PowerFLOW® is built on our proprietary Digital Physics technology based on an extended implementation of the Lattice Boltzmann Method that we have developed over two decades. PowerFLOW differs from competing CFD technology in fundamental ways that make our simulations more useful to our customers.

• Inherently transient: simulates time-dependent phenomena such as turbulent flows;

• Numerically stable: reliable even when used to analyze complex geometries; and

• Highly accurate.

The traditional approach in CFD has been to start with the Navier-Stokes equations, which are a set of partial differential equations that describe the behavior of a fluid. These equations are theoretically sound for many types of flows but are very complex and highly non-linear. Because these equations by their nature cannot be directly solved for any but the simplest scenarios, their application in practice requires the use of numerical techniques to approximate a solution. The main drawbacks of the traditional CFD approach therefore lie not with the Navier-Stokes equations themselves but in the numerical techniques that must be employed to provide solutions to them.

In the traditional CFD approach, the continuous Navier-Stokes equations are discretized, meaning that the flow field is broken up into discrete cells located in three-dimensional space (analogous to the two-dimensional pixels on a computer screen), where the flow properties such as velocity and pressure are solved. Because solving for the fluid properties at all locations in time and space is not mathematically possible, values are computed at these discrete locations, which makes up what is referred to as the computational grid.

There are also significant difficulties with these numerical techniques when simulating flow conditions at the interface where the fluid grid meets a surface. The necessary calculations are computationally intensive and they often become unstable, meaning that no meaningful solution can be provided. In order to address these difficulties and to reduce the computational costs, many traditional CFD approaches use a steady-state solver that simplifies the problem by calculating an average value for each discrete cell, rather than predicting the changing values in time. To improve robustness and stability, traditional approaches also introduce excess numerical dissipation that, while improving stability, works to destroy subtle flow structures that are critical for accuracy.

In contrast, our Digital Physics technology, based on the Lattice Boltzmann Method, describes the fluid flow at the mesoscopic level, between the molecular and the continuum level of Navier-Stokes. As with traditional solvers, the lattice Boltzmann method also discretizes the flow domain. In spite of this discretization, it was proved in the early 1990’s that the lattice Boltzmann method accurately provides solutions equivalent to the highly non-linear Navier-Stokes equations without actually having to solve them. This theoretical proof was the genesis for the founding of Exa.

PowerFLOW provides a further advantage in that it can handle fully complex surface geometry, the details of which are equally important to generate accurate predictions. Another unique component of our core technology is the “discretizer,” which automatically determines the fluid/surface intersection without compromising the geometric fidelity. Other CFD tools offer automatic mesh generation but the limitations of their solvers require that the geometry be simplified, often to such an extent that the accuracy of the results is compromised.

Unlike traditional approaches, PowerFLOW’s extremely low numerical dissipation allows for accurate simulation of time dependent flows enabling sensitive applications such as aeroacoustics. Furthermore, PowerFLOW has been proven to be highly scalable on massively parallel computers to enable shorter turn-around times, a very challenging task with traditional CFD solvers.

For strongly turbulent flows it is not computationally practical to perform direct simulations by resolving all of the scales of motion – from the microscopic ones to large scale flow structures that are the size of the vehicle. Thus, it becomes necessary to incorporate models to account for these unresolved turbulent flow structures. The small turbulent scales are universal in nature and thus can be predicted with theoretical models. The large scales of turbulence are not universal and no model exists that can be relied upon to accurately predict their behavior.

The most common approach to this problem in CFD is to rely on turbulence modeling across the entire scales of turbulence. These models extend far beyond their theoretical basis and are often applied to steady-state solutions where the time dependent nature of the flow is completely ignored. Due to the non-universal nature of the large turbulent scales, this approach has led to a range of user-selected turbulence models with empirical tuning of the parameters in an attempt to be suitable for a specific class of flows. These compromises have enabled traditional CFD approaches to be tailored to specific applications and be deployed in a way that provides some level of insight and understanding of flow behaviors but limit their accuracy and thus their ability to replace the complex experimental testing required to ensure that designs meet their targets.

PowerFLOW leverages the fact that it is a transient solver with high three-dimensional resolution and low dissipation. This allows PowerFLOW to directly simulate the large unpredictable turbulent scales. PowerFLOW uses turbulence theory to model only where it is valid and directly simulates the rest.

These combined attributes of PowerFLOW’s underlying technology enable us to bring simulation solutions to new levels of accuracy and robustness that have not been possible before. This in turn allows our customers to improve their development processes from requiring physical prototype testing at every stage by substituting robust, detailed and accurate digital simulations. This allows the final physical prototype stage to be the one of confirmation rather than discovery.