Multi-Fidelity Machine Learning for Identifying Thermal Insulation Integrity of Liquefied Natural Gas Storage Tanks DOI Creative Commons
Wei Lin,

Meitao Zou,

Mingrui Zhao

и другие.

Applied Sciences, Год журнала: 2024, Номер 15(1), С. 33 - 33

Опубликована: Дек. 24, 2024

The thermal insulation integrity of liquefied natural gas storage tanks is essential for their life-cycle safety. However, perlite settlement (insulation material) can result in leaks and lead to engineering risks. direct measurement difficult due the enclosed structure these tanks. To address this challenge, study presents a data-driven approach based on machine learning real-time monitoring data. This proposes multi-fidelity framework enhance generalizability leverage data effectively. Low-fidelity are readily available but contain systematic errors, while high-fidelity accurate limited accessibility. By combining both types data, enhances generalisability prediction accuracy trained models. results experiments demonstrate that outperforms models solely low- or achieving coefficient determination 0.980 root mean square error 0.078 m. Three algorithms—Multilayer Perceptron, Random Forest, Extreme Gradient Boosting—were evaluated determine optimal implementation. provides reliable method tanks, contributing improved industrial safety operational efficiency.

Язык: Английский

Hybrid physics and data-driven method for predicting existing tunnel lining deformation in twin tunnels construction DOI

Xing‐Tao Lin,

R. H. Chen,

Xiangsheng Chen

и другие.

Computers and Geotechnics, Год журнала: 2025, Номер 179, С. 107019 - 107019

Опубликована: Янв. 10, 2025

Язык: Английский

Процитировано

0
Prediction of Member Forces of Steel Tubes on the Basis of a Sensor System with the Use of AI DOI Creative Commons

Haiyu Li,

Heung‐Jin Chung

Sensors, Год журнала: 2025, Номер 25(3), С. 919 - 919

Опубликована: Фев. 3, 2025

The rapid development of AI (artificial intelligence), sensor technology, high-speed Internet, and cloud computing has demonstrated the potential data-driven approaches in structural health monitoring (SHM) within field engineering. Algorithms based on machine learning (ML) models are capable discerning intricate behavioral patterns from real-time data gathered by sensors, thereby offering solutions to engineering quandaries mechanics SHM. This study presents an innovative approach a fiber-reinforced polymer (FRP) double-helix system for prediction forces acting steel tube members offshore wind turbine support systems; this enables system. as transitional member FRP double helix-sensor were initially modeled three dimensions using ABAQUS finite element software. Subsequently, obtained analysis (FEA) inputted into fully connected neural network (FCNN) model, with objective establishing nonlinear mapping relationship between inputs (strain) outputs (reaction force). In FCNN impact number input variables model's predictive performance is examined through cross-comparison different combinations positions six sets variables. And evaluation costs strain series identified further optimization. Furthermore, variable optimized convolutional (CNN) resulting optimal that achieved accuracy level more fewer sensors. not only improves model but also effectively controls cost. was evaluated several metrics, including R2, MSE, MAE, SMAPE. results CNN exhibited notable advantages terms fitting computational efficiency when confronted limited set. To provide practical applications, interactive graphical user interface (GUI)-based sensor-coupled mechanical tubes developed. engineers predict real time, enhancing SHM systems.

Язык: Английский

Процитировано

0
Prediction of disc cutter wear of shield machines based on transfer learning DOI

Yuxiang Meng,

Qian Fang, Guoli Zheng

и другие.

Tunnelling and Underground Space Technology, Год журнала: 2025, Номер 162, С. 106633 - 106633

Опубликована: Апрель 14, 2025

Язык: Английский

Процитировано

0
A hybrid approach for modifying tunneling-induced response in existing multi-tunnel environment DOI
Hongwei Huang, Tian Gao, Dongming Zhang

и другие.

Computers and Geotechnics, Год журнала: 2024, Номер 179, С. 106921 - 106921

Опубликована: Дек. 2, 2024

Язык: Английский

Процитировано

1
Multi-Fidelity Machine Learning for Identifying Thermal Insulation Integrity of Liquefied Natural Gas Storage Tanks DOI Creative Commons
Wei Lin,

Meitao Zou,

Mingrui Zhao

и другие.

Applied Sciences, Год журнала: 2024, Номер 15(1), С. 33 - 33

Опубликована: Дек. 24, 2024

The thermal insulation integrity of liquefied natural gas storage tanks is essential for their life-cycle safety. However, perlite settlement (insulation material) can result in leaks and lead to engineering risks. direct measurement difficult due the enclosed structure these tanks. To address this challenge, study presents a data-driven approach based on machine learning real-time monitoring data. This proposes multi-fidelity framework enhance generalizability leverage data effectively. Low-fidelity are readily available but contain systematic errors, while high-fidelity accurate limited accessibility. By combining both types data, enhances generalisability prediction accuracy trained models. results experiments demonstrate that outperforms models solely low- or achieving coefficient determination 0.980 root mean square error 0.078 m. Three algorithms—Multilayer Perceptron, Random Forest, Extreme Gradient Boosting—were evaluated determine optimal implementation. provides reliable method tanks, contributing improved industrial safety operational efficiency.

Язык: Английский

Процитировано

0