文献类型：期刊文献

英文题名：Multi-variable operational data fusion and multi-scale hybrid modeling for online fouling monitoring of heat exchangers

作者：Hou, Gang An, Zhoujian Zhang, Dong Du, Xiaoze Ding, Yong Fu, Jian

第一作者：Hou, Gang

通信作者：An, ZJ[1];Du, XZ[1];An, ZJ[2]

机构：[1]Lanzhou Univ Technol, Coll Energy & Power Engn, Lanzhou 730050, Peoples R China;[2]Guizhou Inst Technol, Sch Aerosp Engn, Guiyang 550025, Peoples R China

第一机构：Lanzhou Univ Technol, Coll Energy & Power Engn, Lanzhou 730050, Peoples R China

通信机构：corresponding author), Lanzhou Univ Technol, Coll Energy & Power Engn, Lanzhou 730050, Peoples R China;corresponding author), Guizhou Inst Technol, Sch Aerosp Engn, Guiyang 550025, Peoples R China.|贵州理工学院;

年份：2026

卷号：175

期号：P2

外文期刊名：INTERNATIONAL COMMUNICATIONS IN HEAT AND MASS TRANSFER

收录：;EI(收录号：20261420421129);Scopus(收录号：2-s2.0-105034616166);WOS:【SCI-EXPANDED(收录号:WOS:001731282500001)】；

基金：This work was financially supported by the Science and Technology Project of Gansu province (25CXGA058) ; National Natural Science Foundation of China (52206087; 52130607) ; the Guizhou Provincial Major Scientific and Technological Program (XKBF (2025) 031) ; the Guizhou Science and Technology Innovation Leading Talent Worksta-tion (KXJZ (2025) 024) ; Wuwei City Major Science and Technology Special Projects (WW25A03ZDQ002) ; the China Postdoctoral Science Foundation (2025M770576) ; the Industrial Support Plan Project of Gansu Provincial Education Department (2025CYZC-034) ; the Doctoral Research Funds of Lanzhou University of Technology (061907) ; the Red Willow Excellent Youth Project of Lanzhou University of Technology and the Young Faculty Interdisciplinary Research Cultivation Program of Lanzhou University of Technology.

语种：英文

外文关键词：Plate heat exchanger; Fouling prediction; LSTM model; CNN-BiLSTM-attention model

摘要：As a key heat transfer device, fouling deposition in plate heat exchangers has become a prominent issue that restricts the energy efficiency of the system. This study presents a hybrid time-series modeling framework, Convolutional Bidirectional LSTM with Attention (CNN-BiLSTM-Attention), designed for the prediction of fouling thermal resistance. The framework is integrated with deep neural networks to facilitate real-time monitoring of fouling evolution in heat exchangers. The bidirectional long short-term memory neural network layer (LSTM) of the model captures long-term dependencies and inverse temporal correlations in dust accumulation patterns, thereby enabling the adaptive fusion of multi-dimensional features. By incorporating the sliding window mechanism of the convolutional layer, the model effectively captures local dependencies among adjacent time steps or sensor variables, thereby enhancing its adaptability to complex operational conditions. This approach overcomes the inherent limitation of conventional LSTM models in capturing long-term dependencies, which rely exclusively on recurrent structures for temporal information propagation. The operational data from Plate Heat Exchanger #1 were used for model training, while those from Plate Heat Exchanger #2 were adopted for testing. The prediction accuracy achieved a value of 0.9857, indicating high predictive performance and strong generalization capability. Online monitoring of fouling in plate heat exchangers offers a viable technical solution.