Cancel
Showing results for 
Search instead for 
Did you mean: 

measurment accuracy ?

Valued Contributor
Valued Contributor

I have a part that is around 4500 cubic inches.  I measure the volume of this part, and then use the density to calculate the weight.

There is an accuracy setting, I believe its:    FILE/PREFERENCES/MEASURMENTS/ACCURACY

and the default is  0.99 , I think it will let you change it in the customer defaults up to a maximum of 0.99999 , at least on the package we have at work.

My question is, changing it between these values, changes my weight by up to 50 lbs.   for example, changing the accuracy from 0.99 to 0.999 will give me a drastically different volume.  This is happening to other people with other parts as well.  Any ideas !?

BWT, this is a low density material, so even a large volume change should only result in a several lb difference, not 50

8 REPLIES 8

Re: measurment accuracy ?

Siemens Phenom Siemens Phenom
Siemens Phenom

Hi @Mystic97z !

 

For review, if you have not done so, you can see how this number i used in the TDocs:

 

https://docs.plm.automation.siemens.com/tdoc/nx/12/nx_help/#uid:xid280739

 

When you calculate the properties and hit the toggle for "Show Information Window" what shows at the very bottom for Error Estimates:

 

 Capture.JPG

 

The above is what I see for a simple shape, similar in size (in^3) to yours, with the following properties:

 

Capture2.JPG

 

Changing my accuracy to .999 barely changes my results.

 

My Density (default steel) might be as low as you are using (e.g. my DB says that a material like LDPE (recyclable plastic #4 in the US)  is only ~0.033 lb/in^3), but in that case my error would even worse than you have.

 

When you say "... then use the density to calculate the weight..." are you allowing NX to calculate for you as I did here?

 

Or are you manually calculating Mass outside of NX? If outside of NX, maybe you dropped a 0 or have wrong units, or some other simple math error when the Volume number is getting transposed? In that case, what does NX say?

Re: measurment accuracy ?

Valued Contributor
Valued Contributor
Hi @Mystic97z !

Would you be kind enough to tell us what to you're using?

Thx

Re: measurment accuracy ?

Valued Contributor
Valued Contributor
Thanks for the response guys, I actually dig into this again today. I am using analysis, measure body. Getting a volume from nx, then by hand using density to calculate weight, but the error is in the software. I am getting 2 different volumes. But here is what I learned today:

I have a solid body that I mirror across a plane of symmetry, imagine I modeled the front wheel on a car and I’m mirroring it across a plane to the rear of the vehicle.

I change my accuracy to .99999

I measure both wheels. One has a volume of 1070 in ^3, the other is 1470.

It’s very strange. I have the front wheel modeled. I mirror that one to the back of the car and the volume changes. For an experiment, I then mirrored the mirrored wheel back up to the front, and then mirrored the mirrored wheel back to the rear of the car. All of the wheels on the front (original and mirror) both have volumes of 1070, while both wheels on the back have a volume of 1470. So it’s something about mirroring across this plane that seems to be causing the issue.

Re: measurment accuracy ?

Siemens Phenom Siemens Phenom
Siemens Phenom

Since it appears that you can reproduce the 'problem', I would contact GTAC and provide them with your model(s).  Have them open an IR/PR and development will take a look at this.

John R. Baker, P.E. (ret)
EX-Product 'Evangelist'
Irvine, CA

Re: measurment accuracy ?

Siemens Phenom Siemens Phenom
Siemens Phenom

Hi @Mystic97z !

 

Thanks for the additional information.

 

I noticed you did not answer the question from @FBergeron : what version of NX?

 

I ask that first because I am aware of a recent PR involved advanced weight management on mirrored bodies, that has recently been fixed (fixed in 12.0.2).

 

You might also want to review GTAC for "properties mirror", which is where I went since I could not quickly locate the above PR number.

 

Among the results, I noticed more than 1 query about why the properties of mirror bodies were not exactly the same (and they never will be, due to slight differences in tesselations and/or interaction methods to get properties), but that's talking about a much smaller difference than you described.

 

There are also mentions of some data-specific cases where specific feature geometry (leading to corrupt parts) showed significant differences. You do not say if you can reproduce this with "any geometry" or just your wheel/tire case, so I'll 2nd @JohnRBaker in suggesting you submit the geometry to GTAC, but before that if it's just this geometry, then perhaps running it through Examine Geometry to ensure all is well with both the original and mirror, would be helpful.

Re: measurment accuracy ?

Gears Phenom Gears Phenom
Gears Phenom
@PatMcManus,

Probably better for the OP to respond but since I know not everyone uses the web interface for posting here, the OP tagged his question with NX10.

As a side note, I tried creating a cube and mirroring it in NX11.0.2 and didn't have any volume differences between the 2 solids. I would expect it to have the errors if it were not geometry specific or at least not fixed by NX11.0.2....hope that makes sense.
Tim
NX 11.0.2.7 MP11 Rev. A
GM TcE v11.2.3.1
GM GPDL v11-A.3.5.1

Re: measurment accuracy ?

Siemens Phenom Siemens Phenom
Siemens Phenom

Thanks for pointing out the NX10 tag @TimF ! Not sure why I missed that!

 

I agree - it would be best to see what the OP is actually doing / exact geometry involved.

 

Ata minimum, I think it will require some free-form/B-Splines in the geometry to really check.

 

When you use an analytical shape, like a block, cylinder, etc... it uses an exact method to calculate properties (you'll notice the error estimates are 0.0000), rather than the iterative method that uses the 0.999, etc...

 

For my quick test yesterday, in an attempt to quickly get 4500 in^in volume,  I also started with a block of 9x20x25 (having forgotten the above, until I saw the "0.0000" error :-), so added more features and blends, etc... until I had an unrecognizable free-form , but close to the volume I wanted, that used the iterative method.

Re: measurment accuracy ?

Valued Contributor
Valued Contributor

I did some more testing and found some wierd information.

 

I have a large complicated model , MODEL A  that creates a fairly complex body.  I mirror that body inside model A, and then wave link it into a clean, empty model B.

 

Volume measured in Model A = 1470

Volume measured of mirrored body within Model A = 1170

volume measured of wave linked body in model B = 1170

 

then, inside model A, I mirrod the body, so it now resides in the origional position, and the volume goes back up to 1470. 

 

I did some mirrors OF mirrors, about 5 times, so by the 5th time, it's a mirror of a mirror of a mirror, etc....

each solid is either 1470 or 1170 depending on that side of the plane it's on.

 

I'm mirroring across the global xz plane.  When I mirror across the global xy, and zy planes, I get 2 more unique volumes.  Makes no sense. 

Again, this is happening in lots of completely different, unrelated models.