Downtime Technique Using Artificial Intelligence: A Case Study for an Exposed Berthing Facility

Abstract

Abstract Hydrodynamic modelling is significantly improved in the last decade, however, the coupling of these hydrodynamic models with methods to estimate berth downtime due to environmental conditions (wind, waves, currents) is less developed. The large number of environmental inputs (wind speed, wind direction, wave height, wave period, wave direction, current speed, current direction) and mooring outputs (vessel motions in six degrees of freedom, mooring lines, fender forces) to be considered in a downtime study requires major simplifications as the full wind-wave time series cannot be calculated. Downtime assessment is normally simplified by calculating a limited number of wind-wave combinations. A new approach is developed and presented in this paper by calculating the downtime for the long-term environmental time series using artificial neural networks. Neural networks are some of the most capable artificial intelligence tools for solving very complex problems like the present case. The approach has been applied for a bulk terminal project located in the Arabian Sea. The terminal has no shelter and is totally exposed to wind and swell waves. The wave climate for a 15 years period is established at the project site using spectral wave modelling software MIKE 21 SW. A large set of combined environmental conditions is selected for the dynamic mooring analysis of the vessel at berth. The time domain mooring analysis software MIKE 21 MA is used in this study. An inhouse Matlab code/program is developed using neural network to calculate the downtime for the long-term environmental time series based on the dynamic vessel response for the set of selected environmental combinations. This approach provides a more accurate downtime estimate which is important for the operability of such exposed facilities. The downtime tool is also tested for a different set of environmental combinations and mooring layouts in order to assess the sensitivity of these parameters on the downtime estimate. Up to the authors’ knowledge, this is the first published work applying artificial intelligence techniques for downtime studies.