Need a book? Engineering books recommendations...

Return to index: [Subject] [Thread] [Date] [Author]

RISA-3D Torsion, etc. (Response to Greg Luth)

[Subject Prev][Subject Next][Thread Prev][Thread Next]
I was trying to stay out of this discussion of SAP, RISA-3D, STAAD, etc., but
the post from Greg Luth demands a response.

Greg wrote:

RISA was a very good product when it first came out, but we have had 
 problems with it lately.  The first problem had to do with torsion on a 
 wide flange shape.  We were analyzing a steel structure that relied on 
 torsion in several of the beams for strength and stiffness.  The results 
 from RISA suggested that a wide flange beam was as effective as a tube.  On 
 further investigation, it turned out that this was the case only when you 
 used the RISA database shape.  If we input all the properties, including 
 "J", manually, the results were substantially different.  Our conclusion is 
 that RISA is using warping torsion properties for all torsion problems, 
 even when they are not appropriate, as for long beams where the warping 
 "boundary" affect has little influence on the overall behavior.  This bug 
 is exceedingly dangerous because even with short beams, warping torsion can 
 only be used if the bi-moment is resisted at the support.

We've based our warping calculations on "Torsional Analysis of Steel Members",
published by the AISC. This document is by no means comprehensive; it presents
equations for several idealized cases, and we've extended these equations to
general frame analysis. We treat all warping members (defined by AISC as open
members whose sub-sections do not share a common point, i.e. channels and wide
flanges) as "case 2" members where warping is considered to occur when both
ends of the member are fully fixed and thus considered to be restrained
against warp (all this is in the RISA-3D documentation).

The AISC equations are not "length limited", though the warping stiffness and
stress equations, too complex to present here, are dependent on length and a
warping member becomes much less stiff as it's torque length increases. This
equation is based on several factors in addition to J so just modifying J is
essentially meaningless. When you say "we input all the properties manually",
do you mean you also input Cw, Wno and Sw (in addition to A,I,J), because if
you didn't, of course the results are different.

Inclusion of warping torsion in a general frame analysis is very difficult.
This is why almost every program on the market completely ignores this effect.
RISA-3D, however, does not. I would have to see Greg's model to comment on his
results specifically, but we document quite clearly how we handle warping in
the RISA-3D user's manual and we have multiple verification problems that
prove we do what we say we do, and I'm confident that in Greg's case RISA-3D
was doing exactly what it was supposed to do. Also remember, Greg, you have
the option of turning torsional warping effects OFF altogether, in which case
the RISA-3D beam element will be degraded to the same level of accuracy as
other programs.

I think if you take the time to read and understand RISA-3D's implementation
of torsional warping you'll realize there is no "dangerous bug" and RISA-3D is
doing a good job in modeling an exceedingly difficult effect, which is much
better than just ignoring it completely. Again, if you don't agree with what
RISA-3D is doing, just turn warping OFF.

 We've also gotten strange results when we tried to use the plate elements 
 in RISA3D.

As for your comment on RISA-3D's plate element, could you be a bit more
specific? Again, we've extensively documented and tested this element, and
we've provided verification in the RISA-3D manual. Could it be that you had a
poorly defined model instead (as is usually the case)? As before, it's
difficult for me to answer this specifically with actually seeing these
"strange results".

Also, as much as it pains me to do so, I have to defend STAADIII. Your comment
regarding modeling of diaphragms in STAADIII indicated that you attempted to
do this via "MASTER-SLAVE" connections. This is absolutely the wrong way to
model diaphragms, in STAAD or in any other program.  I'm sure your results
were very "strange" indeed due to the fact that your model was strange (and
 >>My advice is to treat all results from analysis programs as wrong until 
>> proven right.

Finally, something I can agree with.

Bruce R. Bates
RISA Technologies