By Mark Kacik
Edited by Paul Dvorak
CFD or computational-fluid dynamics was once the realm solely of PhDs and aerospace companies. But software developers have done a good job simplifying it so CFD programs can be used by a wider range of engineers and for a variety of tasks. At our company, for example, we use CFD to improve fixtures such as showerheads and valves, and faucets by predicting spray and flow patterns, reducing them in size, increasing their efficiency, and letting them do more. The software lets us pay close attention to fluid dynamics early in a product's development.
CFD software lets engineers refine designs through several iterations in as little as a day. Features such as interactivity with MCAD, easier-to-use meshing algorithms, and user-friendly GUI's have placed the analysis software into a daily-use toolbox for design engineers. Understanding how it differs from the structural analysis of finite elements and knowing what to look for in charts and images lets design teams get more out of the technology.
WHERE CFD WORKS WELL
Optimizing a product is an iterative process involving design, prototyping, testing, refining the design, and reentering the loop. CFD decreases development time and effort by substituting electronic prototypes for physical ones. Although design loops have lengthened with more activity, build-and-test cycles have drastically shortened. Overall, the product-development cycle is much shorter than before even though engineering content has increased.
Structural finite-element analysis, unlike CFD, is an absolute approach. This means numerical returns from FE analyses feed directly into the design decisions, so accuracy is crucial. Although the same often applies to CFD, it's more likely that plots and images will be studied for energy-consuming eddy currents, vortices, large pressure drops at transitions, cavitation, and flow separation. The software has tools to help find each of these. Designs are compared based on incremental or relative improvements to these dynamic phenomena, eventually producing a better design.
But getting the most out of CFD software means scrutinizing results for subtle indications of undesirable phenomena that are sometimes hidden by a feature called autoscaling. A few problems are so obvious they seem to jump off a chart or graph. More often, users have to scrutinize analyses to find them.
PROBLEMS AND SOLUTIONS
As with most products, consumers expect more functions in faucets and shower systems, yet in smaller packages. This more-and-smaller trend places conflicting demands on design teams. Showerheads are an example in which package size conflicts with the need for proper nozzle shape, velocity, and direction. A pleasing or appealing spray often requires fully developed flow at the nozzle exit. This means flow near the nozzle exit should ideally have a near uniform velocity distribution. Tight package constraints, complex geometry, and winding passageways leave little room to follow conventional guidelines, which generally require a minimum of six diameters of straight length to establish a symmetrical, developed flow profile.
CFD provides a way to visualize flow problems by generating velocity and pressure plots, both aligned with and normal to the flow axis. The plots let designers minimize the conditioning section length. For instance, a plot of vorticity (the angular velocity of a fluid) can identify a vortex or strong eddy currents. Based on CFD results, one could insert flow straighteners such as fins or other geometry changes that help correct asymmetric profiles rather than adding a stabilization length. Particle-trace plots (highlighted paths of theoretical individual fluid particles) are also useful in spotting vortices and eddy currents.
Another problem comes in manifolds that feed dozens of nozzles, as in a typical spray head. Manifold-feed design must assure that each nozzle has the same pressure. To assist, CFD software can generate a plot of static pressures to help guide these designs toward an appealing and soothing spray pattern.
Low water-supply pressure presents another challenge. It calls for products with efficient waterways. CFD provides graphic plots of total pressures and velocities that identify energy-consuming phenomena, such as strong eddy currents, geometric transitions with high-loss coefficients, and concentrated points of high velocity and highly asymmetric flow profiles.
Noise is another design issue. Static-pressure plots let designers identify localized negative pressure zones prone to noise-emitting cavitation, a condition in which vapor bubbles form and collapse. A tool that generates so-called ISO surfaces, those at a selected pressure value, highlights areas or surfaces of interest.
Noise is also generated when two or more flow streams converge. The most frequent example of this condition is the blending of hot and cold streams. More noise is created where the streams impact head on, rather than meeting at an angle. Velocity plots identify such areas and guide designers to geometries that soften the blending of multiple flows.
A few undesirable flow characteristics are more difficult to spot because they are so subtle on graphs. Small vortices in a flow channel that feeds a nozzle is one example. Vortices can be spotted in plots of velocity normal to the flow axis. Vorticity (angular velocity) images also provide a sense of swirling liquid. Velocity plots may show these noisemakers, but the velocity values are on a much smaller scale than forward velocity and may be shaded or hidden by an autoscaling feature.
Autoscaling adjusts for high values present in regions of the model that are often not of concern. In doing so, it shades or masks the presence of lower vorticity levels in the area studied. The autoscale feature can be disabled and the scale manually manipulated as needed to gain deeper understanding of the fluid behavior. Additionally, particle-trace plots are effective tools for unmasking vorticity.
Along with displaying fluid behavior, CFD is a subtle but significant learning tool. Engineers and designers can see fluid flow in a way never before possible. Working with the charts and plots, a design team comes to understand flow behavior in complex channels.
ROOM FOR IMPROVEMENT
Although CFD software has developed into a practical tool for daily use, there is room for improvement. For example, to run analyses, one must create models of the waterways. That's not easily done. With solid models, the analyst has to turn solids into voids, and internal spaces into defined volumes using their CAD package. Future versions of the CFD software used here reportedly will have some of this capability. Users then have to recognize and remove certain details such as slivers — small, stagnant volumes adjacent to seals. If left in, the mesher will pack these small volumes with many small elements, generating large models that are prone to failure during mesh generation and take unnecessarily long to solve. Removing slivers and irrelevant details speeds meshing and solving.
Another time-consuming task in the process is finding and correcting mesh-related problems. Areas with intricate geometry, complex fluid dynamics, or both need a finer mesh than other parts of the model. Properly applying a good mesh depends on experience, careful judgment, and finesse. When improperly applied, an analysis will likely fail after running for hours. Then it's necessary to stop the analysis, locate the problem, refine the mesh, and restart the solver. Automating the finding and repair of mesh and solution failures would be a big advance for CFD packages.
Proper model construction is also key to success where intricate geometry exists. Assuring contiguous linking of features and not merging complex geometric shapes will make an analysis more robust.
CFD automation does not replace the need for hand calculations or detailed up-front engineering that pays attention to geometry that minimizes entrance and exit losses, flow straightness, or proper convergence and divergence angles. By supplementing proper engineering practices with the CFD tool, the end result is an optimized design as well as a more reliable product. What's more, test and prototype activities drop off, thereby shortening entire projects.