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

Re: ASD vs. LRFD

[Subject Prev][Subject Next][Thread Prev][Thread Next]
Christopher:

I believe I can clarify my points further.  My comments are inserted.

> I'll agree that seismic loading is a gray area, and that plastic
> response often needs to be assessed. I daresay playing around at the
> margins of  real honest to god non-linear dynamic analysis isn't
> going to help much. Not unless someone chooses to risk his life and
> publically propose time history seismic analysis. So we're left with
> response spectra methods and a sort of quasi-linear approach. Like
> we've always used. 

MTV: Although we're using a "quasi-linear" approach, the real key to
providing a structure that performs well is understanding what will
control the behavior of the structure.  Because the ASD safety factors
are hidden from the end user and they differ for various limit states,
it may not be apparent to the designer whether tension, compression,
flexural yielding, lateral-torsional buckling, connection capacity, or
some other phenomenon is REALLY the controlling condition.  The
designer is often left simply saying, "I'm not sure what controls, but
it meets the code.  So it must be okay."

> The wind loading is puzzling because it seems like you're just
> playing around with the ratio of lateral load to dead load, neither
> of which should be affected by the method of analysis. Hurricane
> Andrew blew just as hard for structures designed by limit methods
> and for elastic analysis. Or have I missed something. Seems to me
> that I should be designing against the same thing with linear
> elastic methods, simply by using the same proportion of lateral to
> dead load.

MTV: My example shows a difference in design requirements (that stem
from the design approach) not a difference in the method of analysis.
(Even when LRFD is used, in most instances an elastic analysis is
performed.)  In the ASD method we attempt to account for the
variability of both demands and capacities by applying a single safety
factor on the capacity side of the equation.  That means that if a
member is controlled by loads that are difficult to predict accurately
(like snow or live load) it will be more likely to fail than will a
member with more easily predicted loads.  As a result of the design
approach (all the "safety" on the capacity side), the designer is
blinded to the fact that his calculation of the dead load is not
perfect and that the wind load will not be exactly the number
predicted; the REAL ratio of lateral to dead load is not certain.  
Because the design specification (ASD) doesn't require that he 
consider the fact that his "precise" dead load calcs can be off (by 
say 10% to 20%) and that the calculated wind loads really are 
uncertain (by say as much as 30%) he misses that fact that the member 
might experience loads of the opposite sense.

You're right that the method of analysis doesn't control the 
magnitude of the loads that nature will apply.  However, taking ASD
load combinations at face value can lead designers to the unfounded
conclusion that they have considered the "real" load that a structure
will experience.  The ASD designer thinks the dead load is 100 kips;
the LRFD designer thinks the dead load is likely between 90 and 120
kips.  Considering the fact that our calculations may not be perfect
can result in more forgiving designs.

> And the comment about service loads really doesn't address 'better'
> as much as 'different.' seems like you could accomplich the same
> thing by doing the design based on service loading and then checking
> for collapse mechanisms afterward, rather than figuring collapse
> mechanisms first and then doing the deflections. Besides the very
> real cumulative effects of service loading, my guts are telling me
> that we're still using all the linear asumptions (notably
> superposition) and pretending that they hold for non-linear
> response. 

MTV: My comments regarding serviceability checks were to dispel the
common fallacy that such checks are only important if LRFD was used. 
You might also note that many aspects of the LRFD approach still
satisfy the linear assumptions (and thus qualify for superposition);
the key difference is where we provide the "safety" or "reliability".
More on that below.

> I'm a little reluctant to get more and more involved in this since
> service loads govern the stuff I do, and it'll be ever thus.
> Moreover the problems I hear about on this list seem service
> based--very few discussions involving ultimate collapse. I wonder if
> my uncle felt the same way when he came back from France in 1919 to
> find a popular majority had somehow passed a law so wildly unpopular
> (and culturally irrelevant) as prohibition.

As I see it, the two strengths of the LRFD method are 1) it allows us
to think about the real capacity, and 2) it forces us to recognize
that our calculations are not perfect.  If all of your designs will
always remain elastic (ie, no earthquake or blast effects), then the
first point may not be important.  However, because the demands and
capacities in real structure do vary from our design assumptions, it
is always important to consider the consequence that our calculations
may not perfectly reflect the REAL service conditions.  In my example,
being slightly off on the dead and wind loads may mean that members
are subjected to forces of a sense that was not anticipated; using a
larger safety factor on tension checks for a member in compression
does not work.

Some design for "ultimate" conditions is allowed in ASD and many
facets of LRFD design revolve around service conditions and linear
response.  The central difference in the two methods is that LRFD
applies Factors to both Loads and Resistances and ASD applies factors
to capacities to arrive at Allowable Stresses.  By "showing" 
designers the factors and separating demand and capacity effects, 
LRFD should provide a better understanding of the likely performance.

-Mike

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Michael Valley                                   E-mail: mtv(--nospam--at)skilling.com
Skilling Ward Magnusson Barkshire Inc.                  Tel:(206)292-1200
1301 Fifth Ave, #3200,  Seattle  WA 98101-2699          Fax:        -1201