An Investigation of the Performance of the ANN Method for Predicting the Base Shear and Overturning Moment Time-Series Datasets of an Offshore Jacket Structure

Abstract

The primary purpose of the current study was to investigate the performance of artificial neural networks to predict the time series of the water surface level (WSL), base shear, and overturning moment using two types of ANN models: Nonlinear Autoregressive models with exogenous inputs (NARX) and Nonlinear Autoregressive models (NAR). After determining the suitable model, NARX, the possibility of predicting the time series of the base shear and the overturning moment data was investigated by considering the water surface level and time as the multivariable model inputs. A jacket model with a height of 4.55m was fabricated and tested in the 402m-long wave flume of the NIMALA marine laboratory. The jacket was tested at the water depth of 4m and subjected to random waves with a JONSWAP energy spectrum. Three input wave heights were chosen for the tests: 20cm, 23cm, and 28cm. The findings showed that using the NARX neural network is a convenient method to predict the base shear and overturning moment values based on the water surface level data as input values. Finally, after suitable neural network determination, using the NARX neural network, the correlation value (R) for calculating water surface level (WSL), Base Shear, and Overturning Moment were obtained as 0.994, 0.97, and 0.94, respectively.