ARTIFICIAL NEURAL NETWORKS FOR OPTIMIZATION PROCEDURES OF LAZY WAVE FLEXIBLE RISER CONFIGURATION

Resumo

Risers are among the most expensive and complex components of an offshore production
system, because they suffer the action of a series of dynamic and environmental loads when connecting
the seabed to the floating unit. An optimal design of such structures can bring more security and cost
savings, which can be achieved by optimization techniques. In this context, this work presents the
development and implementation of Artificial Neural Network (ANN) in optimization procedures of
Lazy Wave flexible riser configuration as an alternative to Finite Element simulations, which has a high
computational cost. For the network selection, a parametric analysis has been done considering the mean
square error, accuracy of two different types of training algorithms and different amounts of neurons
layers. A case study is presented comparing the results of the ANN optimization process with a
simulation by the Finite Element Method (FEM). The results indicate a significant reduction of the
computational cost of all optimization process was achieved using ANN, which was able to predict with
accuracy the magnitudes involved in this type of problem.