A neural network approach to the simulation of load histories by considering the influence of a sequence of rainflow load cycles

A modeller was developed that simulates random load histories by considering a statistical distribution and a sequence of load cycles that correspond to a rainflow algorithm. The random load histories are generated stepwise by putting together series of load cycles, successively predicted by the modeller. The core of the modeller is a conditional probability density function of load cycles, which is modelled by a multilayer perceptron. The theoretical foundations of the modeller and three examples of its application for the simulation of random load histories are presented in the paper. We concluded from the examples that it is possible to simulate new random load histories on the basis of the load cycles, extracted from the original load history with the rainflow method, so that the statistical distribution and the sequence of load cycles from the original load history are preserved in simulated load histories.