In this installment I talk about optimization and analysis. This applies equally to mass production and one-offs, and is really the critical set of skills that sets apart engineers from the many other people involved in successfully manufacturing a product. That said, anyone involved in manufacturing benefits from understanding how these tools are used, even if they won’t use them directly.
Note this isn’t a how-to for FEA, there are lots of those available on Youtube and elsewhere.
Plastic bending calculations
As I discussed in my first post, I used BASF’s guide to write my own calculator. To test it, I compared my results with with finite element analysis performed in Solidworks and Fusion 360. I also confirmed my results with a hand calculation, however that was just to confirm my math was setup correctly.
Finite Element Analysis (not just pretty pictures)
Any engineer reading this can probably think of a time that they’ve seen an FEA analysis that was really just a pretty picture, either because it wasn’t set it up correctly, or because the thing analyzed was not worth the time. The goal of this post is to help you understand how to do FEA and get useful information, not just pretty pictures.
The first step in all this is understanding what data you want to get out of the analysis, and why you can’t get it by other means. The second step is to understand the importance of boundary conditions, mesh size, and loads.
Identifying the problem
I expect the issue here to be a stress concentration where the clip bends away from the base. This is simply a matter of experience, but anyone can detect potential stress concentrations by looking for through geometry with sudden changes in cross section. These are always potential failure points.
In this case, we only have an interest in structural FEA, but there are many other types of FEA for fluid flow, thermodynamics, electromagnetics, etc. Solving problems with them can be approached the same way, but requires a different knowledge base.
The boundary conditions are physical limits that we know or can assume are true. These are absolutely critical to getting a useful solution, and are often the most difficult of any FEA analysis. Today’s example is very simple, mostly because the part is simple. In cases requiring dynamic analysis, or with many components, or with odd physical limits, some or all of any analysis may be garbage regardless of how it is run, and it’s up to the user to identify which parts are useful and which are not. That’s why you pay a professional for this kind of work.
Once I’ve run the simulation and determined the magnitude of the stress, I want to confirm that it won’t cause the design to break. Given that, I need to go back and determine what the stress at the elastic limit is for my material (aka the yield strength). I performed my analyses using ABS, which is a common plastic for both 3D printing and injection molding. One thing to keep in mind is that 3D printed material is anistropic (the strength between layers is significantly lower than the strength of each layer, which is equivalent to injection molded part strength). Basically, it wants to delaminate because the layers aren’t held tightly together.
The yield strength of ABS is quoted at anywhere from 4-6,000 PSI, depending on the test standard (ASTM D638 is the most common) and who performed it. It’s common in the 3D printing world to assume that Z-axis (the direction that layers are stacked in) strength is 30% of the specification yield strength. So I want to stay below 1,300-2,000 PSI in the Z-direction to prevent delamination.
First things first, a mesh is the structure of points that is being analyzed. It looks like, a mesh net or fence, hence the name. It’s built using a series of polygons, usually triangles but there are other options. The actual math is performed at the locations where the triangles meet (the nodes), and the system is basically iterating through until the change in value gets very small relative to the value. If you’re interested in understanding more of the math behind it, I suggest finding a book on numerical methods or contacting your nearest university to take classes.
When it comes to sizing the mesh, we have some areas of interest and some areas that are not interesting. Large, flat or otherwise geometrically identical surfaces usually will not tell us anything of real value (that can’t be calculated by hand relatively quickly, for example). Usually there are a few features of the model that are really of interest, and those require finer meshing. I’ll just jump straight into an example.
Below is the meshed holder in Fusion360. The software has done some automatic optimization of the mesh size in different areas, so you can see the corners where it has made the mesh significantly finer in order to get useful values.
The tradeoff in meshing is time versus the helpfulness of the result. You can make a mesh that has tiny elements which takes forever to solve and gives you high granularity, but it will take longer to run and may not help give you better answers. See the mesh below, which has about 10 times the mesh density of the one above, but most of the mesh is now being solved in areas that are not helpful or interesting.
To make a fairly long story short, the required deflection of this design creates super high stress at that corner. I tried again using relief notches, but they didn’t help much. Ultimately it was quicker and easier to go back to the drawing board and make up a new model.
This switches the mounting side away from the flexure, and it also a little more compact and robust. In addition, it was easier to tweak this to make the holding force closer to the flashlight’s weight. This requires about 1 pound of force to deflect to the point required to allow the flashlight to be retained or removed.
Something wasn’t quite right about my measurements, as round 1 didn’t fit. So, I made some slights adjustments.
Better fit, but I wasn’t happy with how difficult the bevel was making it to insert the flashlight. Time for one more revision.
Much easier to insert, with no retention issue.