“Garbage in, gospel out.” A slightly different take on the old adage, this is many people’s perspective on analysis results. But we engineers know better, right? Simulation results are generally only as good as the inputs to the problem. Or at least, you need to understand the limits of your engineering assumptions, and guide your interpretation of results accordingly. There are however many times you’re certain your results are wrong, and yet you can’t figure out why.
Step One: Do Not Blame The Software
It’s tempting to blame the software, but we find in tech support that less than 1% of so-called “wrong” results actually turn out to be a glitch in the system. Instead, generally the cause is PEBKAC: Problem Exists Between Keyboard And Chair. Here are some common missteps that you can double-check before embarrassing yourself in front of a fellow engineer.
Are You Using the Right Units?
SOLIDWORKS Simulation is great at letting you mix and match units within each dialog box as you set up the problem and examine the results. No more worrying about slugs-per-cubic-inch! However we still see a surprising occurrence of users missing the subtle indication of MPa when they’re expecting to see psi. Verify your units:
- Double-click on the color scale.
- Choose the Definition tab.
- Select desired units.
Also pay attention to the units used within each object on the study tree. Sure you can combine newtons with inches, watts with Fahrenheit, BTU’s with millimeters, etc. – but be careful! If you change the unit type after you’ve typed in a number, SOLIDWORKS does the math to convert the value to the new units. For example, 5 newtons becomes 1.142 pounds, and NOT 5 pounds! After changing from SI to English, type the number again!
Are You Displaying the Desired Result Type?
Years ago, a customer of ours was running validation tests of SOLIDWORKS Simulation versus ALGOR, and he decided that SOLIDWORKS was giving the “wrong answer”. It turns out, ALGOR was reporting First Principal Stress, while SOLIDWORKS was reporting Normal Stress in Z. Surprise! Not the same thing. This mistake happens to the best of us. We forget that by default SOLIDWORKS Simulation gives us Von Mises Stress, which—obviously—is not the same thing as a Principal Stress or Normal Stress or any other kind of stress. Oh, and don’t forget the difference between static pressure and total pressure of a fluid! When we corrected the result types, we found out that SOLIDWORKS matched the theoretical answer within 0.0001%, while ALGOR was off by more than 10%. I’ve always known to question anything Al Gore says! Speaking of theoretical answers…
But It Doesn’t Match My Hand Calculations!
Let me put this bluntly: Your hand calculations are wrong. If hand calculations were sufficient, you wouldn’t need Simulation! A simulation is always more accurate than your hand calculation. Sure, there are times when a singularity or other mathematical anomaly throws your results way off (reason #6 perhaps?). But more importantly, the Simulation software takes into account a lot of pesky phenomena that your hand calculations probably ignore. One of our favorite types of tech support case is trying to figure out which assumption an engineer made in his hand calculation that is simply not true, and also is not neglected by a good simulation study. The most common examples are in Flow Simulation. Here is a partial list of things the solver does correctly, but your hand calculations might not:
- buoyancy effects (hot air rises)
- transient flow (unstable and/or non-steady-state)
- non-idealized geometry
A classic example was a customer who had assumed a steady-state solution for his hand calculations, but was dealing with an event that produced an oscillation.
But It Doesn’t Match My Test Results!
“Nobody believes the test results except the Test Engineer. Everybody believes the analysis results except the Analyst.” Ah, that Holy Grail of simulation: correlation of results with physical testing. So elusive. Results don’t have to correlate to be valuable to the designer! But you should be in the ballpark. And if you’re not? Ignoring all results for a moment, first answer this: Does your simulation match your test? If the results don’t match, it’s usually because the problems don’t match. This happens ALL THE TIME. If you want the computed pressure drop through a system to match the measured pressure drop, you need to take the results at the same locations in the system! That usually means creating models of parts of your test rig to include in the study, especially all the pipes and fittings between the pressure gauges. If your simulation temperatures are very different than your tests, it’s because you’ve neglected or underestimated the effect of a heat source or heat sink in your test environment. For example the large metal table your electronics are sitting on. Or the HVAC above you. Sometimes the mis-match is more subtle, like a small recirculating flow path, but it’s always there if you look hard enough and make fewer assumptions in your study setup.
Double-Check Your Setup
Use the preferred proofreading method of having someone else do it. Get a second pair of eyes to look at the problem. Odds are you made a careless mistake somewhere in your loads, boundary conditions, or mesh settings. But that’s pretty superficial advice, so let me give you some specific “gotchas” to look for in your Simulation study setup:
- Pay attention that Force or Power applied to multiple faces is “Per item” or “Total”.
- Make sure you’re using a High Quality Mesh for final results.
- Initial Temperature needs to be defined for each body for Transient Thermal studies.
- By default all touching faces are Bonded. Is that what you intended?
- Automatically finding Contact Pairs can create a lot of confusion. Verify every one.
- Loads and fixtures should be applied to faces, not edges or vertices (typically).
- A negative Pressure acts in the opposite direction of the arrow symbols.
- For thermal stress, the Zero Strain Temperature is defined in the Study Properties.
- Did you specify the fluid temperature in your Convection boundary conditions?
- In a Thermal study, did you miss selecting and loading any exterior faces?
Sometimes, the results aren’t wrong at all, they’re just not what you wanted to hear. Engineers are much more critical of simulation results that indicate a design flaw than they are of results that show good news. Hopefully, you’re doing analysis using SOLIDWORKS Simulation early and often in your design cycle, so you can identify concerns while they are still cheap and easy to fix! If not, then blame Al Gore.