P O I N T - E S T I M A T E M E T H O D FOR C A L C U L A T I N G

STATISTICAL MOMENTS a
Discussion by Krishnaiyengar Rajagopalan 2

The extension of the point-estimate technique to correlated variables by

related nonnormal variables becomes direct using Li's method via the skew-
ness and kurtosis of the respective distributions. (It may be noted that the
normal distribution has a zero skewness and a kurtosis of 3.) The discusser
would like to ask whether these two parameters (third and fourth central
moments) are sufficient to obtain an accurate value of the reliability. Ex-
ample 6.10 in Ang and Tang (1984) can perhaps be used as a vehicle to
demonstrate the efficacy of the present contribution.
The calculation of the reliability is currently carried out by the advanced
first-order second-moment (AFOSM) method in which the correlated var-
iables are transformed into uncorrelated variables using a technique of ro-
tation of coordinates similar to the calculation of principal stresses at a
material point. This method is, however, correct only for normal variables.
An equivalent normal distribution is then used to transform the problem
into the space of uncorrelated standard normal variables. The reliability
index is then computed from the minimum distance of the failure surface
from the origin.
The current version of point-estimate method (PEM) by Rosenblueth
(1975, 1981) can be used with some of the foregoing transformation ideas.
However the paper by Li seems to indicate that PEM can be directly used
for the calculation of reliability with correlated nonnormal variables. The
discusser would like the author to throw some light on this.
In the last term of (17) of the paper, the double summation appears to
be missing.

APPENDIX. REFERENCE
Ang, A. H. S., and Tang, W. H. (1984). Probability concepts in civil engineering
planning and design. Vol. II, John Wiley, New York, N.Y., 368-372.

Closure by K. S. Li, 3 Member, ASCE

The writer thanks the discusser f o r his interest in the paper. The point-
estimate method (PEM) proposed by the writer is developed for calculating
the statistical moments of a random function. It does not lend itself to direct
application in the advanced first-order second-moment (AFOSM) method
based on the concept of reliability index. However, the PEM can be used
for estimating the reliability of a system using a new procedure completely
different from the AFOSM method. Although the new approach is intended
to be the subject of a future publication, a brief description of the method
is given below to illustrate the application of PEM.
Consider a performance function Z = G(X, Y) of two random variables
"July, 1992, Vol. 118, No. 7, by K. S. Li (Technical Note 1297).
2Assoc. Prof., Oc. Engrg. Ctr., Indian Inst. of Technol., Madras 600 020, India.
3Assoc., Greg Wong and Associates, 8/F, Seabright Plaza, North Point, Hong
Kong.
414

J. Eng. Mech. 1994.120:414-414.