1.中国石油大学(北京)重质油全国重点实验室,北京 102249
2.中国石油大学(北京)人工智能学院,北京 102249
郭建豪(1999—),男,博士研究生,2024310321@student.cup.edu.cn
蓝兴英(1977—),女,博士,教授,lanxy@cup.edu.cn
收稿:2025-12-18,
修回:2026-02-17,
纸质出版:2026-10-25
移动端阅览
郭建豪, 朱春梦, 柳楠, 吉鲁东, 赵云鹏, 石孝刚, 蓝兴英. 融合物理约束与分布式架构的提升管流场重构方法[J]. 化工学报, 2026, 77(10): 5801-5817
GUO Jianhao, ZHU Chunmeng, LIU Nan, JI Ludong, ZHAO Yunpeng, SHI Xiaogang, LAN Xingying. A deep learning framework integrating physical constraints and distributed architecture for flow field reconstruction in CFB risers[J]. CIESC Journal, 2026, 77(10): 5801-5817
郭建豪, 朱春梦, 柳楠, 吉鲁东, 赵云鹏, 石孝刚, 蓝兴英. 融合物理约束与分布式架构的提升管流场重构方法[J]. 化工学报, 2026, 77(10): 5801-5817 DOI: 10.11949/0438-1157.20251428.
GUO Jianhao, ZHU Chunmeng, LIU Nan, JI Ludong, ZHAO Yunpeng, SHI Xiaogang, LAN Xingying. A deep learning framework integrating physical constraints and distributed architecture for flow field reconstruction in CFB risers[J]. CIESC Journal, 2026, 77(10): 5801-5817 DOI: 10.11949/0438-1157.20251428.
面向循环流化床(circulating fluidized bed,CFB)提升管复杂气固流场的快速高精度预测需求,针对传统计算流体力学(computational fluid dynamics,CFD)计算成本高、数据驱动模型易出现守恒性不足等问题,构建经实验验证的多工况CFD数据集,提出融合物理约束与分布式并行编码器-解码器结构的物理信息神经网络(distributed parallel encoder-decoder physics-informed neural network,DPED-PINN)流场重构方法。该方法将基于介尺度曳力模型的质量与动量守恒方程以软约束形式嵌入损失函数,并结合特征解耦、多解码器结构及基于残差的自适应采样策略,以增强对高梯度区域和多尺度流动特征的表征能力。数值结果表明,与标准PINN及常规模型相比,DPED-PINN在已知与外推工况下具有更高的预测精度和物理一致性,能够较为准确地描述壁面颗粒聚团及典型“环-核”结构。此外,固定测点固含率时间序列对比与独立实验测点数据的外部对照进一步验证了模型的时间一致性与外推泛化能力。同时,所提出的方法相对于逐工况CFD模拟,显著提升了计算效率,为CFB提升管复杂气固流场的高效数值模拟提供了一种可行途径。
To address the demand for real-time
high-fidelity reconstruction of gas-solid flow fields in circulating fluidized bed (CFB) risers
this study develops a physics-data fusion deep learning framework. The method targets two main limitations of current simulations: the high computational cost and poor real time performance of traditional computational fluid dynamics (CFD)
and the lack of physical consistency in purely data driven models without explicit physical constraints. An experimentally validated CFD model is first used to construct a multi condition three-dimensional spatiotemporal dataset that covers a wide range of gas velocities and particle fluxes. On this basis
a distributed parallel encoder-decoder physics-informed neural network (DPED-PINN) is proposed. In terms of architecture
a parallel multi decoder structure is adopted to decouple feature extraction and alleviate gradient conflicts in multi task learning. In terms of physical mechanisms
mass and momentum conservation equations based on a mesoscale drag model are embedded into the loss function as soft constraints in order to enforce physical consistency. Furthermore
an adaptive sampling strategy based on residual probability density
combined with a dynamic loss weighting scheme
is introduced to enhance learning in high gradient regions. Numerical results show that
compared with the standard PINN and conventional models
DPED-PINN exhibits higher prediction accuracy and physical consistency under known and extrapolated conditions
and can accurately describe wall particle agglomeration and typical “ring-core” structures. For a representative extrapolated case
the model achieves a solid volume fraction structural similarity index (SSIM) of about 0.887
which is approximately 15% higher than that of a standard PINN
and accurately reconstructs near wall particle clusters and the characteristic core annulus structure. Moreover
the temporal evolution of solids holdup at fixed probes and external comparisons with independent experimental measurements further corroborate the temporal consistency and generalization performance of the proposed approach. Compared with case-by-case CFD simulations
the proposed framework significantly improves computational efficiency
providing a feasible approach for efficient numerical simulation and analysis of complex gas-solid flows in CFB risers.
Chew J W , LaMarche W C Q , Cocco R A . 100 years of scaling up fluidized bed and circulating fluidized bed reactors [J ] . Powder Technology , 2022 , 409 : 117813 .
Razzak S A . Flow characteristic studies on the gas-liquid-solid circulating fluidized bed based on system stability [J ] . International Journal of Multiphase Flow , 2019 , 113 : 279 - 288 .
金涌 , 程易 , 白丁荣 , 等 . 中国流态化技术研发史略 [J ] . 化工进展 , 2023 , 42 ( 6 ): 2761 - 2780 .
Jin Y , Cheng Y , Bai D R , et al . Fluidization research and development in China [J ] . Chemical Industry and Engineering Progress , 2023 , 42 ( 6 ): 2761 - 2780 .
岳光溪 , 吕俊复 , 徐鹏 , 等 . 循环流化床燃烧发展现状及前景分析 [J ] . 中国电力 , 2016 , 49 ( 1 ): 1 - 13 .
Yue G X , Lu J F , Xu P , et al . The up-to-date development and future of circulating fluidized bed combustion technology [J ] . Electric Power , 2016 , 49 ( 1 ): 1 - 13 .
Zhu X L , Xu Y , Tu Q Y , et al . Advanced measurement techniques for gas-solids fluidized beds in the power and energy industry — a review [J ] . Measurement: Energy , 2024 , 4 : 100030 .
徐硕 , 毕可鑫 , 邱彤 , 等 . 智能过程系统工程(iPSE): 当过程系统工程与新一代人工智能相遇 [J ] . 化工进展 , 2026 , 45 ( 4 ): 1877 - 1889 .
Xu S , Bi K X , Qiu T , et al . Intelligent process systems engineering (iPSE): When process systems engineering meets new generation artificial intelligence [J ] . Chemical Industry and Engineering Progress , 2026 , 45 ( 4 ): 1877 - 1889 .
Alobaid F , Almohammed N , Massoudi Farid M , et al . Progress in CFD simulations of fluidized beds for chemical and energy process engineering [J ] . Progress in Energy and Combustion Science , 2022 , 91 : 100930 .
Brunton S L , Noack B R , Koumoutsakos P . Machine learning for fluid mechanics [J ] . Annual Review of Fluid Mechanics , 2020 , 52 : 477 - 508 .
Zhu L T , Chen X Z , Ouyang B , et al . Review of machine learning for hydrodynamics, transport, and reactions in multiphase flows and reactors [J ] . Industrial & Engineering Chemistry Research , 2022 , 61 ( 28 ): 9901 - 9949 .
Upadhyay M , Nagulapati V M , Lim H . Hybrid CFD-neural networks technique to predict circulating fluidized bed reactor riser hydrodynamics [J ] . Journal of Cleaner Production , 2022 , 337 : 130490 .
黄正梁 , 王超 , 李少硕 , 等 . 基于深度学习的气液固三相反应器图像分析方法及应用 [J ] . 化工学报 , 2020 , 71 ( 1 ): 274 - 282 .
Huang Z L , Wang C , Li S S , et al . Development and application of image analysis method based on deep-learning in gas-liquid-solid three-phase reactor [J ] . CIESC Journal , 2020 , 71 ( 1 ): 274 - 282 .
Zhong H B , Sun Z N , Zhu J , et al . Prediction of solid holdup in a gas-solid circulating fluidized bed riser by artificial neural networks [J ] . Industrial & Engineering Chemistry Research , 2021 , 60 ( 8 ): 3452 - 3462 .
Bazai H , Kargar E , Mehrabi M . Using an encoder-decoder convolutional neural network to predict the solid holdup patterns in a pseudo-2D fluidized bed [J ] . Chemical Engineering Science , 2021 , 246 : 116886 .
Raissi M , Perdikaris P , Karniadakis G E . Physics-informed neural networks: a deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations [J ] . Journal of Computational Physics , 2019 , 378 : 686 - 707 .
Raissi M , Yazdani A , Karniadakis G E . Hidden fluid mechanics: learning velocity and pressure fields from flow visualizations [J ] . Science , 2020 , 367 ( 6481 ): 1026 - 1030 .
Buhendwa A B , Adami S , Adams N A . Inferring incompressible two-phase flow fields from the interface motion using physics-informed neural networks [J ] . Machine Learning with Applications , 2021 , 4 : 100029 .
张橙 , 李雪 , 叶茂 , 等 . 物理信息神经网络在两相流中的应用 [J ] . 化工学报 , 2024 , 75 ( 11 ): 3835 - 3856 .
Zhang C , Li X , Ye M , et al . Application of physics-informed neural network in two-phase flow [J ] . CIESC Journal , 2024 , 75 ( 11 ): 3835 - 3856 .
Wang S F , Teng Y J , Perdikaris P . Understanding and mitigating gradient flow pathologies in physics-informed neural networks [J ] . SIAM Journal on Scientific Computing , 2021 , 43 ( 5 ): A3055 - A3081 .
Anagnostopoulos S J , Toscano J D , Stergiopulos N , et al . Residual-based attention in physics-informed neural networks [J ] . Computer Methods in Applied Mechanics and Engineering , 2024 , 421 : 116805 .
Yang M Y , Foster J T . Multi-output physics-informed neural networks for forward and inverse PDE problems with uncertainties [J ] . Computer Methods in Applied Mechanics and Engineering , 2022 , 402 : 115041 .
Yu J , Lu L , Meng X H , et al . Gradient-enhanced physics-informed neural networks for forward and inverse PDE problems [J ] . Computer Methods in Applied Mechanics and Engineering , 2022 , 393 : 114823 .
Ge W , Li J H . Physical mapping of fluidization regimes: the EMMS approach [J ] . Chemical Engineering Science , 2002 , 57 ( 18 ): 3993 - 4004 .
陈恺成 , 田于杰 , 李飞 , 等 . 基于EMMS的循环流化床流域研究 [J ] . 化工学报 , 2020 , 71 ( 7 ): 3018 - 3030 .
Chen K C , Tian Y J , Li F , et al . EMMS-based flow regime study of circulating fluidized beds [J ] . CIESC Journal , 2020 , 71 ( 7 ): 3018 - 3030 .
Gidaspow D , Bacelos M S . Kinetic theory based multiphase flow with experimental verification [J ] . Reviews in Chemical Engineering , 2018 , 34 ( 3 ): 299 - 318 .
Yang N , Wang W , Ge W , et al . CFD simulation of concurrent-up gas-solid flow in circulating fluidized beds with structure-dependent drag coefficient [J ] . Chemical Engineering Journal , 2003 , 96 ( 1/2/3 ): 71 - 80 .
Wang C . High density gas-solids circulating fluidized bed riser and downer reactors [D ] . Ontario : The University of Western Ontario , 2013 .
Garud S S , Karimi I A , Kraft M . Design of computer experiments: a review [J ] . Computers & Chemical Engineering , 2017 , 106 : 71 - 95 .
Li S R , Feng X L . Dynamic weight strategy of physics-informed neural networks for the 2D Navier-Stokes equations [J ] . Entropy , 2022 , 24 ( 9 ): 1254 .
Glorot X , Bengio Y . Understanding the difficulty of training deep feedforward neural networks [C ] // Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics , 2010 : 249 - 256 .
Lu L , Jin P Z , Pang G F , et al . Learning nonlinear operators via DeepONet based on the universal approximation theorem of operators [J ] . Nature Machine Intelligence , 2021 , 3 ( 3 ): 218 - 229 .
Zhou H Y , Zhang S H , Peng J Q , et al . Informer: beyond efficient transformer for long sequence time-series forecasting [J ] . Proceedings of the AAAI Conference on Artificial Intelligence , 2021 , 35 ( 12 ): 11106 - 11115 .
0
浏览量
8
下载量
0
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010102001995号