FATIGUE DAMAGE PREDICTION VIA NEURAL NETWORKS

This paper introduces a new methodology for fatigue evaluation of tubular offshore structures. The accumulated fatigue damage on selected joints, caused by actual wave loading condition, will be assessed through a data acquisition and processing system, based on a PC microcomputer , installed onboard. By measuring the upward jacket displacements and applying feedforward backpropagation trained neural networks, the calculation of fatigue damage is performed in-time. This information, along with inspection reports, could be used to improve further inspection and repair of the platform structural joints. This kind of procedure can also be applied to other types of offshore structures on which the fatigue damage contributes to the theirs failure modes, as FPSO and TLP. Basically, it is necessary to develop an updated dynamical computer model of the structure and to correlate some points to be measured to the fatigue damage on selected places.