文献类型：期刊文献

英文题名：Novel reduced-order framework combining proper orthogonal decomposition and multi-parallel Gaussian process regression: Multi-physics prediction of ethylene cracking furnaces

作者：Hu, Guihua[1];Chen, Mimi[1];Long, Jian[1,2]

机构：[1]East China Univ Sci & Technol, Key Lab Smart Mfg Energy Chem Proc, Minist Educ, Shanghai 200237, Peoples R China;[2]East China Univ Sci & Technol, Engn Res Ctr Proc Syst Engn, Minist Educ, Shanghai 200237, Peoples R China

年份：2025

卷号：305

外文期刊名：CHEMICAL ENGINEERING SCIENCE

收录：;EI(收录号：20250417725494);WOS:【SCI-EXPANDED(收录号:WOS:001406540400001)】；

基金：This work was supported by National Natural Science Foundation of China (62394345, 62273149, 62373155) , Major Science and Technol-ogy Project of Xinjiang (No. 2022A01006-4) , and the Fundamental Research Funds for the Central Universities.

语种：英文

外文关键词：Ethylene cracking furnaces; Computational fluid dynamics; Reduced-order model; Proper orthogonal decomposition; Multiple-parallel Gaussian process regression

摘要：Accurate multi-physics field prediction is important in the design and optimization for ethylene cracking furnaces. However, traditional computational fluid dynamics (CFD) simulations are time-consuming and traditional surrogate models lack information about the physical field. The proposed novel reduced-order model (ROM) framework integrates proper orthogonal decomposition (POD) and multiple-parallel Gaussian Process Regression (mGPR) to predict the multi-physics of an industrial ethylene cracking furnace while significantly reducing computational time and resource requirements. CFD simulations are first conducted to obtain multi-physics data, which are then compared to industrial values. A dataset covering various operating conditions is generated through pairwise experimental design methods. POD is employed to extract modes and coefficients of the physical fields, and mGPR is used to model the nonlinear relationship between the POD coefficients and operating parameters. The results show that the relative error of the outer wall temperature of reactor tube between the POD simulation results and the industrial values is 4.13%. The proposed ROM achieves a global error on the order of 10-3, with minimal truncation degrees of 5, 3, and 6 for flue gas temperature, pressure, and mass fraction of H2O, respectively. The mGPR model outperforms GPR model, demonstrating lower mean squared errors (MSE) (609.667 versus 3718.822) and higher R2 (0.9907 versus 0.9433). In comparison to CFD, the ROM improves computational efficiency by a factor of at least 900 and reduces storage space by approximately 96.3%. The proposed ROM provides reliable technical support for the design and optimization of ethylene cracking furnaces.