From: Roger Turk <73527.1356(--nospam--at)compuserve.com>
Date: Thu, 17 Jun 1999 13:53:37 -0400
Paul Crocker wrote:
[PC] Roger Turk wrote:
[RT] ASD is not based on "empirical" or "arbitrary" stress limits. ASD is
[RT] based on a factor of safety with respect to first yield. LRFD is based
[RT] on a factor of safety with respect to a fully plastic (yielded) section.
[PC] I'm not sure that this shows that ASD stress limits are not arbitrary,
[PC] because while ASD attaches a specific factor of safety to first
[PC] yield, the choice of the factor of safety was probably arbitrary or
[PC] empirical. How were the factors of safety chosen? Why use 1.67 on the
[PC] gross section and 2.00 on the effective section as in your example. What
[PC] rationale does the ASD system have for these choices? If it is based on
[PC] testing, then it is empirical. If it was chosen based on the judgment
[PC] of practicing engineers or academic, then it is arbitrary.
The factor of safety for ASD comes from the same place that the factor of
safety for LRFD (load factors divided by phi for members in pure tension)
comes from: from testing of material.
I can ask you the same questions about the different factors of safety
required in LRFD for tension on gross and effective net sections (1.57 and
1.87 respectively) that you asked about ASD, but I think that the answer is
obvious. In the net section analysis, you have discontinuities, which cause
stress concentrations, which create uncertainties, which require a greater
factor of safety.
Maybe we need to look at the way load factors for concrete USD were
developed. ACI 318-56 has:
U = 1.2 D + 2.4 L or,
U = K(D + L)
D = dead load plus volume change due to creep, elastic action,
shrinkage and temperature, and
K = 2 for columns and members subject to flexural and axial loads.
= 1.8 for beams and girders subject to bending only.
How were these values determined. These were based on "calibration"
(another euphemism) of the design requirements so that the resulting design
was not very much different than what was obtained with WSD. (Wasn't PCA
smart to realize that USD wouldn't be accepted if the resulting design was
very different from what was done previously?)
In ACI 318-63, the load requirements were:
U = 1.5 D + 1.8 L
The 1963 code also introduced phi for the first time and describes it as:
"The coefficient [phi] provides for the possibility that small adverse
variations in material strengths, workmanship, dimensions, control, and
degree of supervision, while individually within required tolerance and the
limits of good practice, occasionally may combine to result in undercapacity."
Now, the factors of safety (load factors) in 1956 were 1.2 for dead load and
2.4 for live load, or, 2 or 1.8 on the total load. In 1963, the factors of
safety were 1.5/phi for dead load and 1.8/phi for live load.
The explanation that was given for the different load factors was that dead
loads can be determined rather well, so a smaller factor of safety can be
applied to the dead load. However, live loads cannot be determined as
accurately as dead loads, so a larger factor of safety is applied to live
This sounds rather logical, until one realizes that the material does not
know, or care, whether it is being loaded with dead or live loads, only what
the total load is that is being applied.
What was the purpose of reducing the load factors? Why, to compete with
steel, of course. Why are the steel load factors the way they are? Why, to
compete with concrete, of course.
[PC] You quite rightly pointed out that you can go back through the LRFD
[PC] calcs and determine what the corresponding ASD factor of safety would
[PC] have to be to get the same results.
No, I didn't go thru the LRFD calcs and determine what the corresponding ASD
factors of safety were. I went thru the LRFD calcs and showed what the
actual factors of safety were in doing LRFD design.
[PC] LRFD calcs ideally lead you to a system with a known (or estimated)
[PC] probability of exceedance rather than a factor of safety with an unknown
[PC] probability of being exceeded. This creates a certain uniformity of
[PC] logic in design, since probability of exceedance is a central principal
[PC] in calculating code wind pressure, seismic forces, rainfall amount, and
[PC] probably many other things that I've never worked with.
The LRFD provisions have *nothing* to do with "probability of exceedance."
They probability that they deal with is demand (based on total load
irregardless of load factors) vs. capacity and that, except for a theoretical
3 percent overlap, the capacity is greater than the demand. The loads,
particularly wind and seismic loads, whether they are used in ASD or LRFD,
can be determined by "probability of exceedance (or non-exceedance)."
Paul Meyer wrote:
[PM] I find that using different load factors for different types of loads
[PM] makes a lot of sense. Why would I assign the same so-called "safety
[PM] factor" to loads as diverse as snow drift loads (which are basically
[PM] wild guesses carried to two significant digits...) and storage tank
[PM] loads where I know the mass of the contents exactly?
Why not use the same factor of safety (load factor) as the material
recognizes only the total load applied to it doesn't care whether it is
composed of dead, live, wind, or seismic loads or a combination of them.
A. Roger turk, P.E.(Structural)