Heave motion prediction of a large barge in random seas by using artificial neural network

• Main Authors: Lee, H.E., Liew, M.S., Zawawi, N.A.W.A., Toloue, I.
• Format: Article
• Published: American Institute of Physics Inc. 2017
• Online Access: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85036649104&doi=10.1063%2f1.5012205&partnerID=40&md5=074ec0777dd88f548f85ceb1d57b1630; http://eprints.utp.edu.my/19903/

This paper describes the development of a multi-layer feed forward artificial neural network (ANN) to predict rigid heave body motions of a large catenary moored barge subjected to multi-directional irregular waves. The barge is idealized as a rigid plate of finite draft with planar dimensions 160m (length) and 100m (width) which is held on station using a six point chain catenary mooring in 50m water depth. Hydroelastic effects are neglected from the physical model as the chief intent of this study is focused on large plate rigid body hydrodynamics modelling using ANN. Even with this assumption, the computational requirements for time domain coupled hydrodynamic simulations of a moored floating body is considerably costly, particularly if a large number of simulations are required such as in the case of response based design (RBD) methods. As an alternative to time consuming numerical hydrodynamics, a regression-type ANN model has been developed for efficient prediction of the barge's heave responses to random waves from various directions. It was determined that a network comprising of 3 input features, 2 hidden layers with 5 neurons each and 1 output was sufficient to produce acceptable predictions within 0.02 mean squared error. By benchmarking results from the ANN with those generated by a fully coupled dynamic model in OrcaFlex, it is demonstrated that the ANN is capable of predicting the barge's heave responses with acceptable accuracy. © 2017 Author(s).