The good news is that SOLIDWORKS Flow Simulation makes it easy for engineers to use computational fluid dynamics (CFD), once an application reserved for specialists—often engineers with advanced degrees who became dedicated analysts. This could also be bad news. If such a powerful tool is used without the understanding of fluid flow fundamentals, it could provide inaccurate results without the design engineer ever realizing it. Much of what dedicated analysts have learned was learned the hard way: by making mistakes. In the previous article, we learned about a few of these mistakes. Here are a few more.
Trying to Replicate Physical Experiments
Referring back to the concept of an analysis plan discussed in the previous article, it’s important to decide if you’re trying to analyze the in-situ performance of your product out in the wild or replicate a physical experiment, as these often have very different requirements.
A frequent source of error is due to users trying to match the results of some test or physical experiment by using boundary conditions that would be more appropriate for the performance of the product under typical usage.
Consider the aerodynamics of a car. Replicating a wind tunnel test may involve a stationary model in a chamber of known size. Replicating the real use case of the car driving down the road may require incorporating effects such as wheel rotation, the relative motion of the ground under the car in an infinitely large open environment.
Another example is for automotive intake accessories or cylinder head geometry as in the figure below. There is a significant difference between trying to simulate the in-situ performance which could require time-dependent flow conditions representing the various strokes of the engine cycle, versus replicating a steady-state “flow bench” test where a fixed amount of vacuum is pulled to determine the resulting flow rate.
Figure 1. Replicating a flow bench test for an intake manifold.
Physical experiments usually only output results at a few key locations, while CFD provides output everywhere. Instrumentation usually involves additional test fixtures and rigging that may influence the device’s performance. If you’re attempting to replicate a physical experiment, representations of these fixtures should likely be included in the analysis. Also ensure that you probe the virtual measurements or define goals in the exact location of any physical test sensors.
Steady State vs Transient Analysis
When developing an analysis plan for thermal problems, it’s important to consider the “thermal mass” of the device and the time period of its intended operation. When a device has a low thermal mass and a long period of intended operation, a steady-state analysis makes a lot of sense. But what if the device has a heavy mass and only operates for short bursts? A steady-state analysis may provide a too-conservative and unrealistic result.
Steady-state analyses don’t reveal how long it takes for peak temperatures to be achieved. The engineer may think temperatures have stabilized quickly but there is no way of knowing whether those temperatures took 30 seconds, 30 minutes or 30 hours to reach a steady state.
When the thermal mass of the part is significant compared to the heat powers and time scale, it can be worth running a transient or time-dependent analysis. While the solve times of a transient analysis are much greater, they can be reduced significantly by taking advantage of solver options such as nested iterations, which perform sub-iterations for each solver timestep and allow specification of a much larger manual timestep size.
Figure 2. Transient thermal natural convection analysis.
In the transient thermal analysis above, the device takes approximately two hours to reach a temperature within a few percent of the steady-state value.
If you run a steady-state analysis and observe fluctuating goals or residual values, it’s possible that your problem may be “unsteady” in nature. This can occur due to vortex shedding or other dynamic effects that can spontaneously appear in the fluid flow at certain Reynolds numbers.
The more unsteady a problem is, the less accurate a steady-state solution will be. Even if you are after an averaged value as an output, it’s best in this case to switch to a transient solver.
To extract a steady-state value from an unsteady transient problem, you can export results from your monitored goals/sensors into Excel or other software and average the results over some relatively steady period.
Figure 3. Configuring averaged results for a transient study.
Alternatively, you may be able to specify in Calculation Control Options “averaged” results for a specific time interval. This will allow viewing time-averaged contour plots and other outputs that should be similar visually to the results you would expect from a steady-state study but with the confidence that the physics are properly supported.
Choosing Internal vs External Analysis
We commonly think of “internal” analyses for problems like manifolds and pipe flow and “external” analyses for flow over a vehicle. But the reality for many products is that the choice is not so obvious and although limiting the calculation to the internal region will almost always solve faster, it may neglect important factors about the outside environment.
Figure 4. Venting of a firebox as internal (left) and external.
If working with a room-scale or larger product, it may be desirable to see flow patterns through inlets or outlets and how they can interact with other geometry such as the floor or other obstructions.
For enclosures, if the model is too difficult to prepare as “water-tight,” we can use an external analysis as a workaround which allows the simulation of leakage through any small openings.
Figure 5. External analysis of an electronics enclosure with leakage from uncapped openings.
For problems where an internal analysis makes sense but there are still important effects to represent around the inlet and outlet, a balance can be achieved by creating a more representative inlet geometry shape. An example would be a hemispherical cap, which helps approximate the inlet air flow direction for devices that feature a rounded opening or velocity stack.
Figure 6. Hemispherical inlet on an internal analysis.
Extending geometry away from inlets and outlets can also help minimize any artificial effects imposed by boundary conditions. Guidelines for CFD typically recommend the length of these extensions in some multiple of the pipe diameter (three times diameter, six times diameter, etc.)
Figure 7. Extended inlets and outlets on a pump.
If you intend to neglect the frictional losses from these extended inlets, then be sure to specify an ideal wall condition on the inner faces.
Unrealistically High Heat Powers
For electronics cooling analysis specifically, a common issue seen is improper definition of the heat powers of electronic components.
Avoid confusing the absolute power rating with dissipated heat power. This can apply to power supplies, inverters, DC converters, etc. The waste heat for these could be estimated by multiplying the power rating by the efficiency – for a 300 W rated power supply that is 90% efficient, we could estimate about 30 W of waste heat that could be applied as an equivalent heat source within the CFD analysis. Applying the 300 W condition would result in some very high temperatures. Light emitting devices like LEDs also emit a portion of their energy as visible light so it’s important to apply their efficiency as well.
A trickier issue is understanding the duty cycle of various components. For many electronics, it may be unlikely that every component on the board will be operating at its maximum rated thermal power continuously 100% of the time. Detailed simulation of duty cycle can be carried out by cycling heat power on and off in a transient analysis but it is more common in practice to overlook minor transient fluctuations in temperatures and instead, simply scale down the heat powers by an appropriate factor.
Neglecting Thermal Radiation
Before neglecting thermal radiation altogether, it’s important to determine whether radiation has only a limited influence on your device’s thermal performance. For forced convection (fan or liquid cooled) electronics devices, the heat transfer tends to be dominated by convection and it is common to neglect thermal radiation.
While conductive and convective heat transfer rates are both proportional to linear difference in temperature, the radiative heat transfer rate is dependent on the difference of each body’s absolute temperature raised to the fourth power. This means that at higher temperature differences and higher absolute temperatures the effects of radiative heat transfer become much more significant.
For passively-cooled electronic devices that rely on natural convection, radiation can be worth investigating. Incorporating radiative heat transfer into your analysis would also be a requirement to analyze effects of different surface finishes and coatings such as black anodize which are known to have high emissivity values.
A quick way to investigate the effects of thermal radiation would be to duplicate your study and enable radiation, observing both the changes in temperature and heat transfer rate due to these newly included effects.
Figure 8. A flux plot showing heat transferred by radiation and convection.
The Flux plot in SOLIDWORKS Flow Simulation is a method to quickly filter by high power components and see where the heat is going. It’s also a useful tool to catch setup issues like lack of contact between components that should be conducting heat to each other.
Alternatively, investigation of radiation can be performed by a quick hand calculation. Incorporating temperatures obtained from your original analysis and measurements of external surface area should give a good idea of whether or not the radiative heat transfer will be significant factor for your problem.
Unsupported Physics
For problems beyond simple thermal and fluid flow, it’s important to ensure your CFD package supports the relevant physics.
For example, consider problems that involve the coupled motion of bodies with fluid flow or other effects such as “free surface” liquid/gas boundaries.
SOLIDWORKS Flow Simulation has the “free surface” functionality which can be enabled which uses the volume of fluid approach to solve simple problems involving sloshing or other behaviors. It also supports rotating components through the definition of rotating regions but doesn’t support any other type of body motion such as reciprocating or oscillating components. It also doesn’t support surface tension or capillary action.
Figure 9. Coupled body motion in SIMULIA XFlow.
For arbitrary motion of bodies with multiple degrees of freedom, coupled two-way FSI and more powerful and varied free surface methods, SIMULIA XFlow and SIMULIA Fluid Dynamics Engineer each have unique advantages.
Figure 10. Mixing tank with headspace in SIMULIA XFlow.
If your problem hinges on chemical reactions, combustion or phase change, you will want to find a solver that can tackle those effects.
Conclusion
It’s easier than ever to use CAD-embedded CFD to predict the performance of your product, with tools like SOLIDWORKS Flow Simulation making the setup process very straightforward.
While it still may require significant research and investment to build a very high accuracy simulation model (the saying, “getting the last 10% takes 90% of the effort” comes to mind), avoiding the mistakes in this article should help you achieve a level of accuracy sufficient to make design decisions and avoid some of the most common pitfalls.
Start by planning out the analysis with defined assumptions, inputs and outputs, then carefully choose where you place your virtual sensors or goals. Ensure your parameters of interest are solved through to their convergence and check that your mesh is adequate. Step back and look at the problem you are trying to solve, and determine if there are any changes required to match a physical experiment or to make sure the proper physics are incorporated.
Lastly, if you aren’t sure if you should make a certain assumption, use a certain approach or just how to proceed in general – reach out to your colleagues or your software provider for assistance! Sharing what you’ve tried so far and any documentation you have for your analysis plan should give them the info they need to quickly advise.