Liquid Pour Simulator

The heart of SPH is estimating continuous field quantities (like density $\rho$) at any point in space by summing contributions from neighboring particles. This is done using a smoothing kernel function $W$.

Here, $\rho_i$ is the density at particle $i$, $m_j$ is the mass of a neighboring particle $j$, $\mathbf{r}$ are positions, and $h$ is the smoothing length that defines the radius of influence. In the simulator, this is why particles only affect those close to them.

Once density is known, pressure is typically calculated using a simple equation of state. The pressure force then drives the fluid motion, combined with viscosity and gravity.

$\mathbf{F}^{pressure}_i$ is the pressure force on particle $i$, and $P$ is pressure derived from density. $\nabla W$ is the gradient of the smoothing kernel. This force pushes particles from high-density to low-density areas. The Gravity parameter in the simulator adds a constant downward force $\mathbf{F}^{gravity}_i = m_i \mathbf{g}$ to this equation.

Automotive Engineering: SPH is extensively used in car crash tests to simulate fuel sloshing in tanks or coolant flow. Engineers use tools like LS-DYNA to ensure fluids don't cause secondary hazards during an impact, which is much cheaper and safer than physical testing.

Visual Effects (VFX): Nearly every large-scale water, lava, or magical fluid effect in movies is created using SPH-based solvers. The method's particle-based nature fits perfectly with rendering pipelines, allowing for realistic splashes and flows around complex shapes.

Civil & Hydraulic Engineering: SPH is used to model extreme fluid-structure interactions, such as dam breaks, wave impact on coastal structures, or debris flows. Its ability to handle large deformations and free surfaces makes it ideal for these chaotic events.

Manufacturing Process Simulation: Engineers simulate processes like liquid metal casting, concrete pouring, or even 3D printing with liquid resins. SPH helps predict potential defects like air bubbles or incomplete filling based on the fluid's viscosity and pour rate.

First, it's a major misconception that more particles always lead to a better simulation. While increasing particle count does improve resolution, the computation time grows at a rate close to the square of the particle number. In practice, identifying the minimum number of particles required to achieve the necessary accuracy is where your skill comes into play. For instance, coarse particles may suffice to observe the overall flow in a container, but capturing fine splashes or the formation of tiny droplets due to surface tension requires techniques like locally refining the particle resolution.

Next, beware of parameter setting pitfalls. Sliders for "viscosity" or "surface tension" in simulators are often dimensionless relative values. This means you must consider scaling (similarity laws) to directly correlate them with values for real-world fluids. For example, simply increasing the viscosity slider using settings that replicate water to mimic more viscous glycerin will produce results deviating from actual behavior. In practical work, a "validation" process is essential, where you calibrate parameters by comparing simulation results against theoretical/experimental values for known simple cases (e.g., pipe flow).

Finally, don't overlook the importance of initial conditions. Whether you start injection with particles arranged in a uniform grid or placed randomly can lead to completely different initial instabilities, significantly affecting the results. Also, setting an injection speed unrealistically high can generate non-physical pressures or splashing, causing the calculation to diverge. The first step is to carefully observe the real-world process and set an initial state that matches it.