Prediction of loading spectra under diverse operating conditions by a localised basis function neural network

To estimate the durability and reliability of a structural component its design loading spectrum must be defined. This spectrum is obtained by combining the loading spectra that correspond to all possible operating conditions. It can happen, however, that some of the loading spectra that should be used for building the design loading spectrum are not measured as a result of cost and time limitations. But if a relationship between the operating conditions and the corresponding loading spectra was known, it would be possible to predict a loading spectrum for an arbitrary combination of operating conditions. In this paper, we present a new approach that makes it possible to empirically model the relationship between the factors of the operating conditions and the corresponding loading spectra. This approach is based on a localised basis function neural network, and we have applied it on simulated and measured load states. The presented examples demonstrate that the new approach is suitable for predicting loading spectra for those combinations of operating conditions for which the load states were not measured.