Using Convolutional Neural Networks in Installation Analysis of Lazy-Wave Flexible Risers

Abstract

Abstract The design phase of offshore installation projects is supported by numerical simulations. These analyses aim to evaluate the mechanical behavior of the equipment involved, such as vessels and flexible pipes, during that operation. Therefore, a common approach is to take the ocean wave loads modeled as deterministic ones (or regular wave approach), which is a simplification that, on the one hand, allows low computational cost, but, on the other one, lacks the representation of the actual behavior of the wave loads, usually better represented by means of an irregular wave modeling. In the way of searching for an irregular wave analysis procedure to be used in the daily design of lazy-wave riser installation analyses, this work proposes an artificial neural network (ANN)-based approach. The proposed model aims to achieve it by training a convolutional neural network (CNN) fed by generated data from short-length finite element-based numerical simulations. This surrogate model can predict quite well the pipe's top tension and approximately the axial tension in the touchdown zone (TDZ) for different configuration stages during the riser's installation operation. Moreover, the proposed model works for different environmental scenarios, which boosts the computational simulation time reduction in this phase of riser design.