I’ve been involved in engineering simulation for 20 years. Not quite sure exactly how that happened, but none-the-less here we are. Back in 1996, when I was studying engineering, a good part of my course looked at the fundamentals of FEA for structural analysis and CFD for flow simulation. We spent an inordinate amount of time manually calculating how a five-element beam would behave. I dread to think how many trees were sacrificed at the expense of my scruffy algebra.
I learned two key things from this exercise. FEA was incredibly useful —I could get an engineering answer to a reasonably realistic problem by using this approach — and that FEA software was a must if I wanted to do this on a more meaningful model.
I guess if I was just looking at varying the load, length and material of the beam I could have created an Excel spreadsheet or something, but it’s certainly not a scalable solution.
The real value FEA software brings is that I can get an answer to an engineering question faster (and cheaper) than any other way.
To truly maximize this though I need to be able to ask the software questions. What about if the hole was bigger, of there were more of them? What if I made it out of aluminum instead of steel, or if the load increased? To do this effectively I need a model that can be setup once and re-run to answer each of these questions.
This is the first of the three P’s: persistence. I need a model that can have loads, materials and geometry changes and be robust enough to cope with the changes and mean I can work productively.
It’s pretty straightforward to build a model that has persistency if you’re just changing the size or magnitude of things, but how about if the topology changes? Applying a constraint to all the holes is tricky if the number of holes changes as you update your design. You need a smarter way to grab them.
The example above is setting up a selection set (highlighted red surfaces) to automatically update if the number of holes changes. We can make use of a cylindrical coordinate system. By setting the selection set to pick up faces between 2 X values (shown with the yellow and blue circles) we automatically capture all the holes if the geometry changes, as in the image on the right.
The second P is to be able to parameterize my structural analysis. If my model is persistent then I can build an automated workflow to look at different designs and loading scenarios. Being able to parameterize the models attributes, geometry values (hole counts, material thicknesses etc), choice of material or the material properties and then loading values will mean that I have the mechanism to automate my analysis study.
You might also want to consider setting up more intelligent parameters with some logic built in. Say for example if the number of holes is 4 or less apply one load and if the number is greater then use another.
It is possible to do this with some though.
The key to doing this is the built in true/false booleans with the ANSYS DesignXplorer expression tool.
(((P1>4)*10)+((P1<5)*5))*(1[N]/1[m^2])
The above expression can be broken down into 2 pieces.
(P1>4)*10 P1 is the parameter that controls how many holes we have on our pattern. If the number is greater than 4, then P1>4 evaluates as 1. The same is true for the P1<5. So for example, 4 holes would result in a pressure of 5 Pa, and for 6 holes, 10 Pa.
1[N]/1[m^2] This is the bit that catches a few people out. DesignXplorer parameters are units sensitive. Since we’re defining a pressure we need to convert our unitless figure into a figure with units.
The third and final P: performance. The previous two P’s, persistence and parameterization are redundant if the whole exercise takes longer than other methods to complete a structural analysis. I need each stage of the analysis to be as fast as possible. Big workstations, clusters and the cloud can all provide the computing power, but can the FEA solver make any use of it? Being able to utilize all of the compute resource means the solver needs to be able to split the job up efficiently and run over all of the CPU’s or cores available to it.
Even with the fastest solver available, if it takes a long time to build a model, my total time to getting results might be restrictive. Being able to set-up an FEA model efficiently and get from geometry to solution as fast as possible depends on everything in between. So, performance definitely relies on solver speed, but also on usability and productivity of the FEA software.
ANSYS Mechanical has been pushing the envelope to deliver a solution to its users in a robust and fast manner for years. The three P’s are a great way for you to think, am I working as productively as I can? Or am I spending too much time driving software and not enough time making the decisions I need from the results of my structural analysis?
If you want to learn how ANSYS Mechanical handles the 3 P’s, why not check out our webinar on the 25th of August.
You’ll discover how persistent, parametric models powered by high-performance solvers produces a model that can be used to get a full understanding, not just a snapshot.
The post 3 P’s of Structural Analysis appeared first on ANSYS.