Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
APPENDIX. REFERENCE
We thank Prof. Hays for his discussion of our paper and give below our
comments to the points raised by him.
Table 3 shows the various steps of calculating the Dd values by our method
for six bridges covering the ranges given in Table 2 by Prof. Hays. Table
3 also lists the values of fraction S/Dd> called LLDF by Prof. Hays, obtained
by our method and those given by SALOD and read from Fig. 16. It can
be seen that the differences between the two sets of values of LLDF are not
as large as shown by Prof. Hays in Fig. 16. With the exception of the 20-
ft-span bridge, these differences are well within the reading accuracy of the
charts used in obtaining the results. The LLDF value for the 20-ft-span bridge
given by our method is considerably smaller than that obtained by analysis
with SALOD; we suspect that this discrepancy is due to the latter analysis
corresponding to a bridge that has much stiffer girders than would normally
be employed in practice for a bridge of such short span. It is not usual to
use AASHTO type II girders for a span of 20 ft. We are pleased to note
that the results of our method are in good agreement with those obtained by
the more rigorous finite element method. As a matter of interest, it may be
noted that the kind of idealization used by Hays et al. (1986) can be analyzed
i i i
I S/2 I S I
{•« * T * *1
The recommendation given in our paper regarding the equivalent span lengths
for obtaining D for moments in the negative moment regions, is based upon
studies reported by Bakht and Jaeger (1985). These studies, which were
conducted through grillage analyses, confirmed that the recommendation,
while being quite approximate, should lead to errors that are on the safe
side. We note that the current load models for obtaining responses in the
negative moment regions are not as well developed as those for positive
moment regions in simply supported spans. Because of this lack of accuracy
in the load model itself, increased accuracy in the live load distribution anal-
ysis of negative moments may not be warranted.
Errata
The following corrections should be made to. the original paper:
p. 1797, Fig. 11(b): 10 should read 0; 20 should read 10.
871
Downloaded 29 Jun 2009 to 131.84.105.87. Redistribution subject to ASCE license or copyright; see http://pubs.asce.org/copyright
APPENDIX. REFERENCE
Ocean structures, situated in exposed areas for years, have to resist mil-
lions of loading cycles produced by the action of wind waves in lifetime.
Structural fatigue is important and there is a close relationship between the
loading time series on structural elements and the time series of sea surface
elevation. A comparison between the methods used to analyze sea wave
records and the methods used by the authors may be useful to evaluate the
results.
The authors generated the loading time series with Rayleigh-distributed
random variables in accordance with a first-order autoregressive technique
with a prescribed correlation of extremes, p. This model is, in some ways,
similar to those presented by Kimura (1980), and Longuet-Higgins (1984).
Medina and Hudspeth (1987) provided an analysis technique of waves in
the complex plane that is useful for analyzing the loading time series used
by the authors. The envelopes of Gaussian time series are Rayleigh distrib-
uted. They are approximately first-order autoregressive processes with pos-
itive parameter, |p|. The time series of extremes used by the authors can be
obtained by sampling, at intervals Af = T01/2, the positive and negative
envelopes of the Gaussian time series, in which T01 is the mean period or
the time for two loading cycles. For the model proposed by the authors, the
time dimensional and dimensionless spectral peakedness parameters (Tv, Qe)
are given by
SA(0) Arr ^ ^ m l\ + | P |\
- y - = 4r v = T01Qe = T01 ^ (15)
"A \ 1 - IPI/
in which SA(0)/a2A = the value of the spectral density function (unity vari-
ance) of the envelope at zero frequency.
The coefficient of variation of the root-mean-square of the time series,
Xrms, can be approximated by