by Combustion engines, propellers, and hydraulic pumps are examples of fluidic devices: instruments that use fluids to perform certain functions, such as generating power or transporting water.
Because fluidic devices are so complex, they are typically developed by experienced engineers who manually design, prototype, and test each device through an iterative process that is expensive, time-consuming, and labor-intensive. But with a new system, the user only needs to specify the locations and velocities at which fluid enters and leaves the device – the computational pipeline automatically generates an optimal design that achieves those goals.
The system could make it faster and cheaper to design fluid devices for all sorts of applications, such as microfluidics-on-a-chip labs that can diagnose diseases from just a few drops of blood, or artificial hearts that could save the lives of transplant patients.
Recently, computational tools have been developed to simplify the manual design process, but these techniques have had limitations. Some required a designer to specify the shape of the device in advance, while others represented shapes using 3D cubes, known as voxels, which result in inefficient, boxy designs.
The computational technique developed by researchers at MIT and elsewhere overcomes these pitfalls. Their layout optimization framework doesn’t require a user to make assumptions about what a device should look like. Also, the shape of the device automatically evolves during optimization with smooth limits, instead of blocky and inaccurate ones. This allows your system to create more complex shapes than other methods.
“Now you can do all these steps seamlessly in a computational pipeline. And with our system, you could build better devices because you can explore new designs that have never been investigated with manual methods. Maybe there are some ways that haven’t been investigated.” explored by experts yet,” says Yifei Li, an electrical engineering and computer science graduate student who is the lead author of a paper detailing this system.
Co-authors include Tao Du, a former postdoc at the Computer Science and Artificial Intelligence Laboratory (CSAIL) who is now an assistant professor at Tsinghua University; and lead author Wojciech Matusik, professor of electrical engineering and computer science, who leads the Computational Design and Manufacturing Group within CSAIL; as well as others at the University of Wisconsin at Madison, LightSpeed Studios, and Dartmouth College. The research will be presented at ACM SIGGRAPH Asia 2022.
Shaping a fluidic device
The researchers’ optimization pipeline begins with a blank three-dimensional region that has been partitioned into a grid of tiny cubes. Each of these 3D cubes, or voxels, can be used to form part of the shape of a fluidic device.
One thing that separates your system from other optimization methods is how it represents or “parameterizes” these tiny voxels. The voxels are parameterized as anisotropic materials, which are materials that give different responses depending on the direction in which force is applied to them. For example, wood is much weaker to forces applied perpendicular to the grain.
They use this anisotropic material model to parameterize voxels as fully solid (such as those on the outside of the device), fully liquid (the fluid inside the device), and voxels that exist at the solid-liquid interface, which have properties of both solid and liquid material.
“When you go in the solid direction, you want to model the material properties of solids. But when you go in the fluid direction, you want to model the behavior of fluids. This is what inspired us to use anisotropic materials to represent the solid-fluid interface. And it allows us to model the behavior of this region very accurately,” Li explains.
Your computational pipeline also thinks of voxels differently. Instead of using only voxels as 3D building blocks, the system can tilt the surface of each voxel and change its shape very precisely. The voxels can then be shaped into smooth curves that allow for intricate designs.
Once your system has formed a shape using voxels, it simulates how fluid flows through that design and compares it to user-defined targets. It then adjusts the design to best meet the objectives, repeating this pattern until the optimal shape is found.
With this design in hand, the user could use 3D printing technology to make the device.
Designs Demonstration
Once the researchers created this design pipeline, they tested it using state-of-the-art methods known as parametric optimization frameworks. These frameworks require designers to specify in advance what they think the shape of the device should be.
“Once you make that assumption, all you’ll get are variations of the shape within a family of shapes,” Li says. “But our framework doesn’t need you to make assumptions like that because we have a high degree of design freedom by representing this domain with many tiny voxels, each of which can vary in shape.”
In each test, their framework exceeded baselines by creating smooth shapes with intricate structures that would likely have been too complex for an expert to specify beforehand. For example, he automatically created a fluidic tree-shaped diffuser that transports liquid from a large inlet to 16 smaller outlets while clearing an obstacle in the middle of the device.
The pipe also spawned a propeller-shaped device to create a twisting flow of liquid. To achieve this complex shape, his system automatically optimized almost 4 million variables.
“I was very pleased to see that our pipeline was able to auto-grow a propeller-shaped device for this fluid tornado. That shape would power a high-performance device. If you’re modeling that target with a parametric-shaped framework, why can’t it grow from such a complex shape, the final device wouldn’t work as well,” he says.
While she was impressed by the variety of shapes it could automatically generate, Li plans to improve the system using a more complex fluid simulation model. This would allow the tubing to be used in more complex flow environments, allowing it to be used in more complicated applications.
This research was supported, in part, by the National Science Foundation and the Defense Advanced Research Projects Agency.
