It occurred to me this week: Who better to explain fixins than a Thanksgiving Turkey! So, I fired up my SolidWorks 2012, and my turkey was done in about 30 minutes. Heh, it’ll take hours on Thursday!

Then, I got his mesh all set up. I used solid elements for his body, shell elements for his wings and tail, and beam elements for his legs. I had to make his legs a little longer to illustrate the point of this blog, but he’s still a handsome turkey I think.

Tech Tip: When using dissimilar mesh types, don’t forget to create Bonded Contact between them.

This is because the equations which define how an element responds to loading are different for each element type. The Bonded Contact tells the Mesh to create equations of compatibility so that the differing element types can pass information back and forth.

My turkey wings bonded automatically to their mounting posts. A few clicks later and I had the tail and legs properly bonded to the body. Mesh completed with the default element size.

Next, my turkey needs something happening to it. How about a 1 psi “tail wind”? Also, maybe another turkey charges his head with 22 pounds lateral force. For a little more color, let’s tug on those wingtips too with a little outward force.

Looks tasty!

For the first run, let’s select the beam joints at the turkey’s ankles and make them Fixed.

A few seconds of solving, and my turkey’s done. Boy does he looked stressed! I guess I would be too if I were a turkey at this time of year. [joke credit goes to Art Woodbury]
Tech Tip: Since beam element stress types are different than solid/shell stress types, you cannot display stresses of both element types at the same time.
I composited two plots together with an image editor to get that pretty picture for you.

Let’s look at a Displacement plot instead, since then I can show you the beams and the solids/shells all at once. Also the color gradient is more turkey-like. Matter of fact (Tech Tip!) you can customize the color legend to your liking (right click the plot and choose Chart Options > Color Options). I also changed the icon color of the fixtures from green to orange, and opted to show the non-deformed shape superimposed (right click the plot > Settings > Deformed Plot Options). Here’s some very turkey-like colors!

Now for the meat of this story [pun intended].

Let’s edit one of the fixtures on the turkey’s ankles and change it from Fixed to Immoveable. This means that the turkey’s ankle, instead of acting like it’s been stuck in concrete by the turkey mafia, it will be able to rotate. But it still can’t move off its spot. This is only true for shell elements and beam elements, because those elements calculate rotations at each node. Solid elements derive rotations by comparing the translation of nearby nodes, and so solids only calculate displacement. So…

Tech Tip: For solid elements only, there is no difference between Fixed and Immoveable fixtures.

Another few seconds in the solving oven, and our turkey looks redder, as his overall displacement increases. If you look at his left leg, you can see it is not as stiff a structure as before. The beam of his left leg rotates freely around the ankle joint, and therefore does not bend like his front leg does. You can also see that the icon representing the fixture is slightly different (arrows don’t have the disks on their ends, indicating that the rotation is not locked).

So the moral of the story is: If you have turkey with all the fixins, you’ll have less mobility than if you have turkey with just some fixins.

Next topic: Pumpkin Pi.

Add another part to the assembly, and build this part as a solid block of material that references the outermost faces of the assembly.  It should completely contain the assembly, and if it is a tad too large in some areas, don’t sweat it.  Once you have a solid that completely envelopes the study geometry, use the SHELL command, and shell this solid TO THE OUTSIDE, by some convenient thickness –maybe .25” or so.  Now, HIDE this solid, and go back to your Flow study set-up.  The assembly is now a closed, “internal” study, and where any air DOES leak out through small gaps in the original assembly, it can’t leak far.

First came making the model.  I took two pictures of my car, one from the front and one from the side (yardstick included for scale):

Then I inserted those pictures into SolidWorks sketches and traced over them:

I used a trick we teach in our Advanced Parts class and made one extrude from the side profile, one from the front, and then “Combined” the bodies to leave only what’s common:

Side Profile Extrude

Front profile extrude, overlapping the side profile

The results of the combine

So the results of the combine aren’t a perfect representation of the shape of my car, but it’s not bad for 10 minutes of work.

I didn’t have a way to take a picture from directly above, so I made my best guess about the top profile, did another combine, and added some wheels and seats:

That’s a good enough car for my purpose.  I added the car to an assembly and downloaded some willing passengers from my favorite free model site, grabcad.com (human models courtesy of GrabCAD contributor Ken Schulze):

No, something still doesn’t look right.  Let’s see…

Okay, that’s better.

There can only be one Flow Simulation project per configuration, so I made one assembly configuration with the top down, one with the top up, and ran an external analysis with the wind flowing at 70 MPH.  (One great thing about Flow Simulation is being able to use whatever units you want, while never having to do a conversion.  The MPH units are just as easy to set up as m/s or any other standard speed unit.)

One thing that makes Flow Simulation different from some other CFD (Computational Fluid Design) packages out there is the setting of goals.  Your goals tell the solver when to stop iterating, so that it doesn’t spend time trying to figure out the turbulence in one corner of your electronics enclosure when all you wanted to know is the velocity through your outlet.  In this case, all we care about is the drag, so my only goal is the force on the entire car in the Z direction (the direction of travel):

One other thing that sets Flow Simulation apart is the ability to watch the solver in action.  This is very different from the FEA (Finite Element Analysis) solver in our Simulation package, where you just hit ‘Run’ and hope.  Here, you can watch the solution converge, and stop the solve if you realize you’ve set-up an input wrong (like having the flow go the wrong way).  Everything seems to be developing fine here, however:

We only see the middle of the car because my preview plane misses both the wheels and the passengers.  Notice in the top left corner of the solver picture, the number of fluid cells is 12,139.  I know from teaching the Flow Simulation class that 20,000 cells with a few goals will usually solve in under four minutes on my laptop, so I should have an answer very soon.  In fact, I wasn’t even able to capture a second screenshot before the solver finished:

Remember, these results are steady-state.  They don’t account for every random eddy the flow could throw off, just where the air goes on average.  Transient flow (non-steady-state) can also be done, it just takes longer to solve.

Back in SolidWorks, I can show almost any result I can dream of.  I always start off with a basic Cut Plot, and I notice a few things wrong right away:

First, our mesh is too coarse.  Any place you see a wide color variation contained in just one or two large mesh boxes means a lot of change is happening at a scale finer than our current mesh can calculate.  It’s guessing, interpolating in the most interesting parts, and we can do better.

Secondly, the car is levitating.  I made the computational domain too tall; zero on the coordinate system wasn’t at the bottom of the tires.

The second problem is an easy fix, just a quick number change.  For the first, there are many ways to go about it, and we teach them all in our classes.  The solution I chose is to make a simple rectangular block to represent where I want the mesh to be smaller, called a “Local Initial Mesh”.  The resulting mesh looks finer in just the areas we care about:

And the result looks a little better, with somewhat less color change in any one mesh box:

This study took 81 seconds to solve and the calculated drag was 107 pounds.  But how do we know if that’s “real”?  How do we know if even the new mesh is fine enough?  This is the crux of any simulation run, in any software.

Until you’ve done enough runs on a given project to develop a sense for it, one way to tell is to increase your mesh fineness and track how your goals change.  Eventually, making the mesh finer will have little effect on your goals, (they will be “mesh-independent” in the lingo) and you will have more confidence that your answers can be trusted.

Also, if you start plotting the time it takes for the solver to finish versus your goals, you’ll see the cost you’re paying for a better answer.  I increased the mesh to around 60,000 total cells, it took 8 minutes to solve, the calculated drag was 96 pounds, and the picture looks even better:

Was that 10% change in my answer worth waiting six times longer?  Maybe.  What about six times longer again to get 5% closer?  Maybe not.

One of the things to realize about simulation is that your answers will never be “right”.  They can only get “closer to reality”.  But the question is, how much time do you want to spend getting that little bit closer to real life?  Simulation is best used comparing two or more CAD models against each other, not comparing idealized mathematical equations against the messy, fuzzy real world.

So, switching to the top up configuration and using the 60,000 cell mesh, the flow definitely looks smoother over the top:

And the numbers bear that out: the calculated drag was 79 lbs- almost 20% less than before!  So my gas mileage gets about 20% worse if I decide to enjoy the sun at 70 MPH?  That’s pretty sobering.

Now with the runs complete, I can plot the data many different ways, including “particle studies” which put finite drops of liquid in the stream and plot where they end up.  I did that on the top down model at 70 MPH.  Then I noticed something interesting and did it again, at 30 MPH.  Compare these two pictures and see if you notice what I did:

Particle study at 70 MPH

Particle study at 30 MPH

I’d love to close the loop and do a physical test with those droplets, but of course I don’t have a wind tunnel big enough.  So there’s no way to prove out those results, right?  Right?

Understanding the effect of dynamic factors like temperature, heat sources and air flow is critical to your product design. SolidWorks Simulation gives you valuable insights into even the most complex designs. It seamlessly integrates simulation, testing and analysis into design, and its intuitive easy-to-use interface quickly makes it part of your process.

Watch as we apply SolidWorks Simulation to a very real-world problem: how long will your brew stay cold at an afternoon barbeque and in the process uncover a potential design problem.  Check out the video, then call for a demo of how SolidWorks Simulation can enhance your design process.

See how SolidWorks Simulation solves tough design problems without complexity.

• Kinematic Systems
  • Movement of part(s) under enforced or constrained motion.
  • Fully controlled and only one possible motion result irrespective of force and mass.
  • Zero degrees of freedom
• Dynamic Systems
  • Movement of part(s) under free motion subject to forces.
  • Partially controlled and infinite number of results depending on forces.
  • Greater than zero degrees of freedom

In motion simulation, you can mix these types of systems, as in the case of a medieval trebuchet system. There are moving components that are part of a fully controlled motion system, like the gear mechanism, and there are free motion object (the projectile). SolidWorks motion calculates 3D contact between objects, incorporates friction, can accommodate springs and dampers and can even show you if any interferences are occurring during the motion simulation run.

In this setup, shown in the video below, we turn the crank of the trebuchet 2 and 3/4 times (990 degrees) for a duration of 3 and 3/4 seconds (pretty fast while in the heat of battle). Then we define contact occurs where two wedge shaped components come together, which triggers the projectile to slide down into the basket. In the basket, we define a virtual spring, which pushes the projectile snug into the correct launching position. Once the 3 and 3/4 seconds are up, we get out of the way of the crank and let it go, to watch the rack and pinion gear system, which is sent back in the opposing direction due to a large counterweight and we see the projectile launch.

Once the simulation has actually run, we can get a rich amount of plot data , including how much torque is required to turn the crank (Do we need Andre the Giant from the Princess Bride and wrestling fame to turn the crank?) or as we mentioned earlier, we want to know how much damage we inflicted on my enemy. Simply choose the plots button in the SolidWorks Motion interface and choose the Category of plot you wish, in our case Displacement/Velocity/Acceleration and then choose a sub-category, in our case linear velocity, and then the component direction, in our case the magnitude. Then click on the projectile inside of SolidWorks and we can then get the speed of the object when it hits the enemy, which is 527 in/sec (around 30 mph). Not bad for a miniature version of the trebuchet projectile at 1/2 a pound which sits on my desk in Meriden, CT.)