2026-06-12

The paper entitled “Precise longitudinal crack detection via continuous texture reconstruction and deep segmentation” has been published in Advanced Engineering Informatics.

Pavement cracks are a major form of road infrastructure degradation, necessitating efficient and accurate detection for timely maintenance. Existing inspection methods rely either on labor-intensive manual surveys or automated systems constrained by high hardware costs and GPS dependency, which limits their flexibility for continuous surface assessment. To address these limitations, this paper introduces a dual-channel crack detection model that integrates continuous pavement texture reconstruction with deep segmentation and high-precision boundary refinement algorithms, enabling on-site implementation and accuracy enhancement for longitudinal crack detection. A feature-based image stitching algorithm is developed to reconstruct continuous pavement textures from high-resolution images, enabling GPS-free crack localization. The proposed method further combines the YOLOv8-seg model with adaptive morphological operations to achieve pixel-level crack reconstruction. Comparative experiments reveal that the hybrid approach achieves superior segmentation performance, with finer boundary delineation and improved branch recovery compared to the baseline model. This paper provides a practical solution for automated inspection and high-fidelity reconstruction of longitudinal cracks, effectively supporting pavement maintenance planning.

Note: Advanced Engineering Informatics is a Q1 journal and a top-tier publication in the field of engineering and technology, with a 2025 impact factor of 9.9. The first author of the paper is PhD student Peng An, and I serve as the corresponding author. This research was supported by the National Key Research and Development Program of China and Shenzhen Science and Technology Program.

2026-06-03

The paper entitled “Numerical investigation of elliptical ribbed manifold microchannel heat sink and multi-objective optimization utilizing machine learning” has been published in Thermal Science and Engineering Progress.

This study addresses the thermal management challenges associated with high-heat-flux chips by focusing on the flow and heat transfer characteristics, as well as structural optimization, of an elliptical ribbed manifold microchannel heat sink (MMCHS). In response to the insufficient understanding of how microchannel height, microchannel width, elliptical rib dimensions, and coolant volumetric flow rate affect pressure drop and heat dissipation performance, extensive numerical simulations were conducted. The study systematically investigated the effects of different channel geometries, rib configurations, and volumetric flow rates on the coupled flow and heat transfer behavior within the elliptical ribbed MMCHS. The results revealed the variation patterns of cooling performance, pressure drop, Nusselt number, friction factor, and thermal enhancement efficiency under different design parameters, and further clarified the underlying mechanisms governing these performance metrics and the associated thermal irreversibility. On this basis, a high-accuracy artificial neural network model was developed to predict the substrate bottom temperature and pressure drop. By integrating the artificial neural network with a genetic algorithm, multi-objective optimization was performed to obtain a Pareto front that simultaneously minimizes substrate bottom temperature and pressure drop. The results show that the optimized design can achieve a 1.36%–3.75% improvement in thermal enhancement efficiency within a lower Reynolds number range, providing valuable engineering guidance for chip thermal management under pumping-power-limited conditions.

Note: Thermal Science and Engineering Progress is a Q1 journal in the field of engineering and technology, with a 2025 impact factor of 5.4. The first author of the paper is doctoral student Shoujun Chen, and Professor Sunwei Li is the corresponding author. This work was supported by the Guangdong Basic and Applied Basic Research Foundation and the Tsinghua Shenzhen International Graduate School–Shenzhen Pengrui Young Faculty Program.

2026-05-29

The "Deep Water Structure Health Monitoring Method and System" has been granted an invention patent certificate (Patent No.: ZL202610409127.1).

This invention discloses a deep-water structural health monitoring method and system. The method includes: establishing a discrete model and constructing an element strain matrix; collecting single-sided surface strain data and establishing a mapping relationship between theoretical surface strain and node displacements; formulating a single-sided inverse finite element data fitting term that incorporates a surface strain fitting component and a transverse shear strain penalty term; constructing a geometry-guided mixed regularization term composed of membrane and bending regularization components; assembling the global equations to solve for full-field displacements; recovering full-field strains and stresses, and outputting monitoring results. The system comprises a data acquisition module, a preprocessing and inverse reconstruction module, a stabilization solving module, and a visualization output module. Based on a full-field response perception scheme combining single-sided inverse finite element analysis with geometry-guided mixed regularization, this invention enables stable and high-precision inversion of full-field displacements, strains, and stresses using only sparse single-sided sensor data, effectively overcoming the dependency on bilateral sensor configurations in traditional methods. 

Note: The inventors of this patent also include Ph.D. student Jiang Chunhao, Ph.D. student Min Yantao, Associate Professor Guo Yutao, and Associate Professor Li Binbin.

2026-05-22

The paper "Assessing coastal groundwater risks from nuclear-contaminated water discharge under extreme drought and sustainable recharge solutions" has been published in Journal of Contaminant Hydrology.

The discharge of nuclear-contaminated water from the Fukushima Daiichi Nuclear Power Plant in Japan has raised concerns about seawater quality and potential environmental risks to coastal groundwater systems, particularly in densely populated metropolitan areas. In response to this pressing challenge, this study developed a radionuclide decay and transport model tailored to the hydrogeological conditions of Shanghai, aiming to comprehensively assess the combined impacts of nuclear discharge, extreme drought, and Managed Aquifer Recharge (MAR) on groundwater quality. The results show that under extreme drought conditions, declining groundwater levels may exacerbate seawater intrusion and intensify seawater–groundwater interactions, resulting in higher radionuclide concentrations than those observed under normal scenarios. Model simulations indicate that, under the assumed input scenario, after 30 years of continuous discharge, the concentrations of ¹³⁷Cs and ⁹⁰Sr at a depth of 160 m in Shanghai's aquifer increase by 33.93% and 46.40%, respectively. Meanwhile, MAR demonstrates significant potential for risk mitigation, reducing the concentrations of ¹³⁷Cs and ⁹⁰Sr by up to 99.58% and 99.10%, respectively. These findings highlight the critical role of proactive, nature-based interventions such as MAR in mitigating nuclear contamination risks and provide important scientific support for enhancing the resilience and sustainability of coastal groundwater resources.

Note: Journal of Contaminant Hydrology is a Q1 journal in the field of Environmental Sciences and Ecology, with a 2025 impact factor of 4.4. The corresponding authors of the paper include Prof. Yaqiang Wei from Shanghai University, Prof. Xinde Cao from Shanghai Jiao Tong University, and Prof. Hui Li from Shanghai University. This research was supported by the National Natural Science Foundation of China.

2026-05-21

Congratulations to Jiang Xiyuan, Su Ziqing, and Feng Yixiao on successfully defending their Tsinghua University Master’s theses!

Jiang Xiyuan's thesis is titled "Rapid Prediction of Jacket Structural Mechanical Response Based on MultiGraph Fusion Network." Focusing on the rapid prediction of jacket mechanical responses, this study abstracts the structure into a multi-dimensional topological graph system and constructs a multi-graph fusion prediction model combining graph neural networks and multilayer perceptrons, enabling the joint learning of marine environmental parameters and structural features. For ultra-large deepwater jackets, a hierarchical graph modeling approach based on local topology is proposed to reduce learning complexity, and an intelligent design platform is developed. This research provides new insights for the efficient analysis and lifecycle intelligent management of offshore engineering structures.

Su Ziqing's thesis is titled "Research on the Abnormal Early Warning of Operation Progress for Cross-Sea Bridges Based on Enhanced Knowledge Graph." This study introduces knowledge graph and ontology engineering technologies to investigate intelligent monitoring and anomaly warning technologies for cross-sea bridge operation progress, with a particular focus on the semantic fusion of multi-source heterogeneous data. Lightweight detection and few-shot learning algorithms are utilized for visual element perception, and automated recognition of operation behaviors is achieved through spatial coupling degree analysis. Meanwhile, a progress warning strategy is established based on the constructed ontology model and dynamic knowledge graph, promoting the transformation of cross-sea bridge operations from “passive response” to “active warning” and “intelligent decision-making”.

Feng Yixiao's thesis is titled "Multibody Dynamics-Based Adaptive Task Scheduling Method for Smart Cities." Addressing the real-time scheduling challenges in smart city edge computing, this study proposes an adaptive task scheduling method based on multibody dynamics principles. By constructing a cloud–edge–device collaborative evaluation model and designing physical field mapping and rigid-body dynamics evolution mechanisms, tasks and resources are mapped into virtual entities to achieve self-organized scheduling in dynamic environments. A grid index is introduced to reduce computational overhead, and a simulation platform is built using real data and a physics engine for validation. This method offers an efficient, robust, and low-overhead solution for real-time resource management in large-scale heterogeneous edge networks. Notably, this thesis was awarded the title of Outstanding Master’s Thesis of Tsinghua University.