STOCHASTIC FINITE ELEMENT METHOD: Response Statistics
4.1 Reliability Theory Background
An important objective of a stochastic finite element analysis of an engineering system should be the determination of a set of design criteria which can be implemented in a probabilistic context. In other words, it is important that any method that is used for the numerical treatment of stochastic systems be compatible with some rationale that permits a reliability analysis. Traditionally, this problem has been addressed by relying on the second order statistical moments of the response process. These moments offer useful statistical in- formation regarding the response of the system and equations will be developed in an ensuing section for their determination. How- ever, second order moments do not suffice for a complete reliability analysis. For this, higher order moments and related information are required. In this context, the response of any system to a certain excitation may be viewed as a point in the space defined by the parameters describing the system. In other words, for every set of such parameters, the response is uniquely determined as a function of the independent variables. When these parameters are regarded as random variables, the response function may be described as a point in the space spanned by these variables, defining a surface that
R. G. Ghanem et al., Stochastic Finite Elements: A Spectral Approach
corresponds to all the possible realizations of the random parameters.
The case where the parameters are modeled as random processes may be reduced, as discussed in Chapter II, to the case of random variables. For the safe operation of a given system, it is usually required that the response, or some measure of it, be confined to a certain region in this space, termed the safe region. Formally, this condition can be expressed as
(4.1)
where (4.2)
is the equation of the failure surface in the e-space. Of special
interest in reliability assessment is the failure probability, Pj. That is, the probability that equation (4.1) is not satisfied. This equals the volume, with respect to some probability measure, of the region in the e-space over which (4.1) is not satisfied. That is,
(4.3)
where le(e) is the joint probability distribution of {6, ""~r}' The
difficulty in using equation (4.3) in engineering practice is obvious. Specifically, the domain of integration is rarely available in an analyt- ical form. Further, the dimension of the ~-space is usually quite large, to the extent that the requisite integration can seldom be carried out either analytically or numerically (Schueller and Stix, 1987). Finally, realizations of the set {~i} are often scarce and do not permit infer- ence about their joint distribution beyond their joint second order moments. Grigoriu et.al (1979) have suggested treating the selection of this joint distribution as a problem of decision theory and proposed a number of optimality criteria to resolve this issue. The problem remains, however, of efficiently implementing these distributions into numerical algorithms. This is a matter that persistently poses severe limitations on the amount of usable probabilistic information. Given these computational and theoretical difficulties, it became desirable to introduce a standard measure for reliability analysis. Such a measure would be based on readily available second moment