Abstract
This paper presents Part 2 in the development of an Artificial Neural Network (ANN) model for detection of mooring line failure of a spread-moored FPSO, global performance analysis used to generate the training and test data for the study.
The development of an ANN model for detection of mooring line failure requires a comprehensive training data that is most practically available from the results of numerical simulations. Time domain analysis is necessary to capture the nonlinear behavior of a moored FPSO system and to represent the behavior of the physical system as accurate as possible. Given the wide range of sea-state conditions, of direction of the sea-states and of draft conditions of the FPSO, the number of time domain simulations is easily larger than 100,000. Therefore, an accurate and numerically efficient tool is necessary for carrying this task.
The FPSO hull motion analysis is performed using MLTSIM, a TechnipFMC in-house, nonlinear time domain floating body motion analysis program. MLTSIM captures various non-linear load and response effects such as mooring stiffness, riser loads, drag and drift forces, as well as various user defined loads. MLTSIM is a numerically efficient and fast time domain solver which can run on both high-performance computing (HPC) system and a single laptop.
Numerical model of a FPSO system has been validated using the results of model tests. In addition, the results of numerical simulations, in terms of hull motions and mooring line tensions, are compared with the results of model tests and a commercial software OrcaFlex. This well-calibrated model is then used for generating the numerical data required for the development of the ANN model.
