In general, the exact probability distribution of a definite integral of a
given non-Gaussian random field is not known. Some information about this u
nknown distribution can be obtained from the 3rd and 4th moment of the inte
gral. Approximations to these moments can be calculated by discretizing the
integral and replacing the integrand by third-degree polynomials of correl
ated Gaussian Variables which reproduce the first four moments and the corr
elation function of the field correctly. The method described (see Ditlevse
n O, Mohr G, Hoffmeyer P. Integration of non-Gaussian fields. Probabilistic
engineering mechanics, 1996) based on these ideas is discussed and further
developed and used in a computer program which produces fairly accurate ap
proximations to the mentioned moments with no restrictions put on the weigh
t function applied to the field and the correlation function of the field.
A pathological example demonstrating the limitations of the method is given
. (C) 1998 Published by Elsevier Science Ltd. All rights reserved.