Mars 2020 Atmospheric Modeling for Flight Mechanics Simulations Soumyo Dutta, David Way, and Carlie Zumwalt NASA Langley Research Center Gregorio Villar Jet Propulsion Laboratory International Planetary Probe Workshop 2018 Boulder, CO June 14, 2018 The technical data in this document is controlled under the U.S. Export Regulations. Release to foreign persons may require an export authorization. 1
Mars 2020 concept of operations Atmosphere modeled within flight mechanics simulations Parameters: Pressure, temperature, and density Winds 2
Example of Atmospheric Effect n Winds have large effect on trajectory dispersions during the long time on parachute Geocentric Latitude (deg) -26.4-26.45-26.5-26.55-26.6-26.65-26.7-26.75-26.8-1342 -1189-1036 -730-577 -424-883 -1189 Touchdown footprint size reduced by 40% -1036-1495 -1342-1495 -1648-1648 -1801-1801 99 Percentile Confidence Ellipse at Parachute Deploy -1954-1954 -2107-2107 -1954 324.9 325 325.1 325.2 East Longitude (deg) Bounding Wind Model MOLA Baseline Eng Wind Model Meso 99 Percentile Confidence Ellipse at Touchdown 3
Atmospheric Need for EDL Simulations n Nominal atmosphere as a function of independent variable ousually function of space altitude, latitude, and longitude n Way to create dispersed atmospheric properties for Monte Carlo runs ohave to capture spatial and temporal dispersions 140 120 100 Altitude (km) 80 60 40 20 0-20 -100-50 0 50 100 150 N-S Winds (m/s) 4
Historical Approaches 1. Engineering Models o GRAM = Global Reference Atmospheric Models are a good example o Useful during early design phases when a flight atmospheric team may not exist 2. Climate Sounder Data-based Profiles o Combine Mars orbiter sounder data with global circulation models Credit: NASA/JPL o Range of profiles created that are queried by flight mechanics simulations in Monte Carlo fashion Credit: D. Kass (JPL) 5
Historical Approaches 1. Engineering Models o GRAM = Global Reference Atmospheric Models are a good example o Useful during early design phases when a flight atmospheric team may not exist 2. Climate Sounder Data-based Profiles o Combine Mars orbiter sounder data with global circulation models o Range of profiles created that are queried by flight mechanics simulations in Monte Carlo fashion 3. Mesoscale Atmospheric Models (Mars 2020 Approach) o Initialized from global circulation models o Model the atmospheric evolution using partial differential equations that govern atmospheric dynamics o Can capture effect of surface interactions and provides good resolution near target sites by using nested grids Latitude (deg) Longitude (deg) Credit: D. Tyler (OSU) 6
Atmospheric Integration into EDL Simulations 1. Look at Mesoscale data along a constant latitude line (or a constant azimuth) Mesoscale data Altitude Notional trajectory Simulation atmosphere table data (Note: Grid spacing decreases at lower altitudes) Longitude 2. Interpolate at every grid point in space for at each time of interest ti where i = 1 n t = t1 t = t2 t = tn Note: Interpolation is done at every height above areoid point where table data is desired. 3. Generate mean and standard deviation at each grid point by looking at the interpolated data at all time points ti where i = 1 n. Note, for statistics, the sample size is just n (number of time points being used for analysis). 7
Notes on the Current Incorporation Method n Has been accepted by the Mars 2020 atmospheric team n Successfully used for Mars Science Laboratory. n Provides means to stress flight mechanics simulation ocaptures model output from Mesoscale models oallows atmospheric property dispersions n Does not preserve hydrostatic equilibrium ohydrostatic equilibrium pressure gradient balanced by gravity; physics in Mesoscale models obroken by current incorporation method due to statistics being taken across the time axis n Does not capture spatial and temporal correlations in atmospheric properties during Monte Carlo analysis Credit: A. Chen (JSR 2014) 8
Alternate Approaches: Standard atmosphere model n Modeling the temperature profiles using constant lapse rates (variation of temperature with altitude) in layers and dispersing lapse rates n Creating a standard atmosphere model tuned to mesoscale data Full Atmospheric Simulation (Mesoscale Models) Gas Table σ s, μ s Layer Definitions 45 40 35 30 Standard Atmosphere Model T, P, ρ, c Altitude (km) 25 20 15 10 5 Discretize temperature profile using waypoints that define layers 0 140 160 180 200 220 240 260 Temperature (K) 9
Alternate Approaches: Pre-generated covariance matrices n Use Mesoscale model data to create covariance matrices a priori n Utilize hydrostatic equilibrium to get pressure based on integrating the ground pressure and temperature; calculate density from equation of state n Preserves physics and spatial correlations n Drawback: Large matrices that need to be handled for each Monte Carlo case o Full temperature grid is (144x144x56) 2 o ~10 TB of RAM Calculate Surface Pressure Grid (off-line) Covariance Matrix Calculate Temperature Grid (off-line) Covariance Matrix Generate Random Surface Pressure from Covariance Matrix Generate Random Temperature Grid from Covariance Matrix Integrate Pressure Grid using Hydrostatic Equilibrium Calculate Atmospheric Density Grid via Equation of State Calculate Wind Grid (offline) Covariance Matrix Generate Random Wind Grid from Covariance Matrix Either off-line on in-line with EDL simulation Adapted from B. Robertson (GT ASDL) 10
Summary n Atmospheric models for flight mechanics simulations need to provide both nominal information and means of dispersions for Monte Carlo cases n Three traditional approaches have been used to provide atmospheric information to EDL simulations in the past MSL and Mars 2020 used Mesoscale model-based atmospheres n Current approach provides ways to incorporate Mesoscale data in flight mechanics simulation and model dispersions but breaks some physical properties like hydrostatic equilibrium n Future projects may want to consider other approaches that preserve the physics omodel temperature profiles as piece-wise continuous profiles governed by lapse rates. Use standard atmosphere approach to calculate pressure and density o Compute covariance matrices of atmospheric properties 11
Acknowledgements n Mars 2020 Council of Atmospheres specifically OSU MMM5 team and SwRI MRAMS team n Georgia Tech Aerospace Systems Design Laboratory Team that supported the CDT project on Mesoscale Atmosphere incorporation 12
Back-Up Slides 13
Example of combo wind profiles created from mesoscale data Height above Areoid (km) 45 40 35 30 25 20 15 10 5 0-3σ Mean North East Syrtis Combo MRAMS MMM5 +3σ MMM5 = Mars Mesoscale Model 5 th Gen (Oregon State University) MRAMS = Mars Regional Atmospheric Modeling System (Southwest Research Institute) Combo = Combined profiles used in Flight Mechanics simulations 5 40 30 20 10 0 10 20 30 40 North South winds (m/s) 14 7
Wind dispersions in EDL simulation: Individual effects of 3+3 winds The combo profiles are used within the flight mechanics simulation in a 3+3 way. First 3 is mean shift and the second 3 is shift in uncertainties. 3 refers to a 3-sigma or 3 standard deviation shift. Effect of 3 + 0: Mean shift is controlled by Monte Carlo input Uniform Distribution Effect of 0 + 3: Additional random noise perturbations are added based on standard deviation data from Mesoscale models to stress the flight mechanics simulation. Note: The black dashed lines are same in the two figures 15
Wind dispersions in EDL simulation: Total effect of 3+3 winds Two effects are independent; therefore, the results is like an RSS. Should expect total variability to be something like 4.24-sigma, rather than 6-sigma. Note: The black dashed lines are the same as the other figures 16
Effect of atmosphere on flight performance statistics n n Atmospheric density affects aerodynamic forces on the vehicle and manifests itself in things like altitude history and parachute loads; closed-loop nature of Mars 2020 entry guidance means density perturbation effects could be amplified Mars 2020 Simulation Winds have large effect on trajectory dispersions during the long time on parachute Touchdown 26.3 26.4 Touchdown footprint size reduced by 40% going from engineering winds to sitespecific mesoscale model 384 576 768 960 1152 1344 1536 1728 1920 2112 10 km radius circle for reference MOLA Engineering Winds HOL Meso HOL Eng Mesoscale Model- Based Winds Geocentric Latitude (deg) 26.5 26.6 26.7 384 576 192 192 384 384 0 192 768 384 0 576 960 768 192 1152 960 1152 1344 1536 1728 1920 2112 2112 10 km 1344 1536 1920 1920 26.9 0 192 0 99 Percentile Confidence Ellipse at 26.8 Parachute Deploy 0 192 384 384 576 768 960 1152 192 384 576 1728 768 960 1344 324.7 324.8 324.9 325 325.1 325.2 325.3 325.4 East Longitude (deg) 1536 13 May 2015 14 HOL 01_V1_MC Mars 2020 6 DoF POST2 EDL Monte Carlo sdutta2 101 1728 1920 99 Percentile Confidence Ellipse at Touchdown 2112 1728 1920 2112 17