Mooring tension prediction based on BP neural network for semi-submersible platform. (1st March 2021)

• Record Type: Journal Article
• Authors: Zhao, Yuliang; Dong, Sheng; Jiang, Fengyuan; Incecik, Atilla
• Is Part Of: Ocean engineering. Volume 223(2021)
• Publication Date: 2021-03-01
• Subjects: BP neural Network -- Mooring design -- Semi-submersible platform -- Tension prediction; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2021.108714
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: The design and optimisation of a mooring system for a floating platform typically necessitates engineering experience and time-consuming numerical simulations. A novel hybrid BPNN–FEM approach, based on the modified BP neural network and FEM is proposed to more conveniently predict the statistic pretension and dynamic tension series of mooring systems with several design variables under irregular sea states. The accuracy of this approach has been validated using several statistical error metrics. This procedure can be used to optimize the mooring system design of the floating platform in a more economical manner than time-intensive numerical simulations. An optional mooring selection is proposed based on the maximum safe operation window and the requirements for the platform drift and mooring line safety performance. A deep-water semi-submersible platform is presented as an example to demonstrate the hybrid approach. This platform comprises twelve mooring lines, and the design variables include the mooring radius, values for the azimuthal spacing of the mooring lines, and the length of the various segments of the mooring lines. The impact of environmental load incidences is also considered in this process. The prediction results are in reasonable agreement with those obtained using the FEM. Highlights: A modified BPNN-FEM approach is proposed to predict dynamic tension response. The ANN-based models can predict static and dynamic response well as FEM. Sigmoid function used as activation function shows slightly better prediction than radbas. Mooring tension with arbitrary design parameters can be easily captured using ANN models. The NARX-BPNN model is feasible to detect mooring performance with platform motion input.