Continuing the discussion of design and CAD tips taken from the GrabCAD design contest, today’s topic will be model preparation related to multiple design analysis iterations.
- A Word about Parametric Analysis
- Model Preparation
- Stress Analysis
In part 1 I related a quick discussion on Parametric Analysis and how cool it is, and how slow it can be. There are a few things we can do to make the model more stable during the constant changes to the basic frame design, and much easier to work with in general.
Top Down Design
I built the basic overall parameters into a skeleton file, along with key axis and planes that represented some of the design limitations given by the contest. This achieves two things: multiple variation can be run by deriving the same basic skeleton into different part files, while maintaining the core feature relationships in each model. The second is that if there is a criteria change, it will be distributed to all prototypes.
There is however a twist here. Since we are running variations in some parameters, such as thicknesses, along with different variations in separate model prototypes, some skeleton parameters will vary within the prescribed limits. If I try and perform analysis on a range of values directly in the skeleton’s parameters, those values will be reflected in ALL prototypes. Once a prototype is developed, I don’t want any feature changes to occur. Very Bad Mojo. How do we get around this?
Rather than driving features directly from the reference (derived) parameter directly, you might set up a parameter that can act as a buffer.
ThickFactor = 1.5 ; ModelMinMatThick = MinMatThicknessFromSkeleton * ThickFactor
There are numerous possibilities with this arrangement, but you get the idea. No all we have to do is adjust ModelMinMatThick or ThickFactor as desired within each prototype, and still maintain the core feature parameter limitations. We could even use Min/Max here to enforce the overall prescribed limitations.
Sketch the Critical Radii
Yes I said it, consider building key structural radii in a sketch. Isn’t that what they teach you not to do? The teachers instruct that it’s better to implement fillets in the model where they can be controlled easier. Yes, that’s what they teach, but those teachers were not shaking the model like a ragdoll.
Sketched fillets offer a certain level of safety where careful planning is concerned. We can see exactly where the fillets will fail when they run out of adjacent faces to work from. What is important is that we find out what the safe range of the fillet is, and then stay within that range. Also, construction geometry can help keep the model more stable. One warning though – stay away from projected geometry if possible.
In this example, all of the dimensions are driven by parameters that will be modified to some degree by the Parametric Analysis. Notice the red arc: I know 12mm is near the upper limit for the adjacent fillets. During the numerous parameter changes, many features failed, but these sketched features never did so long as I stayed below 12mm.
Cut the holes AFTER building difficult geometry !!!!!
I ignored this rule,and as a result paid dearly in time. Over and over I fought to keep features from bulging into holes and recesses, and had to apply very time consuming limitations that would keep features stable between changing surfaces and features. If I had rather cut the holes and recesses last, overlaps would no longer have been a problem. That alone would have been a time savings near 30%.This is probably the worst mistake I made early on.
Fillet the part last
Both the analysis mesh and the overall model ease of design fall victim to the complexities of fillets. Stress Analysis will have to mesh twice as many faces (if not more), which causes time of analysis to build exponentially. Fillets run out of space when plowing through numerous iterations, and the model history will fail, over and over. Remember, when performing parametric dimension analysis, you are not in control, but Inventor instead takes over while you go to sleep.
Build the main structure without fillets first, then place these all at the end of the design. Also, when placing the fillets, place the core structural fillets prior to the small edge rounds in the history. When you proceed to analysis, it’s easy to move the end of part maker above all the fillets while evaluating rough design, and later you can move the end of part marker just below the key structural fillets, and evaluate these without dealing with all the little, less critical curved faces.
Can’t say enough how much I like this feature. I started out modeling the entire bracket, but quickly realized this was not how I wanted to work. The complexity and time involved in caring for twice the difficult tasks, as well as fillets that just would not mirror. No more!
I modeled the design at the plane of symmetry and performed all the remaining tasks on one half. Then, of course, mirror solid was added. At points where I would move the end of part marker for my 3 distinct stages, I added a mirror solid. When I was done, I simply deleted it. You could just drag the mirror around as needed, but it’s so easy to add and delete, why make it more difficult?