Glossary: RealFlow’s Liquid Solver Types
RealFlow is a simulation software with a strong focus on liquids, but it also allows you to perform rigid and soft body simulations, and create open ocean surfaces. All these elements are also often called “solvers”. But, why is a simulation engine called “solver” and what is the difference between Dyverso, Hybrido, and standard particles?
What Is A Solver?
The simulation of natural phenomena like water, the behavior of bodies, or even light is a very complex process. Today we have very efficient mathematical methods to model these effects in a very realistic way. These methods have to be solved by a computer program, hence the name “solver”.
There are many solver types and you might have heard of, e.g. Sudoko, Rubik’s Cube, or slide puzzle solvers. And in RealFlow you will find solvers for fluids, granular substances, and various other materials. The big strength of RealFlow’s solvers is that they are coupled. This means that rigid bodies are able to float on top of ocean surfaces, particle or grid fluids move objects, or water washes away sand.
Newton Fluids & Other Approaches
Maybe you have already came across the term “Newton Fluid”? If yes, you might have asked yourself why certain liquids are called this way. Was Newton perhaps the first “inventor” of a fluid simulator? Well, not exactly, but his work on forces is still the fundament for many applications and Newton’s 2nd law is of particular importance:
Force = Mass • Acceleration
These force-based methods are very accurate and they are actually only valid for particles. Over the years they have been extended to rigid and soft bodies, and many other types of simulations (“Newton-Euler”). Later, these methods have been extended to formulate the famous Navier-Stokes equations to describe the dynamics of viscous fluids.
A different approach is based on the computation of position changes (Position Based Dynamics/PBD). This method is fast, stable, versatile, and highly controllable. PBD simulations are not as accurate as their force-based counterparts, but they are perfectly suited for VFX or interactive applications. With PBD methods it is not only possible to simulate all kinds of liquids, but even hair, wrinkled cloth, or objects with plasticity.
RealFlow’s Dyverso solver is a multi-physics simulation engine for liquids, and granular substances. Furthermore it is possible to mimic the behaviour of rigid and soft bodies. In the near future viscous and viscoelastic materials will be added. Dyverso is GPU-accelerated (OpenCL and CUDA).
- SPH is a rather old technology, but still valuable, because of its accuracy. Dyverso’s SPH implementation is highly optimized, but still requires a high number of substeps compared to other methods.
- PBD is a rather new solver type, extremely fast, and very robust. The idea behind this method is to connect the particles with more or less elastic “constraints”. This concept is the basis for many different materials, not just liquids. PBD simulations are stable – even with a very small number of substeps.
Dyverso: Fields of Applications
Dyverso liquids are mainly used for small- and mid-scaled projects, although the PBD method is also suited for large-scale simulations because of its speed. With SPH, on the other hand, it is possible to create rich fluids with fine structures. This is what you basically need for close-ups or applications where a very high level of detail is required.
Dyverso fluid simulation with tendrils.
Rigid and elastic materials can be used for typical rigid body simulations, and even deformable objects. A single Dyverso domain is able to control a huge amount of rigid/elastic nodes. A special application is the creation of thin, cloth-like objects, as shown below:
RealFlow’s other main technology for liquids is named Hybrido and here we are talking about a FLIP solver. FLIP stands for Fluid IMplicit Particles.
FLIP solvers are a hybrid technology between grids and particles and hence the name “Hybrido”. Like PBD methods, FLIP solvers do not need a high number of substeps. Low substeps makes these solvers very fast, but this can also be a problem in terms of fluid-object interaction. RealFlow provides a wide variety of solution to fix these interaction problems.
In FLIP solvers, the particles are used to carry the fluid’s velocity information. During the simulation, velocity is transferred to the grid, where the physics is done. This transfer can be seen as the core that keeps the simulation stable. The main advantage of the hybrid approach is, aside from speed, that it is possible to add secondary fluids like splashes, foam, or mist.
Hybrido: Fields of Application
Hybrido’s main usage is the simulation of huge water volumes, but in RealFlow you will also find a component that is well-suited for small-scale projects: the viscosity solver. With Hybrido you are able to create destroying floods, stormy seas, impressive rolling waves, and splashy waves on tropical beaches. The viscosity parameter lets you simulate everything from cream to tar or molten glass.
Hybrido wave with approx. 100,000,000 particles.
Standard Particles: Overview
The third main method is the SPH-based standard particle approach. As you have seen, Dyverso also uses the SPH technology, but in an improved form and with GPU support. But despite a huge difference in simulation speed, the standard particle method still has its relevance, because of its deep integration with RealFlow’s Python/C++ interfaces. Almost any aspect of a standard particle simulation can be influenced with Python and C++, and this makes this solver type very versatile.
Standard Particles: Fields of Application
Standard particles are used for small- and mid-scaled projects where a high level of detail or customization is needed. This fluid type can be seen as a legacy solver and we always recommend using Dyverso’s SPH or PBD solvers first, simply because of their speed. With Dyverso you are able to work with much higher particle counts in less time. A decent amount of particles does not only help to improve your particle simulation, but you will also get much better meshes.
Both Dyverso types even provide the same physical parameters as standard particles, but DyPBD and DySPH are much more stable (and faster) in critical simulations, for example high viscosity fluids.
RealFlow’s standard particles solver is no longer under development and can be seen as a “legacy” solver type.