关键词: 仿真数据库, 数字孪生, 深潜球壳, 应力场分布, 优化算法

Abstract:

This paper presents a method for predicting the stress field of deep diving spherical shells based on simulation databases and digital twin technology. By establishing simulation databases of stress field distribution of pressure-resistant spherical shells under different scales and loads, virtual sensing monitoring of stress states in other parts of the vessel is realized through finite sensor layout of pressure-resistant shells on the submersible. Based on the DT(digital twin) technology, a three-level virtual structure layer is constructed. The Level-1 DT layer realizes the spatial mapping and cloud image display from the finite element simulation model to the digital model. The error between the experimental and numerical results of the ultimate bearing capacity of the spherical shell is less than 9.4%. The Level-2 DT layer realizes the data sample deduction of the digital model by create database. The stress field distribution of the spherical shell under the condition that the size and load are not obtained in the simulation database is obtained by the local Lagrange interpolation method. The relative error of the stress interpolation result is 4.8%. The Level-3 DT layer develops a machine learning prediction function for the stress field distribution in the dangerous area of the deep-submersible spherical shell digital model. The BP neural network optimized by the particle swarm optimization algorithm ensures that the error between the prediction result and the simulation result is less than 1%. This method comprehensively considers the material properties, structural dimensions and environmental loads, which can provide a reference for the real-time safety assessment of the pressure hull structure, and realize the dynamic perception, intelligent diagnosis and scientific prediction of the dynamic stress field distribution of all deep-submersible spherical shells on the hull.

Key words: simulation database, digital twin, deep submerged spherical shell, stress field distribution, optimization algorithm