Welcome to MCAD Central

Join our MCAD Central community forums, the largest resource for MCAD (Mechanical Computer-Aided Design) professionals, including files, forums, jobs, articles, calendar, and more.

Register Log in

Generative Structural Anaysis


New member
I an no expert on the GSA module of Catia 5 but I thought I knew enough for some simple analysis. I got however some surprisingly inaccurate results. I therefore tested this by a simple1 piece beam with (a) uniform load & (b) concentrated load at the center. The beam is a 4 x 4 sqare section x 200" long. Steel E=29000000 Poisson Ratio =.266 Density .284lbs_"3 Yield Strength= 36260 psi Not that Catia requires this but the I value for the 4 x 4 beam is 21.333 & the section Modulas is 10.669. The beam is 200" long supported at each end. (a)Using convention calculation a 1000lb uniform load max deflection at the center is 0.168" and the max stress is 2347 psi and (b) with a concentrated load of 1000lbs at the center max deflection is 0.269" and the max stress is 4687psi

Using Catia (a) is .0022" max deflection & max stress is 282psi (b) is .004" max deflection & max stress is 433psi

I expected to see no more than a 5% difference. Could anyone explain why there is this huge error using Catia. Perhaps I am doing this all wrong in Catia so if anyone has the time to model this 4" x 4" beam and do an analysis on it I would appreciate the knowledge that I amor am not using theCatia GSA correctly.


New member
if the cross-section is constant, you can draw a line instead of a 3d bar, then create 1d mesh (you can get more accurate results with shorter time)

the smaller meshes are, the more accurate the result is.

surely, you can draw a 3d bar, then create 3d meshes (tetrahedral shape, higher order (one more node between two vertices), initial size should not be bigger than the bar thickness); after getting results with the initial mesh size, try smaller sizes to get more accurate ones. keep making the meshes smaller and smaller, the results should move to certain values. (convergence)

there is no meaning to compare two cases, if you just use "course meshes" and the results have not yet been "converged".