1988 年 1988 巻 398 号 p. 109-119
Probabilistic safety analysis methods for structural systems involving spatially variable material properties are studied. For the cases to which the first-order second-moment methods cannot be applied well, two different Monte Carlo methods are employed in conjunction with the finite element technique. “Monte Carlo simulation”, which is based on the stochastic field simulation technique and the safety judgment for each trial, is recommended because its accuracy is guaranteed irrespective of the number of random variables. On the contrary, the accuracy of “Monte Carlo integration”, where a numerical integration of the joint density function of the random variables is carried out by means of sample-mean Monte Carlo method, goes down dramatically as the number of random variables getting large. Numerical examples verify the above argument and reveal the effects of correlation parameters on the local and system limit state probabilities.