The main feature of gas dynamics is the
appearance of discontinuities, more precisely, strong gradients. The quality of
computational methods is assessed primarily by their ability to convey this
behavior of a solution as adequately as possible. In our opinion, the
discontinuous particle method [1–3] allows one to cope with these difficulties
better than alternative, traditionally more commonly used difference and finite
element methods. This is achieved because the particle method is based on the
Lagrange approach, and this, in turn, provides automatic mesh generation. In
addition, particle methods have a constructive inclination to parallelization,
economical from the point of view of multidimensionality, ideologically organic
to hierarchical transitions between micro-macro models of the considered
phenomena. It is worth noting that the problems of evaluating the accuracy of
numerical methods on discontinuities currently are very relevant [4, 5].
The Burgers equation [6, 7], viscous and inviscid,
is the simplest model, which shows the key feature for the whole gas dynamics: the
existence of zones with strong gradients of the solution. Having correctly
numerically simulated such zones, one can expect to build effective
computational methods, not only and not so much for macro-models, but also for
meso - levels of the hierarchy in the representation of the Kolmogorov – Fokker
- Planck equations [1, 8, 9].
Using the terminology of Hockney and
Eastwood [10], we divide the whole set of particle methods into 3 subclasses:
particle-grid (PM) methods, particle-particle method (PP) and hybrid
particle-particle - particle-grid methods (PPPM or P3M). In this
classification, the discontinuous particle method can be classified as a
particle-particle method. The smoothed particle hydrodynamics method and its
numerous modifications [11] or the material point method [12] belong to the
same subclass. The particle method is used to solve problems in plasma physics
and gas dynamics [10, 13, 14].
Let us assume there be
material
points located at the initial time in coordinates
and
moving with velocities
. This verbal formulation corresponds to
the Cauchy problem:
|
(1)
|
This problem is
a micromodel of the transport process. Let us show that one can pass from Eq.
(1) to a macromodel of the linear transport process.
We determine the distribution density
:
,
|
(2)
|
.
|
|
The article [2] shows the transition from (1) and
(2) to the generalized transport equation:
,
|
(3)
|
and then to the transport
equation in differential form:
|
(4)
|
That is, if the coordinates of points change
according to the system of equations (1), then the density
is a
generalized solution of the Cauchy problem for the transport equation (4).
Instead of equation (4) we solve the system of equations (1). We can say that
using the representation of the desired function as a set of δ-functions.
What follows is an approximation of the δ-functions by classical
functions, with which the calculations at each time step are performed.
Now write down another ODE system:
|
(5)
|
Making a
transition from (5) similar to that for system (1), we obtain a one-dimensional
quasi-linear transport equation:
|
(6)
|
Let's write down yet another ODE system:
|
(7)
|
Similarly, it
leads to a two-dimensional quasi-linear transport equation:
|
(8)
|
In this section, we obtain the basic equations to
which our method will later be applied.
Let us describe the discontinuous particle method
for one-dimensional quasi-linear transport equation. Let's introduce the
following notations:
is the coordinate of center of i-th particle at k-th moment of time,
is the velocity of particle,
is the height (density) of particle,
is
the area (mass) of particle. In previous works [1–3, 15] we also introduced
.
This
was the width of a particle. But now we consider particles
shapeless, and we will
not use this concept. For better perception, all these values are marked in
Figure 1.
Figure 1: The main parameters used by the
discontinuous particle method.
Let us assume that
is
the area of the trapezium, whose bases are the heights of i-1 and i particles
and whose sides are the distances between the particles, is constant.
.
Let the centers of the particles satisfy the ODE
system (5). We solve the differential part of the system numerically using the
explicit Euler scheme:
.
Since the equation is nonlinear, the particles move
at different speeds, as a result of which the area of the trapezoid between the
particles can either increase or decrease. Taking advantage of the fact that
the area of the trapezoid must remain constant, we write:
.
Let's show this
visually in Figure 2:
Figure 2: Visualizing the condition of constant
trapezoidal area
However, we should keep in mind that by changing the
height of a particle to preserve the area of one trapezium we have implicitly
increased the area of the neighboring trapezium. To avoid the effect of such an
implicit increase, we introduce an additional restriction: if
i-1-th
particle is higher than
i-th, the area of the trapezium must decrease.
And if it increases, the particles should not interact. If
i-1-th
particle is lower than
i-th, the area of the trapezium must increase. And
if the area decreases, the particles also do not interact. We can rewrite the
condition of interaction of particles as follows:
.
Next, we proceed to a description of the algorithm
in the two-dimensional case. Using the aiming parameter from [3], we find a
pair of interacting particles. As shown in Figure 3, we construct a trapezoid
and pass from the solution of the two-dimensional problem to the one-dimensional
one, which has already been described earlier.
Figure 3: Converting two-dimensional problem to a
one-dimensional
It is worth noting that in the two-dimensional case
the comparison with the trapezoidal area in real time is a certain difficulty.
One can either store all data from the first time step and calculate old areas
each time, or calculate all possible pairs of areas in advance, store them in a
two-dimensional array, and then retrieve them as needed. Detailed comparative
calculations have not been carried out, but it can be assumed that for a large
number of particles with a small number of calculation time steps, the first
option is advantageous. In the opposite case, it is preferable to use the
second option.
This section shows the essence of the discontinuous
particle method algorithm without explicitly using the particle form. This is
how the initialization of variables and their subsequent changes are carried out
in the program.
First, we solve the Cauchy problem for the
one-dimensional quasi-linear transport equation (5) with initial conditions:
|
(9)
|
There is an exact solution for such an initial
condition:
Figure 4: Solution of the one-dimensional quasi-linear
transport equation by the particle method.
Figure 4 shows animated visualization of the
numerical solution of equation (5) with initial conditions (9). The red line
shows the exact solution. It can be seen that the shock velocity is calculated
correctly, the shock is smeared on one particle. After the high particle with
higher velocity "collides" with the low particle with lower velocity,
the low particle grows to the height of the high particle. Thus, their speeds
are compared, the low particle stops growing. In its turn it “collides” with the
next particle.
Then we solve the Cauchy problem for the
one-dimensional quasi-linear transport equation (5) with initial conditions:
|
(10)
|
There is also an exact solution for such an initial
condition:
Figure 5: Solution of the one-dimensional
quasi-linear transport equation by the particle method.
Figure 5 shows animated visualization of the
numerical solution of equation (5) with initial conditions (10). The red line
shows the exact solution. It can be seen that the solution in the form of
rarefaction wave is calculated less accurately.
Then we numerically solve the Cauchy problem for the
two-dimensional quasilinear transport equation (7) with initial conditions:
Figure 6: Motion animation of the numerical solution
of the two-dimensional Cauchy problem by the particle method.
The discontinuity propagation in the two-dimensional
problem is represented in the form of animation (Fig. 6). For ease of
perception, the particles in Figure 6 are depicted as circles. It should be
understood that the shape of the particles is not used anywhere in the
algorithm itself. The exact solution is indicated as a thick black line. You
can see in Figure 6 that particles with higher velocity collide with particles
with lower velocity. Those taper, and a one-particle wide gap is formed. This
shows the low approximation viscosity of our method. The particles then
continue to collide with each other, so that the gap moves at the velocity
corresponding to the analytical solution. A visual representation of the moving
process of the shock in the computational domain allows us to evaluate the
properties of the used numerical method.
Thus, we have demonstrated the possibility of a new
version of the discontinuous particle method that allows us to numerically
simulate a system of two-dimensional quasi-linear transport equations with high
accuracy, which is especially pronounced in examples with discontinuous
solutions. The visualization of the solution allows us to see the advantages
and disadvantages of the method more clearly.
1.
Bogomolov S.V., Esikova N.B., Kuvshinnikov A.E.
Micro-macro Fokker–Planck–Kolmogorov models for a gas of rigid spheres // Math.
Models
Comput
.
Simul., Vol. 8, ¹5, 2016, pp. 533–547 (doi:10.1134/S2070048216050069)
2.
Bogomolov S.V., Kuvshinnikov A.E.
Discontinuous particle method on gas dynamic examples // Math. Models Comput. Simul.,
Vol. 11, ¹5, 2019, pp. 768–777 (doi:10.1134/S2070048219050053)
3.
Bogomolov S.V., Filippova M.A.,
Kuvshinnikov A.E. A discontinuous particle method for the inviscid Burgers’
equation // Journal of Physics: Conference Series, Vol. 1715, 2021, 012066 (10.1088/1742-6596/1715/1/012066)
4.
Bondarev A. E., Galaktionov V.A.
Generalized Computational Experiment and Visual Analysis of Multidimensional
Data // Scientific Visualization, Vol. 11, ¹4, 2019, pp. 102–114 (doi:10.26583/sv.11.4.09)
5.
Alekseev A., Bondarev A., Galaktionov V., Kuvshinnikov
A., Shapiro L. On Applying of Generalized Computational Experiment to Numerical
Methods Verification // CEUR Workshop Proceedings, Vol. 2744, 2020 (doi:10.51130/graphicon-2020-2-3-19)
6.
Burgers J. M. The nonlinear diffusion
equation. Asymptotic solutions and statistical problems, D.Reidel Publishing
Company, Dordrecht-Holland / Boston USA, 1974
7.
Amangaliyeva M. M., Jenaliyev M. T., Kosmakova
M. T., Ramazanov M. I. On the solvability of nonhomogeneous boundary
valueproblem for the Burgers equation in the angular domain and related integral
equations, Springer Proceedings in Mathematics and Statistics, Vol. 216, 2017, pp.
123–141 (doi:10.1007/978-3-319-67053-9_12)
8.
Vahid Rezapour Jaghargh, Mahdavi A., Roohi
E. Shear-Driven Micro/Nano Flows Simulation using Fokker Planck Approach:
Investigating Accuracy and Efficiency // Vacuum, Vol. 172, 2020, 109065 (doi:10.1016/j.vacuum.2019.109065)
9.
Gorjia M. H., Torrilhon M. Entropic
Fokker-Planck kinetic model // Journal of Computational Physics, Vol. 430,
2021, 110034 (doi:10.1016/j.jcp.2020.110034)
10.
Hockney R. W., Eastwood J. W. Computer
simulation using particles, New York: Taylor&Francis, 1988
11.
Liu G. R., Liu M. B. Smoothed Particle
Hydrodynamics: A Meshfree Particle Method, World Scientific Publishing, 2003
12.
Alban de Vaucorbeil, Vinh Phu Nguyen, Sina Sinaie, Jian
Ying Wue Material point method after 25 years: theory, implementation and
applications // Advances in Applied Mechanics, Vol. 53, 2020, pp. 195–398 (doi:10.1016/bs.aams.2019.11.001)
13.
Harlow F. H.
The Particle-in-Cell Computing Method for Fluid
Dynamics //
Methods in
Computational Physics, Vol. 3,
ed
B Alder, S Fernbach and M Rotenberg, New York:
Academic Press,
1964,
pp. 319–343
14.
Li S., Liu W. K. Meshfree and particle
methods and their applications // Appl. Mech. Rev., Vol. 55,
¹1,
2002, pp. 1–34 (doi:
10.1115/1.1431547)
15.
Bayev A. Zh., Bogomolov S.V. On the
Stability of the Discontinuous Particle Method for the Transfer Equation //
Math.
Models Comput. Simul., Vol. 10, ¹2, 2018, pp. 186–197
(doi:10.1134/S2070048218020023)