EVOLUTIVE OPTIMIZATION OF A NEURAL NETWORK ASSOCIATED THE DE METHOD APPLIED IN MOORING SYSTEM OF FLOATING PRODUCTION SYSTEMS
Thalita Mongarde Daer1; Edivaldo Ramos Delgado2; Bruno da Fonseca Monteiro2; Juliana Souza Baioco1
1 Universidade Federal Fluminense; 2 POLI/UFRJ
doi:10.20906/CPS/CILAMCE2017-0216
Resumo
Floating production systems (FPS) for oil exploitation are subject to the action of environmental loads, such as waves, wind, and currents in different directions and intensities. The mooring system is responsible for the integrity of the platform throughout its useful life, as well as limiting the displacements caused by environmental loads at convenient levels. One way of evaluating the efficiency of a system is by means of a numerical tool based on finite elements, which defines the resultant tensions from terms of time series. However, this tool requires the execution of static and dynamic analyses that have a high computational cost and a large time interval. Moreover, an array of load cases may include hundreds of combinations. In order to reduce the cost and time of analysis, this article suggests an artificial neural network (ANN) to estimate displacements (offsets), admitting as input a combination of the main parameters that configure a mooring system (radius, angles, applied stress, and the material type of the line). Taking as a principle an already trained neural network, the focus is to find a smaller number of evaluation of the objective function to train the network, through the Matlab software, and, after that, apply them in the DE (Differential Evolution) method, thus optimizing an objective function, reducing the time spent for analysis and computational costs.
Palavras-chave: ANN; MOORING SYSTEM; DE; OFFSHORE SYSTEMS