서지주요정보
(A) stochastic particle Fokker-Planck method for multiscale rarefied gas flows = 멀티 스케일 희박기체 해석을 위한 확률론적 입자 포커-플랑크 방법 개발
서명 / 저자 (A) stochastic particle Fokker-Planck method for multiscale rarefied gas flows = 멀티 스케일 희박기체 해석을 위한 확률론적 입자 포커-플랑크 방법 개발 / Sanghun Kim.
발행사항 [대전 : 한국과학기술원, 2025].
Online Access 원문보기 원문인쇄

소장정보

등록번호

8043563

소장위치/청구기호

학술문화관(도서관)2층 학위논문

DAE 25003

휴대폰 전송

도서상태

이용가능(대출불가)

사유안내

반납예정일

리뷰정보

초록정보

The dissertation addresses the stochastic particle Fokker-Planck (FP) method for rarefied gas flows. Focusing on extending the capability of the FP method for practical engineering applications, various topics are covered in the dissertation. First, the dissertation presents several numerical results for rarefied gas flows using DSMC and FP methods. These results help to better understand the limitations and potential of the FP method for rarefied gas flows. Second, new FP models are devised to describe diatomic gases and gas mixtures. It is demonstrated that the new FP models show reasonable agreement with the DSMC method near equilibrium. Third, a new particle integration scheme is introduced to improve the computational efficiency of the FP method. It is observed that the new integration scheme provides better convergence behavior for cell size and time step than the DSMC method. Finally, the hybrid FP-DSMC approach is studied for efficient simulation of multiscale rarefied gas flows.

희박 기체 유동 해석을 위해 확률론적 입자 포커-플랑크(FP) 방법에 대한 코드 개발 연구와 수치 모델 개발 연구를 수행하였다. FP 방법의 공학적 응용 능력을 확장하는 데 중점을 두면서, 본 논문에서는 여러 소주제를 다루고 있다. 첫째, DSMC 방법과 FP 방법을 적용하여 다양한 케이스에 대한 수치 해석 결과를 제시한다. 이러한 수치 해석 결과들은 현재 FP 방법이 가지고 있는 한계와 앞으로의 가능성을 이해하는 데 큰 도움을 줄 수 있다. 둘째, 이원자 기체 및 기체 혼합물을 모사하기 위한 새로운 FP 모델을 제시한다. 수치 해석을 통해 새로운 FP 모델들이 평형 유동 근처에서 DSMC 해석 결과를 잘 재현하는 것을 확인하였다. 셋째, FP 방법의 계산 효율성을 향상시키기 위해 새로운 시간 적분 기법을 개발하였다. 새로운 수치 기법이 DSMC 방법보다 격자 크기 및 시간 간격에 대해 더 나은 수렴성을 제공하는 것을 확인하였다. 마지막으로, 멀티 스케일 희박 기체 유동 해석을 위해 하이브리드 FP-DSMC 해석 전략을 제시한다.

서지기타정보

서지기타정보
청구기호 {DAE 25003
형태사항 xii, 170 p : 삽화 ; 30 cm
언어 영어
일반주기 저자명의 한글표기: 김상훈
지도교수의 영문표기: Jun, Eun Ji
지도교수의 한글표기: 전은지
Including appendix
학위논문 학위논문(박사) - 한국과학기술원 : 항공우주공학과,
서지주기 References: p.161-166
주제 Rarefied gas flows
Boltzmann equation
Fokker-Planck equation
Langevin equation
Particle Monte-Carlo method
Direct simulation Monte-Carlo
희박 기체 유동
볼츠만 방정식
포커-플랑크 방정식
랑주뱅 방정식
입자 몬테카를로 방법
직접 모사법
QR CODE

책소개

전체보기

목차

전체보기

이 주제의 인기대출도서

Microscopic description of a gas.

Schematic of mean free path.

Classification of the flow regimes by Knudsen number.

Classification of numerical methods by Knudsen number

Stochastic process paths and their distribution at time T

A schematic diagram of the AMR technique on a Cartesian grid

A schematic diagram of the LTS technique when a particle crosses a cell face.

A flowchart of the adaptive particle Monte Carlo simulation.

Simulation cases.

Schematic diagram of a Couette flow.

Molecular parameters of argon for the Couette and Fourier flows

Numerical parameters of the Couette and Fourier flows.

Velocity profiles of a Couette flow: (a) Kn = 0.01, (b) Kn = 0.1, (c) Kn =1, and (d) Kn = 10.

Temperature profiles ofa Couette flow: (a) Kn = 0.01, (b) Kn = 0.1, (c) Kn=1, and (d) Kn = 10.

The relaxation times of different FP models at the midpoint of the Couette flow.

The skin friction coefficients ofthe Couette flow over a wide range of Knudsen numbers.

Schematic diagram of a Fourier flow

Temperature profiles ofa Fourier flow: (a) Kn =0.01, (b) Kn =0.1, (c)Kn=1, and (d) Kn = 10.

The relaxation times of different FP models at the midpoint ofthe Fourier flow

The Nusselt numbers of the Fourier flow over a wide range of Knudsen numbers

Numerical parameters for the sensitivity study of the Fourier flow.

An efficiency comparison between the DSMC and FP methods for different time step sizes: (a) The normalized L2-norm error of the heat fux and (b) the normalized CPU time per iteration per cell.

An efficiency comparison between the DSMC and FP methods for different cell sizes using the time step size of0.02To: (a) The normalized L2-norm error of the heat flux and (b) the CPU time per iteration per cell.

An efficiency comparison between the DSMC and FP methods for different cell sizes using the time step size of T0: (a) The normalized L2-norm error of the heat flux and (b) the normalized CPU time per iteration per cell.

Molecular parameters of argon for the supersonic and hypersonic flows around a cylinder.

Boundary conditions for the supersonic and hypersonic flows around a cylinder

A comparison of numerical parameters between uniform and adaptive simulations.

Adaptive simulation parameters.

The total number of particles as a function of the number of time steps for the uniform and adaptive simulations.

Example of adaptive simulation using ES-FP: (a) grid configuration and (b) local time step size contour.

Validation of the adaptive ES-FP method for Mach 10: (a) The temperature contours around a cylinder and (b) the heat flux on the surface.

Comparative results for various DSMC codes for Mach 10. The numbers in parentheses represent the absolute percentage error with respect to the uniform DSMC method.

A comparison of the computational efficiency between the uniform and adaptive simula- tions.

Temperature contours around a cylinder for Mach 2: (a) Linear-FP, (b) Cubic-FP, (c) ES-FP, and (d) Quad-EFP.

Temperature contours around a cylinder for Mach 10: (a) Linear-FP, (b) Cubic-FP, (c) ES-FP, and (d) Quad-EFP.

Temperature contours around a cylinder for Mach 25: (a) Linear-FP, (b) Cubic-FP, (c ES-FP, and (d) Quad-EFP.

Temperature distribution along the stagnation line: (a)Ma=2, (b)Ma=5, (c)Ma = 10, and (d) Ma =25.

Prandtl number of the ES-FP model along the stagnation line.

Temperature contours for Ma = 10 using the DSMC method. It is again drawn to mark the locations where the velocity distributions are extracted.

Thermal velocity distribution at different locations along the stagnation line for Ma = 10: (a) at Location A, (b) at Location B, (c) at Location C, (d) at Location D, (e) at Location E, and (f) at Location F, as shown in Figure 3.25. The Maxwell-Boltzmann distribution is evaluated with the temperature ofthe DSMC method.

Comparison of total drag and peak heat fux between the DSMC and FP methods. The numbers in parentheses represent absolute percentage errors with respect to the DSMC method.

Comparison of computational efficiency between the DSMCand FP methods. The numbers in parentheses are the normalized CPU times with respect to that of the DSMC method.

Numerical parameters for the sensitivity study of the hypersonic flow around a cylinder.

The percentage error of the peak heat fux for the cell and time step sizes for Ma = 10. The LTS criteria are (a) At = 0.1Tlocal, (b) At = 0.21local, (c) At = 0.47local, and (d) At= 0.8Tlocal.

Pseudo-code for the ESFP and USP-ESFP models.

Molecular parameters of argon.

Shear stress relaxation using different time steps: (a) ESFP and (b) USP-ESFP.

Heat flx relaxation using different time steps: (a) ESFP and (b) USP-ESFP

Numerical parameters in the Couette flow.

Temperature profile in the Couette flow using DSMC, ESFP, and USP-ESFP: (a) sensitivity to cell size and (b) sensitivity to time step.

Wall shear stress in the Couette flow using DSMC, ESFP, and USP-ESFP: (a) sensitivity to cell size and (b) sensitivity to time step.

Numerical parameters in the Poiseuille flow.

Velocity profile in the Poiseuille flow using DSMC and USP-ESFP: (a)Kn=0.01 and (b) Kn= 0.001.

The L2-norm error of bulk velocity in the Poiseuille flow using DSMC and USP-ESFP. The Knudsen number is 0.001.

Comparison of surface properties and computational time in the Poiseuille flow.

Simulation cases for the velocity perturbation problem.

Density profile in the velocity perturbation problem for varying time steps: (a) ESFP (Group F) and (b) USP-ESFP (Group G).

Density profile in the velocity perturbation problem for varying cell sizes: (a) ESFP (Group H) and (b) USP-ESFP (Group I).

The L2-norm error of the density in the velocity perturbation problem using ESFP and USP-ESFP: (a) with respect to time step and (b) with respect to cell size.

Computational domain of the hypersonic flow around a cylinder

Parameters for generating a computational domain in the hypersonic flows around a cylinder.

Numerical parameters in the hypersonic flows around a cylinder.

Temperature contours around a cylinder using DSMC and USP-ESFP: (a) Kn = 0.01 and (b) Kn = 0.002.

Temperature distribution along the stagnation line: (a) Kn = 0.01 and (b) Kn = 0.002.

Shear stress distribution on the surface: (a) Kn = 0.01 and (b) Kn = 0.002

Heat fux distribution on the surface: (a) Kn = 0.01 and (b) Kn = 0.002

Comparison ofsurface properties and computational time in the hypersonic flows around a cylinder. The numbers in parentheses represent the percentage error and relative calculation time with respect to the fine DSMC.

Molecular parameters of argon.

A flowchart ofthe FP-DSMC simulation. Nsteady denotes the number oftime steps required to reach a statistically stationary flow, and Nstopdenotes the number oftimesteps to end the simulation

CPU time consumed by the collision subroutine as a function oftime step: (a) normalized CPU time, and (b) relative CPU time.

Flow and boundary conditions for supersonic flow in a. nozzle.

An illustration ofthe computational domain for a micro convergent-divergent nozzle

Mach number and domain decomposition contours around a nozzle using DSMC and FP- DSMC methods: (a) Mach number contour and (b) FP/DSMC assignments in FP-DSMC simulation, where 0 indicates the DSMC region and 1 indicates the FP region.

Mach number distribution in a nozzle: (a) along the symmetric axis and (b) at the nozzle exit line.

The computational costs ofsupersonic flow in a nozzle. The numbers in parentheses denote the speed-up value with respect to the DSMC method.

An illustration of the computational domain for hypersonic flow around a cylinder.

Flow and boundary conditions for the hypersonic flow around a cylinder

Simulation cases for hypersonic flow around a cylinder.

Comparison oftemperatureand domain decomposition contours around a cylinder using two spatiotemporal resolutions: (a) DSMC temperature contour, (b) FP temperature contour, (c) FP-DSMC temperature contour, and (d) FP/DSMC assignments in the FP-DSMC simulation, where 0 indicates the DSMC region and 1 indicates the FP region.

Temperature distribution along the stagnation line for hypersonic flow around a cylinder: (a) results for (Axoo) At) = (2A.rref, 2Atref), (b) results for (Axoo, At) = (2A.rref, 4Atref).

Heat flxx distribution on the surface for hypersonic flow around a cylinder: (a) results for (2A.rref, 2Atref), (b) results for (2A.rref, 4Atref).

KnGLL contour for hypersonic flow around a cylinder: (a) results for (2Atrer,2At-et), (b results for (2Atrer,44t-et). The white contour line represents the domain decomposition.

Percentage error in the heat transfer rate on the surface: (a) an error curve with respect to the time step, and (b) normalized computational cost with respect to the percentage error.

Comparison of surface properties and computational costs for hypersonic flow around a cylinder. The numbers in parentheses represent the percentage error and speed-up value relative to the reference DSMC solution.

Simulation cases for hypersonic flow around a THAAD-like missile.

The THAAD-like missile configuration.

Flowfield structure around a missile with a lateral jet.

Mach number and domain decomposition contours around a missile: (a) Mach number contour obtained via reference DSMC (flood) and coarse FP-DSMC (black lines), and (b) FP/DSMC assignments in the FP-DSMC simulation, where 0 indicates the DSMC region and 1 indicates the FP region. The black lines also represent the Mach number contour lines from the coarse FP-DSMC solution.

Mach number distribution along 2 = 0.06 m and 乙 = 0.08 m: (a) results at 之 = 0.06m and (b) results at 2 = 0.08 m. The corresponding two lines are marked in Figure 5.14b.

Comparison of surface properties and computational costs for hypersonic flow around a missile. The numbers in parentheses represent the percentage error and speed-up value relative to the reference DSMC solution.

Simulation cases.

Schematic diagram of a normal shock wave.

Molecular parameters of argon for the normal shock wave.

Normalized density over a shock wave for the HS gas: (a) Ma=2,(b)Ma=5,(c)Ma= 10, and (d) Ma= 20.

Normalized temperature over a shock wave for the HS gas: (a) Ma =2, (b)Ma=5,(c) Ma = 10, and (d) Ma =20.

Normalized density over a shock wave for the VHS gas: (a) Ma =2, (b) Ma =5,(c) Ma = 10, and (d) Ma =20.

Shock thickness and asymmetry for the VHS gas: (a) shock thickness and (b) shock asymmetry.

Shock thickness and asymmetry for the VHS gas.

Normalized temperature over a shock wave for the VHS gas: (a) Ma=2, (b)Ma=5,(c Ma = 10, and (d) Ma =20.

Normalized Rxx over a shock wave for the VHS gas: (a) Ma = 2 and (b) Ma = 20.

Normalized 스 over a shock wave for the VHS gas: (a) Ma = 2 and (b) Ma = 20.

Molecular parameters of argon for hypersonic cylinder.

Temperature contours for the HS gas using the DSMC and FP methods: (a) Cubic-FP, (b Quad-EFP with PMax, (c) Quad-EFP with PG13, and (d) Quad-EFP with PG26·

Velocity and temperature along the stagnation line for the HS gas: (a) velocity and (b) temperature.

Temperature contours for the VHS gas using the DSMC and FP methods: (a) Cubic-FP (b) Quad-EFP with PMax) (c) Quad-EFP with PG13, and (d) Quad-EFP with PG26·

Velocity and temperature along the stagnation line for the VHS gas: (a) velocity and (b temperature.

Comparison of drag and peak heat transfer rate for the VHS gas.

Comparison of computational efficiency between the DSMC and FP methods.

Molecular parameters of nitrogen molecules

Isothermal relaxation: (a) normalized rotational energy and (b) normalized vibrational energy.

Relaxation to equilibrium temperature: (a) without using the conservative scheme and (b) using the conservative scheme.

Equilibrium distribution function: (a) thermal velocity, (b) rotational energy, and (c) vibrational energy.

Time correlation functions: (a) shear stress and (b) heat fluxes.

Numerical parameters of the Couette flow.

Couette flow: (a) bulk velocity and (b) translational temperature.

The computational domain of a hypersonic flow over a vertical flat plate

Temperature contours around a vertical flat plate: (a) translational temperature, (b) rotational temperature, and (c) vibrational temperature.

Macroscopic quantities along the stagnation line (y = 0): (a) number density and bulk velocity, and (b) temperatures.

Macroscopic quantities along the top line (y =1/2): (a) shear stress and (b) heat fluxes

Surface shear stress and heat flux: (a) shear stress and (b) heat fux.

A comparison of the computational costs of the DSMC and FPM methods over a wide range of time steps: (a) Total CPU time and (b) CPU time for the collision subroutine.

VHS molecular parameters.

Relaxations in homogeneous flow using the ESFP model: (a) velocity relaxation, (b) temperature relaxation, (c) shear stress relaxation, and (d) heat flx relaxation.

Relaxations in homogeneous flow using the ESFP+ model: (a) velocity relaxation, (b) temperature relaxation, (c) shear stress relaxation, and (d) heat flux relaxation.

Flow and boundary conditions in Poiseuille and Couette flows.

Velocity profiles in Poiseuille flow: (a) ESFP model and (b) ESFP+ model.

Shear stress profiles in Poiseuille flow: (a) ESFP model and (b) ESFP+ model.

Temperature profiles in Couette flow: (a) ESFP model and (b) ESFP+ model

Heat flx profiles in Couette flow: (a) ESFP model and (b) ESFP+ model

Comparison of aerodynamic forces on the wall in Couette flow. The numbers in parentheses represent the percentage error with respect to DSMC.

Flow and boundary conditions in hypersonic flow over a vertical flat plate.

Computational domain for hypersonic flow over a vertical flat plate

Temperature contours around a vertical flat plate at Kn = 0.02: (a) the ESFP model and (b) the ESFP+ model.

Temperature contours around a vertical flat plate at Kn = 0.1: (a) the ESFP model and (b) the ESFP+ model.

Temperature distribution along the stagnation line in hypersonic flow over a vertical flat plate at Kn = 0.02: (a) ESFP model and (b) ESFP+ model.

Temperature distribution along the stagnation line in hypersonic flow over a vertical flat plate at Kn = 0.1: (a) ESFP model and (b) ESFP+ model.

Flow and boundary conditions in hypersonic flow around a cylinder.

Temperature contours around a cylinder using ESFP+ model: (a) mixture temperature (b) helium gas temperature, (c) argon gas temperature, and (d) krypton gas temperature.

Temperature distributions along the stagnation line in hypersonic flow around a cylinder: (a) ESFP model and (b) ESFP+ model.

Heat fux distributions on the surface of a cylinder.

Computational cost comparison between DSMC and FP methods: (a) Total CPU time, and (b) CPU time by the collision subroutine.

Molecular parameters of argon for CFD.

Molecular parameters of argon for kinetic models.

Numerical parameters of Couette flow.

Velocity and temperature profiles between two walls in Couette flow (Kn = 0.001).

Temperature distribution of lid-driven cavity (Kn = 0.01). The black contour line is the DSMC solution.

X-velocity, temperature, and heat fux along the y-direction at the center (Kn = 0.01

Temperature distribution of lid-driven cavity (Kn = 0.1). The black contour line is the DSMC solution.

X-velocity, temperature, and heat fux along the y-direction at the center (Kn = 0.1)

Numerical parameters of hypersonic flat plate.

Temperature distribution over a flat plate (Kn = 0.01, Ma = 5).

Density, x-velocity, and temperature along the x-direction at the cell adjacent to the surface.

Drag and heat transfer rate on the plate.