Using Pink Noise Data to Train Neural Networks for Estimating FPSO Resonant Frequency for Assessing Mooring Line Integrity

This paper proposes a method to extract the system’s characteristic resonant frequency from the horizontal motions of a floating spread moored system based on RNNs (Recurrent Neural Networks). It is assumed that the exciting random force acting on the floating unit does not have a flat Power Spectral Density (PSD) like a white noise; thus, a simplified model for the spectrum is presented and used to produce synthetic data to train the RNN model. Results show that the proposed method performs satisfactorily for the damping interval of interest. A comparison between the RNN method and traditional signal processing techniques is also performed, and the advantages and disadvantages of each are presented. Finally, both methods are applied to the motion obtained from a VLCC-based spread moored FPSO. Since, in this case, the actual resonant frequency is not known a priori, the results of both methods are shown and qualitatively discussed.