Session: 02-05-01 Data-driven and AI-based Models

Submission Number: 179505

A Method for Identification of Impact Load on Stiffened Plate Based on Physics-Informed Neural Networks 

The offshore platform is subjected to complex dynamic loads during its lifetime. As a typical load acting on offshore platforms, helicopter landing impact load is a design scenario for determining the scantling required to resist helicopter landing load. Compared with the helicopter landing on land, due to the harsh environment and the swaying offshore platform, the helicopter landing on the offshore platform has more uncontrollable landing postures, which leads to large impact load with random location. The landing impact loads are difficult to be measured directly, and indirect methods for identifying the loads are more promising, for instance, by estimate the impact loads from the measure stress responses in the helipad or the deck structures. However, to devise such methods, there are many difficulties due to the complexities involved in the problem. A feasible approach is to incorporate physical equations into the loss function of neural networks, thereby ensuring that the trained neural networks obey the fundamental physical laws. In this study, a physics-informed neural network is proposed to identify the helicopter landing loads acting on stiffened plates. The proposed method is validated through numerical simulations for a stiffened plate that is usually used in naval architecture and ocean engineering.

Presenting Author: Yishi Xu Harbin Engineering University

Presenting Author Biography: Mr Yishi Xu is a PhD Student of Harbin Engineering University, and his area of research is structural safety of ships and offshore platforms.

Authors:

Yishi Xu Harbin Engineering University
Man Sun Harbin Engineering University
Jingjing Li College of Shipbuilding Engineering, Harbin Engineering University
Xueqian Zhou Harbin Engineering University
Huilong Ren Harbin Engineering University