A Reduced-Order Modeling Approach to Enable Kinetic ......High Fidelity Modeling: Roadmap...

Preview:

Citation preview

A Reduced-Order Modeling Approach to Enable Kinetic Simulations of Non-equilibrium

Hypersonic Flows

Marco Panesi

AFOSR YIP

Grant No: FA9550-15-1-0132 DEF

Department of Aerospace Engineering

University of Illinois at Urbana-Champaign

1

2

Outline

• Motivation and Background • Master Equation Analysis • Coarse Grained Model

• Conclusions

3

Standard Non-equilibrium Models

• Standard non-equilibrium models for hypersonic flows were mainly developed in the 1980’s and are correlation based:

E.g., dissociation model of Park Multi-temperature model: Average temperature for fictitious Arrhenius rate coefficient

4

Motivation

• A large effort is underway at AIRFOCE which attempts to characterize the microscopic interaction of N2-N2, O2-O2 and N2-O2 from first-principles calculations.

• Ab initio calculations can provide the transition probabilities governing the transfer of energy between the flow and the internal energy modes of atoms and molecules in the gas.

• The large amount of information provided by ab initio calculations has great value, but it must be tailored to fulfill the needs of the problem that is being solved.

• Thus, it is imperative that reduced-order models be developed.

5

Objective

• METHODOLOGY: Developing non-equilibrium models for hypersonic flows based on microscopic theory and applying them to macroscopic scale.

Work at the interface between computational chemistry, experimental data, and CFD.

6

Background: State-to-State Kinetics

MT Models: Conventional methodologies rely on the assumption of Maxwell-Boltzmann distribution:

State to State Models: the internal states are treated as independent species governed by their own kinetics.

Boltzmann Plot

James Clerk Maxwell Ludwig Eduard Boltzmann

7

High Fidelity Modeling: Roadmap

Objectives: To assess the fundamental assumptions adopted in the modeling of hypersonic plasma flows.

Key Relaxation Processes: 1. Energy Transfer: It is crucial to the understanding on the shock layer kinetics 2. Dissociation: critical process governing the redistribution of the kinetic

energy within the internal energy modes and chemistry. 3. Recombination: critical in the boundary layer area, and in the expansion

regions of the flow-field.

Dissociation N2 + N = N + N + N • Rotational equilibrium (T = Trot) • Landau-Teller VT relaxation model • Internal Energy – Chemistry relaxation coupling (e.g., VC) • Existence of a QSS rates • … Rovibrational State-to-state method

8

Analysis of Dissociating and Recombining Flows

Test cases under consideration: 1. Master Equation 2. Flow Behind a normal shock wave 3. Quasi 1D nozzle flow

9

A Novel Approach to the Modeling of Non-equilibrium Flows

First Principles Computation: 1. Quantum chemistry calculations to

generate realistic nuclear interaction potentials (PES)

2. Quasi classical trajectory (QCT) method for the reaction cross-sections

10

Non-equilibrium Flow Behind a Normal Shock Wave

11

Flow-field Quantities

• Rovibrational STS model predicts larger relaxation distance with respect to the vibrational STS model

12

Post-Shock: Rovibrational Populations

• The distribution deviates from the Maxwell Boltzmann distribution • Distribution is dissected into multiple strands for different v

13

1D Shock Tube Problem

Left: rotational and vibrational temperatures Right: population of the first vibrational levels

• Assumption of fast rotational equilibrium is questionable • Dissociation is better described by a unique temperature

14

Master Equation Solution

1. Rate Coefficient is in EXCELLENT agreement with Appleton data 2. Exchange reaction is important for correct estimation of reaction rate constant. 3. In the high temperature region the QSS assumption FAILS!

QSS Rates Estimation

15

Comparison MT and STS Models

Can we use QCT derived rate coefficients and relaxation parameters in the conventional MT models?

NO! Using QCT derived rates based on the QSS assumption (or Boltzmann) are unable to reproduce the STS results.

16

Conclusions of the STS Analysis

MT modeling (QSS based) • Conventional MT models are unable to reproduce the STS results,

because of the invalidity of the QSS assumption.

17

Model Reduction

State To State - RVC

State To State - VC

STS - EC

MT

Accuracy

Complexity

CGM

Increasing Number of Assumptions

T, TR,TV, TE

T, TR,TV

T, TR

T

18

Coarse Grained Method

The methodology of reduction consists of two distinct steps: 1. Local Representation and Reconstruction. It relies on the lumping of the

internal energy levels in macroscopic energy groups and the reconstruction of the population of each grouped state, ni, using macroscopic quantities.

The coefficients and are retrieved using constraints based on the maximum entropy principle and a variational method.

19

Coarse Grained Method

2. Macroscopic Moment Equations and Rate Coefficients. Macroscopic

governing equations (referred to as macroscopic moment equations) are obtained by taking moments of the master equations and by using the reconstructed local representation.

• Zero Order Moment: Uniform Grouping (piece-wise reconstruction).

• First Order Moment: Boltzmann Grouping (linear reconstruction).

Governing Equations

20

Coarse Grained Method

• Novel lumping scheme obtained by sorting the levels by energy and grouping in a bin all levels with similar energies

21

…reconstruction of the population of each grouped state, ni, using macroscopic quantities.

Coarse Grained Method

22

Considerations

1. Grouping strategy is crucial • The choice of the grouping or grouping strategy should be guided by

the physical intuition. • For example, levels characterized by similar energies are likely to be in

equilibrium between each other and should be grouped together.

2. State to state models and Multi-temperature models are a particular case of coarse grained approach.

• For example, conventional TTv model can be obtained by grouping the

vibrational levels in a single bin and prescribing a Boltzmann distribution for each rotational level with T=TRot

23

Conventional Model and Coarse Grain Modeling

MT is a particular case of coarse grain model (1 Group)

Boltzmann distribution (TVib, TRot) Conventional TTVib model if TRot=T

MT

VCR is a particular case of coarse grain model (n Group)

Boltzmann distribution (Trot) Conventional Vib. STS model

when TRot = T

VCR(n)

24

Novel Grouping Strategy

Two groups in the vibrational energy structure

Two different “rotational” temperature for the two groups

VCR(i) Vibrational specific model

Hyb(2,2)

VCR Three energy groups Three internal energy equations MT is a special case (1 internal

temperature)

BC(3)

25

Evolution of the Vibrational Distribution

MT models are unable to predict the distribution function

26

Evolution of the Vibrational Distribution

BC(3) model is in good agreement with the STS model

27

Evolution of the Vibrational Distribution

HyBVC shows excellent agreement with the STS model

28

Evolution of the Vibrational Distribution

VCR2 shows excellent agreement with the STS model

29

Technical Challenges Remaining

• Diatom-Diatom Reactions Given the large number of possible channels the derivation of the “exact” rovibrational STS model is not feasible.

• Analysis of Recombining Flows Challenges are due to the strong deviation from the equilibrium distribution in expanding flows.

• Application to CFD (e.g., US3D) and Validation Validation data should include spatially resolved population measurements of the (ro-) vibrational population and atomic densities. (E.g., S.Sharma, et al. JTHT, Vol. 7, No. 4 (1993), pp. 697-703. )

• Other systems, gas mixtures, higher order reconstruction

30

Publications and Honors

Journal Publications 1. A., Munafo, Y. Liu, M. Panesi “Physics of dissociation and energy transfer in

shock heated nitrogen flows”, Physics of Fluids, Under Review, (2015) 2. Y. Liu, M. Panesi, A. Sahai and M. Vinokur “General multi-group macroscopic

modeling for thermo-chemical non-equilibrium gas mixtures” J. Chem. Phys. 142, 134109 (2015);

3. Panesi, M., Munafo, A., Magin, T. E., and Jaffe, R. L., ”Study of the non-equilibrium shock heated nitrogen flows using a rovibrational state-to-state method”, Phys. Rev. E, Vol. 90, 013009 (2014).

4. Panesi, M. Jaffe, R.L. Schwenke, D.W. Magin, T.E. 2013 Rovibrational internal energy transfer and dissociation of N2-N system in hypersonic flows. J. Chem. Phys. 138, 044312 (2013).

Research Honors 1. 2015 Air Force Summer Faculty Fellowship Program, California, USA. (2015) 2. 2015 Award on Physical Modelling (8th Symposium on Aerothermodynamics

for Space Vehicles - ESA) (2015) 3. 2015 Air Force Young Investigator Award (YIP) (2015)

31

Conclusions

• Using the classical moment method, we introduced a general methodology for modeling thermal and chemical non-equilibrium processes. Based on the maximum entropy principle subject to a series of moment constraints, the logarithm of the distribution function in each energy group is expressed and reconstructed as a power series in internal energy.

• Conventional MT and STS models are “only” particular cases of the more general Coarse-Grain Method.

• These models have been applied to the study of rovibrational energy

excitation and dissociation processes behind strong one-dimensional shock waves in nitrogen flow.

32

NEQRAD Group

33

Acknowledgments

Special thanks to: UIUC

• Dr. A. Munafo (UIUC) • Dr. R. Macdonald (UIUC) - NDSEG • Dr. S. Venturi (UIUC)

NASA

• Dr. R.L. Jaffe (NASA Ames Research Center) • Dr. D.W. Schwenke (NASA Ames Research Center) • Dr. Y. Liu (NASA Ames Research Center) AIRFORCE • Dr. J.L. Cambier (USAF, AFOSR) • Dr. E. Josyula (USAF, AFRL Wright Patterson)

AFOSR YIP

Grant No: FA9550-15-1-0132 DEF

34

Backup Slides

35

Summary (and Conclusions)

36

Summary (and Conclusions)

Significant reduction of the CPU time is obtained with the Bin model

N2-N System: CPU Time in function of the # of BINS

37

Convective Heating

The MT model over-estimate the convective heating by 18 % if the parameters are calibrated using the RVC model

Park Model over-predict by a factor 2

38

Detecting QSS Breakdown

39

Coarse Grained Method

• Novel lumping scheme obtained by sorting the levels by energy and grouping in a bin all levels with similar energies

40

Energy Transfer

Recommended