Characterization of a Hypersonic Low-density
Transcription
Characterization of a Hypersonic Low-density
Characterization of a Hypersonic Low-density Flow for Separated Flow Investigations S. O’Byrne1 , S. L. Gai1 N. R. Deepak1 Y. Krishna1 J. N. Moss2 1 School of Engineering & IT University of New South Wales Australian Defence Force Academy Canberra, Australia 2 NASA Langley Research Center Hampton, VA, USA US Air Force Aerothermodynamics and Turbulence Portfolio Review, July 2013 S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 1 / 27 Introduction Motivation Motivation High Enthalpy, Low-Density Separated Flows Better understanding of viscous flow in thermal nonequilibrium — critical to successful vehicle design Typical flow separation configurations — compression corners, flat-plates with steps, blunt bodies Configurations all have a pre-existing boundary layer at separation — increased complexity and not accounted for in analytical descriptions ‘Tick’ (or ‘Check’) Geometric Configuration Capable of producing separation at leading-edge without pre-existing boundary layer Originally proposed by Chapman et al. (1958) for high Re and low M∞ flows Considered here for low-density laminar high-enthalpy hypersonic conditions Approach to the Problem Comparisons between 3D Navier-Stokes and 2D Direct Simulation Monte Carlo (DSMC) calculations Using a suite of flowfield diagnostics to investigate thermal state of flowfield S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 2 / 27 Introduction Flow Features — Leading Edge Separation Leading-Edge Separation Expansion fan Flow separation Characterized by a strong expansion at the leading edge Reattachment Flow separation very close to the leading edge forming a recirculation region between A, B and C Re-compression shock wave Reattachment on the compression surface Recirculating region S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 3 / 27 Introduction Scope Goal of Research Project Improved Understanding of Viscous Aerothermodynamics A flow configuration without boundary layer development before separation Testing of Chapman’s isentropic recompression theory to estimate the base pressure To aid in designing the experiments based on numerical results Investigate the flowfield using the following techniques Tuneable diode laser absorption spectroscopy of H2 O Time-resolved characterization of freestream temperature and H2 O concentration on each tunnel run. Resonantly enhanced shearing interferometry (RESI) Visualizes the time-evolution of the low-density flow NO PLIF two-line thermometry (rotational and vibrational) and velocimetry Determines rotational, translational and NO vibrational temperatures, and 2 velocity component map. Dual-pump CARS Determines vibrational temperatures of N2 and O2 in freestream S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 4 / 27 Introduction Scope Techniques (a) RESI (Hruschka et al. Appl. Opt. 2008) (c) (b) NO PLIF Thermometry (Hruschka et al. Exp. Fluids 2011) NO PLIF Velocimetry (Hruschka et al. Exp. (d) Fluids 2011) S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) CARS (Fraval et al. (Proc. ISSW, 1999) 5 / 27 Computational Approach Navier-Stokes & Direct Simulation Monte Carlo Numerical Codes & Models Navier-Stokes (N-S) Solver — Eilmer-3 (Jacobs and Gollan, 2010) In-house solver, time-dependent, viscous, chemically reactive Finite-volume, cell-centred, 3D/axisymmetric discretization Second order spatial accuracy: modified van Albada limiter and MUSCL Mass, momentum & energy flux across the cells: AUSMDV algorithm Time Integration: Explicit time integration Direct Simulation Monte Carlo (DSMC) — DS2V (Bird, 2006) Uses probabilistic (Monte Carlo) simulation of physical collisions Models fluid flow using simulated molecules which each represent a large number of real molecules S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 6 / 27 Computational Approach Geometric Configurations Leading-Edge Separation Geometry (a) Geometry (b) Model Configuration Details Surface A ←→ B (S-1; expansion) B ←→ C (S-2; compression) Length, mm 19.730 44.776 Angle θ − 1=30.5◦ θ − 2=23.7◦ Horizontal x distance from A ←→ C = 58 mm Total wetted surface (s) length A 7−→ B 7−→ C = 64.506 mm S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 7 / 27 Computational Approach Geometric Configurations Leading-Edge Flow Conditions T-ADFA Freestream Conditions Flow Parameter Test gas Re [1/m] M∞ u∞ [m/s] T∞ [K] p∞ [Pa] ρ∞ [kg/m3 ] H0 [MJ/kg] Condition E Air 1.34 × 106 9.66 2503 165 290 0.006 3.8 Condition A Air 2.43 × 105 7.25 3730 593 377 0.002 13.3 Two conditions are produced using the same facility and nozzle Condition E Low enthalpy — near perfect gas behaviour Continuum assumptions valid Condition A Higher-Knudsen number and significant slip in temperature and velocity at surface Significant thermal nonequilibrium S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 8 / 27 Computational Approach Geometric Configurations Chapman’s Isentropic Re-compression Theory Chapman et al. (1958) proposed a separated flow model and developed a theory to estimate the base pressure (Pd ) Experimental evidence at high supersonic Mach numbers suggests that the model works remarkably well even for pre-existing boundary layer in estimating base pressure In hypersonic high temperature flows, the efficacy of Chapman’s isentropic recompression model is not “rigorously” verified Here, the same leading edge separation model used by Chapman is considered S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 9 / 27 Computational Results Knudsen number Local Knudsen (Kn) number (a) DSMC: Cond E (b) Navier-Stokes: Cond E (c) DSMC: Cond A (d) Navier-Stokes: Cond A n λ Kn = ∂ρ 2 ∂x + ρ o1/2 ∂ρ 2 ∂y local S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 10 / 27 Computational Results Results: Condition E Pressure, Skin Friction & Heat Flux: Condition E Ho =3.8 MJ/kg Strong expansion at leading edge; followed by flow separation Separation: N-S (s/Le = 0.145); DSMC (s/Le = 0.08) Reattachment: N-S (s/Le = 2.42); DSMC (s/Le = 2.46) S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 11 / 27 Computational Results Results: Condition A Pressure, Skin Friction & Heat Flux: Condition A Ho =13.3 MJ/kg Between 0.05 ≤ s/Le ≤ 0.25 in N-S, rate of pressure reduction decreases with near constant pressure: Indicative of boundary layer growth N-S: Separation at s/Le = 0.56; Reattachment at s/Le = 1.87 DSMC: No indication of separation/reattachment S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 12 / 27 Computational Results Results: Further remarks Differences between N-S and DSMC Significant differences between N-S and DSMC for condition A DSMC predicts almost no separation (except for an infinitesimally small region at the corner) Flow over most of the expansion surface is in slip flow regime (Moss et al., 2012) slip velocity = uw (s) = λw ∂u ∂y slip temperature = Tg − Tw = (∆T )w = M √ Rarefaction parameter(V ) = √ C Res = w λw τw (s) µw 2γ (µw cp )−1 λw k γ+1 ρus ; Res = µ dT dy w ρw µw ;C = ρe µe Criterion for slip flow (Talbot, 1963) for Condition A Location s/Le = 0.25 s/Le = 0.5 s/Le = 1.0 Rarefaction parameter, V N-S 0.143 0.1009 0.0715 S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) Rarefaction parameter, V DSMC 0.332 0.223 0.166 13 / 27 Computational Results Comparison with Champan’s Theory Comparison with Chapman’s Isentropic Re-compression Theory From Navier-Stokes Simulations Average pressure in the recirculation region or dead air region (Pd ) Pressure (P 0 ) and Mach number (M 0 ) downstream of reattachment Mach number at the edge of the mixing layer (Me) M 02 = (1 − ud∗2 )Me2 pd = p0 Flow condition E A and ud∗ = ud /ue )M 02 1 + ( γ−1 2 γ/(γ−1) 1 + ( γ−1 )M 02 /(1 − ud∗2 ) 2 pd /p 0 N-S simulations 0.088 0.08 pd /p 0 DSMC simulations 0.09 - pd /p 0 Theory 0.330 0.345 0 0 0 ReLe = u µρ 0Le 32.88 ×103 8.14 ×103 Simple isentropic flow assumption does not appear to hold in hypersonic flow Streamlines in the shear layer do not recompress isentropically at reattachment, rather extend over a finite region Steep isentropic recompression assumption in theory seems unrealistic in low Re flows with thick shear layers S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 14 / 27 Computational Results Streamlines Streamlines, Separation and Reattachment Angles (a) Condition E (b) Condition A Comparison of measured angle with Oswatitsch (1957) theory dτw /ds v tanθs = limx ,y →0 u = −3 dpw /ds s Angle Separation Reattachment Condition A Theory 37◦ 7.5◦ Condition A Measured 35◦ 10◦ S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) Condition E Theory 47◦ 1.4◦ Condition E Measured 40◦ 4◦ 15 / 27 Computational Visualization Computational Visualization: Resultant Velocity Condition E and Condition A (a) Navier-Stokes (b) Direct simulation Monte Carlo (c) Navier-Stokes (d) Direct simulation Monte Carlo S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 16 / 27 Driver Gas Detection Driver Gas Contamination Measurements Characteristics of Contamination Contamination is strongly dependent on piston motion tuning and interaction of reflected shock wave and contact surface between driver and test gases Test time is significantly shorter than mass flux alone would indicate Test time may fluctuate from one tunnel run to the next Driver gas is argon/helium mixture — no absorption lines Use absorption in water vapour as an inverse measurement Gai (1992) S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 17 / 27 Driver Gas Detection TDLAS arrangement Diode Laser Absorption Arrangement Characteristics Scan rates over 2 transitions at more than 50 kHz, with a single laser Can measure average temperature, velocity and H2 O concentration Even when bottled dry air is used,measurable signal can be obtained Opposed retroreflector geometry ensures precise velocity measurement S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 18 / 27 Driver Gas Detection TDLAS arrangement Diode Laser Absorption Arrangement 1 Nozzle 4 Mirror 2 Pitch fiber 5 Collecting lens S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 3 Collimating lens 6 Catch fiber 19 / 27 Driver Gas Detection TDLAS Measurements Diode Laser Absorption Measurements Assumptions The water vapour concentration is constant in the test gas Reduction in water concentration produces an equivalent increase in driver gas Lines are Doppler broadened At 280–380 Pa, this is a good assumption FWHMDoppler = 3.12 pm, Voigt fit only 3 % different No change in temperature during the test time Not true, but can be corrected for with simultaneous measured temperature S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 20 / 27 Driver Gas Detection TDLAS Measurements Contamination vs time (a) Condition E (b) Condition A Driver gas time There is a region of constant concentration where the test gas passes through the nozzle A gradual decrease followed by an increase in H2 O at later times The increase occurs due to test gas reflecting from the end of the dump tank Test time is considerably shorter than the mass flux estimates of 5 ms and 2.5 ms for conditions E and A respectively S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 21 / 27 Driver Gas Detection Repeatability Repeatability (a) Condition E (b) Condition A Repeatability Condition E: Test gas duration = 1120 ± 60µs — CFD: 700 µs Condition A: Test gas duration = 700 ± 130µs — CFD: 600 µs S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 22 / 27 Driver Gas Detection Velocity Velocity Measurements (a) Doppler-shifted peaks S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) (b) Velocity decay 23 / 27 Conclusions Conclusions Computations Lower enthalpy (higher freestream density) condition E : DSMC predicted a 15 % larger separated region. Pressure, shear stress and heat flux show similar features. Higher enthalpy (lower freestream density) condition A : N-S results predicted a clearly separated region whereas the DSMC indicated a much smaller region DSMC smaller separation for condition A is attributed to the fact that the DSMC calculations take slip effects into account Isentropic recompression theory of Chapman may not be adequate in hypersonic high enthalpy low Reynolds number flows Experiments We have developed a method for measuring H2 O concentration and inferring driver gas contamination using tuneable diode laser absorption spectroscopy. This is the first of a number of techniques to be brought to bear on this problem. Provides on-line measurement on each tunnel run, while other experiments are running Flow duration at condition E is sufficient, but is borderline at condition A. May need to reduce forebody size to compensate for this. S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 24 / 27 Conclusions Challenges Challenges Things to consider. . . Need more collaboration from Navier-Stokes method experts Would be interested in co-operating with larger/higher-enthalpy facilities on this configuration Need to develop good information transfer between experimental and computational investigators Need to know what measured properties are particularly important for testing computational techniques Previous experimental data shows early transition to turbulence in free shear layers Main practical problem is manpower CARS will take a long time to set up and characterize, and requires development of nonequilibrium CARS spectral models. It is a low-yield method. PLIF velocimetry requires lots of tunnel runs S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 25 / 27 Thanks and Acknowledgments Thank you Acknowledgments US Air Force AOARD (Dr I. Wysong) Australian Space Research Program Dr. Peter Jacobs (University of Queensland) — Eilmer Prof. Graeme Bird — DS2V UNSW Silver Star Research Grant For more information Dr. Sean O’Byrne [email protected] Dr. Deepak Ramanath [email protected] Mr. Yedhu Krishna [email protected] School of Engineering & IT University of New South Wales Australian Defence Force Academy Canberra, Australia S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 26 / 27 References References Bird, G. A. (2006), ‘DS2V: Visual DSMC Program for Two-Dimensional and Axially Symmetric Flows’. URL: http://gab.com.au/index.html Chapman, D. R., Kuehn, D. M. and Larson, H. K. (1958), Investigation of Separated Flows in Supersonic and Subsonic Streams with Emphasis on the Effect of Transition, Technical Report 1356, NACA. Jacobs, P. A. and Gollan, R. J. (2010), The Eilmer3 Code, Technical Report Report 2008/07, Department of Mechanical Engineering, University of Queensland. Moss, J. N., O’ Byrne., S., Deepak, N. R. and Gai, S. L. (2012), Simulation of Hypersonic, High-Enthalpy Separated Flow over a ’Tick’ Configuration, 28th International Symposium on Rarefied Gas Dynamics, Zaragoza, July 9-13th, 2012. Oswatitsch, K. (1957), Die Ablosungsbedingung von Grenzschichten, in ‘Boundary Layer Research: International Union of Theoretical and Applied Mechanics Symposium, Freiburg’, Springer–Verlag, Berlin, pp. 357–367. Talbot, L. (1963), ‘Criterion for Slip near the Leading Edge of a Flat Plate in Hypersonic Flow’, AIAA Journal 1(5), 1169–1171. S. O’ Byrne (University of New South Wales, Australian Defence Force Academy) 27 / 27