ESA AO/1-6367/10/NL/AF
“GOCE+ Theme 3: Air density and wind retrieval using GOCE data”
Algorithm Theoretical Basis Document
Version 1.1
Eelco Doornbos – TU Delft Pieter Visser – TU Delft Georg Koppenwallner – HTG Bent Fritsche – HTG
June 20, 2013 2 Contents
1 Introduction 7 1.1 About the project ...... 7 1.2 Purpose of this document ...... 7
2 Overall data flow 9 2.1 Data processing flow chart ...... 9 2.2 Input products ...... 9 2.2.1 EGG CCD (common-mode accelerations) ...... 9 2.2.2 EGG IAQ (attitude quaternions) ...... 11 2.2.3 SST PSO (precise science orbits) ...... 11 2.2.4 AUX NOM (ion engine thrust levels) ...... 11 2.2.5 GOCE mass ...... 11 2.3 Space environment models ...... 11 2.3.1 NRLMSISE-00 ...... 11 2.3.2 HWM07 ...... 12 2.3.3 Planetary ephemerides JPL DE405 ...... 13 2.4 Auxiliary input data ...... 13 2.4.1 Solar and geomagnetic activity data ...... 13 2.4.2 Earth orientation parameters ...... 14 2.5 Output products ...... 14
3 Data calibration 15 3.1 Accelerometer and ion engine calibration using GPS data ...... 15
4 Non-gravitational force modeling 25 4.1 GOCE satellite geometry model ...... 25 4.1.1 Modelling approach ...... 25 4.1.2 Modelling GOCE ...... 28 4.1.3 Reference frames ...... 29 4.2 ANGARA two-phase approach ...... 29 4.2.1 Aerodynamic analysis, Phase 1 ...... 30
3 4.2.2 Aerodynamic analysis, Phase 2 ...... 31 4.2.3 Radiation pressure analysis, Phase 1 ...... 31 4.2.4 Direct radiation pressure, Phase 2 ...... 32 4.2.5 Indirect radiation pressure, Phase 2 ...... 32 4.3 Integral and Monte-Carlo Test Particle methods ...... 35 4.3.1 Monte-Carlo Method ...... 35 4.3.2 Integral Method ...... 39 4.4 Summary ...... 42
5 Reference frame transformations 45 5.1 Reference frames ...... 45 5.1.1 Local ...... 45 5.1.2 Earth-fixed ...... 45 5.1.3 Inertial ...... 46 5.1.4 Orbit-fixed ...... 46 5.1.5 Body-fixed ...... 46 5.1.6 Pseudo body-fixed ...... 48 5.2 Rotation matrices ...... 48 5.2.1 Local to Earth-fixed ...... 48 5.2.2 Inertial to earth-fixed ...... 48 5.2.3 Inertial to orbit-fixed ...... 48 5.2.4 Pseudo body-fixed to body-fixed ...... 48 5.2.5 Inertial to body-fixed ...... 49 5.3 Quaternions, Euler angles and rotation matrices ...... 49
6 NRTDM data importing and (pre)processing routines 51 6.1 Data import ...... 51 6.1.1 SP3FileToOrbitFiles ...... 51 6.1.2 ImportASCIIFormatted ...... 52 6.1.3 PolynomialCalibrationParameters ...... 53 6.2 Basic conversions ...... 53 6.2.1 CalibrateAccelerometer ...... 53 6.2.2 OrbitToGeo ...... 55 6.2.3 WindCorotationSBF ...... 56 6.2.4 QuaternionToEulerAnglesConvert ...... 56 6.2.5 GOCEThrusterAccel ...... 57 6.3 Thermosphere model evaluation ...... 58 6.3.1 GeoToDens ...... 59 6.3.2 GeoToWind ...... 59 6.3.3 WindLocalToSBF ...... 60 6.3.4 WindSBFtoLocal ...... 60 6.3.5 WindVectorProject ...... 61 6.4 Force model evaluation ...... 61 6.4.1 RadiationPressureProduct ...... 61 6.4.2 ANGARA Earth radiation pressure environment model ...... 66 6.4.3 AccelerometerSubtract ...... 68 6.5 Density and wind processing ...... 69 6.5.1 DensityWindFromAccelerometerDirect ...... 69
4 6.5.2 CalibrateWithModel ...... 76 6.5.3 DensityWindFromAccelerometerIterative ...... 76 6.6 Data editing and flagging ...... 79
5 6 Chapter 1 Introduction
1.1 About the project
Accelerometers carried by low-Earth orbiters such as GOCE have the ability to provide highly detailed data for improving our understanding of thermospheric density and winds. Like its predecessor missions, CHAMP and GRACE, GOCE has not been de- signed for studies of the thermosphere. Nevertheless, the application to thermosphere studies of these earlier missions has resulted in density and wind data sets containing information at unprecedented levels of precision and coverage. The algorithms for pro- cessing the data of these these earlier missions have been adapted for GOCE, and will be presented here.
1.2 Purpose of this document
The purpose of this document, according to the project’s Statement of Work, is as follows:
“Algorithm Theoretical Basis Documents (ATBD): This document shall de- scribe into details all the algorithms, methods and models implemented for each Theme. The report shall include also all related data sources, process- ing steps and output data. In particular, the ATBD shall provide a complete description of all the algorithms, methods and models (both theoretical and technical) and corresponding input/output data flows, respectively. In ad- dition, this document shall report a scientific analysis of the results driving to specific development choices and trade-offs for all the algorithms imple- mented for developing the whole suite of target products. Technical consid- erations justifying the selected methodologies shall be also provided.”
7 8 Chapter 2 Overall data flow
2.1 Data processing flow chart
Figure 2.1 shows an overview of the data processing. The processing starts with the import, pre-processing and calibration of the data. This results in four main data sources, on which the further processing is based:
• Orbit
• Attitude
• Acceleration
• Thruster actuation
The processing involves the use of thermosphere models and satellite surface force mod- els, in combination with custom algorithms for the purpose of determining density and wind. This will be described in full detail in the remainder of this document.
2.2 Input products
The following sections describe the origin of the GOCE input products that are used in the algorithms.
2.2.1 EGG CCD (common-mode accelerations)
Common-mode acceleration measurements from the gradiometer instrument are used. The so-called sensitive pairs are used for each gradiometer axis. This data is obtained from the TU Delft HPF system, but is also readily available to GOCE users via the goce.esa.int website.
9 GOCE L1B and GOCE thruster L2 data actuation data
Pre-processing and calibration
Thruster Orbit Attitude Acceleration actuation
Radiation Empirical density Empirical wind pressure model model evaluation model evaluation evaluation
Modelled Modelled radiation composition, Modelled wind pressure temperature accelerations
Compute initial Remove non- relative velocity aerodynamic accelerations
Relative velocity Observed in SBF frame aerodynamic acceleration
Aerodynamic Density and wind model evaluation processing
Modelled density Density data Wind data
Figure 2.1 Overview of the data processing in a single flow chart. Final and interme- diate data products are shown in yellow boxes. Standard data processing steps are blue, the use of thermosphere models is shown in green and the use of non-gravitational force models in brown.
10 2.2.2 EGG IAQ (attitude quaternions)
Reconstructed attitude quaternion from the star cameras and gradiometer instrument are obtained from the TU Delft HPF system. This data is also readily available to GOCE users via the goce.esa.int website.
2.2.3 SST PSO (precise science orbits)
The precise science orbit ephemeris is based on GPS satellite-to-satellite tracking data. It is obtained from the TU Delft HPF system, but is also readily available to GOCE users via the goce.esa.int website.
2.2.4 AUX NOM (ion engine thrust levels)
The ion engine thrust level data is provided directly to the project team’s FTP site by ESA. The data come in XML files, which are parsed so that only a text file with a timeseries of
2.2.5 GOCE mass
The GOCE mass as a function of time is available in a file named GOCE-Mass-Properties- COG-data.GOC, available for download from the goce.esa.int website.
2.3 Space environment models
The NRTDM software incorporates several external models for the space environment. The following such models are applied in the GOCE data processing:
• NRLMSISE-00 for atmospheric composition and temperature
• HWM07 for in-track winds
• Planetary ephemerides DE405
• Solar radiation pressure environment model (eclipse, varying Sun-Earth distance) from the ANGARA phase 2 code
• Earth radiation pressure environment (Earth albedo and IR fluxes at the spacecraft) model adapted from the ANGARA phase 2 code
2.3.1 NRLMSISE-00
The NRLMSISE-00 model is the latest iteration of the MSIS series of models [Hedin et al., 1977a,b, 1979, Hedin, 1983, 1987, 1991] developed at NASA’s Goddard Space Flight Cen- tre. The model development has been adopted by the Naval Research Laboratory, leading to the current version NRLMSISE-00 [Picone et al., 2002].
11 NRLMSISE-00 inputs The model takes the following inputs:
• Day of year
• Time of day
• Altitude
• Geodetic longitude
• Geodetic latitude
• Local solar time • 81-day average of F10.7, centred on the current day • Daily F10.7 for the previous day • 3-hourly ap for current time • 3-hourly ap for 3 hours before the current time • 3-hourly ap for 6 hours before the current time • 3-hourly ap for 9 hours before the current time • Average of eight 3-hourly ap indices from 12 to 33 hours before the current time • Average of eight 3-hourly ap indices from 36 to 57 hours before the current time
Outputs Outputs of the model are:
• Exospheric temperature
• Temperature at requested altitude
• Total mass density • Number densities for H, He, N, O, N2,O2, Ar and anomalous oxygen
2.3.2 HWM07 HWM07 is the latest version of the Horizontal Wind Model, which is a companion model to the MSIS series of models. This model was also originally developed at Goddard Space Flight Center [Hedin et al., 1988, 1991, 1996]. The latest version [Drob et al., 2008] was further developed at the Naval Research Laboratory and adds a separate component for storm-induced winds [Emmert et al., 2008].
12 Inputs The inputs to the HWM model are identical to those for the NRLMSISE-00 model (see above), with two exceptions. The F10.7 solar activity proxy input is not used by the model, and the code accepts only a two element array for the geomagnetic activity index ap,of which only the second element, containing the current value, is used by the model.
Outputs The outputs of the HWM07 model are: • Meridional wind (northward) • Zonal wind (eastward) This wind vector output is in the local frame (see Section 5.1), although the upwards component is always zero, and the north- and eastward axes are switched in the model subroutines arguments, with respect to the reference frame definition.
2.3.3 Planetary ephemerides JPL DE405 The JPL planetary ephemerides DE405 [Standish, 1998] are implemented in NRTDM for the computation of the position of the Sun with respect to the Earth and the satellite. This is necessary in the calculation of radiation pressure accelerations.
Inputs The only input to the ephemerides calculation is the time.
Outputs As output, the position of the Sun with respect to the Earth in the inertial frame is re- turned.
2.4 Auxiliary input data
The NRTDM software requires two types of auxiliary data: • Solar and geomagnetic flux data used in calling the thermosphere density, temper- ature and wind models (see Section 2.3). • Earth orientation parameters for the conversion between inertial and Earth-fixed reference frames (see Section 5.1).
2.4.1 Solar and geomagnetic activity data
The NRLMSIS and HWM models use the F10.7 solar flux proxy and ap geomagnetic activ- ity indices, to represent variations in the energy input from the Sun into the atmosphere. The data are downloaded from NOAA’s National Geophysical Data Centre server, at the following URL: • ftp://ftp.ngdc.noaa.gov/STP/GEOMAGNETIC DATA/INDICES/KP AP • ftp://ftp.ngdc.noaa.gov/STP/SOLAR DATA/SOLAR RADIO/FLUX
13 2.4.2 Earth orientation parameters The conversion between inertial and Earth-fixed coordinates (see Chapter 5) requires the use of Earth-orientation parameters. These are downloaded from the following location:
• http://hpiers.obspm.fr/iers/eop/eopc04/eopc04 IAU2000.62-now
2.5 Output products
The main outputs of the algorithm, to be provided as a function of time, are
• Density (kg/m3)
• Crosswind (m/s), vector with east, north, up components
Additional fields from the NRTDM data storage can be provided alongside this data. The most important candidate data fields for inclusion in the final product are:
• Geodetic latitude, longitude, height, local solar time
• Quality flags
• Modelled density
• Modelled wind
• Modelled wind projected on crosswind direction
14 Chapter 3 Data calibration
3.1 Accelerometer and ion engine calibration using GPS data
Ideally, the non-gravitational accelerations ang experienced by GOCE are equal to the so-called common-mode accelerations acm observed by its gradiometer plus the thrust T from the ion engines divided by the mass m of the satellite:
= + ang acm T/m (3.1)
Equation 3.1 indicates that (1) the common-mode accelerations acm and (2) the thrust vector T need to be properly calibrated. Ad (1). The common-mode accelerations are obtained by taking the average of ob- servations by pairs of accelerometers. In fact, the three most sensitive combinations are taken, one for each direction of the gradiometer reference frame (GRF), which is pair 1&2 for the X-axis, pair 3&4 for the Y-axis, and pair 5&6 for the Z-axis in the GRF. The center of the gradiometer almost perfectly coincides with the satellite’s center of mass. By tak- ing these common-mode combinations, rotational terms are almost perfectly eliminated, provided the accelerometers are properly calibrated and aligned. Based on the precision requirements for the GOCE satellite gravity gradient (SGG) observations, it is assumed that the scale factors for the accelerometers are known very well and that in fact these scale factors can be considered to be equal to ONE (1.00) for the Level 1b common-mode accelerations (”EGG CCD” data, see HPF [2010]). Also, the requirements for the alignment of the accelerometers are very stringent. However, the requirements for the SGG observations are valid for the measurement bandwidth (MB) of 0.005-0.1 Hz, meaning that the accelerometer observations can and in fact are offset by (drifting) biases. These biases need to be known before the common-mode accelerome- ter observations can be used for thermospheric density and winds determination. These biases can be estimated by precise orbit determination (POD) by the same methodology applied for calibrating e.g. the CHAMP and GRACE accelerometers Bruinsma and Bian- cale [2003], Bruinsma et al. [2004], Visser and van den IJssel [2003], where the common- mode accelerations represent the non-gravitational accelerations in the GOCE dynamic
15 force modeling for the POD. Moreover, the POD provides a validation of the assumption that the scale factors are equal or close to ONE as well. For GOCE, this methodology has been implemented and is used operationally in the framework of the GOCE High-level Processing Facility (HPF). The architecture of this implementation is described in HPF [2007] (pp. 51-53) and Bouman et al. [2011], for which the GOCE standards were adopted Gruber et al. [2010]. Please note that time series of kinematic orbit positions are used as the observation type. These kinematic position estimates can be considered a condensed set of GPS Satellite-to-Satellite Tracking (SST) observations. For this project, a few enhancements were made to the HPF implementation. These enhancements are:
• Use of GOCO02S gravity field Goiginger et al. [2011] model instead of EIGEN-5S Gruber et al. [2010]: the post-launch GOCO02S model provides a better represen- tation of the gravitational forces acting on GOCE than the pre-launch EIGEN-5S model leading to smaller aliasing of gravitational errors into estimated common- mode acceleration calibration parameters;
• Use of the kinematic Precise Science Orbit (PSO) instead of the Rapid Science Orbit (RSO). The claimed accuracy of the PSO product is of the order of a few cm in terms of position Bock et al. [2011]. The PSO solutions have a length of 30 hr with a 6-hr overlap between consecutive days. For the estimation of calibration parameters, the full PSOs are used, i.e. an arc length of 30 hr is adopted;
• Use of star tracker mounting matrices from the GOCE gradiometer calibration (AUX EGG DB): the star tracker observations are used instead of the combined quaternion product EGG IAQ HPF [2010] because of better continuity (less data gaps): the accuracy of the orientation of the GRF is crucial for estimating calibration parameters for the common-mode accelerations.
Because of the quality of the gravitational force modeling, it was found that 3-dimensional kinematic orbit fits of around 15 cm could be obtained by estimating only the begin posi- tion and velocity plus three common-mode acceleration biases for every 30-hr arc (fixing the scale factors to ONE). Thus only 9 parameters were estimated for each arc. Figure 3.1 displays the estimated biases for a period covering November 2009 - Septem- ber 2011. It can be observed that the estimates for the X axis (predominantly the flight direction) are very stable with an Root-Mean-Square (RMS) of fit of 0.15 nm/s2 to a linear model representing a bias of -187.5 nm/s2 (at start time 1 November 2009) and a drift of 0.0010 nm/s2/day. For the Y-axis (predominantly cross-track) the RMS value is equal to 12 nm/s2 for a bias value of -310 nm/s2 and a drift of 0.14 nm/s2/day. For the Z-axis (predominantly radial direction), the results are less reliable with an RMS of 66 nm/s2,a bias of 97.8 nm/s2 and a drift of -0.02 nm/s2/day. Ad (2). It is assumed that the thrust magnitude and direction of the ion engines are well known and accurately calibrated. However, a calibration of the ion engine thrust magnitude can be done (to some extent) as well by POD considering the fact that regu- larly maneuvers are carried out, where the GOCE ion engines have different thrust levels. To this aim, the same implementation as for the estimation of the common-mode acceler- ation biases can be used with the following changes:
• The common-mode accelerations are no longer used;
16 Figure 3.1 Bias estimates for sensitive common-mode combination, i.e. accelerometer pair 14 for the X-, pair 25 for the Y-, and pair 36 for the Z-axis.
17 • Piecewise linear empirical accelerations are defined with time interval of 15 min be- tween the notes for the X, Y and Z axes of the GRF. These accelerations are estimated and represent the GOCE non-gravitational accelerations.
Please note that the empirical accelerations are defined in the GRF allowing a direct comparison with the observed common-mode accelerations. For a 30-hr arc, the total number of estimated accelerations is equal to (30 × 4 + 1) ∗ 3 = 363. With the begin position and velocity, the total number of estimated parameters is then equal to 369. The above approach has been successfully implemented for other missions van den IJssel and Visser [2005, 2007] and has been tested for a few selected GOCE days, see Figures 3.2-3.4. Jumps in the non-gravitational acceleration level can especially be observed for the X- axis. For 15 June 2009 (during commissioning phase) and 7 March 2010, these jumps are relatively large: of the order of 1700 and 950 nm/s2, respectively. At the end of 14 August 2011, the jump is relatively small, about 400 nm/s2. It is interesting to see that similar jumps can be observed for the Z axis, but at a much smaller level. The differences between the provided ion engine thrust levels can be compared with these jumps. A similar comparison can also be done with the common-mode accelera- tions, provided that these are properly calibrated. Time series of accelerations derived from the ion engine thrust values, the common-mode accelerometer observations and those estimated by POD are displayed together in Figure 3.5 for the X-axis for a few hours before and after the change of ion engine thrust level. Especially for the selected day during the commissioning phase (15 June 2009), the jump is clearly visible in all time series. As the GOCE mission progresses, so does the atmospheric drag due to increased levels of solar activity. The ion engines nominally compensate the atmospheric drag and the ion engine thrust values thus display the larger variations of the atmospheric drag for the later stages of the GOCE mission. This can be observed clearly in the middle and bottom plots of Figure 3.5. These increased levels of atmospheric drag fluctuations make it more difficult to clearly identify jumps, and thus match different ion engine thrust levels with different levels of the estimated non- gravitational accelerations or common mode observations. Since the most dominant aerodynamic drag fluctuations occur at a frequency of 1 cycle per orbit revolution, a comparison has been made between the averaged values of the different time series for periods equal to one orbital revolution before and after each maneuver. A margin of 5 min before and after the maneuver was used in order to eliminate the transition period of ion engine thrust levels. The orbital period was taken − equal to 5400 s. Thus, for each maneuver time tman the period before covers tman 5700 − − − to tman 300 s and the period after covers tman 300 to tman 5700 s. The averaged ac- celeration levels for the different time series are displayed in Table 3.1 for the selected maneuvers. For the first two selected days, 15 June 2009 and 7 March 2010, the consis- tency of the jumps in the mean values (red and blue) is of the order of a few percent: respectively 1690.8, 1677.3, 1696.7 nm/s2 and 960.1 , 985.3, 977.7 nm/s2. For the latest se- lected day, 14 August 2011, this jump derived from the ion engine thrust values is equal to 423.5 nm/s2 compared to 352.0 and 340.9 nm/s2 for the jumps derived from the POD and common-mode values (green), a discrepancy of up to 25%. The ion engine thrust values are available with a time interval of 10 s compared to 1 s for the common-mode accelerations and 15 min for the values estimated by POD. Com- paring with values determined by POD thus requires an averaging interval of at least 15 min, during which the aerodynamic drag variations would not average out. Using an
18 Figure 3.2 Non-gravitational accelerations: common-mode vs. estimated by POD. Time is relative to 14 June 2009 00:00 (note the difference in scale).
19 Figure 3.3 Non-gravitational accelerations: common-mode vs. estimated by POD. Time is relative 6 March 2010 00:00 (note the difference in scale).
20 Figure 3.4 Non-gravitational accelerations: common-mode vs. estimated by POD. Time is relative to 14 August 2011 00:00 (note the difference in scale).
21 15 June 2009 10:16:35
7 March 2010 14:00:33
14 August 2011 23:58:33
Figure 3.5 Non-gravitational accelerations for the X-axis: common-mode, estimated by POD and derived from the ion engine thrust values. The approximate maneuver time is indicated above the plots.
22 Table 3.1 Averaged acceleration values for the orbital revolution before and after the selected maneuvers (values in nm/s2).
Before After Difference Date Ion engine thrust 15 June 2009 834.9 2526.8 1690.8 7 March 2010 2397.1 3355.5 960.1 14 August 2011 2600.8 3025.8 423.5 Precise Orbit Determination 15 June 2009 -60.5 1616.8 1677.3 7 March 2010 -2.2 983.1 985.3 14 August 2011 -0.7 351.3 352.0 Common mode 15 June 2009 108.0 1804.7 1696.7 7 March 2010 186.4 1164.1 977.7 14 August 2011 188.5 529.4 340.9
averaging interval of 1 orbital revolution only partially takes care of this. Of course, the quality of the averaging depends on the evolution of the orbit and of atmospheric condi- tions, which is the reason of the larger discrepancy found for the selected day later on in the GOCE mission. A comparison with common-mode accelerations would be more straightforward, pro- vided the ion engine thrust values would be known with a smaller time interval as well. Figure 3.6 provides a zoom-in of the associated acceleration values close to the maneu- ver. The jumps caused by a change of thrust level can be clearly observed, but again a comparison between ion engine thrust values and common-mode accelerations is not straightforward for 14 August 2011 with the given sampling rates.
23 15 June 2009 10:16:35
7 March 2010 14:00:33
14 August 2011 23:58:33
Figure 3.6 Zoom-in of non-gravitational accelerations for the X-axis: common-mode and derived from the ion engine thrust values. The approximate maneuver time is indicated above the plots.
24 Chapter 4 Non-gravitational force modeling
4.1 GOCE satellite geometry model
4.1.1 Modelling approach There are different possibilities to model a spacecraft for surface force calculations. Their complexity ranges from spheres or simple rectangular shapes up to a detailed overall modelling of the local geometry. It depends on the application which degree of com- plexity is best suited. It is obvious, for example, that it will not be possible to model lift with spheres, or torques with symmetrical polyhedrons. For the present contract it was decided to use a detailed modelling of the geometry, for several reasons:
• A detailed model is most capable of determining also detailed force and torque coefficients in all coordinate directions.
• Using appropriate methods for integrating the local coefficients also mutual shad- owing of different parts of the geometry can be taken into account quite generally, without the need of a case-by-case analysis as for simplified panel methods, for example.
• The availability of a software for non-gravitational force calculations (ANGARA) which uses such a detailed geometry modelling.
In ANGARA [Fritsche et al., 1998] the geometric modeling of the surfaces of a spacecraft is based on the principle of element-by-element definition. In Fig. 4.1 the structure of the geometric modeling is given. Each spacecraft is represented as a combination of separate elements. The initial geo- metric elements are called geometric primitives. This set of geometric primitives includes first-order elements such as rectangle, polygon, circle, and ellipse; second-order elements such as cylinder, cone, sphere, ellipsoid, and paraboloid; and also sets of cross sectional
25 Geometric model
GM1.1
GM1.2
GM1.3 Compound Level
C6
C4 C5
C2 C1 C3
Primitive Level
Cone Sphere Rectangle Triangle Figure 4.1 Structure of the geometric modelling.
26 Y
L R2
R1 X Phi1
Z Phi2
Figure 4.2 Example of a primitive shape and its parameters
elements whose surfaces are determined by points in parallel sections. Each primitive has its own parameters, which define the shape and size of the primitive. On the second level, a combination of primitives forms a compound. For compounds, a set of parameters can be declared, which influence the compound shape and size. This process is named parameterization. Compounds can be also used in other compounds. A compound with a fixed set of parameters transforms into a Geometric model. From one compound several models can be made, where each of them has its own fixed set of parameters (like solar array rotation angle). Geometric primitives are defined by several geometric parameters in its own local coordinate system. For each kind of primitive a special subroutine exists which can provide the transformation from the analytical sur-
Parameter Range Explanation R1 0 < R1 < ∞ Radius in the first cross section R2 R1 < R2 < ∞ Radius in the second cross section L 0 < L < ∞ Length of the cone Phi1 0 < Phi1 < 360 Angle (in degrees) between the OY axis and the segment origin in the OYZ plane Phi2 0 < Phi2 < 360 Angle (in degrees) of the segment in the OYZ plane. step1 0 < step1 Partition step size along the OX axis step2 0 < step2 Partition step size along the OY axis normal ” + ”or” − ” Normal sign (inner/outer surface = +/−)
Table 4.1 List of the primitive parameters
27 Figure 4.3 ANGARA GOCE geometry
Figure 4.4 ANGARA GOCE panelized geometry
face description to the panelized description. An example of a primitive definition is shown on Figure 4.2 and Table 4.1. The surface of each geometric element is split into triangular panels approximately equal in size. The sizes of partition steps in two quasiorthogonal directions are defined as element parameters. Thus, a panel model is composed of plane triangular panels approximating the surfaces of all the elements included in this object.
4.1.2 Modelling GOCE The GOCE geometry model was created from CAD drawings at a high level of detail. This level of detail in the geometry should translate into higher accuracy in the force coefficients, and consequently in the density and wind results. In fact, at the current level of detail of geometric modelling, it is likely that other aspects, such as uncertainties in the gas-surface interaction, dominate the force coefficient error. Figure 4.3 shows the GOCE geometry modelled with ANGARA. The surface consists
28 of 209 primitives. Figure 4.4 shows the ANGARA panel model of GOCE. The number of panels is 5402.
4.1.3 Reference frames There are three different reference frames to consider when a satellite is modelled for the calculation of non-gravitational forces and moments:
1. Geometric reference frame. This is the coordinate system in which the satellite ge- ometry is modelled. In principle its coordinate directions are arbitrary, but its rela- tion to the body-fixed axes (see below) has to be known. In ANGARA this relation is fixed.
2. Body-fixed reference frame. The geometric reference frame (see above) is body- fixed as well, but the so-called Spacecraft Body-Fixed (SBF) frame has a fixed re- lation to the aerodynamic reference frame (see below). In ANGARA the relation between the axis of the geometric (index g) and body-fixed system (index b) are: • = − xg xb • = − yg zb • = − zg yb The spacecraft body-fixed frame for GOCE, and its relation to other frames is fur- ther described in Section 5.1.5. Figure 5.1 shows the body-fixed coordinates for GOCE.
3. Aerodynamic reference frame. This frame is actually determined only by the in- verse wind vector. The x-axis of this coordinate frame is given by the opposite direction of the free-stream direction (wind) relative to the satellite. Therefore the aerodynamic attitude of a satellite is determined by only two angles instead of a full set of three e.g. Euler angles. The attitude is undetermined relative to a rota- tion about the wind vector. The two angles determining the attitude of the satellite relative to the wind vector (or more exactly: the direction of the wind vector in the spacecraft body-fixed frame) are the angle of attack α and the side-slip angle β. The ( ) aerodynamic angles can be computed from the directions Vx, Vy, Vz of the inverse wind vector by: • = α β Vx cos cos • = β Vy sin • = α β Vz sin cos
4.2 ANGARA two-phase approach
The traditional approach to address the problem of modeling non-gravitational effects is to consider a simplified geometry of a given spacecraft on a given orbit and to compute the force and torque acting on the S/C, considering the fluxes of photons and molecules as function of the orbit phase. Disadvantages of this approach are the restricted repre- sentation of the real geometry and a time-consuming orbit propagation.
29 Figure 4.5 Body-fixed coordinate system of GOCE
In the approach used in ANGARA, the process of computing perturbing forces and torques acting on a specific spacecraft has been organised in two phases. In Phase 1, the spacecraft model is established, and the coefficients of force and torque for aerodynamic and radiative actions are computed in advance for the expected range of flow/radiation attack angles and speed ratios/wavelengths by means of test particle or integral methods. The calculation results are stored as matrix coefficients in a database. The contents of the database can be distributed as data file. In Phase 2, the aerodynamic and radiative environment of the spacecraft are estab- lished for the actual state vectodr and attitude of the satellite at a given epoch. The resulting forces and moments are computed by interpolating the coefficients in the data matrix computed in Phase 1 for reference conditions, and by multiplication of these coef- ficients with corresponding normalisation factors derived from the actual enviromental conditions. In this way, the major part of the whole work is concentrated in Phase 1. In Phase 2, the results of Phase 1 only have to be adjusted for user-defined environmental conditions. The effort for force and torque calculation becomes independent of the complexity of the spacecraft geometry, which allows to calculate aerodynamic and radiation forces on complex spacecraft in near real-time.
4.2.1 Aerodynamic analysis, Phase 1 For the free-molecular flow each gas species can be considered independently, and aero- dynamic coefficients of forces and moments depend only on the free stream molecular speed ratio S∞ and wall temperature TW of a spacecraft for fixed gas/surface interaction properties. If a similar law of gas/surface interaction is used for all gas species, it is only nec- essary to perform the aerodynamic computations for different S∞i, covering the whole range of free-stream velocities U∞, molecular masses mi, and ambient temperatures T∞.
= U∞ S∞i · · ∞ 2 k/mi T i Then, the following procedure can be used for Phase 1:
30 1. Determination of the range of all possible values S∞ for all atmospheric species;
2. Calculation of aerodynamic coefficients for a set of S∞i and for a specified range α β re f of angles of attack , side slip angles , and reference wall temperatures Tw of a spacecraft. The result of the computation is a matrix of coefficients; C 3. Storage of the following aerodynamic coefficients of force ( a) in the results database: Cinc Cout + Cmr a , a a
where ‘inc’, ‘out’ and ‘mr’ stands for incident, reflected and multiply reflected.
4.2.2 Aerodynamic analysis, Phase 2
1. Determination of S∞i for user defined free-stream velocity U∞, temperature T∞ and atmospheric composition ni (number density of gas species ’i’); 2. Interpolation of aerodynamic coefficients of forces and moments for all values of α β S∞i and for a user defined spacecraft orientation , ; Cout + Cmr 3. Correction of a a for a user defined wall temperature Tw of the spacecraft Cout + Cmr = Cout + Cmr · re f a a a a re f Tw/Tw (4.1) Tw Tw (4.2)
4. Computation of the aerodynamic force coefficient from all contributions for the gas species.
C = Cinc + Cout + Cmr a,i a a a (4.3)
5. Determination of the effective aerodynamic force coefficient vector, by summation over the atmospheric constituents, weighted by their partial contribution to the to- tal mass density. N ρ C = i C a ∑ ρ a,i (4.4) i=1
6. Determination of aerodynamic forces 1 2 inc out mr F = ρU∞ · A · C + C + C (4.5) a 2 ref a a a
4.2.3 Radiation pressure analysis, Phase 1 The calculations are conducted similarly to the calculations of aerodynamic coefficients for Phase 1.
1. Calculation of the radiation coefficients of force C in the body-fixed coordinate system for a specified range of α, β (spacecraft view angles towards the radiation source), for molecular speed ratio S = ∞, for the visible spectrum, and for the IR spectrum.
31 y x
z g
North pole Z A a i i z r i R s R 0i yg cos ( a ) > 0 O Y i if the element of Earth surface is visible from the spacecraft
x g Figure 4.6 Integration of indirect radiation over the Earth surface
2. Calculation of the self-radiation coefficients of force for the illuminated part of the sh spacecraft (Csel f ) and for the shadowed part (Csel f ) at reference wall temperature re f α β Tw and given direction towards the sun and . 3. Storage of the radiation coefficients in the results database: