Control-‐Oriented Model of a Hypersonic Vehicle (MASTRIM) -‐
Containing an Integrated, Advanced Propulsion Sub-‐Model (MASIV)
Jim Driscoll, Sean Torrez, Derek Dalle University of Michigan,
Michael Bolender, Jonathan Muse, AFRL
1 September 2012
A. MoNvaNon
B. Results prior to last year
C. Results for last year
D. Next year’s plans
Outline
2
3
A. MoNvaNon Develop an improved “control-‐oriented” model by significantly modifying Doman-‐Bolender model -‐ hypersonic vehicle (MASTRIM) with integrated propulsion (MASIV) -‐ compute liD, drag, thrust, moments in 1 sec on PC using ROMs
-‐ including realisIc shock interacIons in the inlet, -‐ combustor finite rate chemistry, 3-‐D mixing -‐ nozzle expansion waves to compute lower edge shear layer -‐ add operability limits: ram-‐scram, unstart, flame-‐out
4
B. MoNvaNon Ascent trajectory of MAX-‐1 trimmed vehicle 1. OpImizaIon study -‐ how to minimize fuel ? 2. MDO study – opImize geometry (surrogate vs. collocaIon) 3. Delivered to MIT -‐ linearized A, B matrices, poles and zeros
Generic hypersonic vehicle geometry (GHV, HiFIRE6) 8. Add 3-‐D shock interacIons to inlet (3-‐D Riemann problem) 9. Add ethylene fuel “lookup tables”
Operability Limits 4. When does ram-‐scram occur ? 5. How much do forces jump at ram-‐scram ? controllability ? 6. When does unstart occur ? 7. Can a “befer” trajectory avoid unstart ?
5
MoNvaNon
A Reduced Order Model (ROM) is required Suppose vehicle ascends along a constant dynamic pressure path
We ask: when does ram-‐scram transiIon occur ?
AlItude (km)
Flight Mach Number
acceleraIon q =
0.6 atm 0.8 atm 1.0 atm
Computed ram-‐scram transiIon boundaries
To generate this plot: liD, drag, thrust, moments were computed 1500 Imes = (guess 15 AoA’s to trim) X 5 (q values ) X 20 alItude values /curve
MAX-‐1
6
MoNvaNon ROM is best when thousands of force computaIons are needed -‐ when opImizing the trajectory -‐ when generaIng operaIng maps -‐ for geometry opImizaIon (MDO) ROM provides understanding of Operability Limits -‐ Unstart Margin -‐ understand what factors help or hurt -‐ Ram-‐scram transiIon -‐ esImate large jump in forces -‐ Engine Flameout -‐ need finite rate chemistry
ROM is useful for a “first look” at the design space -‐ First idenIfy a small number of problem cases -‐ Understand what the problem is -‐ Run high fidelity CFD just for those cases
7
MoNvaNon: Operability Limits – when does unstart occur ? Unstart Margin = L1 / L2 L1 = distance between leading edge of shock train and entrance to isolator L2 = total length of isolator
station: 2 3 4
inlet isolator combustor nozzle
(c)
(b)
(a)
L2
L1
1. First must properly compute the thermally-‐choked (ram) case
2. Then MASIV predicts isolator back pressure (pback)
3. Length of shock train then computed from experimental relaIon
(L2 – L1) / H = funcIon (pback / pinlet )
8
For a short isolator, MASTRIM shows that as you accelerate too fast (2 m/s2) à you unstart before you reach ram-‐scram !
MASIV tells us one way to avoid unstart !
Unstart Margin
ßunstart
Flight Mach Number
MASIV predicts the proper acceleraIon path = lower ER path à reduce acceleraIon to 1 m/s2 for the predicted Ime period
MAX-‐1
MoNvaNon: Operability Limits = Unstart, Flameout, Ram-‐Scram Issues
AlItude (h)
Flight Mach Number M∞
insufficient liD, insufficient air flow to trim
flame out at low pressure
insufficient mixing (low Re)
engine unstart
ram -‐ scram thermal limit of wall cooling
4 8 12
9
B. Results Prior to Last Year
1. MASTRIM -‐ Developed a 6-‐DOF Flight Dynamics Model of a Hypersonic Vehicle -‐ based on Doman-‐Bolender trim code
2. Completed Propulsion/Vehicle integraNon
a. SAMURI -‐ reduced order inlet & nozzle code addedà realisIc inlet losses from mulIple shock/expansion interacIons.
rapidly computes lower edge of exhaust plume
b. MASIV -‐ reduced order code for combustor method: lookup tables from high fidelity 3-‐D combusIon, 3-‐D mixing added: finite-‐rate chemistry of hydrogen and ethylene fuel,
gas dissociaIon losses
3. Operability Limits -‐ began methodology -‐ ram-‐scram transiIon, engine unstart, flame out at low pressure
10
4. Published three examples:
a) Level Flight Generic X-‐43, Mach 8 b) Ascent OperaIng maps (δCE, α, ER) vs. (M∞ , h) c) Turn Poles, zeros, Ime to double amplitude
5. Disseminated the MASTRIM code
B. Prior to last year (con’t)
Delivered MASTRIM to AFRL Demos to AFRL Air Vehicles and Propulsion
Boeing (Bowcuf) HunIngton Beach March 11 NASA AMES (Soloway) Ohio State
Demos at MIT and NASA Glenn are planned Michael Smart UQ @ HyTASP Conference
6. ValidaNon tests vs. high-‐fidelity 3-‐D CFD & some experiments (6-‐10%) reduced run Ime to ~ 1 sec
11
Developed a 6-‐DOF Flight Dynamics Model MASTRIM control parameters: 1. δCE elevon combined deflecIon angle 2. δER equivalence raIo change 3. Δxc cowl lateral translaIon distance 4. Δyc cowl transverse translaIon distance 5. Δθc cowl flap angle 6. εFUEL fracIon of fuel injected at locaIon 2 for a turning trajectory: 7. δDE elevon differenIal deflecIon angle 8. δCR rudder combined deflecIon angle 9. δDR rudder differenIal deflecIon angle Trims the vehicle Computes poles, zeros, Ime to double amplitude
Δxc
Δyc
12
Prior to Last Year – SAMURI inlet ROM
13
Track discrete waves, no CFD grid used Shocks: exact 2-‐D oblique shock relaIons with real gas properIes Expansions: discreIze conInuous fan into 2-‐4 waves Wave interacIons: exact Riemann jump condiIons Boundary layers: displace wall by displacement thickness
inlet losses: PRF = (stagnaIon) pressure recovery factor ~ 60% inlet compression raIo: p2/p∞ = 40
Moo = 10, α = 0 , q = 2040 lbf/D2
Dalle, J. of Propul. &Power, 2010
6.6% error in our ROM compared to CFD++ for stagnaIon pressure raIo
14
Prior to last year: for nozzle, SAMURI predicts lower edge
engine flow
Plume lower edge
Riemann solver -‐ interacIon of many expansion waves
Dalle, sub. to J. of Propul.& Power, 2011
15
MASIV Reduced Order Model of Combustor
Pre-‐computed lookup tables of finite-‐rate chemistry of jet-‐in-‐cross flow mixing 15 species, 22 reacIons ethylene or hydrogen fuel Fuel can be added anywhere Area is variable Wall fricIon, heat transfer Real gas properIes
bow internal isolator const area diverging external shock inlet burner burner nozzle
∞ 1 2 3 4 5 e
Wake of jet includes:Recirculation zoneVortex entrainment
Torrez, J. of Propul. & Power, 2011
16
Prior to last year – flight dynamics of climbing flight
acceleraIon from Mach 7 to Mach 11, constant q & mass flow air trajectory minimizes total fuel consumpIon
stable – to – unstable transiIon at Mach 9.5
due to increased α
Prior to last year = one journal paper, 12 AIAA conference papers
“Reduced-‐Order Modeling … Inlets”, J. of Propul. &Power, 2010.
Last year = 4 journal papers, 5 AIAA conference papers 1. “Reduced Order Modeling … Scramjet”, J. of Propul. & Power, 2011. 2. “Reduced Order Modeling.. .Nozzles”, J. of Propul. & Power, 2012. 3. “Ram-‐Scram TransiIon…” J. of Propul. & Power, 2012. 4. “Minimum-‐Fuel Ascent…Using Surrogate OpImizaIon”, submifed to JSR, 2012.
“Ram-‐Scram TransiIon…on Ascent Trajectory”, to be submifed, 2012.
“Performance Analysis of Variable-‐Geometry Scramjet Inlets…” AIAA Paper 2011-‐5756 “Design of …Flowpaths …”, AIAA 2011-‐2380 “Turn Performance of a ….Hypersonic Vehicle”, AIAA 2011-‐6300
“Hypersonic Vehicle Flight Dynamics… ”, AIAA 2010-‐7930 “Flight Envelope CalculaIons of a Hypersonic Vehicle…” AIAA Paper 2011-‐2368
Uncertainty PropagaIon in Integrated Airframe-‐Propulsion Systems”, AIAA 2011-‐2394
PublicaNons, Awards
17
Sean Torrez won AIAA Airbreathing Propulsion Graduate Student Award at San Diego AIAA JPC
18
C. Last Year -‐ Results Ascent trajectory of MAX-‐1 trimmed vehicle 1. OpImizaIon study -‐ how to minimize fuel ? 2. MDO study – opImize geometry (surrogate vs. collocaIon) 3. Delivered to MIT -‐ linearized A, B matrices, poles and zeros
Generic hypersonic vehicle geometry (GHV, HiFIRE6) 8. Add 3-‐D shock interacIons to inlet (3-‐D Riemann problem) 9. Add ethylene fuel “lookup tables”
Operability Limits 4. When does ram-‐scram occur ? 5. How much do forces jump at ram-‐scram ? controllability ? 6. When does unstart occur ? 7. Can a “befer” trajectory avoid unstart ?
19
C. Last Year: Ascent trajectory – opNmizaNon -‐ minimize fuel
Dalle, Torrez, Driscoll, Bolender, Bowcuf, sub. to JSR
24 km Altitude 32 km
minimum fuel trajectory: q = 1 atm.
acceleraIon varies as:
24 km Altitude 32 km
ER varies as:
surrogate method: q = constant compute operaIng maps of: (ER, , α, δCE, …) = fcn (M, )
20
Last Year: MDO study -‐ opNmize geometry
Torrez, Dalle, Driscoll, AIAA 2011-‐5757
Study #1: vary locaIon of fuel injecIon, angle of wall divergence locaIon of wall divergence
Minimize fuel along a constant q = 1 atm. path
21
Study #2: Displace cowl verIcally (Δy) – change compression raIo Displace cowl horizontally (Δx) – change spillage, mass flow Rotate cowl lip (Δθ) -‐ shiD shock interacIon locaIons What geometry provides best compression ? Lowest fuel required ?
Last Year: MDO study -‐ opNmize geometry
Dalle, Torrez, Driscoll, AIAA 2011-‐575
Result for opNmizaNon study: opNmize flowpath
“Performance Analysis of Variable-‐Geometry Scramjet Inlets…” AIAA Paper 2011-‐5756
Varied: δ xcowl M∞ δ ycowl α δ Θflap
Maximized: Inlet recovery Factor (p02/ p01)
“Design of Dual-‐Mode Engine
Flowpaths …”,
AIAA 2011-‐2380
Thrust (normalized)
p3 in 105 Pa
downstream fueling
upstream fueling
Varied: fuel injector locaIon Maximized: thrust
“Design of Dual-‐Mode Engine Flowpaths …”, AIAA 2011-‐2380 22
Last year: Operability Limits: added ram-‐scram transiNon
Where is the thermal choking locaIon ? How to solve ODEs near the singular point ?
L’Hospital’s Rule: choking is where:
area change (given)
~ heat added by combusIon
Torrez, S., AIAA 2011-‐2380 23
24
Last Year: Operability Limits – when does Ram-‐Scram occur ?
AlItude (km)
Flight Mach Number
acceleraIon q =
0.6 atm 0.8 atm 1.0 atm
Computed ram-‐scram transiIon boundaries
MAX-‐1
If you require larger acceleraIon – you need larger ER -‐ stay thermally choked longer -‐ so: larger Flight Mach Number at ram-‐scram transiIon
25
-‐ at ram-‐scram, affecIng controllability
Torrez, S., AIAA 2011-‐2380
Last Year: DerivaNves of performance curves are disconNnuous
vehiclealtitude(km)
flight Mach number M
scramram
Thrust in kN
ram-‐scram boundary
26
Last Year – how large are the sudden jumps in forces at ram-‐scram ?
a challenge to the control system
before ram-‐scram
aDer ram-‐scram
inlet shocks jump during ram-‐scram
1
0. 5
wall pressure (atm)
x/H
x/H
before
aDer
wall pressure jumps down during ram-‐scram
27
Last Year: Operability Limits – when does unstart occur ? Unstart Margin = L1 / L2 L1 = distance between leading edge of shock train and entrance to isolator L2 = total length of isolator
station: 2 3 4
inlet isolator combustor nozzle
(c)
(b)
(a)
L2
L1
1. First must properly compute the thermally-‐choked (ram) case
2. Then MASIV predicts isolator back pressure (pback)
3. Length of shock train then computed from experimental relaIon
(L2 – L1) / H = funcIon (pback / pinlet )
28
For a short isolator, MASTRIM shows that as you accelerate too fast (2 m/s2) à you unstart before you reach ram-‐scram !
MASIV tells us one way to avoid unstart !
Unstart Margin
ßunstart
Flight Mach Number
MASIV predicts the proper acceleraIon path = lower ER path à reduce acceleraIon to 1 m/s2 for the predicted Ime period
MAX-‐1
Last year: added the flame-‐out limit
Our finite-‐rate chemistry goes “out” if :
combustor gas pressure or temperature is too low
combustor gas velocity is too large (scalar dissipaIon rate)
inlet provides insufficient compression raIo
lookup tables: chemical reacIon rate = funcIon (dissipaIon rate)
29
D. Next Year
1. Add Generic Hypersonic Vehicle:
2. Add 3-‐D inlet: 3-‐D Riemann wave interacIons, inward-‐turning add Tyler’s AFRL Vehicle Model Generator
3. Ethylene fuel: run GHV on ethylene fuel, speed up code
30
4. Uncertainty: apply uncertainty analysis to compute data
needed by Anu’s adapIve control model 5. IntegraNon of Cesnik’s aero thermo elasIc model with:
MASTRIM – trim code, external forces MASIV – propulsion code
6. InteracNon: with Anu Annaswamy’s AcIve-‐AdapIve Control Laboratory at MIT to couple MASTRIM to her adapIve control model
7. OpNmizaNon of more of the control variables
(cowl displacement, elevon size, fuel type)
D. Next Year (conNnued)
31
8. Operability limits -‐ flame out limits 9. ValidaNon: determine uncertainty of our unstart, ram-‐scram predicIons
10. MDO: opImize inlet panels , elevon size, combustor length for GHV
32
Next Year -‐ Provide Uncertainty -‐ to MIT for adapNve control
“Adaptive Control of Hypersonic Vehicles in the Presence of Thrust and Actuator Uncertainties”, T. Gibson, Anu Annaswamy, AIAA 2008-6961 “Uncertainty in Analysis of Integrated Airframe-Propulsion for Hypersonic Vehicles” N. Lamorte, D. Dalle, S. Torrez, J. Driscoll, AIAA 2011-2394
Λ = effect of thrust uncertainIes on the B matrix
Rs = rectangular saturaIon funcIon = ui (if ui < umax) d = disturbance term u = control input δthrust, δelevon
33
Next Year: 3-‐D Inlet -‐ wave interacNons for GHV
“Compound Compressible Flow for Hypersonic Inlets”, Mark Lewis, AIAA 2009-7304 “3-D Analysis of … Inward-Turning Inlets”, Malo-Molina, Gaitonde, AIAA J. 48, 2010 “…Hypersonic Inlet Rectangular to Ellipical…”, M. Smart, C. Trexler, JPP 20 2004
3-‐D inlet
“Compound Compressible Flow” and “Streamline Tracing”
Discrete stream tubes; Shapiro influence coefficients; Enforce cons. mass, mom, energy + surface tracing of 3-‐D shocks
34
IntegraNon with Cesnik’s aero thermo elasNcity model
-‐ Begin with X-‐43-‐like MAX-‐1 vehicle
-‐ Revisit Doman-‐Bolender problem – effect of longitudinal bending on inlet shocks, spillage, loss of thrust
-‐ Add more complex geometry and thermo elasIcity model
-‐ Add mulIple shock interacIons that were omifed in Doman-‐Bolender
-‐ Consider intense heat transfer where bow shock reflects from engine cowl
Next year: compute controllability
AlItude (h)
Flight Mach Number M∞
ram –scram*
structural – excessive forces on elevons
4 8 12
gas density too low for elevons
*At ram-‐scram limit: thermal choking b.c. suddenly disappears, Sudden change in wall pressures, forces, moments DerivaIves of thrust performance curves are disconInuous
vehiclealtitude(km)
flight Mach number M
scramram
Thrust in kN
Metric: Time-‐to-‐double amplitude, which varies along trajectory
engine unstart
35
36
Flight Dynamics of a High Speed Turn Trajectory
increasing the Mach number shortens the period of each mode
Pole Time to Double -‐2.24 0.31 1.84 0.38 -‐0.0058 120.0
Long term impact
MASTRIM is a validated flight dynamics plahorm upon which to build an adapNve control law
Anu Annaswamy (MIT): adapIve control methods to avoid unstart, flame out, and issues during ram-‐scram transiIon
-‐ needs a validated flight dynamics model
-‐ needs uncertainty analysis
-‐ flight dynamics model must be sufficiently fast, robust, and have sufficient control variables
-‐ “fundamental” aspect of our model is important – we can explain why the control system adapts in a certain way
37
Payoffs to the Air Force
• Research addresses important component of control design and analysis development
• Novel modeling approaches to support G&C design and analysis beginning in the vehicle conceptual design phase
• State-‐of-‐the art techniques in reduced-‐order and first-‐principles modeling
• Codes have been transiIoned to NASA AMES, AFRL, Boeing, and
Raytheon
38