How to Think Like a Simulation Engineer
I like to talk about six tools of the trade when I talk about setting up simulation models that I have affectionately named: the Chainsaw, Virtual Reality, the Transmogrifier, the Reverse-3D Glasses, the Doctor and the Jedi Mind Trick. Knowing what these methods are, what they do and why and when you would use them will allow you to think like a simulation engineer. Seasoned simulation engineers (people who run simulations almost 24/7) are constantly thinking about how to go about setting up, troubleshooting and post-processing models. Let’s talk about these tools in detail.*
(*Most simulation engineers don’t work 24/7.)
The Chainsaw is the tool you use to cut models in half and then only solve that half. This is applicable whenever parts have symmetry, which they usually do, sometimes even two points or axis symmetry. I know people tend to think of cutting models in half as extra work, but it really isn’t. Consider this: Imagine a model of symmetrical shape, like a rectangle, that has 50,000 nodes. If all loads and fixtures are the same, then the displacements and reactions on the 25,000 nodes on one side will be identical to the other. However, the solver doesn’t know that, so it will assume you need all nodes solved. This means while the program is crunching through the last half of the nodes it actually already has that data solved! Imagine applying that to anything else. I’m asking you, would you like more solve time, or less? Pick less! As for setup, making an assembly cut could not be simpler, and then you just apply a symmetry condition as a fixture on the cut face. Done.
Virtual Reality is the tool you use to simplify difficult-to-mesh connecting hardware such as bolts, bearings, springs or any parts you are not interested in but need the mass effect (remote mass). Virtualizing parts has many benefits. Firstly, you don’t have to mesh them, so that saves time. Secondly, virtualization removes interferences. It is common for things like bolts, pins, bearings and other connectors to interfere due to being modeled larger than the holes they fit into. This makes simulation difficult, as meshing with interference is not possible. Also of benefit is the ability to add in virtual loads, such as a bolt preload or a spring K factor. Lastly, bolts and pins are actually nonlinear in their factors of safety, and thus the most accurate way to calculate them is with the polynomial equation inherent in the virtual connector.
The Transmogrifier is the tool you use to make a model work for finite element analysis (FEA). This means making modifications, configurations and simplifications in order to make simulation feasible. The motto is, “If a simulation model doesn’t work . . . MAKE it work!” For many users, this appears to be a huge obstacle. However, simplification is your friend and is actually quite easy. The trick is mastering the use of configurations, as it will make less of a headache for you, the user. Remember that configurations of parts can be used in a configuration of an assembly, so an FEA configuration of an assembly, referencing FEA versions of parts with modified dimensions, is really all you need. As for the modifications themselves, they usually involve eliminating interference that is created as a result of stacking tolerance in virtual parts, or the opposite (gap). While gaps can be handled easily, interference cannot. Other things include simplifying unnecessary features, such as cosmetic features, or features that serve a purpose other than what is currently being simulated (for example, heatsinks don’t serve any purpose structurally, just suppress them for your strength test and deal with them at a later time).
The Doctor is the tool you use to check on the health of your study. The Doctor comes in many forms, but mostly when simulation engineers talk about health, they mean mesh health. This can easily be determined with the mesh aspect ratio plot, found in the mesh tools. An ideal aspect ratio is 1, which means all sides are equal. Stretched elements, such as tall, skinny ones or short, wide ones, lead to inaccurate force calculations. However, no simulation can have a uniform aspect ratio of 1 or it would be too mathematically complicated to run; instead, we look for ratios ranging between 1 and 10 for good accuracy, 1 and 5 for great accuracy and 1 and 3 for near-perfect accuracy. Another tool you can use is the Energy Norm Error plot, which is a standard stress plot. This result shows where probable errors in the iterations are accumulating and usually coincide with poor mesh; the difference is that this also shows you how the forces are reacting, even in areas with mesh that looks good but is still not refined enough for the amount of energy transfer. The Doctor should realistically always be utilized.
The Reverse-3D Glasses is the tool you use to simplify 3D models into a 2D or even a 1D version that retains their accuracy, while making meshing much, much simpler. Many elements of designs, such as sheet metal, have very thin cross-sections over large spans. Since we have to mesh by the lowest common denominator, making meshes for sheet metal in solid is overly complex. To solve this unnecessary complexity, we have the option to remove a dimension from constant cross-section models and solve only their spans. We also have the option to turn things like beams into 1D models, only solving for length. These assumptions are almost always valid; sheet metal and beams are very often constant cross-sections, and as long as we virtualize it properly, the results will be the same as if it was solid. Also, we have in Sim Pro and above the 2D Simplification tool, which allows us to analyze an entire 3D model as only a 2D cross-section and extrapolate the results; this has the obvious advantage of minimizing mesh complexity. The idea behind it is very similar to the mentality of the Chainsaw, in that if the cross-section is constant, and one slice of it is solved, they really all are solved and you are just repeating the same number crunches over and over.
The Jedi Mind Trick is the tool you use to remove unnecessary parts and replace them with virtual forces. This is different than virtual reality, which deals with things like bolts and other connectors; instead, this is a question of how deep down the rabbit hole do you want to go? For example, we could analyze a design on a laptop satchel to make sure it has a good factor of safety. To do this, we could model the entire satchel, with the laptop, and the person’s shoulder as the strap hangs from it, apply gravity and let it run, but that would take forever to set up and get parameters for. Instead, we can simply run a simulation on one connector, with a fixture, and a force that is close to the force from the scenario above. The cascade model can also be used here. The idea behind the Jedi Mind Trick is to use, for lack of a better term, telekinesis — invisible forces as opposed to complicated contacts and assembly interactions, which will result in the same boundary conditions. Also, I’m aware this is not technically a Jedi Mind Trick (it’s actually Force Move Object or Force Push), but this was a better name. I’m willing to bet you’ll remember it now.
Let’s see some examples of these tools:
Here we have a simple bolted plate test. We want to know how much force can be applied to that single plate as it is put into tension by a fixture on one side and the plates on the other side tearing it apart. We are assuming the bolts, which are grossly oversized, are not of interest. Having said that, we can use a number of tools on this: the Chainsaw, the Doctor, the Reverse-3D Glasses and the Jedi Mind Trick. The Chainsaw is obvious. We can just cut it in half lengthwise and save on simulation time. We can also take away the thickness of the plate using the Reverse-3D Glasses so we do not have to mesh across the thin cross-section. Lastly, we can eliminate the bolts entirely and just replace them with a force via the Jedi Mind Trick.
Mesh of the bolted plate
You can see how simple this model turns out to be. Notice the symmetry fixture on the edge that is cut; this will make it so results can be extrapolated later and also prevent the model from violating that line of symmetry. Another thing to be wary of when using the Chainsaw is using forces; you have to cut forces in half, as the program will not automatically update like a pressure load would. This is not the case if you use the “Force per Item” option, however, because you’ll have fewer items and thus the same amount less force. As for the Doctor, let’s see a comparison of health versus error.
Checking aspect ratio on a low-quality mesh
Following is a low-quality mesh, with some high aspect ratios. Compare it to the percent error.
High error induced by poor aspect ratios in the mesh elements.
The error plot shows a lot of elements with large errors near that hole. For comparison, our high-quality error looks like the following. Note that the scales are the same, 0 to 15 percent.
Error plot for a higher quality mesh with more appropriate aspect ratio.
Much better. We can trust these stress results because of the low error. As a bonus, the high number of nodes allows us lots of probe points, and thanks to our tricks, the solve time is still only four seconds (the high-quality solid, with mesh one-quarter of this density, took 30 seconds).
Let’s look at a more complicated example.
Thread pull test
In this case, we are interested in seeing if these threads will survive a pull out test. However, this model is not fit for simulation as of now. The threads don’t actually touch each other, and the inside of the female thread has fillets whereas the stud does not, which will cause interference. Let’s use the Transmogrifier tool to help that. It’s not so complicated, really. Just make an FEA configuration.
Adding an FEA configuration
Then change the features and dimensions in any parts you need. In this case, we only need to change one.
Focusing on the right features
Simplifying the threads to 2D
Now that we have our model correctly made for FEA, we will use our Reverse-3D Glasses to make a 2D model, as 3D threads are a real nightmare.
Using the 2D simplification
We use the 2D simplification tool with axi-symmetric selected. This will give us an infinitely thin slice to work with. Now we need to assign no penetration contacts between the teeth; bonded will result in too strong a part to accurately compare it to real life.
Since the parts are directly touching, we can just set component contacts to global. Let’s see the results from this test.
Results from the 2D simplified mesh
This test looks good. No tooth is at or above yield. However, let’s use our Doctor tool to verify.
Error on the simplified 2D thread mesh
The error in the top and bottom tooth sections is very high, over 25 percent. Clearly, we need more mesh. Let’s look at it again with a fine mesh.
Thread results with a more appropriate mesh
Using this fine mesh, we can see the tip of the bottom tooth is above yield, and the majority of its body is without a good factor of safety. It is very likely this tooth will break, leading to the second tooth becoming the next in line (minus the strength of one tooth), which will also break, and so on. This bolted connection is not safe and is subject to cascade failure. This is a result we did not see in the original and illustrates the importance of a health check in a study.
Error map for the final results
The errors on the tip of the thread are removed (errors on the fixed body and underneath the threaded connection are unimportant).
Now, let’s see an example using all these tools at once.
This model has a number of built-in interferences due to the hardware. Eventually, we will have to get rid of them, but remember that the Chainsaw should always be your first tool. Reason being . . .
Going to town with the Chainsaw
Now we only have two pins, one bolt and a limited number of interferences to deal with, in addition to the fact that we have less to mesh and define. The interferences are still an issue, and while we could eliminate all of them, we should try to virtualize using Virtual Reality as much as possible. Let’s start with those bolts.
Replacing bolts with a virtual connector
This bolt is now suppressed and replaced with a virtual connector. In SOLIDWORKS Simulation 2015 and above, this can be done automatically by right-clicking the Connections folder and selecting Toolbox Fasteners to Bolts. This eliminates all the interferences the bolt was causing, as well as includes the ability to add preload (something not possible in static analysis without virtualization).
Setting up virtual fasteners
The pins are defined in a similar fashion. Here, I virtualized their stiffness and resistance to movement (they have infinite translation resistance but 50 N*m/rad rotation). That takes care of the fasteners. Now let’s fix the body. Just like before, we’ll make a configuration and adjust dimensions.
Configuration for the body
More configuration setup
This also applies to subassemblies.
Taking care of the shaft
Transmogrification complete. Now to reduce as many thicknesses as possible, use Reverse-3D Glasses.
Mesh of the reflex engine
Notice we did not make one of the blocks a shell. That’s because if you check the above screenshot, you will see the plug part fits into a recess. That recess can’t be modeled with shells (everything has to be constant, remember), so we bite the bullet and run it as a solid. As opposed to putting in complicated pressure loads from flow on the inside of the pipe, instead, we will use thermal expansion to control the displacements from a worst-case scenario.
Calculated deformations for the reflex engine
Here’s our super-deformed model with displacements. Let’s do a health check using the Doctor.The Doctor visits the reflex engine mesh
Our shell mesh looks mostly good, but some elements in that plug are very bad. However, this is acceptable. The reason why is that plug is significantly far away from the areas of large displacement, and in reality, we could probably get rid of it entirely. (It does have heat load applied to it; switch the heat load instead to the faces it directly touches — there’s our example of the Jedi Mind Trick.) We kept it in this case because it has the other hole for the bolt and thus is necessary for our pre-load.
Losing what’s not critical to the analysis
Our error is low. Lastly, thanks to virtualization, we can get results from those pins.Final results on the reflex engine
This is a result that is not easy to get from other studies. It can even predict failure from a factor of safety criteria. This type of operation we just performed is typical for simulation engineers, and these tools along with practice will help you think just like they do about how a problem should be handled in FEA. These are the tools you’re looking for.
About the Author
Michael Kalin has been teaching simulation since 2012. He regularly does presentations for simulation at conventions such as SOLIDWORKS World and SOLIDWORKS User Groups. In his off time, he enjoys building his own 3D printers, playing Shogi, traditional archery and Ghostbusting.