Session: Structural Design-2

Paper Number: 164970

164970 - Efficient Structural Analysis Method for Floating Wind Platforms Through Potential Theory and Neural Networks 

Abstract: 

Objective

Cost-effective and reliable floating offshore wind technology requires rigorous structural analysis of platforms, encompassing both ultimate and fatigue limit states. This analysis involves a multi-stage process: integrated simulations to determine environmental loads, followed by detailed Finite Element Method (FEM) analysis to assess structural integrity.

For ultimate limit state analysis, critical load instants are extracted from integrated simulations for FEM evaluation, while fatigue analysis necessitates the transfer and processing of complete time-series load data. However, two primary challenges hinder the accuracy and efficiency of this process. First, existing integrated simulation tools often provide only aggregated forces, lacking the detailed pressure distributions necessary for accurate FEM modeling of submerged platform surfaces. Second, the computational demands of fatigue analysis, requiring FEM simulations and rainflow cycle counting across numerous load cases, pose a significant bottleneck.

This paper introduces a structural analysis methodology that overcomes these challenges. First, we present Pandilo, a time-domain pressure distribution computation tool based on potential theory. This tool enables accurate load transfer to FEM analyses. We demonstrate its practical application through a water tightness analysis of a concrete platform with no pretension. Second, we employ a trained neural network for efficient stress prediction, enabling a full fatigue analysis of the platform.

Methodology

Pandilo utilizes linear potential and requires a pre-processing step involving a hydrodynamic frequency-domain analysis of a mesh representing the platform's wet surface, using a potential panel code such as WAMIT. Based on this analysis, Pandilo calculates the contribution of each panel to the added mass, potential damping and force RAO (Response Amplitude Operator). Then, using the results of an integrated simulation, specifically the platform's motions (position, velocities and accelerations) and the wave height, Pandilo computes the time-domain hydrodynamic pressures acting on each panel of the mesh. This post-processing step can be efficiently executed on a standard laptop within minutes for a typical 3-hour simulation.

Pandilo was used to verify the water tightness of the internal tanks of BERIDI hexagonal platform, under extreme loading conditions. Integrated simulations, conducted according to guidelines, were performed. Critical instants, identified from these simulations and corresponding to extreme tower base loads, mooring tensions, and hydrodynamic loads, were post-processed using Pandilo. Pandilo mapped hydrodynamic pressures onto the platform's submerged surfaces using the concurrent wave height and platform motion data. These pressure distributions, along with concurrent mooring line, tower base, and inertial loads, were then applied to a finite element model enabling the structural evaluation.

For fatigue analysis, two sequential neural networks were employed to efficiently determine the distributed pressures on the hull and the subsequent internal material stresses, significantly reducing computational time. This acceleration enabled the application of rainflow cycle counting to all time series across hundreds of load cases, as specified by relevant guidelines, for every structural element. The neural networks were trained using a dataset of 25.000 s of simulation for the distributed pressures and 3000 representative structural states for the stresses. A subsequent validation with an independent set of states data obtained by FEM was performed.

Results

Based on the results of the FEM model, an analysis of the bending moments in the walls of the platform’s water-tight tanks is performed. In these walls it is observed that, in general, the bending law varies in magnitude but works always in the same direction Therefore, water tightness of the internal walls is guaranteed.

The validation of the neural network demonstrated stress prediction compares very well to FEM calculations, with computational times extremely efficient. The structural solution of a three-hours case can be provided in seconds, compared with a huge computational effort to solve it with a traditional approach.

Conclusions

An enhanced structural analysis methodology is introduced for floating platforms. Pandilo's efficient hydrodynamic load transfer to FEM, compared to CFD, enables detailed extreme load analyses, as demonstrated on the BERIDI platform. Neural networks significantly accelerate fatigue analysis, allowing comprehensive assessments. These advancements improve the feasibility and reliability of floating wind platform designs.

Presenting Author: Jon Cerrada-Garcés CENER

Presenting Author Biography: Jon Cerrada is a mechanical-industrial engineer with both bachelor's and master's degrees from the Public University of Navarre (UPNA). His master's thesis focused on calculating distributed hydrodynamic pressures on floating platforms. Currently, he is a researcher at the National Renewable Energy Centre (CENER) in the mechanical area of the wind energy department. He has also worked with artificial intelligence solutions.

Authors:

Jon Cerrada-Garcés CENER
Jose Azcona-Armendáriz CENER
Alvaro Olcoz-Alonso CENER
Amaia Marco CENER
Clara E. Acosta BERIDI