I have been having some trouble on an analysis I am running, let me explain:
I have an assembly whose basic geometry adds up to about 1/3 the total weight of the actual assembly. I have added another 1/3 of weight in point masses which represent the largest components which are inside the assembly. The last 1/3 of the weight is small components, wires, weld material, etc. (things that I do not want to model in as geometry or point masses).
My solution was to add a pattern of point masses along the largest surfaces of the assembly and link them to the rest of the assembly to artificially increase the weight, but still distribute it evenly throughout the system (thus avoiding local stress errors from localized masses).
Without this added blanket mass, my modal response is approx. 100Hz (this is a reasonable value but I need to include the other 1/3 of the total mass). With the blanket mass, the analysis returns values of approx. 1400Hz. This is an unrealistic value which I believe has something to do with how I linked the masses to the surfaces of the assembly (I used rigid links).
So here are my questions:
1. What is the best way to apply rigid links to a pattern of point masses and a surface.
2. Is there a better way to artificially increase the weight of the assembly evenly (I would really prefer to not mess with the densities of the materials I'm using, that will create issues in our system)
Thanks for any feedback,
If you add mass and see your eigenfrequency increase, then something is wrong.
The rigid links add (infinite, local) stiffness to the model, so that is indeed a likely explanation.
1. Best is to add masses without adding stiffness, so either do not use rigid links, of if you have to, tie the rigid links to a very small area of your model so as to add as little stiffness as possible.
2. Increasing density is by far the easiest option I think. Another idea that springs to mind is to add surfaces to the outside of the model, and apply shells with density and low stiffness to those surfaces. But this may well cause other problems, if it is possible at all. I've only tried something similar in Optistruct (=Nastran) to find surface stresses for fatigue in a dynamic response analysis, and at some point it started to complain about it, after which I abandoned this method.
I agree with Patrick to some extent: increasing density is for this matter by far the easiest option.
When working with this kind of model (where a lot needs to be adapted with respect to the actual CAD model, it is best to make a copy of the assembly, preferably on a local disk on your computer. De-attach it from you PLM system. I do this using the "file-save as- save a backup" option. Then you have the possibility to change everything, without getting issues in your system.
A slightly different, but possibly also useful technique is to introduce a new material for your base model, a fictitious material with the higher density.
Second: Patrick is absolutely right about the rigid links , they easily mess up you stiffness (the area you connect becomes rigid).