Journal of Wind Engineering and Industrial Aerodynamics, 12 (1983) 385--387 385
Elsevier Science Publishers B.V., Amsterdam -- Printed in The Netherlands
Short Communication
A SIMPLIFIED CALCULATION METHOD FOR GUST FACTORS
c.J. WOOD
Department of Engineering Science, University of Oxford, Parks Road, Oxford (Gt.
Britain)
(Received March 25, 1983; accepted May 16, 1983)
Summary
The gust-factor analysis of Greenway is reviewed and empirical formulae are proposed
to replace two tabulated functions in the analytical prediction of the probability distribu-
tion for the largest short-time-averaged gust in a specified observation period. The result-
ing method is suitable for rapid use with a hand calculator.
A familiar characteristic of the turbulent wind is that the highest short-
duration gust speed V is greater than the mean speed V by an amount which
increases as the gust-averaging time Tar becomes short and the high-frequency
peaks are not filtered off. This difference is also increased if the observation
period Tobs becomes large, because the peak value V is itself a randomly
fluctuating variate and an extended observation period increases the prob-
ability that a higher value of V may occur. In fact, V can be described only
through an extreme-value probability function which may be expected to
have the Fisher--Tippett Type 1 form
= u + (I/a) {-In[-In(l-P)] } (I)
where P is the probability that a given value V will be exceeded at least once
in any future observation period Tobs.
An analytical study of the problem of predicting gust factors has been
published in this journal by Greenway [1]. In that analysis, the windspeed
V(t) was represented as a variable having a Gaussian probability distribution
with variance aw 2 and a v o n Karman power spectrum 8w (n) given at any
frequency n by
nSw(n)/ow 2 = 4 ~/(1 + 70.8 ~2)s/s (2)
where fi = n X L u / V
and xLu is the longitudinal integral length scale of the turbulence.
When the windspeed signal V(t) is filtered with an averaging time Tar (or
a 3 db cut-off frequency, 0.44/Tar), the more rapid fluctuations disappear
and the von Karman power spectrum Sw (n) is modified accordingly. In
particular, the zero-crossing frequency of the filtered signal is reduced to a
0304-3908/83/$03.00 © 1983 ElsevierSciencePublishers B.V.
386
value v given by
v X L u / V = F(TavV/x Lu) (3)
and the variance of the signal is reduced from Ow 2 to o 2, where
O/Ow = V(TavY /XLu) (4)
The principal contribution made by Greenway [ 1] to this analysis was
the numerical computation of the complicated functions F and G from the
filtered yon Karman spectrum, and their tabulation for a range of values
of the dimensionless averaging time TavV/x Lu*. Then, using the earlier work
of Davenport [2], he was able to derive expressions for the mode u and the
dispersion 1/a (the defining constants in the Fisher--Tippett probability
function (eqn. 1)) as follows:
u / V = 1 + ( o w / V ) V [ 2 Ln (FVTobs/XLu)] ~'4 (5)
1/a V = ( o w / V ) V / [ 2 In (FVTobs/X Lu) ] 1/2 (6)
Clearly, given suitable data for the turbulence intensity a u / V and the
longitudinal length scale x Lu as published for example by ESDU [ 3], these
expressions are very easy to evaluate using Greenway's tabulated values for
F and G. Also, being correctly formulated in terms of dimensionless obser-
vation- and averaging-time parameters, the results are superior to the gust-
factor data currently offered by ESDU [4].
In order to simplify the calculation still further and eliminate the need
to interpolate the rather sparse tabulations of Greenway [ 1 ], the present
short note offers two empirical formulae to replace the tabulated values of
F and G. These are
F = 0.0066 + 0.2130 (TavV/XLu) -°'6s4s (7)
V = 1-0.1925 [(XLu/TavV ) + 0.1] -°'6792 (8)
These functions are compared with the exact computations of Greenway
in Table 1. This comparison shows that the error in using the proposed
empirical formulae is less than 2% over the range of averaging times con-
sidered. Further refinement of eqns. (7) and (8) is not justified in the con-
text o f gust prediction, because the Greenway theory is already limited by
the somewhat unrealistic assumption of a Gausaian probability distribution
to represent the parent windspeed.
The proposed functions for F and G, when used in conjunction with eqns.
(5) and (6), provide all that is necessary for the direct calculation of the ex-
treme gust probability function in terms of the intensity and length scale of
the wind turbulence.
*Greenway actually expressed the functions F and G in t e r m s o f the parameter TavV /
4.5XLu, for r e a s o n s w h i c h are n o t i m p o r t a n t here.
387
TABLE 1
Comparison of eqns. (7) and (8) with Greenway's exact computations of the functions F
and G
TavV/XL, F F(eqn. (7)) G G (eqn. (8))
0.002 12.608 12.432 0.997 0.997
0.004 7.956 7.901 0.995 0.006
0.006 6.079 6.061 0.994 0.994
0.010 4.335 4.341 0.992 0.992
0.015 3.316 3.331 0.989 0.989
0.02 2.743 2.761 0.987 0.987
0.04 1.740 1.757 0.979 0.978
0.06 1.335 1.349 0.973 0.972
0.10 0.958 0.968 0.961 0.960
0.15 0.738 0.744 0.949 0.948
0.2 0.614 0.617 0.938 0.936
0.4 0.395 0.395 0.901 0.899
0.6 0.306 0.304 0.871 0.869
1.0 0.222 0.220 0.820 0.820
1.5 0.172 0.170 0.767 0.769
2.0 0.143 0.142 0.723 0.728
4.0 0.093 0.093 0.597 0.607
6.0 0.072 0.073 0.517 0.528
10.0 0.054 0.054 0.420 0.426
15.0 0.043 0.043 0.350 0.350
20.0 0.037 0.037 0.307 0.302
References
1 M.E. Greenway, An analytical approach to wind velocity gust factors, J. Ind. Aerodyn.,
5 (1979) 61--91.
2 A.G. Davenport, Note on the distribution of the largest value of a random function
with application to gust loading, Proc. Inst. Civ. Eng., 28(1964) 187.
3 Engineering Sciences Data Unit, Characteristics of atmospheric turbulence near the
ground, Data Item 74031, Eng. Sci. Data Unit, London, 1974.
4 Engineering Sciences Data Unit, Characteristics of wind speed in the lower layers of
the atmosphere near the ground, Data Item 72026, Eng. Sci. Data Unit, London, 1972.
