From 93dfae5bf8f69ff5ffb9a9a7093dd90c1aa68833 Mon Sep 17 00:00:00 2001 From: Quentin Wach Date: Sat, 29 Jun 2024 00:24:45 +0200 Subject: [PATCH] Update. --- src/Eulerian Fluid in 2D.md | 14 +- .../Eulerian_Fluid_in_2D/eulerian_fluid.py | 249 ++++++++++++++++++ 2 files changed, 258 insertions(+), 5 deletions(-) diff --git a/src/Eulerian Fluid in 2D.md b/src/Eulerian Fluid in 2D.md index 7bb17b3..a81ebdf 100644 --- a/src/Eulerian Fluid in 2D.md +++ b/src/Eulerian Fluid in 2D.md @@ -1,6 +1,6 @@ # Eulerian Fluid Simulator -We will look at a 2D simulation here first, mainly following the work of Matthias Müller [^1]. Though moving to 3D is quite trivial. It is _Eulerian_ because we use a grid rather than points for the computations. +We will look at a 2D simulation here first. Moving to 3D is quite trivial. It is _Eulerian_ because we use a grid rather than points for the computations. Below, we will be mainly following the work of Matthias Müller [^1] [^2] with added details and insights from various other sources. We assume that: 1. Water is an [incompressable fluid](https://en.wikipedia.org/wiki/Incompressible_flow). @@ -47,7 +47,7 @@ First, we compute the divergence. We can then handle obstacles or walls by fixing those velocity vectors. So for static object, that point of the border would be zero. But if it is moving this will of course impact the velocity and we can simulate how the fluid is being pushed around! ### General Case -We define the scalar value \\(s^{i,j}\\) for each cell, where objects are zero and fluids 1. We update it as +It is useful to define the scalar value \\(s^{i,j}\\) for each cell, where objects are zero and fluids 1. We update it as $$ s \leftarrow s^{i+1. j} + s^{i-1, j} + s^{i,j+1} + s^{i,j-1} @@ -62,7 +62,7 @@ v_y^{i,j} \leftarrow v_y^{i,j} + d \cdot s^{i,j+1}/s \\\\ v_y^{i,j+1} \leftarrow v_y^{i,j+1} + d \cdot s^{i,j+1}/s. $$ ->**Question**: What is \\(s\\) ? +Why? ### Solving the Grid Naturally, we want to solve the whole grid. One, and possibly the simplest method here is to use the [Gauss-Seidel method](https://en.wikipedia.org/wiki/Gauss%E2%80%93Seidel_method): @@ -130,7 +130,7 @@ $$ And to calculate the velocity at any arbitrary position within the grid, hence, to interpolate, we can calculate a **weighted average** of the surrounding values $$ - \bar{v} = w^{00} \cdot w^{10} \cdot v^{i,j} + w^{01} \cdot w^{10} \cdot v^{i+1,j} + w^{10} \cdot w^{11} \cdot v^{i,j+1} + w^{11} \cdot w^{11} \cdot v^{i+1,j+1} + \bar{v} = w^{00} \cdot w^{10} \cdot v^{i,j} + w^{01} \cdot w^{10} \cdot v^{i+1,j}\\ \qquad \qquad+ w^{10} \cdot w^{11} \cdot v^{i,j+1} + w^{11} \cdot w^{11} \cdot v^{i+1,j+1} $$ where \\(w^{00} = 1 - x/h \\), \\(w^{01} = x/h \\), \\(w^{10} = 1 - y/h \\), and \\(w^{11} = 1 - y/h \\) for the height \\(y\\) within the points and the width \\(x\\) within the points. @@ -140,5 +140,9 @@ We can store the density value at the center of each cell and _"advect"_ it just Calculating streamlines is rather straightforward as well. We start at the position \\(x_1\\) and step forward with the step-size \\(s\\) to get \\(x_2 \leftarrow s \cdot v(x_1) \\) and so on for \\(n\\) steps, each time sampling the velocity \\(v\\) at the corresponding position. +--- ## References -[^1]: [Matthias Müller, _"How to write an Eulerian fluid simulator with 200 lines of code.", Ten Minute Physics, 2022](https://matthias-research.github.io/pages/tenMinutePhysics/17-fluidSim.pdf) +[^1]: [Matthias Müller, _"How to write an Eulerian fluid simulator with 200 lines of code."_, YouTube: Ten Minute Physics, 2022](https://matthias-research.github.io/pages/tenMinutePhysics/17-fluidSim.pdf) +[^2]: [Matthias Müller, _"Notes On Eulerian Fluid Simulations."_, GitHub (PDF), 2022 ](https://matthias-research.github.io/pages/tenMinutePhysics/17-fluidSim.pdf) +[^3]: https://gist.github.com/vassvik/f06a453c18eae03a9ad4dc8cc011d2dc +[^4]: https://jamie-wong.com/2016/08/05/webgl-fluid-simulation/ diff --git a/src/code/Eulerian_Fluid_in_2D/eulerian_fluid.py b/src/code/Eulerian_Fluid_in_2D/eulerian_fluid.py index e69de29..a84f69f 100644 --- a/src/code/Eulerian_Fluid_in_2D/eulerian_fluid.py +++ b/src/code/Eulerian_Fluid_in_2D/eulerian_fluid.py @@ -0,0 +1,249 @@ +""" +Eulerian Fluid in 2D. +author: @QuentinWach +date: June 28, 2024 +""" +import numpy as np +import matplotlib.pyplot as plt + +U_FIELD = 0 +V_FIELD = 1 +S_FIELD = 2 + +class Fluid: + def __init__(self, density, numX, numY, h): + self.density = density + self.numX = numX + 2 + self.numY = numY + 2 + self.numCells = self.numX * self.numY + self.h = h + self.u = np.zeros(self.numCells, dtype=np.float32) + self.v = np.zeros(self.numCells, dtype=np.float32) + self.newU = np.zeros(self.numCells, dtype=np.float32) + self.newV = np.zeros(self.numCells, dtype=np.float32) + self.p = np.zeros(self.numCells, dtype=np.float32) + self.s = np.zeros(self.numCells, dtype=np.float32) + self.m = np.ones(self.numCells, dtype=np.float32) + self.newM = np.zeros(self.numCells, dtype=np.float32) + + def integrate(self, dt, gravity): + """ + For every point i,j in the grid, we update the velocity field + by adding the gravity force to the y-component of the velocity + making sure to differentiate between fluid and solid cells. + """ + n = self.numY + for i in range(1, self.numX): + for j in range(1, self.numY - 1): + if self.s[i * n + j] != 0.0 and self.s[i * n + j - 1] != 0.0: + self.v[i * n + j] += gravity * dt + + def solveIncompressibility(self, numIters, dt, overRelaxation): + """ + Solve the incompressibility problem by solving the Poisson equation + + """ + n = self.numY + cp = self.density * self.h / dt + + for iter in range(numIters): + for i in range(1, self.numX - 1): + for j in range(1, self.numY - 1): + if self.s[i * n + j] == 0.0: + continue + + s = self.s[i * n + j] + sx0 = self.s[(i - 1) * n + j] + sx1 = self.s[(i + 1) * n + j] + sy0 = self.s[i * n + j - 1] + sy1 = self.s[i * n + j + 1] + s = sx0 + sx1 + sy0 + sy1 + if s == 0.0: + continue + + div = self.u[(i + 1) * n + j] - self.u[i * n + j] + self.v[i * n + j + 1] - self.v[i * n + j] + p = -div / s + p *= overRelaxation + self.p[i * n + j] += cp * p + + self.u[i * n + j] -= sx0 * p + self.u[(i + 1) * n + j] += sx1 * p + self.v[i * n + j] -= sy0 * p + self.v[i * n + j + 1] += sy1 * p + + def extrapolate(self): + n = self.numY + for i in range(self.numX): + self.u[i * n] = self.u[i * n + 1] + self.u[i * n + self.numY - 1] = self.u[i * n + self.numY - 2] + for j in range(self.numY): + self.v[j] = self.v[n + j] + self.v[(self.numX - 1) * n + j] = self.v[(self.numX - 2) * n + j] + + def sampleField(self, x, y, field): + n = self.numY + h = self.h + h1 = 1.0 / h + h2 = 0.5 * h + + x = max(min(x, self.numX * h), h) + y = max(min(y, self.numY * h), h) + + dx = dy = 0.0 + + if field == U_FIELD: + f = self.u + dy = h2 + elif field == V_FIELD: + f = self.v + dx = h2 + elif field == S_FIELD: + f = self.m + dx = dy = h2 + + x0 = min(int((x - dx) * h1), self.numX - 1) + tx = ((x - dx) - x0 * h) * h1 + x1 = min(x0 + 1, self.numX - 1) + + y0 = min(int((y - dy) * h1), self.numY - 1) + ty = ((y - dy) - y0 * h) * h1 + y1 = min(y0 + 1, self.numY - 1) + + sx = 1.0 - tx + sy = 1.0 - ty + + val = (sx * sy * f[x0 * n + y0] + + tx * sy * f[x1 * n + y0] + + tx * ty * f[x1 * n + y1] + + sx * ty * f[x0 * n + y1]) + + return val + + def avgU(self, i, j): + n = self.numY + u = (self.u[i * n + j - 1] + self.u[i * n + j] + self.u[(i + 1) * n + j - 1] + self.u[(i + 1) * n + j]) * 0.25 + return u + + def avgV(self, i, j): + n = self.numY + v = (self.v[(i - 1) * n + j] + self.v[i * n + j] + self.v[(i - 1) * n + j + 1] + self.v[i * n + j + 1]) * 0.25 + return v + + def advectVel(self, dt): + self.newU[:] = self.u + self.newV[:] = self.v + + n = self.numY + h = self.h + h2 = 0.5 * h + + for i in range(1, self.numX): + for j in range(1, self.numY): + if self.s[i * n + j] != 0.0 and self.s[(i - 1) * n + j] != 0.0 and j < self.numY - 1: + x = i * h + y = j * h + h2 + u = self.u[i * n + j] + v = self.avgV(i, j) + x -= dt * u + y -= dt * v + u = self.sampleField(x, y, U_FIELD) + self.newU[i * n + j] = u + if self.s[i * n + j] != 0.0 and self.s[i * n + j - 1] != 0.0 and i < self.numX - 1: + x = i * h + h2 + y = j * h + u = self.avgU(i, j) + v = self.v[i * n + j] + x -= dt * u + y -= dt * v + v = self.sampleField(x, y, V_FIELD) + self.newV[i * n + j] = v + + self.u[:] = self.newU + self.v[:] = self.newV + + def advectSmoke(self, dt): + self.newM[:] = self.m + + n = self.numY + h = self.h + h2 = 0.5 * h + + for i in range(1, self.numX - 1): + for j in range(1, self.numY - 1): + if self.s[i * n + j] != 0.0: + u = (self.u[i * n + j] + self.u[(i + 1) * n + j]) * 0.5 + v = (self.v[i * n + j] + self.v[i * n + j + 1]) * 0.5 + x = i * h + h2 - dt * u + y = j * h + h2 - dt * v + + self.newM[i * n + j] = self.sampleField(x, y, S_FIELD) + + self.m[:] = self.newM +""" +# Example usage: +if __name__ == "__main__": + density = 1.0 + numX, numY = 50, 50 + h = 1.0 + fluid = Fluid(density, numX, numY, h) + + # Simulation parameters + dt = 0.1 + gravity = -9.8 + numIters = 10 + overRelaxation = 1.9 + + # Example simulation step + fluid.integrate(dt, gravity) + fluid.solveIncompressibility(numIters, dt, overRelaxation) + fluid.extrapolate() + fluid.advectVel(dt) + fluid.advectSmoke(dt) + + # Visualization + plt.imshow(fluid.m.reshape(fluid.numX, fluid.numY), cmap='gray') + plt.colorbar() + plt.title("Smoke Density") + plt.show() +""" + +def initialize_smoke(fluid): + # Initialize some smoke in the center of the grid + centerX, centerY = fluid.numX // 2, fluid.numY // 2 + for i in range(centerX - 5, centerX + 5): + for j in range(centerY - 5, centerY + 5): + fluid.m[i * fluid.numY + j] = 10.0 # Higher density of smoke in the center + +# Example usage: +if __name__ == "__main__": + density = 1.0 + numX, numY = 50, 50 + h = 1.0 + fluid = Fluid(density, numX, numY, h) + + # Initialize smoke + initialize_smoke(fluid) + + dt = 0.1 + gravity = -9.8 + numIters = 10 + overRelaxation = 1.9 + + # Simulate one step + fluid.integrate(dt, gravity) + fluid.solveIncompressibility(numIters, dt, overRelaxation) + fluid.advectVel(dt) + fluid.advectSmoke(dt) + + # Visualize results + smoke_density = fluid.m.reshape((fluid.numX, fluid.numY)) + + plt.imshow(smoke_density.T, origin='lower', cmap='gray') + plt.colorbar(label="Smoke Density") + plt.quiver(np.arange(fluid.numX), np.arange(fluid.numY), + fluid.u.reshape((fluid.numX, fluid.numY)).T, + fluid.v.reshape((fluid.numX, fluid.numY)).T, color='r') + plt.title("Smoke Density and Velocity Field") + plt.xlabel("X") + plt.ylabel("Y") + plt.show() \ No newline at end of file