Can You Use Generative Design for Internal Fluid Flow?

UPDATED June 25, 2021. Added link to a detailed generative design manifold recently made available.

With generative design, the computer can finally become an active ally in the design of a product and be considered the first true computer aided design. Note the emphasis on “aided.”

“Rather than asking if this shape meets the requirements, we are asking which shape best meets the requirements,” said Keith Meintjes of CIMdata in a 2017 blog post.

Looking critically at what has passed as computer aided design, we realize that its middle name is something of a misnomer. CAD replaced our drafting tables, but like drafting tables, it only helped with the documentation of a design. The real design—the transformation of an idea into a shape—has already occurred in our heads. The drafting board helped the designer lay out the design on vellum. CAD helped to define the design more neatly, more precisely and in three dimensions. Little design, if any, was taking place on either medium.

And so, to the present day, design continues as a mental exercise, perhaps helped by sketching on the side, using whatever is handy—a cocktail napkin, a tablecloth (why restaurants don’t take reservations for groups of engineers), the walls (why the modern offices have erasable wall surfaces or whiteboards), or, if the engineer is unabashedly cool, an iPad or its equivalent.

Until now, generative design has primarily been demonstrated for structural optimization, where a part is usually optimized for mass while subject to loads and restraints based on the limits of the strength of its material. Centuries of study, testing and forensic examination, as well as a generation of simulation, has left engineers with a fairly good intuition of the mechanical behavior of parts. An experienced mechanical engineer has a good chance of determining a strong-enough, or too-strong, structural part by considering how forces pass through it, maximizing material along those paths and minimizing material elsewhere.

An intuitive grasp of fluid flow, however, is further behind, and faulty fluid flow is something we continue to propagate. We excuse ourselves because fluid flow is invisible, dynamic. We can’t see it, so how can we correct it? Examples abound. A cargo container is dragged down the highway behind a truck. The underside of a car is a clutter of cavities and corners, pipes and brackets. A bus is as aerodynamic as a tool shed. The air ducts in our homes and offices turn corners, as do the water pipes. If we could see the buffeting, the drag, the eddies, the recirculation, the streamlines that break up into turbulent flow, then, surely we would not let any of those things happen. 

Without aerodynamic scrutiny, fluid flow is subject only to other, admittedly practical considerations of commonly available stock material or manufacturing operations: containers are rectangular to accommodate efficient stacking and packing of their contents. The corrugated sheet metal used for shipping containers is necessary for strength despite the havoc it plays on the airstream. Fluid flows around shapes constructed from flat stock that is joined to form corners cause separation in external flow and in internal flow (tubes, pipes and ducts), eddies and recirculation.

In the world of fluid flow, external airflow is the poster child of computational fluid dynamics (CFD) programs. Our walls are adorned with multicolored streamlines over F1 racing cars and our bookshelves with pictures of airfoil’s angle of attack.

Still, external fluid flow simulation and testing is reserved for the most glamorous and expensive products—our airplanes, rockets and bullet trains. They are the ones most often depicted with visualized flow in CFD programs or in wind tunnels using smoke trails or fluttering ribbons on surfaces. Lesser products that could benefit from flow testing or simulation can’t afford it. A wind tunnel can cost tens of millions of dollars to build and hundreds of dollars an hour to rent. CFD, thought to be a savior of fluid flow and a less expensive and more practical alternative to wind tunnels, remains in the hands of expert practitioners, still not mainstream in availability or ease of use for the typical product engineer.

Less glamorous than external fluid flow but far more common is internal fluid flow, the flow inside vehicle and aircraft cabins, gas and liquid manifolds, automatic transmissions, exhaust systems, heat exchangers, gas turbines, rocket engineers, air ducts, pipes, dishwashers, pumps and compressors, to name a few. But because it is inside—even more hidden from view—internal fluid flow is even less likely to be considered. Out of sight, out of mind.

Let us look at what generative design could do for fluid flow, at least in theory.

Generative design for fluids, when integrated with other design and manufacturing applications in a product development platform, would be intended for use by product engineers. It would not require and not be limited to skilled analysts and specialists like stand-alone optimization applications. Therefore, when used in a design context, an integrated generative design application would be in a good position to prevent flow problems before they are baked into a design. Similar to structural generative design, generative design for fluids early in the design phase would allows for more design exploration, or more attempts of getting it right.

The traditional linear approach to the product development cycle has simulation follow design. If simulation rejects the design, the design is modified. If a product passes simulation, it gets manufactured. Allowing generative design for fluid flow to operate in the design phase is asking “what if?” over and over again, as in “what if we route the flow this way?” If that doesn’t improve flow, generative design will try another way, with an incremental change in the flow path. Structural generative design will  make changes rapidly and without pause, over and over again, stopping only when performance criteria is met, and then only to try a different tack to reach an alternate solution. It will find as many solutions as you will give it time for. You will be left with many solutions—all meeting design and performance goals you set. You can flip through them like a picture book and select one that please you the most, knowing all of them will work and that one or more could be far superior to its traditionally designed forbearer. Can generative design for fluid flow do that?

The modern age of engineering, called the Fourth Industrial Revolution by the World Economic Forum and Industry 4.0 by others, includes the generative design and 3D printing, both of which make the potential to make a dream come true for both fluid flow specialist and design engineers. Whereas the design engineer could have dismissed curvy fluid paths as dreamy and impossible to manufacture, along come 3D printers that say, “We can do that.” Indeed, a curvy shape can be made just as easily with 3D printing as can a shape with the sharp corners or bends of subtractive manufacturing or conventional construction. The marriage of the two—generative design for imagining the optimum shape and 3D printing for manufacturing it—should open up a brave new world for improving fluid flow, elevating fluid flow to its proper place at the initial phase of product development, the place in which 80 percent of total product cost is determined, rather than be relegated to last place, backwardly driven by methods of manufacture convenient to the machinist.

Take the designs of hydraulic manifolds. To withstand high pressures, they are made of metal. The easiest material to start with is a rectangular block cut to the right length. The most convenient method to make fluid paths are holes drilled from one face that meet with holes drilled from another face. The holes are of different size to create different hydraulic forces and mass flows from the output port. Drilling holes produces straight holes, of course. When holes intersect at right angles, as they would when the outlet port is on an adjacent face, the pressure drop is severe. There is also pressure drop when the holes abruptly change diameter. In both cases, smooth, linear flow gives way to recirculation, eddies, turbulence—all energy-robbing factors of pressure drop.

Fluid manifolds designed as those above are created with total regard to convenient methods of manufacture and with no regard to optimum fluid flow, which would have no sharp bends and gradually changing diameters. But a design engineers can only dream of optimum fluid flow. The real world has only the fluid paths the machinist has created. The design engineer is indeed taking a back seat to the machinist in the design of the part.