Description of statistical dependencies of parameters of random load states (dependency of random load parameters)

In order to predict fatigue life of structures under random loading, the pr obability distribution of parameters of random load states of structures sh ould be determined, given that the material endurance is known and a proper damage rule is chosen. In this article a statistical dependency of two par ameters of random load states (the amplitude and the mean of load cycles) i s estimated by means of the multivariate p.d.f. of rainflow load cycles tha t are represented as vectors S=(S-a,S-m). The multivariate p.d.f of load cy cles is modelled with a mixture of multivariate normal functions. A compari son of different methods for an estimation of the unknown parameters of the normal mixture is carried out. Agreement of numerically modelled and exper imentally obtained data is checked by comparison of corresponding marginal probability distributions of load cycle amplitudes and means. The maximum l ikelihood method was recognised to be the most appropriate for the determin ation of the multivariate p.d.f. of load cycles. (C) 2000 Elsevier Science Ltd. All rights reserved.