffit enclosure configuration with windscreen model. The VRCW1 venting pattern closes the level 1 and 2 vents on the right windward side of the enclosure as seen in Figure 35b. The effect of closing these vents results in significantly improved flow quality near the primary mirror surface and lower velocities in the region of the secondary mirror. Closing of the level 3 vent on the right side of Figure 35b would likely further reduce the flow gradients between the primary and secondary mirrors. 2.5 Time-Accurate Flow Modeling All computed results presented thus far have been from RANS, or steady-state flow solutions. The solutions obtained from this approach, when sufficiently converged, provide adequate predictions of basic flow characteristics of the various configurations. This type of analyses has been useful for top-level design considerations such as site location, soffit type, and ventilation configurations. However for more accurate and detailed flow predictions of more evolved telescope designs, the time-dependent approach should be utilized. In particular this allows estimations of unsteady pressure loads possible which help size the structural elements to provide the desired stiffness needed for the optical performance of the telescope. If the energy equation is active and temperature effects are included, the time varying thermal behavior of the enclosure with the chosen ventilation scheme can also be investigated. To exercise this analysis mode in the FLUENT program, a test case was computed of the closed-soffit configuration with a 75th percentile 90-degree crosswind case and the windscreen model included. The turbulence model employed is the Detached Eddy Simulation (DES) model that directly resolves turbulent eddies away from viscous walls and returns to the RANS model near the wall where direct turbulent behavior cannot be resolved without adding extremely high grid resolution and resorting to the costly Large Eddy Simulation in these regions. The same computational grid that was used for the previous RANS analyses was used for this initial time-dependent simulation but typically the mesh is refined to properly resolve the variety of turbulent length scales in the problem. The time step chosen for the simulation was 0.01 seconds based on the 10 m/s nominal flow velocity and smallest grid sizes above the telescope near 15 cm. Full solution saves were made every 20 time steps or every 0.2 seconds to create animations such as isocontours of qcriterion of which a single frame is shown below in Figure 36. The q-criterion is defined as the second-invariant of the deformation tensor and is written as Q = 0.5(W*W - S*S) (2) where W is the vorticity magnitude and S is the mean rate-of-strain. This allows an effective visualization of the turbulent flow structures by selecting a small positive value for the contour and coloring according to a scalar function, in this case static gauge pressure. The figure reveals good resolution of small turbulent structures near the telescope where the grid cells are as small as 15 cm but poor resolution of the structures near areas of large (1m) grid spacing (as expected) such as near the windward edge of the sharp enclosure doors at the top of the enclosure. Figure 36. Isocontours of q-criterion colored by gauge pressure baseline site, closed soffit, 75th percentile winds, right cross wind, porous windscreen Proc. of SPIE Vol

19 As the flow solution evolves in time, statistical parameters such as time-mean and Root-Mean-Squared (RMS) quantities can be computed such as surface pressure shown in Figures 37a and 37b respectively for the telescope and floor surfaces. These quantities are based on approximately 60 seconds of simulated time. While the variance in average surface pressure on the primary mirror is relatively small at 5 Pascals, this variance has a bearing on the design of the adaptive optics controlling the primary mirror shaping. The RMS pressure variations in Figure 37b show elevated levels near the edge of the primary mirror and even higher levels on the secondary mirror (white contours) indicating the largest pressure fluctuations are in these areas. The unsteady pressure environment on the secondary mirror is perhaps one of the most important design considerations since its stability is paramount to optical performance. (a) Average Surface Pressures (b) RMS Surface Pressures Figure 37. Computed contours of time-averaged telescope surface pressures and RMS surfaces pressures baseline site, closed soffit, 75th percentile winds, right cross wind, porous windscreen 3. FUTURE WORK After initial successful RANS analyses of the most major design considerations for the GMT, future work will focus on higher fidelity time-dependent analyses using a more detailed telescope model as shown in Figure 38. Unsteady CFD analyses including these detailed components will provide much needed data for understanding structural design requirements. Inclusion of the energy equation will provide spatial and temporal predictions of temperature and density variations to aid design of thermal management techniques and ultimately the predicted optical performance of the GMT using Optical Path Integration (OPD) procedures used routinely for aircraft turret analyses to determine optical performance. Figure 38. Computer Aided Design (CAD) model of detailed primary and secondary mirror components for CFD Proc. of SPIE Vol

20 CONCLUSIONS CFD analysis of the GMT and enclosure is performed to study the impact of the enclosure soffit design on the quality of the air flow entering the enclosure. Simulations for 75 th percentile winds with the telescope pointing into the wind, away from the wind, pointing 90 degrees to the wind, and pointing 45 degrees in to the wind for both closed and open soffit configuration concepts were completed. Since temperature gradients in the flow are not directly computed in this CFD simulation, the minimum height above terrain is defined as a proxy to characterize the quality of air as undesirable entering the enclosure. Both qualitative and quantitative analyses of the CFD results indicate that more of the undesirable quality air enters the enclosure with the closed soffit than enters with the open soffit, a trend that was seen regardless of GMT site location. This is due to an increased area of flow stagnation at the base of the enclosure with the closed soffit, which pushes more of the lower-elevation air into the enclosure opening. The smallest amount of lowlevel air entering the enclosure for the open soffit occurs at the Alt1 site location, while the smallest amount of low-level air entering the enclosure for the closed soffit occurs at the Alt2 site location. The effect of venting configurations is predicted to be most significant for the 90-degree and 45-degree crosswind cases where the approach flow compresses against the enclosure and air is forced through the ventilation slots and flows above and below the primary mirror. Closing the lowest level L1 and L2 vents (VRCW1 venting) provides a more favorable flow field for the cross-wind cases by causing air at the higher L3 location to exit out the top of the enclosure rather than directly over the telescope. Velocity gradients and flow recirculation over the telescope are also reduced significantly with the VRC1W1 venting scheme. Future work will focus on time-accurate CFD simulations using higher-fidelity geometry modeling and more refined grid resolution to be able to predict figures of merit directly applicable to aiding the design of many of the GMT components including primary mirror actuation, secondary mirror stability, enclosure and door unsteady loading, and windscreen considerations such as thickness, porosity, and deployment schemes for different pointing directions. Inclusion of the energy equation in the CFD will allow spatial and temporal variations of temperature and density so that Optical Path Distortion (OPD) estimates can ultimately be obtained to predict potential image quality. ACKNOWLEDGEMENTS This work has been supported by the GMTO Corporation, a non-profit organization operated on behalf of an international consortium of universities and institutions: Astronomy Australia Ltd, the Australian National University, the Carnegie Institution for Science, Harvard University, the Korea Astronomy and Space Science Institute, the São Paulo Research Foundation, the Smithsonian Institution, the University of Texas at Austin, Texas A&M University, the University of Arizona, and the University of Chicago. REFERENCES [1] Marcum, D.L., Anisotropic Solution Adaptive Unstructured Grid Generation Using AFLR, Final Report, NASA Grant No. NNL04AA91G, (March 2007). [2] [3] Kelecy, F.J. Coupling Momentum and Continuity Increases CFD Robustness, Ansys Advantage, Vol 2, Issue 2, (2008). [4] Chen, D., and Cochran, L., CPP Wind Tunnel Test Report Topography, CPP, Inc., CPP-EF-DOC , Rev A, (1 May 2011). [5] Hardie, K., and Trancho, G., GMT Environmental Conditions, GMT-REF-00144, Rev C., (24 July 2015). [6] Vogiatzis, K., and Thompson, H., On the Precision of Aero-Thermal Simulations for TMT, Thirty Meter Telescope (USA), SPIE , (2016). [7] Teran, J., Burgett, W., and Grigel, E., GMT Site, Enclosure, and Facilities Design and Development Overview and Update, SPIE , (2016). [8] McCarthy, P., Fanson, J., and Bernstein, R., Overview and Status of the Giant Magellan Telescope Project, GMTO Corp., SPIE , (2016). [9] Danks, R., Smeaton, W., Initial Computational Fluid Dynamics Modeling of the Giant Magellan Telescope Site and Enclosure, SPIE , (2016). [10] Farahani, A., Kolesnikov, A., and Cochran, L., GMT Enclosure Wind and Thermal Study, GMTO Corp., Proc. 8444, 1-13 (2012). Proc. of SPIE Vol