Patran Accuracy 1

[johnhors]

A colleague of mine brought the following to my attention:-

"Actually, because Patran is single precision, accuracy is only 6 sig figures (even though can write out 7 figures, the last is not necessarily accurate)

A workaround to problems with significant figures/accuracy is to scale the model larger in Patran so that the coordinate values are defined accurately with 6 significant figures"

MSC Software knowledge database
Product Line: Patran
Solution#: 2900


Am I missing something here ?

 

[EdR]

johnhors:

You may be right....I haven't convinced myself yet that scaling can't buy some improved accuracy if its done correctly and on all applicable values.

Ed.R.

 

[jetmaker]

I ran into this problem just recently while improting *.bdf files back into Patran. This is what was found.

Importing the geometry from CATIA into PATRAN left me with a detailed and accurate geometry which was a couple inches in length. The origin of the reference coordinate system was sufficiently far away from the geometric structure (approx 200"). This geometry was then meshed in PATRAN, and input files (*.bdfs) written for NASTRAN. The *.bdf files were transfered to another engineer, who ran the analysis and then improted the *.bdf back into PATRAN for post-processing. What was found was that the once smooth mesh was now textured, and subsequent runs of the file failed due to bad element geometry. I was puzzled.

My colleague pointed out that to solve this problem, the nodes need to be in a ref. coordinate system which is local to the geometry as the precision is lost due to the limitied number of significant figures being carried in the *.bdf. With very small global edge lengths, (approx 0.01 - 0.05), what was happening is the loss of precision caused nodes to be move relative to eachother due to rounding. This in turn altered the element shape, causing the error messages during analysis.

jetmaker

 

[jbaylor]

If you are modeling an aircraft component with Patran and got the geometry in aircraft coordinates and in millimeters, the lack of significant figures would matter a great deal. With the first four or five significant figures common across the entire model, then there can only be two or three that define your mesh. Scaling does not fix this problem, but translating the geometry to a local coordinate system for modeling does. I generally try to use a coordinate system with the origin inside or on the boundary of the model so all of the significant figures have meaning to the analysis.