Dynamic analysis of mooring line systems under various environmental conditions including waves, currents and winds requires computationally expensive time domain analysis. In this paper, Nonlinear Autoregressive with Exogenous input (NARX) model was used to predict the time histories of a mooring line top-tension. The dataset for network training was obtained from the nonlinear analysis of the mooring system under random ocean waves of different sea states. To check the validity of the proposed method, the top-tension of a mooring line under the variety of different ocean wave environment with different significant wave heights and modal periods was predicted using the proposed method and compared with the nonlinear time domain analysis results. The predicted time series of the top-tension of a mooring line in different sea states has good correlation compared with the direct analysis results This method can be used to predict the top-tension of a mooring line without the computationally demanding nonlinear time domain analysis.
Due to the increasing demand in energy, exploitation of oil and gas in deep waters has been gained more importance. Exploration of oil in deep waters needs floating production systems connected to mooring lines and risers such as FPSO (Floating Production, Storage and Offloading) and semi-submersible platforms. Mooring lines that keep a floating offshore structure in position are required to comply with safety limits such as tensions. Engineers use nonlinear time domain finite element analysis to obtain the tensions in different sea states. Short term dynamic analysis which is approximately 3 hours for each sea state takes too much time because of considering the hundreds of wave spectra. To have meaningful solutions in rapid mode with high accuracy, researchers develop prediction method for time domain analysis. Sagrilo et al. (2000) focused on the development of a high effective practical approach to assess the short-term extreme response statistics of flexible risers excited by the first-order heave motion of a floating unit. The analysis of three different flexible risers configurations illustrated the accuracy and the robustness of this approach to calculate the extreme response statistics. Billings and Wei (2005) proposed a new class of wavelet networks for nonlinear system identification. In the new networks, the model structure for a high-dimensional system was chosen to be a superimposition of a number of functions with fewer variables. Mazaheri and Downie (2005) used a response-based method as a reliable alternative approach. In order to perform the calculations faster using large databases of sea states, ANN was designed and employed. The response based method was applied to a 200000 dwt FPSO and obtained results were discussed in the case study. Mahfouz (2007) described a new method to predict the Capability-Polar-Plots for offshore platforms using the combination of the artificial neural networks and the Capability Polar Plots Program (CPPP). The estimated results from a case study for a scientific drilling vessel were presented in his study. Guarize et al. (2008) indicated a very efficient hybrid ANN-Finite Element Method (FEM) procedure to perform a nonlinear mapping of the current and past system excitations (inputs) to produce subsequent system response (output) for the random dynamic analysis of mooring lines and risers. Yasseri et al. (2010) proposed a predictive method for identifying the range of sea-states considered safe for the installation of offshore structure. They made finite element analysis for different sea-states with characterization of significant wave height and mean zero-up crossing wave periods. They identified the range of sea-states suitable for safe pile-driving operation as a case study. Pina et al. (2013) presented a new surrogate model based on artificial neural networks (ANN) to evaluate the response of mooring lines and risers expeditiously. Their aim was to get result as accurate as finite element method based dynamic analysis in shorter time. Queau et al. (2014) aimed to test the robustness of their previous researches and extend the ranges of the input parameters for steel catenary riser systems under static loading, by means of numerical simula-tions. An approximation using a series of neural networks was presented; it successfully approxi-mates over 99% of the cases of the database with an accuracy of ±5%. Jacob et al. (2014) proposed an approach based on Wavelet Networks (WN) - a combination of the feed-forward neural network architecture with the wavelet transform. The goal is to obtain dramatic re-ductions in processing time, while providing results nearly as good as those from non-linear dynamic fi-nite element methods Kim (2015) predicted the dynamic response of a slender marine structures under an irregular wave with NARX based quadratic Volterra series. He compared the predicted time series of the response of structure with quadratic Volterra series and nonlinear time domain simulation results. Volterra series results coincided with simulation results.
