Artificial Neural Network-Based Fatigue Life Prediction for Fixed Offshore Platforms

Objectives / Scope This paper aims to present an approach utilizing Artificial Neural Networks (ANN) to predict fatigue life of tubular joints in fixed offshore platforms. The objective is to speed up the time-domain fatigue analysis of fixed offshore by training ANNs with smaller datasets derived from time-domain finite element method (FEM) analyses, thereby avoiding impractical, time-consuming complete time-domain simulations. Methods, procedures, process The methodology involves conducting shorter time-domain finite element analyses of a fixed platform using a FEM-based software for just a few sea states. Key structural responses related to fatigue life prediction are extracted from these analyses and used to train ANN models. ANN's training focuses on establishing accurate correlations between input environmental conditions and, resulting fatigue damage, with the goal of providing a reliable, fast-response predictive tool. The ANN is designed to perform accurate predictions for several sea states from a single training session, potentially enabling its use for fatigue damage estimation in the design and re-evaluation of such structures. Results, Observations, Conclusions The results show that the neural network approach accurately predicts fatigue life and dynamic responses for marine structures, with significant time savings compared to FEM simulations. The model successfully generalized across various sea states, providing reliable fatigue life estimates. In conclusion, this hybrid FE-ANN model offers an efficient and precise tool for fatigue assessment in offshore engineering. Novel, Additive Information The novel contribution of this work lies in the development of a hybrid FEM-ANN model capable of accurately predicting fatigue life and dynamic responses across diverse sea states with minimal computational cost. By leveraging a single trained neural network, the model demonstrates exceptional generalization, which is critical for real-world offshore engineering applications. This approach significantly enhances efficiency while maintaining high accuracy, offering a promising solution for large-scale fatigue assessments of marine structures.