Dead Simple Flow Simulation Gets New Optimization Features

One of the simplest flow simulation workflows has added an adjoint solver to perform optimizations.

Early in development cycles, engineers need quick and easy answers to make informed decisions. Otherwise, they risk discovering issues late in the design process—when it’s costly to fix.

To produce this simulation, engineers traditionally need an air-tight model; AirShaper doesn’t. (Image courtesy of AirShaper.)

This reality presents a particular challenge for those who can benefit from aerodynamic and hydrodynamic data. Computational fluid dynamics (CFD) simulations are traditionally challenging to produce without significant technical knowledge. They also require fluid tight models. At the early stages of design, engineers are unlikely to have these unless they spend weeks making them.
“The industry standard is to close everything so the mesher doesn’t complain. And if the mesher doesn’t complain, the solver won’t crash. That’s the basic workflow,” said Wouter Remmerie CEO of AirShaper. “In our case, we accept the holes. We allow the mesh to leak inside the model. So, just like in reality, you have a gap that is small enough that it doesn’t influence the aerodynamics on the outside, but it’s big enough for the [air] to leak inside. Which means that during the simulation you will have a static pocket of air that is standing inside the model.” 
This pocket of static air represents an extra computational penalty for the web-based AirShaper CFD tool. But by accepting this penalty, the tool can offer the engineering community one of the simplest flow simulation workflows available. On top of that, this flow simulator now offers access to optimization tools. 

How to Offer Simple Flow Simulations Early in Development

AirShaper is a web-based flow simulator that is powered by OpenFOAM technology. It’s designed to be dead simple to operate. However, it doesn’t compromise on the quality of the results. The target audience of this flow simulation technology are: 

• Design engineers who are not simulation experts
• Engineers who need answers early in the development process

The AirShaper user interface is designed for ease of use. (Image courtesy of AirShaper.)

“What you see on the interface are the only things you need to set up. We ask for things that you know as a designer, without having to have simulation knowledge,” said Remmerie. To that end, to start a simulation, you just have to:

1. Upload geometry
2. Define if the geometry is: on/above ground, static/moving, and in air/water
3. Set the velocity of the object (if it’s moving) or of the fluid (if the object is static)
4. Set the orientation
5. Define if there are any wheels
6. Set the units
7. Set the scale of the model

“We can work with open surface models and non-watertight geometry,” said Remmerie. “Which is kind of a big plus for the target audience we have. Because they’re usually not set up to close every possible gap in the model.”
Finally, the user defines the fidelity of the simulation. This selection will determine the price AirShaper charges for the computation. 

• Basic simulations have about 1 million cells and come with online results and OpenFOAM data. This level of fidelity is good for those looking to get an idea of the global airfield. 
• Regular simulations come with everything in the Basic package, except that the model utilizes 10 million cells. The user also receives a full simulation report. At this fidelity, engineers can get a good concept evaluation of the model. 
• Advanced simulations come with everything in the Regular package, except that the model is made up of about 1 hundred million cells. This level of fidelity is good for small organizations that need in-depth results without access to traditional CFD technology.

“When you are looking at the smaller customers that we have, they typically do not do aerodynamics themselves, either because it’s too expensive, too much of a hurdle to start doing, or they have to hire a consultant. We enable them to get into aerodynamics because we have a very simple pay-per-use model,” said Remmerie.

The Optimization Capabilities of AirShaper

An optimization tool, in the form of an adjoint solver, is now available on AirShaper. Remmerie explained, “Adjoint is a way of calculating all the gradients on the surface to know if we should push the surface inward, or outward, to get close to a certain goal.”
To pick an adjoint element, engineers need to click on whatever they want to optimize. It could be a part or a morphing space that is represented by a box. In this case, the mirror is selected. Like the definitions of the wheels during the simulation step, the software automatically recognizes the part.
 

The adjoint solver options in the AirShaper user interface. (Image courtesy of AirShaper.)

Engineers then define the goal of the optimization. It could be to maximize or minimize lift, drag, or lift over drag. Finally, they determine the number of cycles to perform. After these simple steps, the optimization is ready to run.

The adjoint solver optimizing the shape of a mirror to reduce drag. (Image courtesy of AirShaper.)
As seen in the optimization of the mirror, the adjoint solver decreased the cross section by reducing its thickness at the back end—creating a teardrop shape.
To optimize the front splitter, it’s better to utilize the morphing space method. A model was uploaded with a box that represents this space. The user clicks on the box, as they did previously with the mirror, and sets up the simulation objectives. 

The adjoint solver optimizing the shape of a front splitter to increase downforce. (Image courtesy of AirShaper.)
As seen in the optimization of the front splitter, the adjoint solver lowers the gap between the car and the ground. This inverted wing shape creates a venturi-like airflow that speeds up the flow under the car and increases the downforce.
“Not only does [the adjoint solver] predict stuff; it also inspires the designers because it does things that we don’t expect. But then you dig into the results and you see how it makes sense,” said Remmerie. “It helps you speed up the design cycle because you go from weeks of design cycles to a fully optimized sequence.” 

How AirShaper Fits into the Flow Simulation Ecosystem

AirShaper aims to take the complexity out of flow simulations and optimizations. The best example of this is its ability to utilize non-watertight geometry. The software pays a premium on computational resources, but its customers get an easy user experience that provides accurate results.

The Table of Contents from an automatically generated AirShaper simulation report. (Image courtesy of AirShaper.)

A similar conservative approach was used to set up the adjoint solver. “Normally in a professional CFD package, where the user is responsible, you have to tweak a lot of settings to get it to work in a stable way,” said Remmerie. “You can automate this, but it’s still your responsibility.” 
“In our case,” he added, “we pay extra for more conservative settings that are more robust. So, if you’re asking about relaxation factors, schemes, [custom algorithms] and correctors used during the simulation, they all make it more expensive to get to the same result. But if you don’t apply these robustness parameters, in some cases, the calculation will diverge and blow up. We can’t afford to have any simulation blow up. Which means that on all the simulations, we pay extra to make sure they are stable.”
Continuing this simplicity theme, users don’t choose their own turbulence model. K-omega SST is used for all simulations as it blends speed, accuracy and cost. The adjoint solver uses the Spalart-Allmaras turbulence model because it has fully derived adjoint solver turbulence equations. 
Clearly, AirShaper isn’t the tool you need when it comes to the nitty-gritty details in a product’s design. But it isn’t trying to replace those tools.
“I have been the engineer who had been trained with the software to tweak all the settings. That’s still very valuable,” said Remmerie. “Our message is we don’t want to replace that software. We’re actually very complementary. Imagine you are a train designer—those models are huge. Making a watertight geometry to do a first check on aerodynamics is a huge task; it will delay the whole project for weeks. You cannot afford that. What we offer is a tool you can use, early-stage design, to check and benchmark aerodynamic performance.”

To learn more about AirShaper, read Fresh Air: Flow Simulation Couldn’t Get Much Easier.