Zhenzhong Hu

    He received both his BE and PhD degree in the Department of Civil Engineering at Tsinghua University, China. He was a visiting researcher in Carnegie Mellon University.
    He is now the associate professor in Shenzhen International Graduate School, Tsinghua University, and also the secretary general of the BIM Specialty Committee of the China Graphics Society.
    His research interests include information technologies in civil and marine engineering, building information modeling (BIM) and digital disaster prevention and mitigation.
  • 2025-04-17

    2025-04-17

    The paper Vision-based adaptive cross-domain online product recommendation for 3D design models has been published in the journal Computer-Aided Civil and In Infrastructure Engineering.

    Three-dimensional (3D) digital design is extensively adopted in the architecture, engineering, consulting, operations, and maintenance (AECOM) industry to enhance collaboration among stakeholders. Although recommendation systems are commonly employed to facilitate purchasing in e-commerce websites, none involves recommending online products to users from 3D building design models due to dimensional and stylistic discrepancies. This study proposes a visionbased adaptive cross-domain online product recommendation method, VacRed, for 3D building design models. First, a cross-domain approach is proposed to transform design models into e-commerce images, addressing discrepancies in dimension and style between them. Second, an adaptive mechanism is introduced to solve the issue of image quality instability caused by variations in generator weights during the training process of generative models. Third, a cross-domain product recommendation system is developed based on deep learning to recommend the top k relevant online products for a given building design product. Finally, experiments were conducted to ascertain the effectiveness of the VacRed method. The experimental results of this method demonstrate its excellent performance, achieving a precision rate (PR) of 87.20% and a mean average precision of 83.65%. This study effectively connects two main stages in the AECOM industry, design and purchasing, and two large communities, design and e-commerce.

    Note: Computer-Aided Civil and Infrastructure Engineering belongs to the Q1 zone of engineering and technology journals, with an impact factor of 8.5. The first author of the paper is Zhou Xiaoping from Beijing University of Civil Engineering and Architecture. The research results were supported by the National Natural Science Foundation of China.

  • 2025-04-12

    2025-04-12

    Thesis AI-based prediction of seismic time-history responses of RC frame structures considering varied structural parameters has been published in the Journal of Building Engineering.

    In this paper, an end-to-end framework for Intelligent Seismic Response Prediction, ISRPnet, is introduced. ISRPnet comprises a structural parameter module for discretizing reinforced concrete frame structures into a series of static features and an encoder-decoder architecture for encoding seismic loads and autoregressively predicting seismic responses. The model is trained on a data set of 16,544 cases generated through validated fibre-based finite element models. ISRPnet achieves promising performance on both frequent and rare earthquakes. ISRPnet rapidly and highly precisely predicts temporal responses for frequent earthquakes. The peak displacement predictions remain accurate for rare earthquakes. The superiority of the physical loss and the advantages of gated recurrent unit over long short-term memory are analysed in comparative experiments. Verification with unseen seismic waves beyond the training data shows the robust generalization and extrapolation capabilities of the framework. The proposed model accomplishes efficient surrogate computation of the full-process seismic response for a class of RC frame structures.

    Note: Journal of Building Engineering, a journal in the field of engineering and technology, belongs to the SCI journal in Q1 region, with an impact factor of 6.7 in 2024. The first author is Ge Kang,  a master's student in 2022, and the corresponding author is Wang Chen, an assistant professor in the Department of Civil Engineering at Tsinghua University. The research results were supported by the National Natural Science Foundation of China and the Cross-disciplinary Research and Innovation Fund Research Plan of Tsinghua Shenzhen International Graduate School.

  • 2025-03-31

    2025-03-31

    "Development of CAE courses for civil and architectural engineering to cultivate strong cross-disciplinary talents in the new era" has been first published on the China National Knowledge Network in the Journal of Architectural Education in Institutions of Higher Learning.

    In the current rapidly developing economic and technological environment, the civil and architectural engineering industry is facing new challenges and opportunities. To meet the demands of the new era, cultivating talents with interdisciplinary knowledge and comprehensive abilities has become an important direction for education reform in this industry. Based on an analysis of the demand for talents in the new era of civil engineering and architecture, this article explores how to cultivate strong cross-disciplinary talents through the construction of computer-aided engineering (CAE) courses and summarizes the basic experience of CAE course for civil and architectural engineering construction and reform.

    Note: I am the first author of the paper. The authors include Zhu Shiyi, a member of the research team, and Lin Jiarui, an associate researcher at the Institute of Disaster Prevention and Mitigation, Department of Civil Engineering, Tsinghua University.

  • 2025-03-30

    2025-03-30

    The citation count of "A self-learning dynamic path planning method for evacuation in large public buildings based on neural networks" on Google Scholar is 100. 

    Evacuation path planning is of significant importance to safely and efficiently evacuate occupants inside public buildings. Current computer simulation methods carry out evacuation analysis and then provide emergency education and management with a vivid virtual environment. However, efficient evacuation path planning approaches for evacuation guidance still meet the challenges of generating the analysis models, and lacking of real-time analysis methods under dynamic circumstances. In this study, a dynamic path planning approach based on neural networks is proposed for evacuation planning in large public buildings. First, an automatic process to develop the evacuation analysis model with simplified but sufficient information is presented. Then a path generation algorithm is proposed, together with an evaluation process, to generate a number of training sets for policy neural networks. When the primary policy neural network is preliminarily trained, it falls into a self-learning iteration process. Finally, the approach embeds a dynamic algorithm to simulate the mutual influences among all occupants in the building. The neural network was trained according to a real large public building and then the approach managed to provide rapid and feasible evacuation guidance for both occupants to escape in multiple scenarios and managers to design the evacuation strategy. Test results showed that the proposed approach runs 8–10 times faster than existing software and traditional search algorithms.

    Note: Neurocomputing is a journal of engineering technology in Q2 area. I am the corresponding author of the paper.

  • 2025-03-21

    2025-03-21

    A Multi-Factor-Fusion Framework for Efficient Prediction of Pedestrian-level Wind Environment Based on Deep Learning Published in IEEE Access.

    Efficient and accurate assessment of the Pedestrian-Level Wind Environment is essential to maintain a healthy and safe urban living environment. Numerical simulations, such as computational fluid dynamics and multi-scale modeling techniques, are commonly used for wind environment analysis. However, they are computationally intensive and time-consuming, particularly when dealing with the complexities of urban landscapes. This study proposes a novel Multi-Factor-Fusion (MFF) framework that leverages deep learning techniques. This framework integrates Graph Convolutional Networks and Long Short-Term Memory networks to extract and fuse multiple factors and create an end-to-end neural network model capable of directly predicting wind fields. By avoiding the need for grid division and iterative calculations, the framework significantly enhances the efficiency of wind environment analysis. Furthermore, multi-scale simulation data is used to train the model and correct the predictive results, ensuring the accuracy of the final results. This innovative approach has the potential to revolutionize the Pedestrian-Level Wind Environment prediction by achieving a trade-off between efficiency and accuracy.

    Note: IEEE Access is an important SCI journal in the field of engineering technology. I am the first author and corresponding author of the paper. The research results were supported by the National Key Research and Development Program of China, and the Guangdong Basic and Applied Basic Research Foundation and the National Natural Science Foundation of China.