0

  I wanted to simulate the motion of a fluid (continuously filling the entire space) in a given space, say $3$-dimensional Euclidean space. To calculate the dynamics of fluid motion, I used the Navier-Stokes equation and the mass continuity equation.

  Let $\vec{v}$ be a vector of $\mathbb{R}^3$. And let $$\frac{D}{Dt} = \frac{\partial}{\partial t} + \vec{u} \cdot \nabla$$ be the material derivative. Also, assume the following conditions for the fluid.

  • The fluid is compressible.
  • The fluid is viscous.
  • The fluid is isothermal.
  • Dynamic viscosity $\mu$ is a constant in the whole spacetime.

Then, the Navier-Stokes equation is given as $$\rho \frac{D \vec{u}}{Dt} = -\nabla p + \mu \nabla^2 \vec{u} + \frac{1}{3} \mu \nabla (\nabla \cdot \vec{u}) + \rho \vec{g}.$$ And the mass continuity equation is given as $$\frac{\partial \rho}{\partial t} + \nabla \cdot (\rho \vec{u}) = 0.$$   Here, we can categorize existing variables. The independent variables are obviously time $t$ and position $\vec{x}$, and the dependent variables are:

  • unknown variables: $\rho (t, \vec{x})$, $\vec{u} (t, \vec{x})$, $p(t, \vec{x})$
  • given variables: $\vec{g} (t, \vec{x})$, $\mu$

In other words, we have $5$ unknown dynamic fields to be simulated: $\rho$, $u_x$, $u_y$, $u_z$, and $p$. However,

  The Navier-Stokes equation gives $3$ time-dependent equations for the time evolution of $\vec{u}$;
  The mass continuity equation gives $1$ time-dependent equation for the time evolution of $\rho$.

That is, there is a lack of $1$ time-dependent equation for the time evolution of $p$. (Given the initial conditions, the future shape of the system must be determined, so there must be an equation relating $p$ or its time derivatives to $\vec{u}$, $\rho$, and known quantities.) Hence, my question is,

  To time-dependently simulate (compressible, viscous, isothermal) fluids, how can we calculate the pressure $p$ at each point in spacetime? What is the relation between $p$ and $\vec{u}$, $\rho$, or the time evolution equation?

  Any help will be appreciated.

Modular Discriminant
  • 51
  • Is this for educational purposes? Is this your first fluid simulation? You might consider starting with a 2D incompressible simulation to get familiar. Jos Stam's paper http://graphics.cs.cmu.edu/nsp/course/15-464/Fall09/papers/StamFluidforGames.pdf has long been a nice introduction for people starting with fluid dynamics – AccidentalTaylorExpansion Aug 03 '23 at 18:54
  • @AccidentalTaylorExpansion This simulation isn't for educational purposes, I just want to make it because I saw the simulation videos and wanna try the simulation myself. In fact, I failed in simulating a compressible fluid because of this problem, so I tried simulating a 2D incompressible fluid. I could use the vorticity-stream function formulation, which has the advantage of automatically considering the pressure, and the simulation actually succeeded. (Although it has many flaws, I actually observed Kármán vortex street.) But what I need is a (2D) compressible fluid simulation. – Modular Discriminant Aug 03 '23 at 19:47
  • @AccidentalTaylorExpansion Also, thank you for giving me the paper! I will read it carefully. – Modular Discriminant Aug 03 '23 at 19:47
  • This question here is basically the same thing, though OP in that linked question is specifically concerned with Euler hydro equations. The answer is still the same: you need an equation of state. – Kyle Kanos Aug 03 '23 at 20:32
  • @KyleKanos Wow, thanks for the interesting answer! I saw the equation $\frac{\partial p}{\partial t} + \vec{u} \cdot \nabla p + \gamma p \nabla \cdot \vec{u} = S$ for the first time. Umm, I guess I'll have to think about this a bit more... Thank you! – Modular Discriminant Aug 03 '23 at 22:25

1 Answers1

0

As you have figured out, you will need an extra equation relating the variables $p$, $\rho$ and $\mathbf{u}$. For compressible fluids, usually one takes a state equation of the form $\rho(p)$ (as $T$ is held constant in your system). See for example the models here for real gases and here for water in isentropic flows.

Woe
  • 388
  • Thank you for answer! Perhaps I don't need a real gas, so the Stiffened equation of state $p = \rho (\gamma - 1) e - p^*$ in http://www.sklogwiki.org/SklogWiki/index.php/Stiffened_equation_of_state or the Cole equation of state $p = B ((\frac{\rho}{\rho_0})^\gamma - 1)$ would be sufficient. However, I have questions about the formulas above: – Modular Discriminant Aug 03 '23 at 20:47
  • (1) The relationship between $p$ and $\rho$ in the above formula is obtained not from pure classical fluid dynamics, which assumes that the fluid is a continuum, like in the NS equation or the mass continuity equation; but from thermodynamics, which considers the fluid as a set of molecules. And depending on the situation, I will have to use different formulas. Is this true..? This looks a bit strange to me. (2) Just for checking, is $p$ in the above formulas mechanical pressure $p_{mech}$? I referenced https://physics.stackexchange.com/a/427067/363082. – Modular Discriminant Aug 03 '23 at 20:47
  • (3) In the above formulas, $p$ depends only on $\rho$. Is it correct that $p_{mech}$ does not depend on $\vec{u}$? I looked at https://en.wikipedia.org/wiki/Navier%E2%80%93Stokes_equations#Pressure_recovery, and I didn't understand this formula, but thought that $\nabla p$ had something to do with (the derivative of) $\vec{u}$. Thank you! :) – Modular Discriminant Aug 03 '23 at 20:47
  • That's not really my area of expertise, but maybe I can help somehow. (1): I don't think different equations are needed as long as the continuous assumptions holds; (2): I would guess that, yes, $p$ is the pressure due to the mechanical entities, since other effects can be inserted into the body force term of the N-S equation; (3) The pressure dependence on $\mathbf{u}$ is included in the N-S equation itself. – Woe Aug 04 '23 at 00:21