Prediction of dynamic responses of floating structures using NARX with mirroring technique

Abstract

Displacements, velocities and accelerations of Six Degree of freedom of a single floating structure was predicted using Time Series NARX feedback neural Networks. The nonlinear autoregressive network with exogenous inputs (NARX) is a recurrent dynamic network, with feedback connections enclosing several layers of the network is based on the linear ARX model, which is commonly used in time-series modelling is used in this study. Time series data of displacements of a single floating structure was used for training and testing the ANN model. In the training stage, this time series data of environment parameters was used as input and dynamic responses was used as target. Benchmarking result and error prediction was compared between two techniques of Neural Network training. The prediction result of the model responses can be concluded that NARX with mirroring technique increase the accuracy and can be used to predict time series of dynamic responses of floating structures.

Data-Driven Modeling Approach for Mistuned Cyclic Structures

“…The applicability of data-driven methods, such as machine learning and artificial neural networks (ANNs), has received growing interest in recent years for modeling structural dynamic systems due to their ability to approximate nonlinear functions with good accuracy [18][19][20][21]. Previous studies have focused on predicting time series data for structural health monitoring and fatigue analysis of structures, such as wind turbines subject to nonlinear loading [22], buildings subject to base excitation similar to earthquakes [23], floating structures [24], or engine vibrations [25,26]. These studies have used recurrent neural networks (RNNs) [22], and more specifically long short-term memory networks (LSTMs) [25,26], nonlinear autoregressive networks with exogenous inputs [24], or generalized regression neural networks [23] that use states and/or parameters at previous states to predict the current state.…”

“…Previous studies have focused on predicting time series data for structural health monitoring and fatigue analysis of structures, such as wind turbines subject to nonlinear loading [22], buildings subject to base excitation similar to earthquakes [23], floating structures [24], or engine vibrations [25,26]. These studies have used recurrent neural networks (RNNs) [22], and more specifically long short-term memory networks (LSTMs) [25,26], nonlinear autoregressive networks with exogenous inputs [24], or generalized regression neural networks [23] that use states and/or parameters at previous states to predict the current state. More recently, convolutional neural networks traditionally used for pattern and object identification [27][28][29] have been applied in the time domain for modeling the structural dynamics of single-DOF and small multi-DOF lumped parameters systems [30].…”