on numerical studies of explosion and implosion in air · 2.4 test 4: exact solution (solid line)...
TRANSCRIPT
On Numerical Studies of
Explosion and Implosion in Air
FU Sau-chung
A Thesis Submitted in Partial Fulfilment
of the Requirements for the Degree of
Master of Philosophy
in
Mathematics
© T h e Chinese University of Hong Kong
June 2006
The Chinese University of Hong Kong holds the copyright of this thesis. Any per-
son(s) intending to use a part or whole of the materials in the thesis in a proposed
publication must seek copyright release from the Dean of the Graduate School.
/1:./ - i “ � V i 圖 V:r\
• 11 oci w )i: 參^ ^ ^ ^ L 丨 , ‘ : : � � T Y一 扇 j
Thesis/Assessment Committee
Professor ZOU Jun (Chair)
Professor XIN Zhou Ping (Thesis Supervisor)
Professor AU Kwok Keung Thomas (Committee Member)
Professor WANG Xiao Ping (External Examiner)
Abstract of thesis entitled:
On Numerical Studies of Explosion and Implosion in Air
Submitted by FU Sau-chung
for the degree of Master of Philosophy
at The Chinese University of Hong Kong in June 2006
Explosion and implosion in air are rigorous fluid dynamics phenomena and
hence, those studies are highly restricted in a laboratory. Computational fluid
dynamics becomes an important tool to investigate such problems. Numerical
calculations of hyperbolic conservation laws play an important role in such
problems. The resolution of such numerical schemes is highly improved, since
the introduction of a class of non-oscillatory explicit second-order accurate
finite difference schemes by using the concept of total variation diminishing
(TVD). Recently, the development of a class of numerical schemes so called
relaxation schemes contributed the field on its simplicity, without using either
Riemann solvers spatially or a nonlinear system of algebraic equation solvers
temporally. The leading order approximation of the relaxation schemes, so
called the relaxed scheme when applying to the one-dimensional scalar conser-
vation law was proved to be TVD.
In this research, the problem of explosion and implosion in air with spher-
ically or cylindrically symmetric flow is studied numerically by using the re-
laxed scheme. The governing equations are basically the Euler equations with
a geometric source term. A splitting method is employed to deal with the inho-
mogeneous equation. The numerical dissipation of the relaxation scheme is to
a large extent determined by a matrix A which is constructed when applying
the scheme to the problem. The appropriate choice of matrix A is critical in
i
the accuracy and practicality of the scheme. An algorithm to choose the ma-
trix and technique in applying the relaxed scheme is developed in addition to
maintaining the good features of the relaxed scheme. Some numerical tests are
carried out to test the reliability of the algorithm to the Euler system. Numer-
ical simulation of four cases namely spherical explosion, cylindrical explosion,
spherical implosion and cylindrical implosion are discussed. The numerical
results are presented and compared with some previous numerical, analytical
and experimental results.
ii
摘要
空中向外爆炸及向內爆炸是很劇烈的流體動力學現象,所以,在實驗室內做有
關的硏究是有一定程度的困難,於是,計算流體動力學便成爲了硏究有關課題
的一個重要的工具,而雙曲守恆律的數値計算在此類問題中擔任了重要的角
色。自從引進了全變差下降(TVD)的槪念而發展的一類非振還、顯式、二階精
度差分格式後,數値計算格式的分解能力大大地提升。最近,一類名爲鬆弛格
式(Relaxation Scheme),它免去運用空間上的黎曼解算子或是時間上的非線性代
數方程組解算子,這簡單性給予數値計算這範疇上很大的意義。當應用在一維
單守恆律時,鬆弛格式的首階近似,又名放鬆格式(Relaxed Scheme),已證明是
TVD 的。
在本硏究中,放鬆格式將用來探討球面對稱及圓柱對稱的空中向外及向內爆炸
的問題。問題的控制方程爲歐拉方程帶有一個幾何有關的源項,這非齊次方程
將用分裂法來處理。鬆弛格式的數値耗散很大程度基於運用這格式時所構造的
一個矩陣‘A’,格式的準確度及實用性取決於能否適當地構造它。在保持放鬆格
式的優點這前提下,本論文提供了構造‘A’的算法及應用放鬆格式的技巧,並使
用實際的數値測驗來證實了本算法的可靠性,並作了四個數値模擬:球面對稱
的向外爆炸、圓柱對稱的向外爆炸、球面對稱的向內爆炸及圓柱對稱的向內爆
炸。計算結果將會跟以前數値計算的、分析的及實驗的結果作一比較。
Acknowledgement
I would like to take this opportunity to express the deepest gratitude to my
research supervisor Prof. Z. P. Xin for his guidance and generous support
to me. Many thanks are given to him for giving me the chance to study
in this field. Also, thanks must go to my colleagues in the office for their
encouragement and inspiring discussion.
This research has been supported by The Institute of Mathematical Sci-
ences (IMS), The Chinese University of Hong Kong. I am much grateful to
IMS for the award of a postgraduate studentship to me.
iii
“The heavens declare the glory of God;
the skies proclaim the work of his hands. ”
Psalms 19.1 (NIV)
iv
Contents
Abstract i
Acknowledgement iii
1 Introduction 1
1.1 Background of Explosion and Implosion Problems 1
1.2 Background of the Development of Numerical Schemes 2
1.3 Organization of the Thesis 5
2 Governing Equations and Numerical Schemes 6
2.1 Governing Equations G
2.2 Numerical Schemes 8
2.2.1 Splitting Scheme for Partial Differential Equations with
Source Terms 8
2.2.2 Boundary Conditions 9
2.2.3 Numerical Solvers for the ODEs - The Second-Order,
Two-Stage Runge-Kutta Method 10
2.2.4 Numerical Solvers for the Pure Advection Hyperbolic
Problem - The Second-Order Relaxed Scheme 11
3 Numerical Results 29
3.1 Spherical Explosion Problem 30
3.1.1 Physical Description 32
V
3.1.2 Comparison with Previous Analytical and Experimental
Results 33
3.2 Cylindrical Explosion Problem 46
3.2.1 Physical Description 46
3.2.2 Two-Dimensional Model 49
3.3 Spherical Implosion Problem 52
3.3.1 Physical Description 52
3.4 Cylindrical Implosion Problem 53
3.4.1 Physical Description 53
3.4.2 Two-Dimensional Model 53
4 Conclusion 65
Bibliography 68
vi
List of Figures
2.1 Test 1: Exact solution (solid line) and numerical solution (square
symbol) for pressure, density, velocity and temperature at time
t = 0.2 22
2.2 Test 2: Exact solution (solid line) and numerical solution (square
symbol) for pressure, density, velocity and temperature at time
t = 0.15 23
2.3 Test 3: Exact solution (solid line) and numerical solution (square
symbol) for pressure, density, velocity and temperature at time
t = 0.012 24
2.4 Test 4: Exact solution (solid line) and niimeriral solution (square^
symbol) for pressure, density, velocity and temperature at time
t = 0.035 25
2.5 Test 5: Exact solution (solid line) and numerical solution (square
symbol) for pressure, density, velocity and temperature at time
t = 0.8 26
3.1 Test 6: Comparison of the smearing effect at the contact discon-
tinuity among exact solution (solid line) and numerical solutions
(rela:xed scheme - square symbol, Harten's scheme - dashed line,
modified Harten's scheme - dash-dot line) at time t 二 1.0 31
3.2 Numerical simulation of the explosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t = 0.2
to t = 0.4 35
vii
3.3 Numerical simulation of the explosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time 力= 0 . 6
to t 二 1.0 37
3.4 Numerical simulation of the explosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t 二 1.15
to t = 1.30 39
3.5 Numerical simulation of the explosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t 二 1.65
to t 二 1.90 41
3.6 Numerical simulation of the explosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t 二 2.0
to t = 2.4 43
3.7 Numerical simulation of the explosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t = 2.5
to t 二 2.9 45
3.8 Comparison between the relaxed scheme and the (a) analyti-
cal (Erode), (b) experimental (Boyer) results of the spherical
explosion 47
3.9 Comparison between the spherical and the cylindrical explosion. 48
3.10 Comparison between the 2D model and the quasi-ID model of
the cylindrical explosion at time (a) t 二 1.65 and (b) t = 2.2. . . 50
3.11 Three-dimensional plot of the pressure contour of the cylindrical
explosion at time t = 2.5 51
3.12 Numerical Simulation of Implosion for (a) Pressure (b) Velocity
(c) Density and (d) Temperature from time t = 0.05 to t = 0.15. 55
3.13 Numerical simulation of the implosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t = 0.18
to t 二 0.27 57
viii
3.14 Numerical simulation of the implosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t = 0.28
to 力二 0.35 59
3.15 Numerical simulation of the implosion problem for (a) pressure
(b) velocity (c) density and (d) temperature from time t = 0.4
to It 二 1.0 61
3.16 Comparison between the spherical and the cylindrical implosion. 62
3.17 Comparison between the 2D model and the quasi-ID model of
the cylindrical implosion at time t 二 0.4 63
3.18 Three-dimensional plot of the pressure contour of the cylindrical
implosion at time t = 0.5 64
ix
List of Tables
2.1 Data for Five Riemann Problem Tests 21
3.1 Data for the 6th Riemann Problem Test 29
3.2 Initial Data for Explosion Problem 32
3.3 Initial Data for Implosion Problem 52
X
Chapter 1
Introduction
1.1 Background of Explosion and Implosion Problems
Studies of spherical blast waves have long been interested by many researchers
([1], [2], [3], [4], [5]). An experimental investigation of the explosions of a glass
spheres under high internal pressure was carried out by Boyer [2] in 1960. A
moving outward shock wave and a moving inward rarefaction wave generates
at the time the glass sphere broken instantaneously. The former is called
the main shock. Due to the dimensionality effect, a shock, also called tlie
second shock, follows from the tail of the rarefaction wave and increases with
strength. The second shock reflects from the origin of the sphere and starts
propagating outward. Theoretical investigation predicts that once the second
shock interacts with the contact discontinuity, similar to the formation of the
second shock, a third shock generates and moves inward. Similarly, fourth,
fifth, sixth shocks would be formed provided that the time is long enough and
the strength of the explosion is large enough. However, hardly can the third
shock be observed in experiments carried out in a laboratory ([2], [5]). Some
analytical results were studied in order to find the explicit formulae for the
location of the shocks and the contact discontinuity ([1], [3]).
For the implosion problem, since it is very difficult to break a spherical
diaphragm instantly ([2], [5]). The resulting flow of the experiment was com-
1
CHAPTER 1. INTRODUCTION 2
pletely nonuniform and the experimental results were not satisfactory. Because
of the difficulty in carrying out an experiment in the laboratory, computa-
tional fluid dynamics becomes important to investigate the problem. Sod [6
carried out a numerical experiment on a converging cylindrical shock by using
Glimm's method and operator splitting. Sod successfully captured the second
shock produced by the interaction between the main shock and the contact
discontinuity.
Due to the limitation of the experimental study, numerical simulations pro-
vide a new approach to investigate the fluid dynamics problem. For example,
Liu et al. [7] simulated the explosion and implosion problems by using a mod-
ified Harten's scheme. They successfully captured the main features of the
problems. Recently, high resolution and robust numerical methods and nu-
merical technique are developed quickly. It would be valuable to carry out
some numerical simulations by using different classes of numerical methods to
test the practical use of the recently developed methods. Also, the physics of
the explosion and implosion problems would understand more thoroughly by
the simulations of different numerical schemes.
1.2 Background of the Development of Numerical Schemes
Cylindrical and spherically symmetric wave motion arises naturally in the the-
ory of explosion in air. In these situations, the multidimensional equations may
be reduced to essentially one-dimensional equations with a geometric source
term to account for the second and third spatial dimensions. The govern-
ing equations become a nonlinear system of hyperbolic conservation laws with
source terms. One of the approach to tackle such inhomogeneous system is
to split for a time step, into the 'advection problem' which is a homogeneous
hyperbolic problem, and the 'source problem' which is a system of ordinary
differential equations (ODEs). The solvers for the two split problems can be
chosen independently and the numerical methods for both fields are highly
CHAPTER 1. INTRODUCTION 3
developed. There is a vast literature on ODEs and its numerical methods (see
8] for reference). For the pure advection (homogeneous) hyperbolic problem,
two requirements on numerical methods, namely high-order (at least second-
order) of accuracy and absence of spurious oscillations have been competing
in the history of computational fluid dynamics. A prominent class of nonlin-
ear method, total variation diminishing (TVD) method, is one of the most
significant achievements in the development of numerical methods for partial
differential equations in the last two decades.
A class of explicit second-order finite difference schemes for computation
of weak solutions of hyperbolic conservation laws was developed by Hart en [9
in 1983. These highly nonlinear schemes, which were proved to be TVD, are
obtained by applying a non-oscillatory first-order accurate scheme to an ap-
propriately modified flux function. Following his work, other robust numerical
methods or techniques were developed, e.g., essentially non-oscillatory scheme,
subcell resolution, etc ([10], [11], [12], [13]). In simulating the explosion and
implosion problems, Liu et al. [7] improved the Marten's TVD scheme to obtain
a high resolution on a contact discontinuity by modifying the scheme with the
technique of the artifical compression method (ACM) ([14], [15]). The above
schemes are one-dimensional in nature and they can be used to calculate a
higher dimensional problem with the help of a dimensional splitting method.
However, they are not easy to extend to the multi-dimensional version. So, it
would be useful if a class of numerical schemes can be developed, for which
can generalize to a higher dimensional version easily and keep its simplicity
and explicit nature.
Later, a class of numerical schemes called relaxation schemes for systems
of conservation laws in several space dimensions was introduced by Jin & Xin
16] in 1995. The second-order schemes were shown to be TVD in the zero
relaxation limit for scalar equations. The advantage of the relaxation scheme
is that neither spatially Riemann solvers nor temporally nonlinear system of
algebraic equation solvers are required. Also, it keeps its simplicity for general-
CHAPTER 1. INTRODUCTION 4
ization to a higher dimensional space. Such advantages are interested by many
researchers to further investigate and develop the scheme recently ([17], [18],
19], [20], [21], [22], [23], [24], [25], [26], [27], [28]). The idea of the relaxation
scheme is to use a local approximation by constructing a linear hyperbolic sys-
tem with a stiff lower order term that approximates the original system with a
small dissipative correction. The leading order approximation of the relaxation
scheme in the small dissipative correction limit is called the relaxed scheme.
During the construction process, a constant matrix ‘A’ is to be constructed.
In order to ensure the dissipative nature of the leading order approximation
of the relaxation system, a dissipative condition is required. Although the
dissipative condition can always be satisfied by choosing a sufficiently large
A. However, larger A introduces more numerical viscosity, so, it is desirable
to obtain the smallest A to meet the numerical stability condition. Evje and
Fjelde [24] showed that the relaxation scheme produces a poor approximation
for a typical mass transport example which involves transition from two-phase
flow to single-phase flow. This is due to the 'over-estimate' in choosing the
matrix A causing the excessive 'smearing out'. They solved the problem by a
flux splitting method so that the choice of the matrix A was possible to become
'reasonably' small. So, the numerical dissipation of the relaxation scheme is
to a large extent determined by the matrix A and the choice of A becomes
critical in the application of the scheme.
In applying the relaxation scheme to the explosion and implosion problems,
we encounter a similar difficulty in choosing the matrix A. As suggested by Jin
& Xin [16], the matrix A can be roughly estimated by the characteristics of the
original Euler equations such that the characteristics of the relajcation system
interlace with those of the Euler equations, or simply take the largest eigenvalue
of the Euler equations over the whole space-time domain. However, when the
shocks reflect at the origin, the physical quantities will inflate dramatically
and causing a 'sudden inflate' of the eigenvalue at this specific time. Surely,
the matrix A estimated by the above method will be too large for most of
CHAPTER 1. INTRODUCTION 5
the calculation time. It would be a main contribution for this research to find
an algorithm or at least, some guidelines to estimate the matrix A for the
explosion and implosion problems.
1.3 Organization of the Thesis
The objective of this research is to understand the physics of the explosion and
implosion problems through numerical simulations; meanwhile, the technique
of the numerical scheme for the problem is developed. More specifically, an
algorithm of using the relaxed scheme on the problems is developed. In chapter
2, the mathematical model are stated. The governing equations of a spherically
or cylindrical symmetric flow are basically the Euler equations with a geometric
source term. A splitting method is employed to deal with the inhomogeneous
equation and the Euler system is mainly solved by the relaxed scheme. The
appropriate choice of the matrix 'A' is critical in the scheme. An algorithm
to choose the matrix is also developed in chapter 2. The numerical results
are presented in chapter 3. Four cases namely spherical explosion, cylindrical
explosion, spherical implosion and cylindrical implosion are discussed. The
numerical results are compared with some previous numerical, analytical and
experimental results. Also, two-dimensional models are calculated in order to
compare with the approximated cylindrical symmetric models. Finally, some
concluding remarks are made in the last chapter.
Chapter 2
Governing Equations and Numerical Schemes
The main interest in this research is to simulate the explosion and implosion
problems numerically, and to analyze the numerical method involved. In this
chapter, we develop the mathematical model for the problems. After choosing
a suitable numerical method for the problems, the details and techniques of
the method will be discussed.
2.1 Governing Equations
In the theory of explosion or implosion in air, the problem is normally regarded
as an inviscid flow (Sod [6], Liu et al. [7]). By dropping the effects of the
body force and the thermal conduction, the equations for an unsteady, two-
dimensional, compressible inviscid flow, also called the Euler equations are as
follows:
f + T + T =。’ (2.1)
6
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 7
where
( \ I \ ( \ p pu pv
pu + p puv U 二 , F肌二 , G ( U ) = . (2.2)
pv puv pv"^ + p
V E ) \ {E-Vp)u ) \ {E^j))v )
Here p is the density, p is the pressure. E is the total energy, u and v are
the velocity components with respect to the x and y-direction respectively.
Similarly, the one-dimensional (ID) Euler equations can be written as,
f + T =。’ (2.3) where
( \ I \ P pu
U= pu , F { U ) = 叫 h p . (2.4)
V E ; V {E^p)u
The equation of state for a perfect gas is required for the closure of the
problem,
E 二 + Ifm'. (2.5) 7 —丄 2
where 7 is the ratio of specific heats, which is a constant for a calorically
perfect gas and dependent on the temperature for a thermally perfect gas. In
our model, we assume the gas is calorically perfect and having 7 = 1.4 (i.e.
for air).
Wave motion in the explosion and implosion problems is normally cylin-
drical or spherically symmetric in nature. The multi-dimensional Euler equa-
tions may be reduced to essentially one-dimensional equations with a geometric
source term vector, S{U), to account for the second and third spatial dimen-
sions. It is usually written in the form dU dF{U) ^ 广
1 + (2.6)
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 8
where
( \ ( \ ( \ p fm pu
U 二 pu,刚二 fm'+p , 聊 二 - d pu^ . (2.7)
\ E ) \ {E + p)u y \ {E+p)u
Here now x is the radial distance from the origin and u is the radial velocity.
For a. 二 1, (2.6) returns to our one-dimensional Euler equations (2.3) and
it represents a plane one-dimensional flow. For a 二 2, (2.6) represents the
cylindrical symmetric flow which approximates a two-dimensional flow. For
a = 3, (2.6) represents the spherically symmetric flow, an approximation to a
three-dimensional flow.
A difficulty occurs in solving (2.6) due to the arising of the singularity at
the origin, x = 0. In order to remove the singularity, the following form of the
conservative equations is adopted to ensure that the resulting finite difference
numerical scheme employed is compatible with the original equations at the
origin (Liu et al. [7]). At the origin,
dp I dfjii 瓦 小 二 。 ,
dE di^E + p)u ^ ( �
i + = 0, (2.幻
u = 0.
2.2 Numerical Schemes
2.2.1 Splitting Scheme for Partial Differential Equations with Source
Terms
The governing equations (2.6) are non-linear system of hyperbolic conserva-
tion laws with source terms. Such inhomogeneous system can be calculated
numerically by a splitting method. We split the governing equations (2.6) into
the pure advection hyperbolic system, which is a homogenous system, same as
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 9
(2.3),
dU dFiU) , �
i + ~ ^ : 。 , (2.9)
and, a system of ordinary differential equations (ODEs),
I " 二风 [ / ) . (2.10)
By using the data at 力二 n M , where A力 is the step size of time, and n is
an integer, as the initial condition and solving the pure advection hyperbolic
problem to find the state at 力二 (n+1) A力,we get an intermediate state. Then,
using this intermediate state as the initial condition and solving the ODEs. We
get the result at 力=(n + I)At, at last.
By re-written the procedure above as follows:
U{t = {n+l)At) 二 二 (2.11)
Here, 幻 and 龙)can be any numerical solvers with the time step At for
the ODEs and the pure advection hyperbolic system respectively. If S 脚 and
are at least second-order accurate solution operators, Strang [29] showed
that the splitting scheme (2.11) is second-order accurate in time.
2.2.2 Boundary Conditions
At the origin, the equations (2.8) are discretized as follows:
, � A 力 - 购 )
仇 二 仇 ^ h ‘
pn+l — p n 力 ( 五 ? + P S 胸 - 必 鍾 风 - 税 - a ^ , (2.12)
PS+1 = (7-1)丑?+1,
= 0.
Here and after, the subscript 'j' indicates the spatial positions at x = jAx*,
where Ax is the step size for space. So subscript '0', T and '2' indicate the
points at the origin, x = Ax and x 二 respectively. Also, the superscript
'n' indicates the temporal positions at 力二 nAt.
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 10
Reflective boundaries are employed at the origin. This boundary conditions
can account for the coming and interaction of waves from the axi-symmetric
direction. A fictitious state ITlj is defined from the known state U言 inside the
computational domain,
二 P], 二-u�and (2.13)
For the end side of the domain, open-end boundary conditions, or so called,
transmissive boundaries are used so as to allow the passage of wave without
any effect,
p]+i - P], = u] and (2.14)
2.2.3 Numerical Solvers for the ODEs - The Second-Order, Two-
Stage Runge-Kutta Method
In (2.11), the numerical solvers S � � and 龙) can be chosen independently.
We first concentrate on solving the ODEs (2.10). For simplicity, it is preferred
to choose an explicit method. However, if the system of ODEs is stiff, nu-
merical difficulties arise especially for explicit methods. Even using implicit
method. Leveque [30] observed that large overshoots and non-physical wave
speeds occurred for hyperbolic conservation laws with stiff source terms. So
checking the stiffness of the ODEs is upmost important. As stated by Lambert
8], a nonlinear system of the form (2.10) is said to be stiff if
(i) Re{\i) < 0, i = 1 ,2 , . . . , m, and,
(ii) � � m i n j \ R e [ X i ) .
Here Re{Xi) denotes the real part of the complex number Ai, which is the
eigenvalue of the Jacobian A{U)三藉.In our ODEs, the Jacobian is found to
be
f 0 0 ^
A{U) = 2u 0 . (2.15)
^ -jEu/p + (7 - 1)仏3 jE/p - 3 ( 7 - 1)1/72 -fu j
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 11
So, the eigenvalues are u and ya, and then the stiffness ratio is
maXj\Re{Xi)\ : minj\Re{Xi)\ 二 7,
which is, for air, only 1.4 in our case. Hence, our system of ODEs is not stiff.
The second-order, two-stage Runge-Kutta explicit method is chosen to use
due to its simplicity and high accuracy. For time step At, when the system
evolutes from time t介 to f^+i,where f^ — nAt, the Runge-Kutta method is
described as
u � = I T +
U � = [ / � + A:i5,(f +ALL/�), (2.16)
[T+i 二 - + [/⑵-. 2 L
Here and after U几 and [/时丄 represent the solutions at time f^ and t 几 r e -
spectively.
2.2.4 Numerical Solvers for the Pure Advection Hyperbolic Prob-
lem - T h e Second-Order Relaxed Scheme
A relaxation scheme based on the corresponding relaxation systems was devel-
oped by Jin & Xin [16] a decade ago. The advantage of the relaxation scheme
is that neither spatially Riemann solvers nor temporally nonlinear system of
algebraic equation solvers are required. Also, it keeps its simplicity for gen-
eralization to a higher dimensional space. Higher dimensional systems can be
treated in the same way as in the one dimension because of the constant linear
characteristic fields. Such advantages are interested by many researchers to
further investigate and develop the scheme recently.
Our pure advection hyperbolic problem (2.9) is a system of conservation
laws in one space variable. A corresponding relaxation system can be intro-
duced, dU dV ^ r.3
1 二 0 V g K a t 十 彻 , ‘
dV dU 1 i + A - = e〉0’ (2.17)
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 12
where e is a positive constant, calling the relaxation rate. The 3 x 3 matrix A,
A = diag{ai,a2,a3), ai,a2, as > 0, (2.18)
will be determined later.
The idea of the relaxation scheme is to solve the relaxation system (2.17)
instead of the original hyperbolic system (2.9). The special feature of the linear
characteristic fields and localized lower order terms of the relEDcation system
results in the advantages of the scheme mentioned before. The discretization
of the relaxation system is called the relaxing scheme. The leading order
approximation of the relaxing scheme in the small e limit is called the relaxed
scheme.
For simplicity, we may choose
A = al, a>0, (2.19)
where I is the identity matrix. The reduced system, i.e. the zero e limit of the
relaxation system is dissipative provided that
— < 1. (2.20) a
Here A = maXi\Xi{U)\ where A are the genuine eigenvalues of F'(U). For ID
scalar conservation laws,
专t等“鳴 (2.21)
where / is a scalar valued function. We have a corresponding relaxation sys-
tem,
du dv n ‘ + T 二 0, u,v G R , at ox
芸+ 二 - - { v - f { u ) ) ^ e > 0 . (2.22)
ot ox €
The dissipative condition now is referred to as the subcharacteristic condition
by Liu [31], < f{u) < v ^ for all u. (2.23)
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 13
It will be shown later that this condition plays an important role in the numer-
ical stability. Also, the dissipative structure of the first-order correction to the
original system implies that the numerical solution to the relaxation system
converges to the entropy solution of the original system [16 .
As observed in the numerical tests by Jin k Xin, for the case of compressible
Euler equations, the relaxing scheme and the relaxed scheme produce essen-
tially the same result. So, in our problem, we employ the second-order relaxed
scheme, instead of the relaxing one, for the pure advection hyperbolic part.
We describe below the one-dimensional version of the relaxed scheme we are
going to employ. The following second-order explicit Runge-Kutta is applied
in the time discretization.
"⑵ 二 " ⑴ — 巧 ( " ⑴ ) — ⑴ ) ) ’ (2.24)
� + 1 = 1 [[/- + � - . Z ‘- �
Here Fj^i is the numerical flux detailed below, and we use uniform girds for
both spatial and temporal space with step sizes Ax and At respectively.
In order to obtain a second-order accuracy in spatial discretization, van
Leer's MUSCL scheme [32] are used. The MUSCL scheme uses the piecewise
linear interpolation to achieve the second-order accuracy. For the p-component
of the flux, it is written as
+ & < —。、 ), (2.25)
where
cjf 二 (i^⑷O/JVi) 土 的 ⑷ ( W ) 干 「严n)树疗),(2.26)
没士 = 士 树 、 R - 伸 干 y/^ryff广
] — 土 v / ^ y j ? 广— F ( P 卿干 v^yj办". “
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 14
(/)(0f) is called the slope limiter function. Sweby [33] gave out a general con-
dition of the slope-limiter function for the scheme to be TVD,
0 < ^ < 2 , and —e
0 < _ < 2 (2.28)
In particular, we use the one introduced by van Leer [32] in this research,
m 二 (2-29)
We want to remark here that due to the explicit and the special structure of
the linear characteristic field of the relaxation scheme, the components of the
solution vector can be obtained in parallel computing without any modification
of the scheme.
Theorem:
For the slope-limiters satisfying (2.28), by choosing the matrix A as (2.19),
the corresponding second-order relaxed schemes (2.24) to (2.27) for the ID
scalar conservation law (2.21) are TVD provided that the subcharacteristic
conditions (2.23) and the following CFL condition are satisfied,
辟:^ \ (2.3。)
Proof of the Second-Order Relaxed Scheme to be T V D for ID scalar
Conservation Law
Our proof essentially follows the same as Jin & Xin's [16] analysis with only a
slight modification.
A scheme is called TVD if its total variation TV{u) :二 — Uj
decreases in time,
< T V �
The relaxed scheme with the method of line takes the form
At )
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 15
:二 - ‘ (,(略 — ^ K - i ) ) + 盖 ( � 1 - +
+ 4 ( 7+1 — + - < ) • (2.31)
By the subchar act eristic condition (2.23), we have
C] — ( ^ + 权 4 ) ^ 0, (2.32)
At ( / ( 略 1) - m)\ D] :二 — ( • 一 : j - 0. (2.33)
By substituting Cj and Dj, the relaxed scheme (2.31) becomes
二 u] - - uU) + - U-)
+ 4 (^7+1 - 斤 + - 〜+) • (2.34)
Then we have
- (1 - Q - - u]) + ~ u]^,)
( 以 ? + 巧 ( 2 . 3 5 )
where
^i+l 二 去[( 7+2 - - (S-+1 - ^7)
- O + « - < i ) ] . (2.36)
By rewriting cr, using C] and D”
= — & 2 ) ^ + 2 ( 4 3 — 4 2 )
(2.37)
and observing that
= C〕_等 (2.38)
GJ D八 u ; � u ] � ) = (1), (2.39)
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 16
we get
切 - “ 广 = [ i - ( i - m ) + c ^ i 一
+ ( 1 - 嘲 C M 抖 1) \ 乙 }
+ f l + 没 州 ) — — 略 1). (2.40)
By using the CFL condition (2.30) and the condition for slope limiter (2.28),
we can prove that the three coefficients on the right-hand side of (2.40) are
non-negative,
1 + 崎 + 1 ) - 榮 (2.41)
We then obtain
- <
1 — (1 - m ) + c ] - ( 1 + I ) . ] I略1—以?I
+ + _ 辦 、 - 略 2 -略 i l . (2-42) V �"j+2 /
By (2.39), the slope limiter condition (2.28) and the fact that
_ 二 0, V6> S 0,
we have
等 C h W ] - 二 树灯 )略 1 -
= (2.43)
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 17
Now we obtain
K i i -
1 - ( 1 — I m ) ) c] - ( i + \ < m ) ) 巧 ] —
+ ( 1 - 臺 州 ; i K - �
+ ( 1 + 臺 術 + 1 ) ) 1 ^ 州 1 < + 2 — 略 1 |
- l m + 2 ) D j + 2 \ u ] ^ 3 — (2.44)
By summing up over all j yields,
= ^%-ooKtl - y^iI < S告—⑴I略1 = TV{u-) (2.45)
Up to now, we have proved that the MUSCL relaxed scheme with one-step
time discretization is TVD.
In order to get second-order accuracy in time, the second-order Runge-
Kutta splitting scheme is used: ⑴ 7/几 (1)
We have already proved that
释 ⑴ ) 二
= - + ^^ (^K+i) - ^ K ) ) I
By using the result of Shu k Osher [12], we can show that the second-order
Runge-Kutta splitting scheme is also TVD,
T 寧 + 1 ) =
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 18
= 頌 — u]\ + 全 - t^i) + At — I
< 全 卿 , + 全 歸 ⑴ )
= T V { u ^ ) ,
and we complete the proof of the theorem.
We remark that no rigorous proofs are available for the TVD property
for either the multi-dimensional scalar case or ID system conservation laws.
However, we are going to develop an algorithm to choose the matrix A and
setup some criteria for somewhat similar to the CFL condition (2.30) in the
next section. Then, a number of numerical test will be carried out to give
confidence to our algorithm, before applying to the explosion and implosion
problems in the next chapter.
Algorithm for the Relaxed Scheme
We want to solve the ID Euler system (2.9) by using the relaxed scheme
ineiiLioiied before. For siiiipliciLy, we choose the inauix ds in (2.19) for the
system. In order to ensure the relaxed system to be dissipative, the condition
(2.20) must be satisfied. In fact, this can always be satisfied by choosing a
sufficiently large A. However, larger A introduces more numerical viscosity,
so, it is desirable to obtain the smallest A to meet the numerical stability
condition.
Although we cannot obtain any CFL condition other than the case of ID
scalar for the scheme to be TVD, we use the condition (2.30) as a reference.
We define the CFL number as
CFL number := (2.46) Ax
For returning to the scalar case, the CFL condition (2.30) is expressed as
CFL number < (2.47)
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 19
Since the eigenvalues of the ID Euler equations are u and w 土 c where
c = (2.48)
is the sound speed. In order to obtain the minimum A, and fulfill the dissipative
condition (2.20), one can take
a = max{sup\u - sup\u\^, sup\u + (2.49)
here the supremum is taken over all t and x in the whole space-time domain.
However, as evidenced by the previous researches ([2], [6], [7]), we know that
the quantities of the eigenvalues would have a large variation along time, espe-
cially when the shocks reflect at the origin and the physical quantities would
inflate dramatically. If we simply take the supremum over all time and space,
the resulting A would become too large relative to some time intervals. Exces-
sive numerical viscosity would be injected during such time intervals arid the
accuracy of the calculation would be lower in general. So, we propose to choose
a 'dynamic' matrix A such that it is tailor-made for each time step interval.
Let lis consider a system evoliitins: from time f^ to 力 I n s t e a d of taking
the supremum of eigenvalues over all time, Lhe idea is to lake I lie supremum
over this time step interval only. Now, we want to estimate the maximum
eigenvalues inside the time interval The estimate follows the spirit
of the Godunov scheme. We assume the state at 力“as a piecewise constant.
The state keeps constant along every spatial interval i, x^^ i]. Evoluting
the system to the time t奸i is equivalent to solve a Riemann problem with the
initial condition of the state at f . The state at the position x] will remain
unchanged inside the time interval At provided that maXi | A | < 盡 where Xi
is the eigenvalue of the system. Then the constant ''a' is chosen as the same
maimer as (2.49), but now, the supremum is taken over t G and for
all X in the spatial domain. By using this method of estimate, the constant
'a will be different for each time step. For each time step interval, we suggest
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 20
the CFL number to be,
CFL number == v ^ ^ ^ = maXi \ —^ < (2.50) A x A r c — 2
which matches with the CFL condition (2.47) when returning to the scalar
case. By using the CFL condition in the scalar case as a reference, we impose
no additional constraint on the CFL number for the method proposed in this
section.
We remark that the above method provides only a rough estimate since
the Godunov scheme is only first-order accurate. So, this method in no way
guarantees that the estimate always satisfies the dissipative condition, however,
a close estimate to the minimum A is expected to obtain. In practical use, we
set the time step in the computer program by
At = (CFL number) x A x 剛
J(5 X max{\u + c|, \u\, \u — c\)
By changing the adjusting factor j3 as long as the numerical stability is still
maintained, we believe that the solution is reliable.
Numerical Test for the Algorithm
As mentioned in the previous section that we can only prove the T V D property
of the ID scalar conservation law. However, in this research, what we want
to calculate instead, is the system of equations. In order to apply the relaxed
scheme on the Euler system with the algorithm developed in the previous
section, we have done a series of numerical tests to support it.
It should be noted that we are not going to test the accuracy of the numer-
ical scheme in this section, instead, we want to give evidence that the scheme
do converge to the 'correct' solution and it is numerically stable. So in the
following parts, instead of comparing with other numerical methods, we com-
pare our numerical solution with the exact solution. Once the result shows
that the scheme converge to the exact solution, we conclude that the scheme
is numerically stable and hence reliable.
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 21
Test PL UL PL PR UR PR Xo
1 1.0 0.75 1.0 0.125 0.0 0.1 0.5
2 1.0 -2.0 0.4 1.0 2.0 0.4 0.5
3 1.0 0.0 1000.0 1.0 0.0 0.01 0.5
4 5.99924 19.5975 460.894 5.99242 -6.19633 46.0950 0.4
5 1.0 2.0 0.1 1.0 -2.0 0.1 0.5
Table 2.1: Data for Five Riemann Problem Tests.
Five tests are selected to test the performance of the numerical scheme.
They have been used by Toro [34]. Relaxed scheme with the algorithm devel-
oped in the previous section are tested. In all the tests, the spatial domain is
X £ [0,1]. The ratio of specific heats is 7 二 1.4. The step size is Ax 二 0.005
and the CFL number: 0.3. For each test, different step size, CFL number
and adjusting factor (3 in (2.51) have already been tested repeatedly to ensure
that the result is numerically stable. The tests are Riemann problems with
the initial data summarized in table 2.1, in terms of the primitive variables.
All quantities are dimensionless. Subscript 'L' and ‘R, represents the left and
right constant values of the initial data discontinuous at the position x = xq.
The first test is similar to the 'Sod test problem' [35]. It is, in fact, a
modification of the Sod test and the solution consists of a right shock wave, a
right traveling contact wave and a left sonic rarefaction wave. This test is very
useful in assessing the entropy satisfaction property of numerical methods.
The second test is called '123 problem' and its solution consists of two
strong rarefactions and a trivial stationary contact discontinuity. The solution
region between the non-linear waves is close to vacuum, so this test is useful
in assessing the performance of numerical methods for low density flows [36].
The Richtmyer (or two-step Lax-Wendroff) method fails to give a solution to
this problem.
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 22
1 \ i \ 0J9 r \ 0.9 i- \ OS i- \ 0.8 — \ 0.7 r \ 0.7 ;• \
\ ^ ^ i o 5 ^ \ io.6 L • i 1 o
0.4 — I CW 二 _• Li
03 - 0.3 02 r 0.2 -0.1 r ^ 0.1 -
0 r I I I I I—I I I—I—I—I—I—i—I—\ \ i I \ ~ F I I 1 t I t I • I I I I I • > , , I , I 0 0.25 0.5 0.75 1 0.25 0.5 0 75 1
K X 1J6 r- 3S r-1.4 L
/ i ^ / … r
: / - r • 、‘「 / = • 奋 : / 专 ^ - / S2S ^ 5 I
06 fc 2J8 -: 2.4 L X
D.4 7 X
。” " l \ J 一 。 r . . , . I _ , _ , _ I _ I _ _ I _ , _ , _ , _ _ I _ I _ I _ I _ _ 1 _ _ 1 _ _ I 1 8 _ . _ _ . I _ _ , _ , _ , _ , I _ , I , 1 0.25 0.5 0.75 1 0 D.25 0.5 075 1
* X
Figure 2.1: Test 1: Exact solution (solid line) and numerical solution (square symbol)
for pressure, density, velocity and temperature at time t = 0.2.
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 23
LZ i i I _ I I I I I I I I I I I I I I 1 I 1 I I 1 c I I I I I I . I I I . I . . I I I , I I
0 0 . 2 5 0 . 5 0 . 7 5 1 0 0 . 2 5 0 . 5 0 7 5 1 X X
I:丨 / 1 \ y -2 r I < . . • . . _ I 0 h • . • . I . • • • I . . . . I . . , . I
0 0 . 2 5 0 .5 0 .75 1 0 D 2 5 0 5 0 . 7 5 1 X X
Figure 2.2: Test 2: Exact solution (solid line) and numerical solution (square symbol)
for pressure,density, velocity and temperature at time t = 0.15.
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 24
t 8 r 门 誦 r \
9 0 0 - \ - ‘
; \ . 800 7 \
[ •. I 迎 r X i 3 :
a 'TOO ^ ^ 3 0 0 — 2 - -
200 — } - • 仰 p ^
0 一 L — — :__I I I I I—I—I—I—J—I—I—I—I—I I I 1 1 I I pj r . « . . I . . . . I • . • . t . . . . I
0 0 . 2 5 0 . 5 0 . 7 5 1 0 0 . 2 5 0 . 5 0 . 7 5 1 X X
: [ r — t 16 1. / 2000 -
,4 ^ / • -
, 1 2 - / = 1 5 0 0 -
t ; / ^ • I '0 r / I • ^ 3 r / i 1000 -
- / ^ : :
” Z 5 0 0 - • :l/ t__ s - . . • • _ I _ _ I _ _ I _ _ I _ 1 _ I _ 1 _ _ I _ 1 _ _ 1 _ I _ _ I _ I _ _ 1 _ _ I _ _ I 0 h . . . . I . . , I ' • , , . . I
0 : 2 5 0 5 0 7 5 1 0 0 25 0 .5 D.75 1 Jf A
Figure 2.3: Test 3: Exact solution (solid line) and numerical solution (square symbol)
for pressure, density, velocity and temperature at time t 二 0.012.
Test 3 is the left half of the 'blast wave problem' by Woodward and Colella
37], which is a very severe test problem. The solution consists of a left rarefac-
tion, a contact discontinuity and a strong right shock with Mach number^ 198.
The Richtmyer (or two-step Lax-Wendroff) method fails to give a solution to
this problem.
Test 4 is a very severe test, too. Its solution represents the collision of two
strong shocks produced by the above test (Test 3) and consists of a left facing
shock traveling very slowly to the right, a right traveling contact discontinuity
and a right traveling shock wave. So, all three discontinuities are traveling to
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 25
1800 r- � 30 : r r ^
1600 - , - • •
1鄉 7 . 25 - ‘ I I • • 1200 - : •
- 2 0 - I w t= 10C0 - ^ i ; : S 800 - 2 � i CL , O . I'
鄉 r i L : 5 -20D 7 , -
�0 0.25 0.5 0.75 1 °0 0.25 ' ' ' ' 05 ' ' ' ' 075 ' ' ' ' 1 X X
25� 3DD r- •� _
如: 250 i 15 - ‘
: 200 -
2 : •t - ^ I : •_ g : si 150 -
• li : • 10D -D - ;
( ' � � • .10 r . • • • I__1 I_1__I I__1 I_1_I 1__1_I n ~• • • • I . . . . I . 薩. . . I
0 025 D5 075 1 "o 0 25 CI .5 075 1 * -A
Figure 2.4: Test 4: Exact solution (solid line) and numerical solution (square symbol)
for pressure, density, velocity and temperature at time t 二 0.035.
the right.
The fifth test simulates the collision of two uniform streams. Its solution
consists of two strong symmetric shock waves and a trivial contact discontinu-
ity. In this test, the adjusting factor [3 should be at. least 6.5 for stable solution,
while P ^ I is large enough for all other tests.
In figures 2.1 to 2.5, it is shown that the relaxed scheme with the algorithm
in the previous section passes all the five tests above. Although we cannot
prove the TVD property of the scheme when applying to the Euler system,
the numerical experiments above show the applicability of our scheme to the
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 26
6 J- 6F : 5: ^ 5 r
- _ ‘ 4 . 5 i- I
41 , 4 i-
I 丨 r . ” I 3 - 讓 3 —
I ' = 2 . 5 ;.
2 - 2 ^
: 1.5 i-1 - 1 : ^ 丨、
; 0 . 5 r
Ci ^^^^^^ L I I I I I I I I I I I I I I ^^^^^^ n t I 1 I I I I I I I—I—I—I—1—1—I—I—L—I I I 0 0 . 2 5 0 . 5 0 . 7 5 1 0 0 . 2 5 0 . 5 0 . 7 5 1
>C X
, 2 . 5 厂
-
2 ^ f ] 2 -
I -S 1.6 -
•f , • : 1 ® 0 - g
- ; I 1 - I •1 -
D.5 --2 一 I' ‘ J 丨塞
- I I • i l l I I I I I I 1 i l l • • I I 11 • • • f 1 . . . . I • . . . I . • . . I 0 0-25 0.5 0.75 1 "o D.25 05 0.75 1 X X
Figure 2.5: Test 5: Exact solution (solid line) and numerical solution (square symbol)
for pressure, density, velocity and temperature at time t 二 0.8.
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 27
Euler system.
Two-Dimensioned Case
One of the advantage of the relaxation scheme is its simplicity for generaliza-
tion to a higher dimensional space. Since the governing equation (2.6) is an
approximation to the multi-dimensional case. So, it is worthwhile and con-
venience to calculate the original multi-dimensional system by the relaxation
scheme in order to analyze the effectiveness of the approximated equation. We
choose to calculate the two dimensional case for comparison with the cylindri-
cal symmetric approximated model.
Considering the two-dimensional Euler System (2.1) and (2.2), similar to
the ID case, a corresponding relaxation system can be introduced,
dU dV dW ^ TT, ot ox oy
dV . dU 1 T /rTW + 乂 石 二 ⑴),
BW dlJ 1 ^ + = —{W-G{U)), e > 0 . (2.52) at oy e
For simplicity, as in (2.19), we consider the 3 x 3 matrices /I 二 a I and B : hi,
where a, 6 > 0 are to be chosen. Jin & Xin [16] showed that the reduced system
(2.52) is dissipative provided that \2 " 2
- + ^ < 1 . (2.53) a 0
Here A 二 ma:Ci|Ai((7)| and ji = max^\|2i{U)\ where 入 and yu are the gen-
uine eigenvalues of F'{U) and G'{U) respectively. The relaxed scheme can be
expressed as follows:
U⑴二『-f加、U, - F]事))
U � = u⑴ — 芸 ( F j + “ U � ) — F ) _ “ U � ) )
CHAPTER 2. GOVERNING EQUATIONS AND NUMERICAL SCHEMES 28
『+1 二 ; + (2.54)
where At, Ax and Ay are the step size for time, for spatial x-direction and
spatial y-direction respectively. Indices 'n', and ‘j, are corresponded to 't',
‘X,and ‘y,respectively.
For the p-component of the flux,it is written as
+ 全 - 《 力 , (2.55)
绍!(『)二 I (作。)+ G(PnU�办 I麵焚—(4,)
+ 去 《+1), (2.56)
where the slope limiters are
一 二 F�P�m) 土 • 增 " - F � ( 化 1 , � 干讽 恕 说
‘‘一沙 ) ( " r + i,》士 沙)(喝)干 v/^ 吻)", )
二 ⑷(「;士 錢 ( 恋 - 干 v [yg),n) (2.59)
一 士 坊 - G � 干 彻 愁 _
We define the CFL number as
CFL number max v ^ ^ ^ ] , (2.61) V k Ay; 、 乂
so that it would naturally return to (2.47) for degenerating to the case of ID
scalar conservation law.
Similar algorithm as the previous section in order to obtain a rough esti-
mation of matrix A and B are used. However, numerical stability should be
taken more carefully by the adjusting factor P because of the more uncertainty
in this 2D case now.
For the next chapter, we would use the numerical scheme stated and the
algorithm developed in this chapter to calculate the explosion and implosion
problems.
Chapter 3
Numerical Results
In this chapter, results of numerical simulations of explosion and implosion
problems are presented. The numerical method developed in the previous
chapter is used. The results are shown to be reliable by comparing with the
data calculated by the modified Harten's scheme [7]. The modified Harten's
scheme attempts to have a high resolution for contact discontinuities in addi-
tion to maintaining the good features of Harten's TVD scheme. A technique
in the same spirit of the artifical compression method (ACM) is used in the
modified Harten's scheme. The original ACM [14] is capable to maximize
the resolution of a steady progressing profile. In addition to the five tests in
the previous chapter, in order to test the smearing problem of our numerical
scheme at contact discontinuities, we carry out the sixth test as shown in table
3.1.
This Test 6 corresponds to an isolated contact discontinuity moving slowly
to the left. In figure 3.1, the relaxed scheme (square symbol) is compared with
the Harten's scheme [9] (dashed line) and the modified Harten's scheme [7
Test PL ul PL PR Ur PR Xo
6 1.4 -0.1 1.0 1.0 -0.1 1.0 0.5
Table 3.1: Data for the 6th Riemann Problem Test.
29
CHAPTER 3. NUMERICAL RESULTS 30
(dash-dot line). It is seen that the resolution for contact discontinuities of the
relaxed scheme is not as high as the modified Harten's but better than the
Harten's scheme. In the following presentation of the explosion and implosion
problems, the results of the relazxed scheme and the modified Harten's scheme
would be shown on the same graph. This comparison is significant because
these two numerical schemes are based on completely different ideas.
3.1 Spherical Explosion Problem
For the spherical explosion problem, for purpose of comparison, we choose
the same model as such experimentally investigated by Boyer [2],analytically
analyzed by Erode [1] and numerically discussed by Liu et al. [7]. A glass sphere
with radius Xq = 1 in. initially filled with air at pressure p' j = 326 p.s.i. and
temperature T' = 299 K is exploded in air at the atmospheric pressure and the
same temperature inside the sphere. The superscript 'prime' {') denotes the
dimensional quantity. The dependent variables are non-dimensionalized via
P 二 丨fL P 二 PV(PoCO^), u - (3.1)
Here Cg is the initial sound speed calculated by (2.48),in which the correspond-
ing density is Pq 二 p'L and the pressure is determined by the ideal gas law.
The independent variables are non-dimensionalized via
力=力7 ( 4 4 / 4 ) ’ x = x'l{Ax',). (3.2)
As mentioned in chapter 2, in our model, we assume the gas is calorically
perfect and having 7 : 1.4 (i.e. for air). The initial values corresponding
to the non-dimensionalized quantities are calculated and summarized in table
3.2.
At time ^ = 0, the glass sphere with radius x 二 0.25 is broken and it is an
analogue to a Riemann problem. In the following numerical results, A x = 0.05,
CFL number二 0.3,the adjusting factor (3—1.
CHAPTER 3. NUMERICAL RESULTS 31
间 1 广 • Rela>«<i Scl-»em«
“ ———-Marten's Scheme - Modified Marten's Scheme
i^l ^ x-i
- C, 1.3 - r,
a> - 〜 : •
I 4 f … ^ OS ^
OS I 11 - p 1 L
Ojb -O � G .25 0.75 1 。.岛丄• • • • ^ ^ • • • • ols • • • • 0 .75 ' ' ' ' 1 X K
D � 2J0 -;- pT;
: 2 .4 L >f 。叶 23 L
^ : ' •f 1 2.1 ^ t i I - i
1 JO - r. E*act
n 1 s ^ Harten'sTclieme ⑴ _ Mocified Marten's Scheme • I 7 —
1 j6
J" • • • • 0i5 • _ I _ • • • '0.l6 • • 0 “ • • • • 0.5 0 I _ _ _ ;
X K
( b ) Exact Solution
• R e l a x e d S c h e m e P~ — 一 一 — Marten's S c h e m e f —- — fVIoclified H arte n s S c h e m e
1 4 • • • • • m-m — 辑 m — _ _ • _ •
- — � � . - 、、-\
1 . 3 - \
• f J
1 . 1 - \ •\ : \
一 \ _ \ - • •
1 _ ^
O Q I I 1 I I I I I 1 I I 1 i 1 1 0 . 3 0 . 4 0 . 5
X
Figure 3.1: Test 6: Comparison of the smearing effect at the contact discontinuity
among exact solution (solid line) and numerical solutions (relaxed scheme - square
symbol, Hapten's scheme - dashed line, modified Harten's scheme - dash-dot line) at
time t = 1.0.
CHAPTER 3. NUMERICAL RESULTS 32
Internal Density p'H 二 26.1886 kgm—3 pfj 二 21.733
Internal Velocity u丨h = 0 m/s uh = ^
Internal Pressure p'fj = 326 p.s.i. pn 二 15.514
External Density p' = 1.205 kgm—3 p乙 二 1.0
External Velocity u'l = 0 m/s ul = 0
External Pressure p'L = 15 p.s.i. pi 二 0.715
Unit Time t' = 293 /is 力二 1
Sphere Radius j/q 二 1 in. xo = 0.25
Table 3.2: Initial Data for Explosion Problem.
3.1.1 Physical Description
The present numerical results are in fairly good agreement with the results
obtained by Liu et al. [7]. Immediately after the broken of the glass sphere at
力二 0, a spherical shock (main shock) is generated and moves outward in the
air. The compressed gas is expari(ieci through a spherical rarefaction wave and
a contact discontinuity separates the expanded gas from the air compressed
by the shock. Different from an one-dimensional shock tube problem, due to
the three-dimensionality of the flow, the region between the tail of the rar-
efaction wave and the main shock wave is not a steady-state region. The high
pressure gas upon passing through a spherical rarefaction wave, due to the
increase of volume from the dimensionality, must expand to a lower pressure
than that reached through an equivalent one-dimensional expansion. This
'over-expansion' must be compensated by a compression wave or shock. This
is the physical explanation of the formation of the second shock in a spherical
explosion (see [2],[3] for reference). This second shock is rather weak and
propagates outward initially. Then its strength increases with time and the
second shock stops moving outward, but starts reversing to the inward direc-
CHAPTER 3. NUMERICAL RESULTS 33
tion at about t = 0.6. The contact discontinuity propagates outward slowly
from X 二 0.25 to 0.7 and starts reversing the direction at about t = 0.8 (see
figures 3.2 and 3.3). It is seen from the figures that the results calculated from
the relaxed scheme is highly agreed with those calculated by the modified
Harten's scheme.
The second shock reflects from the origin at about t = 1.2 and interacts with
the contact discontinuity at about t = 1.8. It propagates through the contact
surface and continues moving outward. The interaction causes the contact
discontinuity propagates slightly outward again and; meanwhile, ari inward
rarefaction wave is generated during the interaction. Similar to the formation
of the second shock, a rather weak, inward moving third shock follows (see
figures 3.4 and 3.5).
Similarly, the third shock reflects from the origin at about t = 2.7 (see
figures 3.6 and 3.7). It can be expected that a similar situation as stated above
would repeatedly occur resulting in the formation of the fourth, fifth and sixth
shocks, but they are too weak to be observed or formed in our simulation. The
third shock seems a bit more sharp when calculated by the modified Harten's
scheme than by the relaxed sclieine. Any the main features of the explosion
problem are captured by the relaxed scheme successfully.
3.1.2 Comparison with Previous Analytical and Experimental Re-
sults
Since the shocks and contact discontinuity especially the second and third
shock waves, are rather weak, the positions are obtained based on a rather
objective observation. However, for the purpose of comparison of the main
features of this problem, this rough observation is enough. In figure 3.8(a),
the analytical result predicted by Erode [1] are compared. The relaxed scheme
shows a good agreement with Brode's result. Figure 3.8(b) shows the compar-
ison between the result of the present numerical scheme and the experimental
result obtained by Boyer [2]. There are two main deviations for the results.
CHAPTER 3. NUMERICAL RESULTS 34
(a) 1 6 �
I u 容 t> t—0.2. Re laxed S c h e m e
X • t=0 .3 Re laxed S c h e m e 二i • t=0 .4 Re laxed S c h e m e
12 —f t=0 .2 Modrf ied Har ten 's S c h e m e 幻 - t t=0 .3 Modi f ied Har ten 's S c h e m e
I f — — _ _ 1=0.4 Modrf ied Har ten 's S c h e m e .2 10 - t 2 : 1 岩 ^ - ^ Rarefact ion W a v e
t “ \ in i £ : I £ 4 - \
— % S e c o n d Ma in S h o c k : \ S h o c k
n I I I I I I I I I — I — I — I — I 1—I 0 0 . 5 1 1 . 5 2
Radial Distance (Dimensionless)
(b) 2 r~
— t> t= 0.2 Re laxed S c h e m e — D t=0.3 Re laxed S c h e m e — o t=0.4 Re laxed S c h e m e _ t=0.2 Modi f ied Harten 's S c h e m e — t=0.3 Modi f ied Harten 's S c h e m e
^ 5 _ — — — — t=0.4 Modi f ied Harten 's S c h e m e a> -
I : 遍
I 1 - i h r c i r
Li/ -邏 1 ^ ^ ^ Main Shock
M 丨 t 0 ^ 1 ‘ 1 1 ~ ~ L u l l丨 " I I I i i • III 丨丨 1 i J ^ i i i w i J ! — 細 _ l i i i .•丨 _ j r _ . _ . i i . " _ . » _ . “ . _ i _ i i i i J
0 0 . 5 1 1 . 5 2
Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 35
(c)
\ t> t=0.2 Relaxed S c h e m e 冗 V D t=0.3 Relaxed S c h e m e ^ ^ T O t=0.4 Relaxed S c h e m e
_ \ t=0.2 Modi f ied Marten's S c h e m e _ r t=0.3 Modified Marten's S c h e m e
t 一 一 — — t=0.4 Modif ied Harten 's S c h e m e
^ 1 5 - 1
g - t Rarefact ion wave
•i _ \ ^ > Contact Discontinuity
r � \ X 0> _ \ S e c o n d /
: ^ ^ f e l ^ Main Shock
f) 1 I I I I I I I I I I I I I 1 1 1 1 1 1 0 0 . 5 1 1 . 5 2
Radial Distance <Dimensionless�
(d)
2 . 5 —
- > ^ l ^wip®*®*^ Main S h o c k
0 Sfe I ^ ^ ^ '1 1 ^ 、二
$ ~ \ ^ J ^ G o n t a c t Discontinuity
•色 - \ ? OJ I \ 1 ' i t C> t=0.2 Relaxed S c h e m e 3 . ' \ i> J • t=0.3 Relaxed S c h e m e 2 J I • t=0.4 Relaxed S c h e m e ^ — ^ N a - ^ j ^ ^ t=0.2 Modif ied Marten's Scheme g" - t=0.3 Modrfied Marten's Scheme g . — — — - t=0.4 Modif ied Marten's Scheme
: S e c o n d Shock
n 1 1 1 1 1 1 1 1 I I I I I I 1 1 1 1 1 1 0 0 . 5 1 1.5 2
Radial Distance (Dimensionless)
Figure 3.4: Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 36
(a)
1 . 2 5 -
一 丨 / I 瓦 1 r y r t ^ ^ ^ M a i n S h o c k
H i / y ^ ^ / > t = 0 . 6 R e l a x e d S c h e m e
J^ / m a t = 0 . 8 R e l a x e d S c h e m e i W o t = 1 . 0 R e l a x e d S c h e m e
^ - 1=0.6 M o d i f i e d M a r t e n ' s S c h e m e 0 . 2 5 — S e c o n d ^ 1 = 0 . 8 M o d i f i e d H a i t e n ' s S c h e m e
f) J luiijiMmiiiwi nur “ • I 1 I ‘ • I I I I • 1 I I I I 0 0 . 5 1 1 . 5 2
R a d i a l D i s tance ( D i m e n s i o n l e s s )
(b) •> t = 0 . 6 R e l a x e d S c h e m e n t=0. 8 R e l a x e d So h e m e
一 • t= 1 .0 R e l a x e d S c h e m e 1 _ t = 0 .6 M o d i f i e d Mar ten ' s S c h e m e
— t = 0 . 8 M o d i f i e d Hapten 's S c h e m e _ Jf — — — — t = 1 . 0 M o d i f i e d Mar ten ' s S c h e m e
— : / _ ^ ^ ^ S e c o n d S h o c k
r \ Shock
I " ^ '' 0 5 — I y ^
_ 1 I I I I I I I—I—I—I—I—1—;—I—I—I—1—i—I—I 0 0.5 1 1.5 2
R a d i a l D i s t a n c e ( D i m e n s i o n l e s s )
CHAPTER 3. NUMERICAL RESULTS 37
(c) t> t=0 .6 Re laxed S c h e m e • t=O.S Re laxed S c h e m e
: A o t=1 .0 Re laxed S c h e m e _ » t=0 .6 Modrf ied Har ten 's S c h e m e _ P t=0.8 Modrf ied Har ten 's S c h e m e
2 J — — — — t=1 .0 Modrf ied Har ten 's S c h e m e
f 轉 Contact Discontinuity
I:: 7 £ f l ^ S e c o n d S h o c k
— — _ L
I ••••••••叫•»«••,•••••• Hn r\ OPJpHHyMBfflg I • I I ' I I I I ' I 丨 ‘ ‘ I I
0 0 . 5 1 1 .5 2 R a d i a l D i s t a n c e ( D i m e n s i o n l e s s )
( d )
M a in Shock
I : ^ ^ t/> 匿 _ 丨_ — __ Contact Discontinuity
i : I a> 1 — > t=0.6 Re laxed S c h e m e 3 Ir a t=0.8 Relaxed S c h e m e % 孤 J ^ • t=1.0 Relaxed S c h e m e ^ — ^V.a^T'^^f i r t=0.6 Modif ied Harten's S c h e m e g - I ^^iSafeBfi^ t=0.8 Modrfied Harten's S c h e m e = 0 5 _ - - - - t=1.0 Modrfied Harten's S c h e m e
卜 . : ^ I I ‘ r “ S e c o n d Shock
(MWHIlWMIWWMMWhMBWit � I I • I I 1 J I I I I I I I I 1 1 1 1 1
0 0 . 5 1 1 . 5 2
Radial Distance (Dimensionless)
Figure 3.4: Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 38
(a)
Y -, ^ t=1.15 Relaxed S c h e m e -丨 ° t=1.20 Relaxed S c h e m e
2 , • t=1.30 Relaxed S c h e m e
'trt" ‘ p t=1.15 Modif ied Marten's S c h e m e t=1.20 Modif ied Haiten's Scheme
- g 2 ^ — — — — t=1.30 Modif ied Marten's S c h e m e
•i ^ g T I c 1 ^ —L
Q ^ Reflected Second Shock
一 - , f rom Origin Main Shock
1 I ; u
0.5 1 I",
Shock n Li_r I • I I I I I 1 1 1 1 I I I I I I I I
0 0 .5 1 1.5 2 Radial Distance (Dimensionless)
(b)
1 厂
- Main Shock 0 . 5 ^
o Second Shock ^ ^ "uT I Reflected from Origin „
!9L 丨 > t=1.15 Relaxed Scheme 公 口 t=1.20 Relaxed Scheme
l i V M • t=1.30 Relaxed Scheme - / t=1.15 Modified Haiien's S c h e m e
^ t=1.20 Modified Marten's Scheme _ — — — - t=1.30 Modified Harten's Scheme
-1.5 - f \
- Second Shock
1 1 1 1 1 I I ‘ I I I � I I I I I I I 1 1 “ 0 0 .5 1 1.5 2
Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 39
(c)
10 IP
^ > t=1.15 Relaxed S c h e m e 卜 • t=1.20 Relaxed S c h e m e r o t= 1.30 Relaxed S c h e m e
^ [“ t = 1 . 1 5 M o d i f i e d M a r t e n ' s S c h e m e > t=1.20 Modified Marten's S c h e m e
§3 r - — - - t=1.30 Modified Marten's S c h e m e
OJ P E I 0 [
4 U S e c o n d S h o c k Its |- Ref lected from Origin g Ji C o n t a c t D i s c o n t i n u r t y
O Main Shock
•丨丨丨丨I_丨丨"•‘丨旧丨川•賺丨"II丨1丨II
- e CO nd S h o c k r \ l n m n i ^ * ^ 1 I t I • ' I • I I I i I I I I 1 1 1
0 0 .5 1 1.5 2 R a d i a l D i s t a n c e ( D i m e n s i o n l e s s )
(cl)
6 r^
5.5 — » t二1.1 5 Relaxed Schem e 二 • t=1.20 Relaxed S c h e m e
_ — - 5 • t= 1 . 3 O R e l a x e d S c h e m e ^ S t=1.15 Modified Marten's S c h e m e _aj 4.5 fr t=1.20 Modified Marten's S c h e m e "E ^ — — — - t=1.30 Modified Marten's S c h e m e .2 4 0-绘 日 1 3.5 Ef 己 3 婆 Second Shock Reflected from Origin ^ Main Shock 3 2 . 5 ^ ^ C o n t a c t D i s c o n t i n u r t y ^^^
t•; o.s.L、^:^^^^^^^'.™
二 Second Shock n I I I I I I I I 1 1 1 1 1 1 1 1 1 1 1 1
0 0 .5 1 1.5 2 Radial Distance (Dimens丨on丨ess)
Figure 3.4: Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 40
2 ~
> t=1 .65 Re laxed S c h e m e ^ ^ ° t=1 .80 Re laxed S c h e m e ^ ^ ^ o t=1 .90 Re laxed S c h e m e - t=1 .65 Modi f ied Har ten 's S c h e m e
t=1 .80 Modi f ied Harten 's S c h e m e ITT 1 5 _ ^ ^ — — _ _ t = 1 . 9 0 Modi f ied Har ten 's S c h e m e 0 - 丨 nter€i.ction between
'5) — S e c o n d S h o c k and Contact Discontinuity g _ F o r m i n g Third S h o c k M o v i n g T o w a r d Or ig in
1 — 、 丨 知
1 : 芝 0 . 5 一 M a i n S h o c k
- S e c o n d S h o c k
f) I I I I I I I I I I t I I I I I I I I I_I I 1 I 1__I I I I I_I 0 0 . 5 1 1.5 2 2 . 5 3
Radial Distance (Dimensionless)
(b)
Q 5 _ S e c o n d S h o c k Interacting with
- Contact Discontinuity Ma in Shock forming Third Shock
一 — ^
I :
- ci . c . , > t= 1.65 Re laxed S c h e m e u_o .5 - S e c o n d S h o c k • t=1.80 Re laxed S c h e m e
• t=1.90 Re laxed S c h e m e > — t=1.65 Modi f ied Harten's S c h e m e
- t=1.80 Modi f ied Harten's S c h e m e — — — _ t=1.90 Modi f ied Harten's S c h e m e
—I—I—I—I~I—I—I—1—I_I—I—I—I I_I I I_I I_I I I I I_I—I—I—I—I_I 0 0 . 5 1 1.5 2 2 . 5 3
Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 41
(c)
6 一
- > t=1 .65 Re laxed S c h e m e L 口 t=1 .80 Re laxed S c h e m e
5 H o t=1 .90 Re laxed S c h e m e -一._ I t=1 .65 Modi f ied Harten's S c h e m e W I t= 1.S0 Modi f ied Harten's S c h e m e 2 [ # — — t=1 .90 Modi f ied Harten's S c h e m e
M " f m
£ f S e c o n d S h o c k
I 已 p B W S e c o n d S h o c k Interacting with >% F & W Contact Discontinuity
2 I f % fo rming Third Shock
S p Ma in S h o c k
1 -
n I r I I I I I I I I I I I__I I__I_I 1__I 1__I__I__1_I_I_I__I_I 1 1_I 0 0 . 5 1 1 .5 2 2 . 6 3
Radial Distance (Dimensionless)
(d)
> t=1 .65 Relaxed S c h e m e - Q t=1 .80 Relaxed S c h e m e J • t=1.90 Relaxed S c h e m e
^ ]_ t=1 .65 Modif ied Harten's S c h e m e r t=1 .80 Modif ied Haiten 's S c h e m e
_ _ _ _ t=1 .90 Modif ied Harten's S c h e m e K I § ^ S e c o n d Shock Interacting with Contact Discontinuity
4 forming Third S h o c k
i . 4 h ^ = � \ Ma in Shock
I 1 r M \ a> I te^ Contact Discontinuity
卜 L ——^ 0 5 ?" S e c o n d Shock
n r I I I I I I I I I I I I I I I I I 1—I—I—I—I—I—I—I—I—I I I I 0 0 . 5 1 1 . 5 2 2 . 5 3
Radial Distance (Dimensionless)
Figure 3.4: Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 42
(a)
2 r— ‘ t=2.0 Relaxed S c h e m e
- � t=2.2 Relaxed S c h e m e - o t=2.4 Relaxed S c h e m e - t=2.0 Modif ied Marten's S c h e m e - — ~ ^ t=2.2 Modif ied Marten's S c h e m e
1 5 一 — _ — _ t=2.4 Modif ied Marten's S c h e m e
'vi - ^
I , - K I X ~ 1 — V l ^ T _ S e c o n d S h o c k
I 0 5 Main S hock
Q- ) Third Shock — ^ — ^ Moving Toward Origin
n • I '丨 I_I I I I I_I I I I—I_1—I—I—I—I_I—I—I—I—I_I—1—I—I—I_I 0 0.5 1 1.5 2 2.5 3
Radial Distance (Dimensionless)
<b)
> t=2.0 Relaxed Soli erne 0.5 厂 口 t=2.2 Relaxed Scheme
_ • t=2.4 Relaxed Scheme t=2.0 Modified Marten's S c h e m e
- t=2.2 Modified Marten's S c h e m e _ — — — _ t=2.4 Modrfied Marten's Scheme
l/T _ Third Shock o> 售 0.25 -
^ ^ Main Shock
_n OR I I—I—I_I—I—I—I—I__I—I—I—I—I_I—I I I I_I I I I I_I—I—I—1—I_I 0 0.5 1 1.5 2 2.5 3
Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 43
<c)
> t=2 .0 Re laxed S c h e m e ® r " o t=2 .2 Re laxed S c h e m e
�: • t=2 .4 Re laxed S c h e m e L t=2.0 Modi f ied Marten's S c h e m e r t=2.2 Modified Haiten's S c h e m e
^ r — — — _ t=2.4 Modi f ied Marten's S c h e m e
f ;
•H 4 r T h i r d S h o c k
I - ^ ^ E 3 , C o n t a c t D i s c o n t i n u i t y
警 2 斷 Z 0 S e c o n d S h o c k Ma in S h o c k
爆 ^ ^ “ f) • I I I I I I I • I _ I I I I • I I I I I I I I I I I i I I I
0 0.5 1 1.5 2 2.5 3 Radial Distance (Dimensionless)
(d)
> t=2.0 Re laxed S c h e m e - • t=2.2 Relaxed S c h e m e
3 “ • t=2.4 Relaxed S c h e m e t=2.0 Modif ied Hai ten 's S c h e m e t=2.2 Modi f ied Marten's S c h e m e
- . _ __ — — t=2.4 Modified Haiten's S c h e m e iS 2.5 it
^ I ___ 囊 S e c o n d Shock
1 2 | 一 啣 二
Oi ‘ , I Ma in S h o c k
I \ O) S. Contact Discontinuity
n L_i—I I—I I I 1 I I I I I I I I I I I I 1—I—I—I——I—I——i——I—I—I—I 0 0.5 1 1.5 2 2.5 3
Radial Distance (Dimensionless)
Figure 3.4: Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 44
<a)
1 �
I ^ Second Shock
^ y ^ Refleotion of Third Shock .E 0 . 5 from Origin
3 — j n ' t=2.5 Relaxed Scheme ijy r w - t=2.7 Relaxed Scheme
n � “ t=2.9 Relaxed Scheme Q- 一 ^ t=2.5 Modified Marten's Scheme
_ t=2.7 Modified Marten's Scheme 一 — — — — t = 2 . 9 Modified Marten's Scheme
n I I I • I I I 丨 I I I I I 1 1 1 1 1 1 1 1 0.5 1 1.5 2
Radial Distance (D imens ion less�
( b )
o t=2.5 Relaxed Scheme _ a t=2.7 Relaxed Scheme
• t=2.9 Relaxed Scheme - t=2.5 Modified Haiten's Scheme
t=2.7 Modified Marten's Scheme 0.25 - — — — — t=2.9 Modified Marten's Scheme
<D -
/ Second Shock
Third Shock
n OR ‘ I I I I I I I I I 1 1 1 1 1 1 1 1 1 fi _。」。0 0.5 1 1.5 2
Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 45
(c )
6 I — - > t=2.5 Re laxed S c h e m e - • t=2.7 Re laxed S c h e m e - • t=2.9 Re laxed S c h e m e
5 广 t=2.5 Modi f ied Marten's S c h e m e � t=2.7 Modi f ied Marten's S c h e m e
^ ^ “ — — — — t=2.9 Modi f ied Marten's S c h e m e K �
I 寸 cu -£ 3 -
己 [p Contact Discontinuity
I 2 � ^ ^ ^ ^ ^ ^ ^ S e c o n d S h o c k
1 ^ ^ ^ 細 麵 丨 丨 " 丨 丨
n � I I • I 睡 _ _ ‘ • • 丨 I I 1 1 1 1 1 1 1
0.5 1 1.5 2 Radial Distance {D imens ion less�
(d) t> t=2.5 Relaxed S c h e m e
p Q t=2.7 Relaxed S c h e m e 2.5 — • t=2.9 Relaxed S c h e m e
J- t=2.5 Modif ied Ha i ten 's S c h e m e 1 t=2.7 Modrfied Marten's S c h e m e
^ J \ _ _ _ _ t=2.9 Modified Harten's Scheme
% 2 1\ _ _ _ _
i l r — ^ ― c -|<fc. % fe^ S e c o n d S h o c k
I I V \ S 广 Contact Discontinuity E >
门 I I t I I I I I I I I I I I I I I ‘ ‘ I I 0 0.5 1 1.5 2
Radial Distance (Dimensionless)
Figure 3.4: Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 46
First, the time for the second shock to reflect at the origin is later in the ex-
periment. Second, before the interaction with the second shock, the contact
discontinuity in the experiment do not indicate any inward motion. Our de-
viations obtained are the same as those calculated by the modified Harten's
scheme. Liu et al. [7] suggested three reasons for the deviations. First, the glass
sphere may not have broken instantaneously and it results in a not completely
spherically symmetric flow. Second, the moving fragments of the broken glass
diaphragm causes the energy loss in the flow system. The adjustment by in-
creasing the pressure may partly compensate the loss of the kinetic energy
residing in each fragment, but the loss of the thermal energy is not considered.
Third, the effect of the slower moving fragments, which have much greater
inertia due to their higher density than air,in the flow and across the contact
discontinuity may greatly contribute to the asymmetry of the flow.
3.2 Cylindrical Explosion Problem
For the cylindrical explosion problem, same numerical settings are used as in
the spherical case except that a = 2.
3.2.1 Physical Description
The physics of the cylindrical flow is similar to the spherical flow. Figure
3.9 compares the movement of shocks and contact discontinuities for these
two flows. The main shock is moving faster in the cylindrical flow than in
the spherical flow. Both second shocks propagate outward at first and then
reverse backward to the origin. The cylindrical second shock reverses at a
later time (about t = I) than the spherical case (t = 0.6). So, the reflection
of the second shocks from the origin is later for the cylindrical case (t : 2.05)
than the spherical case(^ = 1.2). Although it is not shown in the figure, it is
expected that the third shock of the cylindrical case would be formed at about
t 二 3.2 when the second shock interacts with the contact discontinuity.
CHAPTER 3. NUMERICAL RESULTS 47
(a)
5 p - > Ma in S h o c k - Re laxed S c h e m e I D S e c o n d S h o c k - Relaxed S c h e m e
4 <3 Third S h o c k - Re laxed S c h e m e 二 • Contact Discontinuity - Re laxed S c h e m e - Ma in S h o c k - Brode Analytical
g 4 — > S e c o n d S h o c k - Brode Analytical • ^ 三 • Contact Discontinuity - Brode Analyt ical .2 3.5 二
E 3 二 Z S E . ^ 2 . 5 - Z 0 = M a i n S h o c k C • r>> ra 2 - -17? 二 ; 口
••口 口口
1 = ’、>、斤 口 z �
« 1 厂 一 C o n t a c t Discontinuity •口•口口口口 Third S h o c k
r S e c o n d S hock • •二•口口f" , � � < 〈工…了 f) • I I I I ^ I I I 1__*L,UniU • I_I I I I I__I I I I I_T iHr rtl__I I I 0 0.5 1 1.5 2 2.5 3
T i m e (D imens ion less )
(b)
5 r— > fAain Shock - Relaxed S c h e m o 二 u S e c o n d S h o c k - R e l a x e d S c h e m e
4 ^ _ c T h i r d S h o c k - R e l a x e d S c h e m e 二 o Contact Discontinuity - Relaxed S c h e m e 3 » Ma in S h o c k - Boyer Exper iment
沿 4 — • S e c o n d Shock - Boyer Exper iment • ^ 三 • Contact Discontinuity - Boyer Exper iment >>
.S 3.5 -= = z E 3 二 Z
— r>> Q I g 三 Ma in Shock
•它 = z ••口
•2 : - Contact Discontinuity n•口口
"CJ . - • 4 4 • • •门 6口 9 • • 2 ^ ~ • • 一 • ••口 - Third Shock
S e c o ” d S h o c k • / o - , /-I 1~ I I I I r I 1 I I I I 口印口 I I I I 产 fa I I I I I I __I I__I
O 0 . 5 1 1 .5 2 2 . 5 3 Time (D imension less)
Figure 3.8: Comparison between the relaxed scheme and the (a) analytical (Brode),
(b) experimental (Boyer) results of the spherical explosion.
CHAPTER 3. NUMERICAL RESULTS 48
5 > Main Shock - Spherical : • Second Shock - Spherical : < Third Shock - Spherical
4.5 - o Contact Discontinuity - Spherical 一 -_ V Main Shock - Cylindrical i? 4 L • Second Shock - Cylindrical a> - • Contact Discontinuity - Cylindrical
I 3 . 5 L
E 3 二 Q : Main Shock
I . 少 Z Z
一 - •口
夸 = ••口•• Third Shock
n ~ I I I [ I I I I I I I Ofn^ 1 I I I I I I t , ^ • T I I I I I 0 0.5 1 1.5 2 2.5 3
Time (Dimensionless)
Figure 3.9: Comparison between the spherical and the cylindrical explosion.
CHAPTER 3. NUMERICAL RESULTS 49
3.2.2 Two-Dimensional Model
As mentioned before, the convenience of the generalization to a higher dimen-
sional space of the relaxation scheme is one of its advantages. We take this
advantage to calculate the two-dimensional model of the explosion problem in
order to compare with the cylindrical approximation model (quasi-ID model).
The 2D scheme has been described in the last section of the previous chapter.
As a 2D 'Riemann problem,,the two sets of initial data are separated by a circle
with radius 0.25. For the points adjacent to the quadrilateral cells cutting the
initial discontinuity, the initial data is modified by assigning the area-weighted
value of the two sets of data. This 'smoothing' process avoids the formation of
small amplitude waves created at early times by the staircase configuration of
the data. In the following numerical results, CFL number二 0.3, the adjusting
factor j3 = 1, Ax = Ay = 0.1
In figure 3.10(a), the quantities of pressure, density, velocity and temper-
ature between the quasi-lD and 2D models are compared at 力=1.65. The
radial distance in the 2D model is extracted from the positive x-a:x:is. At this
particular time position, the second shock has been reflected from d i e � n g m
but has not interacted with the contact discontinuity yet. Figure 3.10(b) shows
the situation at time t 二 2.2, where the third shock has just been formed and
propagating inward. Agreement is good between the two models except for
the position of the contact discontinuity and the position very near the ori-
gin. We also observe that there is a strange 'trough' for the velocity profile
at about the position of the contact discontinuity. Since the diversities stated
above are much worse if we don't carry out the 'smoothing' process of the
area-weighted for the initial values, or using a coarser gird. So, it is believed
that the rectangular grid of our finite difference scheme which results in a
non-circular geometry of the initial data causes those diversities stated above.
Although the density and temperature profiles at the position very near the
origin are different for the two models, however, the highly agreement of the
CHAPTER 3. NUMERICAL RESULTS 50
(a) 2厂
2 D M o d e l : 2 D M o d e l I . ” o Quasi 1D Model 1.9 二 o Model
i i r y 0.2 ^__:
ofc=f , • . , I I I • , . • . I . • • i I ' ' -I • ' ' ' j ' ' ' ' 3 ' ' ' ' i Distance (nimeitsinnless Ratlml D istancc < Dimension less)
2.5 r-2Q Model • 20 Model OJ3 L o Quasi-1 • Modal o QuasMD Model
-0.4 - •
: I . _ . _ ~ « — • ~ I — ' ~ • • , I ~ ~ — ' ~ I ^ 2 3 A 0 1 2 3 4 R dUil Distance (D linenslunlesst Radial D»tarice mirrieris'iciriles )
(b) 1 " r on … � . 2D Mendel
: - M o d e . ; 。 Q u a m D Model
3.5 -
l i I ” t o e S 2 ]
of • • I • j I • I I ” I , I j I • •,丄 � 4 ' ' ' ' -i • ' ' ' j ' ' ' ' i ' ' ' ' A R.Klial Disrance 4 Dimension less) Rartkil O isrtancR <r>inu?i»s«>nless>
0-4 r 2D Model F 2D Model
o Quasi-1D Modal ^ ^ ^ = Quas-iD Model
I � : A X t ~ . _ � f
t � � X i :
- o . e : . . . . . , . . • • • • • . . • : . • • _ . . ‘ . I . _ . _ . _ . _ I _ . . _ . . _ I -OB^‘ ‘ ‘ ‘ :: ‘ 2 3 4 Q 1 2 3 ^ RaOMI Distcirice |D imensioriless) Rodiol D istance (Diinen ioi l'evs)
Figure 3.10: Comparison between the 2D model and the quasi-ID model of the
cylindrical explosion at t ime (a) t 二 1.65 and (b) t = 2.2.
CHAPTER 3. NUMERICAL RESULTS ,5]
Figure 3.11: Three-dimensional plot of the pressure contour of the cylindrical ex-
plosion at time i — 2.5.
shock position away from the origin shows that the quasi-ID model is a good
approximation to the cylindrical symmetric problem. Our reflective boundaries
in (2.13) and the compatible equations at the origin (2.8) are acceptable for
the simulation of the reflection of shock at the origin. Also, negligible or even
none dimensional effect are observed in affecting the waves interaction and re-
flection on the numerical scheme. Figure 3.11 shows the three-dimensional plot
of the pressure contour at time t 二 2.5. The roughly circular shape indicates
the cylindrical symmetric of the flow.
CHAPTER 3. NUMERICAL RESULTS 52
Internal Density 二 1.0
Internal Velocity ul = 0
Internal Pressure pl — 1-0
External Density = 4.0
External Velocity uh 二 Q
External Pressure pH = 4.0
Table 3.3: Initial Data for Implosion Problem.
3.3 Spherical Implosion Problem
The initial condition of the implosion problem are summarized in table 3.3.
The radius of the sphere is 0.25. In the implosion problem, the numerical
settings are Ax = 0.05, CFL number= 0.3,the adjusting factor (3=1.
3.3.1 Physical Description
At the time of the glass sphere broken (t 二 0): a moving inward shock wave
(main shock) and a moving outward rarefaction wave form. A contact discon-
tinuity, which is moving inward separates the shock and the rarefaction waves.
Different from the explosion problem, there is no second shock follows at this
moment (see figure 3.12). The main shock then reflects from the origin and
interacts with the contact discontinuity at about t = 0.2. Due to the inter-
action, the contact surface stops moving inward. The contact discontinuity
propagates slightly outward for a short time and then stays stationary at the
position X 二 0.175. A second shock which is moving inward is generated due to
the interaction. The main features of the implosion problem are captured by
the relaxed scheme, as effectively as the modified Harten's scheme. Pressure
is imploded at about t = 0.3 that indicates the reflection of the second shock
at the origin (figure 3.14). We don't observe the formation of the third shock
CHAPTER 3. NUMERICAL RESULTS 53
when the second shock encounters the contact discontinuity. It may because
it is physically too weak to observe.
3.4 Cylindrical Implosion Problem
For the cylindrical implosion problem, same numerical settings are used as in
the spherical case except that a = 2. For the initial setting stated above, it is
exactly the same model as Sod [6] and Liu et al. [7 .
3.4.1 Physical Description
As in the explosion case, the physics of the cylindrical implosion is similar
to the spherical implosion. Figure 3.16 compares the movement of shocks and
contact discontinuities for these two implosion flows. The main shock interacts
with the contact discontinuity at about the same position in both cases. Unlike
the case of the explosion problem, the main and second shocks are moving a
bit slower in the cylindrical flow than in the spherical flow. No third shock is
observed in both cases and it may because the third shock is too weak to be
observed
3.4.2 Two-Dimensional Model
Similar to the analysis in the explosion problem, a comparison between 2D
model and the quasi-ID model is made. The numerical settings of the 2D
model are, CFL number= 0.3, the adjusting factor 二 1 and Ax = Ay = 0.05.
Good agreement is shown in figure 3.17 between the 2D model and the quasi-
ID model except at the position near the origin. Due to the reasons stated
before in the explosion section, same conclusion can be drawn in this implosion
case. The quasi-ID model as an approximation model to the cylindrical flow
is acceptable. Figure 3.18 shows the three-dimensional plot of the pressure
contour at time t 二 0.5. The roughly circular shape indicates the cylindrical
symmetric of the flow.
CHAPTER 3. NUMERICAL RESULTS 54
(a)
OQ > t = 0 . 0 5 R e l a x e d S c h e m e ‘ r � ° t = 0 . 1 0 R e l a x e d S c h e m e
i io o t =0 .13 Re laxed S c h e m e 1 3 J- I <3 t = 0 . 1 5 R e l a x e d S c h e m e
4 ; j\ t=0 .05 Modi f ied Ha i ten ' s S c h e m e 1 a L. {' t = 0 . 1 0 Modif ied Marten's S c h e m e
"vT r ; I t=0 .13 Modi f ied Ha i ten ' s S c h e m e trt 4 /中< t=0 .15 Modi f ied Marten's S c h e m e
1 14 . 2 \ c 12 V" Ref lected M a in S h o c k f rom Or ig in <U “ : �
I 8 二 cn - <3> •ij t/t -6 - \ \ Q- : M a i n S h o c k
I riiiii'iiiiiiii imin iin ih 11 ill 11 r IM r 1 1 1 — i — i — i — i — I — i — i — i — i — I — i — i — i — i — I O 0.1 0 . 2 0 . 3 0 . 4 0 . 5
Radial Distance (Dimensionless)
(b) o t=0 .05 Relaxed S c h e m e
� a t=0 .10 Relaxed S c h e m e : o t=0 .13 Relaxed S c h e m e
. _ - t = 0 . 1 5 R e l a x e d S c h e m e 1 . o 一
: t=0 .05 Modi f ied Ha i ten 's S c h e m e = t=0 .10 Modi f ied Marten's S c h e m e
- H 1 二 — — — — t=0 .13 Modi f ied Marten's S c h e m e ^ - — "1=0.15 M o d i f i e d H a r t e r V s S c h e m e
. 1 0.5 - 一C工 Reflected Ma in S h o c k f r o m Origin
I I ' V Eo.5 > -1 ri: M a i n S h o c k
-
二 • -1.5 - o
r I I I I I I I I I I I I I I I I I I I I—I—I—I—I—I 0 0.1 0 . 2 0 . 3 0 . 4 0 . 5
Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 55
(c) 1Q > t = 0 . 0 5 R e l a x e d S c h e m e
p o t=0 .10 Re laxed S c h e m e 1 \ o t=0 .13 Re laxed S c h e m e
'"j ^ <1 t=0 .15 Re laxed S c h e m e g t t=0 .05 Modi f ied Ha i ten 's S c h e m e
o 丨, t =0 .10 Modi f ied Harten 's S c h e m e 一 S t=0 .13 Modi f ied Ha i ten 's S c h e m e Jj) ^ — t=0 .15 Modi f ied Harten 's S c h e m e
考 : j \ A .O < 经 ^ \ Ref lected M ain S h o c k f rom Orig in ^ J / V � �
i — o :
“ y S i — Contact Discontinuity 2 一
~ p Main Shock I r V “ r , L . i i U r > “ _ , , - M < f * ^ I I I 1 1 1 ~ I I I — — I — — I ~ I ~ ~ I 1——I 1——J
O 0.1 0 . 2 0 . 3 0 . 4 0 . 5 Radial Distance (Dimensionless)
<d)
p > t = 0 . 0 5 R e l a x e d S c h e m e � D t=0 .10 Relaxed S c h e m e
o t=0.1 3 Relaxed S c h e m e � ^ t = 0 . 1 5 R e l a x e d S c h e m e _ ^ t=0 .05 Modi f ied Harten's S c h e m e
t=0 .10 Modi f ied Hai ten 's S c h e m e - \> - — —— — t=0 .13 Modi f ied Harten's S c h e m e
-S fi _ t=0 .15 Modi f ied Hai ten 's S c h e m e
§ k \ 巧 ^^^^^ Reflected Ma in Shock from Origin ��
a -4 K x ^ z Contact Discontinuity
f 2 - Ma in S h o c k ^ ^ ^ ^ ^ ^ ^
I I 1 t I I I I I I I t I I I I t I I I I 1 1 1 1
O 0.1 0 .2 0 . 3 0 . 4 0 . 5 Radial Distance (Dimensionless)
Figure 3.12: Numerical Simulation of Implosion for (a) Pressure (b) Velocity (c)
Density and (d) Temperature from time t — 0.05 to t = 0.15.
CHAPTER 3. NUMERICAL RESULTS 56
(a)
� > t=0 .18 Re laxed S c h e m e : ° t =0 .20 Re laxed S c h e m e
o t=0 .24 Re laxed S c h e m e : < t =0 .27 Re laxed S c h e m e “ t =0 .18 Modi f ied Har ten 's S c h e m e
.^^_ _ t =0 .20 Modi f ied Har ten 's S c h e m e 办 — — — - t = 0 . 2 4 Modi f ied Har ten 's S c h e m e
‘ ^ ^ t=0 .27 Modi f ied Har ten 's S c h e m e
: M o v ' i n g TowaJd O
O • I I I • I I I I 窗丨 I I I I I I I I I O 0.1 0.2 0.3 0.4 0.5
Radial Distance (D imensionless)
(b) 1 厂 > t=0 .18 Relaxed S c h e m e
- ° t=0 .20 Re laxed S c h e m e - o t=0 .24 Re laxed S c h e m e _ < t=0 .27 Re laxed S c h e m e _ t=0 .18 Modi f ied Harten 's S c h e m e _ t=0 .20 Modi f ied Harten 's S c h e m e
"vT 0.5 — — — — - t=0 .24 Modi f ied Harten 's S c h e m e ^ _ ^ ^ ^ t=0 .27 Modi f ied Harten 's S c h e m e
"cu > - 0 . 5 - S e c o n d Shock
- Mov ing Toward Origin
一 1 I 1 I I I 睡 I f • I I 丨 I I I I I I I I I I I I I I O 0.1 0.2 0.3 0.4 0.5
Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 57
(C)
[T > t=0 .1S Re laxed S c h e m e _ 口 t =0 .20 Re laxed S c h e m e - • t=0 .24 Re laxed S c h e m e — < t=0 .27 Re laxed S c h e m e _ ! — t=0 .18 Modi f ied Har ten 's S c h e m e ~ t=0 .20 Modi f ied Har ten 's S c h e m e
— 一 一 一 t=0 .24 Modi f ied Harten 's S c h e m e gj - M a i n S h o c k t=Q.27 Modi f ied Har ten 's S c h e m e •Si 丨 n teractwith g - Contact Discontinu'rty •JJ5 - to fo rm S e c o n d S h o c k g - M o v i n g T o w a r d Or ig in
^ � M • S h k
Ihock*^ Contact Discontinuity o I I I I I I I I 1 1—I 1 1 I • I I I I I I I I I I I
O 0.1 0.2 0.3 0.4 0.5 Radial Distance (Dimensionless)
(d)
> t = 0.1 S Relaxed S c h e m e _ ° t=0.20 Relaxed S c h e m e
• t=0 .24 Relaxed S c h e m e _ < t=0 .27 Relaxed S c h e m e
. . 广 - t=0 .18 Modi f ied Harten 's S c h e m e g ^ - t=0 .20 Modi f ied Harten 's S c h e m e h — 一 一 一 t=0 .24 Modi f ied Harten 's S c h e m e "c - t=0 .27 Modif ied Harten 's S c h e m e .O -c 5 — Main Shock 丨irteract with 髮 _ tp p. ^ ^ ^ ^ Contact Discontinuity .E _ V i ^ t o form S e c o n d S h o c k & - Mov ing Toward Origin
I 4 -
oJ - r \ I 鬼货 _ 一 一一一――…Contact Discontinuity
# ^ | i Shock Main S h o c k
9 ll I I I I I I I I I I I I I I I I ' I ' I 1 1 1 1 J 0.1 0.2 0.3 0.4 0.5 Radial Distance (Dimensionless)
Figure 3.4: Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 58
(a)
^ r~ Imp los ion of > t=0 .2S Re laxed S c h e m e 二 P r e s s u r e due to ° t =0 .30 Re laxed S c h e m e - Ref lect ion • t =0 .32 Re laxed S c h e m e : o f S e c o n d S h o c k < t = 0 . 3 5 Re laxed S c h e m e
_ “ at O r ig in t=0 .2S Modi f ied Har ten 's S c h e m e _一“ ' — t=0 .30 Modi f ied Har ten 's S c h e m e Jg / — — — - t = 0 . 3 2 Modi f ied Har ten 's S c h e m e Qi ' ^ t=Q.35 Modi f ied Har ten 's S c h e m e
i 發 , M a i n S h o c k
^ - Ref lected ^ ^ " ‘ 4 ^ ^ ~ —
r j I I I I I I I I I I I I I I I I I I 1 1 I O 0 . 2 5 0 . 5 0.75 1
Radial Distance (D imension less)
(b)
t=0. 28 Re laxed S c h e m e � t二0.30 Re laxed S c h e m e
- o t=0 .32 Re laxed S c h e m e 0.2 一 民弹 < t=0 .35 Re laxed S c h e m e
- r p ^ t=0 .28 Modi f ied Harten 's S c h e m e - r S ^ iJv t=0 .30 Modi f ied Ha i ten 's S c h e m e - a 7 — _ t=0 .32 Modi f ied Harten 's S c h e m e
Jg - f / / ] / 1 t=0 .35 Modi f ied Harten 's S c h e m e
‘ : 旧 ^ ^ I 1 I—I—I I I I I I i ' • I ' 1 ' I ' ' I
0 0.25 0.5 0.75 1 Radial Distance (Dimensionless)
CHAPTER 3. NUMERICAL RESULTS 59
(c)
7 n-^ t=0 .2S Re laxed S c h e m e 口 t=0 .30 Re laxed S c h e m e • t=0 .32 Re laxed S c h e m e < t=0 .35 Re laxed S c h e m e
^ - t=0 .28 Modi f ied Marten 's S c h e m e — b t=0 .30 Modi f ied Marten 's S c h e m e "t/T M — — — - t=0.32 Modified Marten's Scheme 2 I t=0 .35 Modi f ied Marten 's S c h e m e
•1 ,,丨 Ma in S h o c k
11 K ^ ^ o 3 \
Reflected Contact Discontinuity _ S e c o n d _ S h o c k
O I I I I I I I 1 I I I I I I I I I I 1 1
0 0.25 0.5 0.75 1 Radial Distance (Dimensionless)
(d)
5 厂 Reflected > t=0.2S Relaxed S c h e m e - S e c o n d 口 t = 0 . 3 0 R e l a x e d S c h e m e
Shock • t=0 .32 Relaxed S c h e m e - < t=0 .35 Relaxed S c h e m e
一 急 ^ ^ t=Q.28 Modi f ied Marten's S c h e m e ^ t=0 .30 Modif ied Marten's S c h e m e a> — — — - t=Q.32 Modif ied Marten's S c h e m e
"c ^ t=0 .35 Modif ied Marten's S c h e m e 0 4 -f g ^ 1 1 :: 1 Q ‘> ^ ^ ^ Contact Discontinuity
¥ B -<’ f
I � 卜 . : ^ ^ 一 脚
I I I 1 I I I I ) I I I I I I I I I 1 I
0 0.25 0.5 0.75 1 Radial Distance (Dimensionless)
Figure 3 . 4 : Numerical simulation of the explosion problem for (a) pressure (b) ve-
locity (c) density and (d) temperature from time t 二 1.15 to t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 60
(a )
6 � t = 0 . 4 R e l a x e d S c h e m e _ ° t = 0 . 6 R e l a x e d S c h e m e - • t = 0 . 8 R e l a x e d S c h e m e - < t = 1 . 0 R e l a x e d S c h e m e
5 .5 — t = 0 . 4 M o d i f i e d H a r t e n ' s S c h e m e _ --.. “ t = 0 . 6 M o d i f i e d H a r t e n ' s S c h e m e JJ5 I — — — - t = 0 . 8 M o d i f i e d H a r t e n ' s S c h e m e at fe t = 1 . 0 M o d i f i e d H a r t e n ' s S c h e m e
.O - S e c o n d 爹、 g - S h o c k 量
.i 4.5 二 S h o c k
I ~ o r . I [ I I I I I I I • I t I I_I_I_I_I_I_I_I_I_I_I_I_I_I_I_I_I_I_I_I_I 0 0.25 0.5 0.75 1 1 . 2 5 1.5 1.75
Rad ia l D is tance ( D i m e n s i o n l e s s )
( b )
- > t=0 .4 R e l a x e d S c h e m e _ ° t = 0 . 6 R e l a x e d S c h e m e
Q 2 _ • t = 0 . 8 R e l a x e d S c h e m e - < t=1 .0 R e l a x e d S c h e m e - 1=0-4 M o d r f i e d Ha r ten ' s S c h e m e _ t = 0 . 6 M o d i f i e d Har ten ' s S c h e m e - S e c o n d 铁 — — — - t = 0 . 8 M o d i f i e d Ha r ten ' s S c h e m e
g - S h o c k k \ t = 1 . 0 M o d i f i e d Ha r ten ' s S c h e m e
I I I I I I I I I I I I I I_I__I__1_I__I_I__I__I_1_I_I_I__I_I_I~I~I~I~I~I~I 0 0 . 2 5 0 . 5 0 . 7 5 1 1.25 1 . 5 1 . 7 5
Radia l D istance (D imens ion less )
CHAPTER 3. NUMERICAL RESULTS 61
<c)
6 � > t=0 .4 Re laxed S c h e m e ° t=0 .6 Re laxed S c h e m e o t=0 .8 Re laxed S c h e m e < t=1 .0 Re laxed S c h e m e
t=0.4 Modrf ied Harten 's S c h e m e t=0 .6 Modrf ied Harten 's S c h e m e
5 - t=0 .8 Modi f ied Harten 's S c h e m e </> ^ t=1 .0 Modi f ied Harten 's S c h e m e
•i ^ .Q 1 ^ ^ ^ ^ M a i n S h o c k I S "> u:' ^ ^ ~ Contact Discontinuity
。!J � 1 1 1 1 1 1 1 1 i I 塞 1 1 1 1 1 1 1 _ I _ _ I _ I _ I _ I _ I _ I _ I _ I _ I _ I _ _ I _ I _ I ~ I ~ I ~ I ~ I
0 0.25 0.5 0.75 1 1.25 1.5 1-75 R a d i a l D i s t a n c e ( D i m e n s i o n l e s s )
( d )
4 5 � ^ t=0.4 Relaxed Sc ! ‘ eme _ ° t = 0.6 Relaxed S c h e m e - • t=O.S Relaxed S c h e m e -爽 < t=1.0 Relaxed S c h e m e
^ . t=0.4 Modi f ied Harten's S c h e m e ^ 4 - a t=0.6 Modi f ied Harten's S c h e m e S 容^ f e — — — - t=O.S M o d m e d Harten's S c h e m e
"E , ^ ^ t=1.0 Modrf ied Harten's S c h e m e I
S 3.5 - I £ ] C o n t a c t D i s c o n t i n u i t y
a i:
3 ^ :7 g “ , Main Shock
I 2.5 ;- ^^Tl^^S^^;[;
� • • I I I _ , I I I I I 1 丨_I_I_I_I 1_I_i_1 1_I_I_I——I_I 1_I~I~i 1~I~I 0 0.25 0.5 0.75 1 1.25 1.5 1.75
R a d i a l D i s t a n c e ( D i m e n s i o n l e s s )
F i g u r e 3.4: Numer i ca l s imulat ion of the explos ion p r o b l e m for (a ) pressure (b ) ve-
l o c i ty ( c ) density and (d ) t e m p e r a t u r e f r o m t i m e t 二 1.15 t o t 二 1.30.
CHAPTER 3. NUMERICAL RESULTS 62
p > Main Shock - Spherical
- • Second Shock - Spherical 1> _ o Contact Discontinuity - Spherical _ Main Shock - Cylindrical &
2 1 一 • Second Shock - Cylindrical / 广 S • Contact Discontinuity-Cylindric^il X • 0) j> Z —薩 0 - 7 m 0.75 - tx X ^ / •i — y z <U Main Shock > / Z 门 � 1 0 . 5 - Z / Z Second Shock
I : 1 -符 ^ ^ i jn Contact Discontinuity
:V八 0 I I ^ I I—I—I—I—I—I—I—I—I—I—I—I—I—I
0 0.25 0.5 0.75 1 Time (Dimenstonless)
Figure 3.16: Comparison between the spherical and the cylindrical implosion.
CHAPTER 3. NUMERICAL RESULTS 63
5 r- m 5 -2D Wtottel - 2D Model
a QLHSi-ID Model a ° Quasi-1D MoOei
I ; I
r| \f~ W • • _ _ I , • • • • . . , • I . , , , I 2 5 I I •‘ III_I I I_I_I—I—I—t—I_I——I——I——I
2 5� R d D rstancefDunensionilss) ‘ � Rad 'l fDimensionllss, ‘
5 p
^ I r I : "|��l : I / 竺 - : \
1.2: 1 / 1 : : \ • 善 � . 3 — V y I 3 : \
-0.5J I I I I • ‘ I ‘ ‘ ‘ I ' ‘ ‘ ‘ ‘ J 0.25 0.5 0.75 1 � Rmllll Distancefoimensionl!..) R miial D bt. nce (Dimensionless)
Figure 3.17: Comparison between the 2D model and the quasi-ID model of the
cylindrical implosion at time t 二 0.4.
CHAPTER 3. NUMERICAL RESULTS 64
Figure 3.18: Three-dimensional plot of the pressure contour of the cylindrical im-
plosion at time t 二 0.5.
Chapter 4
Conclusion
The present development provides a demonstration on the numerical simula-
tion of explosion and implosion problems by using the relaxed scheme.
Cylindrical and spherically symmetric explosion and implosion in air are
numerically simulated. The governing equation is a nonlinear systems of hy-
perbolic conservation laws with source terms. The inhomogeneous system is
split for a time step, into the advection problem which is a homogeneous hy-
perbolic problem: and the source problem which is cl system of ordinary differ-
ential equations (ODEs). The ODEs is solved by the second-order, two-stage
Runge-Kutta method. The pure advection hyperbolic problem is calculated
by the second-order relaxed scheme developed by Jin & Xin [16]. A constant
matrix is constructed in the application of the relaxed scheme. The nu-
merical dissipation of the relaxation scheme is to a large extent determined by
the matrix A and the choice of A becomes critical in applying the relaxation
scheme. An algorithm to choose the matrix is developed. The matrix is cho-
sen 'dynamically' such that it is tailor-made for each time step interval. The
matrix constructed as A = al where the constant a defined as the square of
the maximum eigenvalue inside the time interval for all spatial do-
main. We impose no additional criteria on the CFL number for the algorithm
proposed in practical use. Though we cannot prove the TVD property of the
scheme in system of conservation laws, a series of numerical tests are carried
65
CHAPTER 4. CONCLUSION 66
out to exam the applicability of the algorithm on the Euler system. The test
results are satisfactory.
By applying the relaxed scheme in the explosion and implosion problem,
four cases namely spherical explosion, cylindrical explosion, spherical implo-
sion and cylindrical implosion are presented. For the explosion problem, at the
time immediately after the broken of the spherical or cylindrical diaphragm,
the main shock is generated and moves outward in the air. The compressed
gas is expanded through a rarefaction wave and a contact discontinuity sepa-
rates the expanded gas from the air compressed by the main shock. Different
from an one-dimensional shock tube problem, due to the three-dimensionality
of the flow, the region between the tail of the rarefaction wave and the main
shock wave is not a steady-state region. The high pressure gas upon passing
through a spherical rarefaction wave, due to the increase of volume from di-
mensionality, must expand to a lower pressure than that reached through an
equivalent one-dimensional expansion. This 'over-expansion' is compensated
by a second shock. This second shock is rather weak and propagates outward
initially. Its strength increases with time and the second shock stops moving
outward, but starts reversing to the inward direction and reflects from the
origin. Also, the contact discontinuity propagates outward slowly and starts
reversing the direction until it interacts with the reflected second shock. The
contact discontinuity propagates slightly outward again and a third shock gen-
erated similarly. The numerical simulation has successfully captured all the
main features of the explosion problem including the propagation of the main
shock, the reflection of the second shock, the interaction with the contact dis-
continuity and the formation of the third shock, as well. Comparison is made
between previous numerical results [7], analytical results [1] and experimental
results [2]. Also, thanks to the advantage of the relaxation scheme to gen-
eralize to a higher dimensional space easily, a 2D model with circular initial
condition is calculated to check the cylindrical symmetric model. The results
of our quasi-ID model is found acceptable.
CHAPTER 4. CONCLUSION 67
For the implosion problem, at the time of the spherical or cylindrical di-
aphragm broken, an inward moving shock wave (main shock) and an outward
moving rarefaction wave form. A contact discontinuity, which is moving inward
separates the shock and the rarefaction waves. Unlike the explosion problem,
there is no second shock follows at this moment. The main shock then re-
flects from the origin and interacts with the contact discontinuity causing the
formation of the second shock (similar to the formation of the third shock in
the case of the explosion problem). Due to the interaction, the contact surface
stops moving inward but propagates slightly outward for a short time and then
stays stationary. The main features of the implosion problem, the reflection of
the main shock, the movement of the contact discontinuity and the formation
of the second shock are captured by the relaxed scheme, as effectively as the
modified Harten's scheme. We don't observe the formation of the third shock
when the second one encounters the contact discontinuity. It may because it
is physically too weak to observe. A 2D model has been presented so as to
compare to our cylindrical symmetric quasi-ID model.
The present research contribnteR a picture to understand the phvsics of the
explosion and implosion problems through the numerical siniulaLion by the
relaxed scheme. Meanwhile, the technique of the application of the relaxed
scheme is developed especially on the choosing of the important matrix A,
which plays the critical role on the accuracy and practicality of the scheme.
Although the matrix A chosen by the algorithm developed in this research
has shown to be practically capable for most Euler problems by the numeri-
cal tests, it is still a rough estimate and in no way guarantee the dissipative
condition to be always satisfied. Based on the present finding, it would be
valuable for the future to find a procedure for choosing A that will rigorously
be proved to fulfill the dissipative condition. Moreover, noted the smearing
effect on the contact discontinuity (figure 3.1(b)),equipping the relaxation
scheme with ACM technique can also be considered as a future development
so as to eliminate the smearing effect.
Bibliography
1] H. L. Erode, "Theoretical Solutions of Spherical Shock Tube Blasts," The
RAND Corporation, RM-1974, 1957.
2] D. W. Boyer, "An Experimental Study of the Explosion Generated by a
Pressurized Sphere,” J. Fluid Medi., vol. 9,pp. 401-429, 1960.
3] M. P. Friedman, "A Simplified Analysis of Spherical and Cylindrical Blast
Waves," J. Fluid Meek., vol. 11, pp. 1-15, 1961.
4] W. E. Baker, Explosions in Air^ Wilfred Baker Engineering, San Antonio,
1973.
5J I. I. Glass, Nonstationary Flows and Shock Waves^ Oxford Science Publi-
cations, Oxford, 1994.
6] G. A. Sod, "A numerical study of a converging cylindrical shock," J. Fluid
Mech., vol. 83, pp. 785-794, 1977.
7] T. G. Liu, B. C. Khoo and K. S. Yeo, "The Numerical Simulations of
Explosion and Implosion in Air: Use of a Modified Harten's TVD Scheme,"
Int. J. Numer. Meth. Fluids, vol. 31, pp. 661-680, 1999.
8] J. D. Lambert, Computational Methods in Ordinary Differential Equations,
John Wiley and Sons, 1973.
[9] A. Hart en, "High Resolution Schemes for Hyperbolic Conservation Laws,"
J. Comput. Phys., vol. 49, pp. 357-393, 1983.
68
BIBLIOGRAPHY 69
10] A. Hart en and S. Osher,"Uniformly High-order Accurate Non-oscillatory
Scheme. I,” SIAM J, Numer. Anal., vol. 24, No. 2,pp. 279-309, 1987.
11] A. Harten, B. Engquist, S. Osher and S. Chakravarthy, “Uniformly High-
order-accurate Essentially Non-oscillatory Scheme, III," J. Comput. Phys.,
vol. 71, pp. 231-303, 1987.
12] C. W. Shu and S. Osher, "Efficient Implementation of Essentially Non-
oscillatory Shock-Capturing Schemes," J. Comput. Phys., vol. 77, pp. 439-
471, 1988.
13] A. Harten, "ENO Scheme with Subcell Resolution," J. Comput. Phys.,
vol. 83, pp. 148-384, 1989.
14] A. Harten, "The Artificial Compression Method for computation of
Shocks and Contact Discontinuities: I. Single Conservation Laws," Comm.
Pure Appl. Math” vol. 30, pp. 611-638, 1977.
15] A. Harten, "The Artificial Compression Method for Computation of
Shocks and Contact Discontinuities; Til. Self-adjusting Hybrid Schemcs,"
Math. Comput, vol. 32, pp. 363-389, 1978.
16] S. Jin and Z. P. Xin, "The Relaxation Schemes for Systems of Conserva-
tion Laws in Arbitrary Space Dimensions," Comm. Pure Appl. Math.) vol.
49, pp. 235-276,1995.
17] S. Jin and Z. P. Xin, "Numerical Passage from Systems of Conservation
Laws to Hamilton-Jacobi Equations, and Relaxation Schemes," SIAM J.
Numer. Anal” vol. 35, No. 6, pp. 2385-2404, 1998.
18] A. Chalabi, "Convergence of Relaxation Schemes for Hyperbolic Conser-
vation Laws with Stiff Source Terms," Mathematics of Computation, vol.
68, No. 227, pp. 955-970, 1999.
BIBLIOGRAPHY 70
19] S. Jin, M. A. Katsoulakis and Z. P. Xin, "Relaxation Schemes for
Curvature-Dependent Front Propagation," Comm. Pure Appl. Math., vol.
52, pp. 1587-1615, 1999.
20] A. Klar, "Relaxation Scheme for a Lattice-Boltzmann-type Discrete Ve-
locity Model and Numerical Navier-Stokes Limit," J. Comput. Phys., vol.
148, pp. 416-432, 1999.
21] S. Jin and L. Pareschi, "Discretization of the Multiscale Semiconductor
Boltzmaim Equation by DiflFusive Relaxation Schemes," J. Comput. Phys.,
vol. 161, pp. 312-330, 2000.
22] R. J. Leveque and M. Pelanti, "A Class of Approximate Riemann Solvers
and Their Relation to Relaxation Schemes," J. Comput. Phys., vol. 172,
pp. 572-591, 2001.
23] A. Chalabi and D. Seghir, "Convergence of Relaxation Schemes for Ini-
tial Boundary Value Problems for Conservation Laws," Computers and
Mathematics with Applications, vol. 43,pp. 1079-1093, 2002.
24] S. Evje and K. K. Fjelde, ''Relaxation Schemes for the Calculation of
Two-Phase Flow in Pipes," Mathematical and Computer Modelling) vol.
36, pp. 535-567, 2002.
25] A. Jiingel and S. Tang, "A Relaxation Scheme for the Hydrodynamic
Equations for Semiconductors," Applied Numerical Mathematics, vol. 43,
pp. 229-252,2002.
[26] X. G. Li, X. J. Yu and G. N. Chen, "The Third-order Relaxation Schemes
for Hyperbolic Conservation Laws," J. Computational and Applied Math-
ematics, vol. 138, pp. 93-108,2002.
27] A. Klar, L. Pareschi and M. Seaid, “Uniformly Accurate Schemes for
Relaxation Approximations to Fluid Dynamic Equations," Applied Math-
ematics Letters, vol. 16, pp. 1123-1127, 2003.
BIBLIOGRAPHY 71
28] R. Burger, K. H. Karisen and N. H. Risebro, "A Relaxation Scheme for
Continuous Sedimentation in Ideal Clarifier-Thickener Units," Computers
and Mathematics with Applications^ vol. 50, pp. 993-1009, 2005.
29] G. Strang, "On the Construction and Comparison of Difference Schemes,,,
SIAM J. Numer. Anal., vol. 5, No. 3, pp. 506-517, 1968.
30] R. J. Leveque and H. C. Yee, "A Study of Numerical Methods for Hyper-
bolic Conservation Laws with Stiff Source Terms," J. Comput. Phys., vol.
86, pp. 187-210, 1990.
31] T. P. Liu, "Hyperbolic Conservation Laws with Relaxation," Commun.
Math. Phys., vol. 108, pp. 153-175, 1987.
32] B. Van Leer, "Towards the Ultimate Conservative Difference Scheme. V.
A Second-Order Sequel to Godunov's Method," J. Comput. Phys., vol. 32,
pp. 101-136, 1979.
33] P. K. Sweby, "High Resolution Schemes Using Flux Limiters for Hyper-
bolic ConservB.tion Laws." SIAM J. Nvmer. AnnJ,.. vol. 21. No. 5. dd.
995-iOll, 1964.
34] E. F. Toro, Riemann Solvers and Numerical Methods for Fluid Dynamics:
A Practical Introduction, 2nd ed., Springer-Verlag Berlin Heidelberg, New
York, 1999.
35] G. A. Sod, Numerical Methods in Fluid Dynamics, Cambridge University
Press, 1985.
36] B. Einfeldt, C. D. Munz, P. L. Roe and B.Sjogreen, "On Godunov-Type
Methods near Low Densities," J. Comput. Phys., vol. 92, pp. 273-295,
1991.
37] P. Woodward and P. Colella, “The Numerical Simulation of T w o
Dimensional Fluid Flow with Strong Shocks," J. Comput. Phys., vol. 54,
pp. 115-173, 1984.
C U H K L i b r a r i e s
004359385