2024-05-22

Congratulations to Ning Houchun, Zhang Xiaobing, and Zhang Jiahong on successfully passing the thesis defense for their Master's degrees at Tsinghua University!

Ning Houchun's thesis is titled "Research on attitude monitoring and optimization technology of offshore floating wind turbines based on digital twins." This study investigates real-time monitoring and analysis techniques for the attitude of offshore floating wind turbines, with a particular focus on accurately capturing the posture of the column. Real-time data collection and cleansing are performed using the Internet of Things and data cleansing technology, while a digital twin platform enables three-dimensional dynamic visualization of the operational state. Additionally, finite element and non-linear regression methods are employed to analyze column deformations and derive precise mathematical models for capturing the deformation characteristics under various wind and wave conditions.

Zhang Xiaobing's thesis is titled "Highway Engineering Safety Management Based on Knowledge Graph and Data Template." This research innovatively explores the field of highway engineering safety management using techniques such as ontology, knowledge graph, and data templates. By dividing data dimensions, constructing semantic models, and establishing a standardized knowledge repository, deep extraction and management of knowledge in the engineering domain are achieved. Ultimately, a safety management template is designed to transform ontology into a user-friendly format, assisting managers in accurately identifying operational scenarios and acquiring necessary knowledge.

Zhang Jiahong's thesis is titled "Code-Based Knowledge Extraction and Application in Subway Engineering." This study employs deep learning, large language models, ontology, and knowledge graph techniques to systematically analyze, extract, and manage knowledge in the field of subway engineering. Through the analysis of 143 engineering specification documents, a knowledge extraction method is developed to automatically identify entity elements and construct a knowledge graph of subway engineering specifications. Furthermore, an enhanced model architecture that integrates knowledge graph and large language models is proposed, along with the development of a platform with knowledge navigation, management, and question-answering capabilities.

2024-05-09

Improving interoperability between architectural and structural design models:  An industry foundation classes-based approach with web-based tools has been cited more than 100 times..

Medium to large construction projects often involve multiple structural consultants who use a variety of structural analysis applications. Interoperability between these applications and technologies is insufficient, and there is still not enough research on interoperability issues in the field of structural engineering. In this paper, a new method combining unified information model and algorithm based on IFC is proposed. The method is designed to achieve bidirectional transformation between architectural and structural models, as well as accurate transformation between multi-structural analysis models, and direct export of IFC architecture tools through apis, and has been successfully applied to four large-scale practical projects. The test results show that the method not only successfully completes the bidirectional conversion between the four structural analysis tools, but also compares the conversion effect with that of the direct link tool, which proves its accuracy and effectiveness. In addition, the platform supports centralized and remote collaboration and can run in both Client Server (C/S) and Browser Server (B/S) environments, providing strong technical support for construction projects. This paper has been widely concerned and cited because of its solution to practical problems, technological innovation, empirical testing, wide applicability and advancement of structural engineering research.

Note: Automation in Construction is the top issue in the field of engineering technology, and the current impact factor is 10.3. I am the first author and corresponding author of the paper.

2024-05-06

The paper titled "Fine-tuning vision foundation model for crack segmentation in civil infrastructures" has been published in the journal Construction and Building Materials.

Large-scale foundation models have become the mainstream deep learning method, while in civil engineering, the scale of AI models is strictly limited. In this work, a vision foundation model is introduced for crack segmentation. Two parameter-efficient fine-tuning methods, adapter and low-rank adaptation, are adopted to finetune the foundation model in semantic segmentation: the Segment Anything Model (SAM). The fine-tuned CrackSAM shows excellent performance on different scenes and materials. To test the zero-shot performance of the proposed method, two unique datasets related to road and exterior wall cracks are collected, annotated and open-sourced, for a total of 810 images. Comparative experiments are conducted with twelve mature semantic segmentation models. On datasets with artificial noise and previously unseen datasets, the performance of CrackSAM far exceeds that of all state-of-the-art models. CrackSAM exhibits remarkable superiority, particularly under challenging conditions such as dim lighting, shadows, road markings, construction joints, and other interference factors. These cross-scenario results demonstrate the outstanding zero-shot capability of foundation models and provide new ideas for developing vision models in civil engineering.

Note: The first author of this paper is Ge Kang, a master's student in 2022, and Assistant Professor Guo Yutao is the corresponding author. The research findings were supported by  the National Natural Science Foundation of China and tthe Cross-disciplinary Research and Innovation Fund Research Plan of Tsinghua Shenzhen International Graduate School.

2024-03-24

Recently, Huaxia Construction Science and Technology Award Committee released the "2023 China Award for Science and Technology in Construction" award project list, by my participation in the completion of the "Key technologies and applications of digital construction for urban metro driven by digital twins" project, won the 2023 "China Award for Science and Technology in Construction" second prize.

China Award for Science and Technology in Construction is one of the most influential science and technology awards in the field of housing and urban and rural construction in China, aiming to recognize organizations and citizens who have made significant contributions to the construction industry, promote the application of scientific and technological achievements and productivity transformation, and improve the overall technical strength of the housing and urban and rural construction industry. Our team won this award, which fully reflects our technical strength and innovative value in the field of digital twin technology and urban subway construction. In the future, we will continue to improve our technologies and methods, promote the process of digital transformation and upgrading, and contribute to the development of urban transport infrastructure.

Note: The project was led by myself, Wang Wei, Lin Jiarui, Zou Dong, Zhang Jianping, Guo Yutao, Wang Hongdong, Cui Libo, Xue Zhigang, Chen Xiangxiang, Wu Zhen and Zhang Bangchao were the main practitioners. The main completion units include Guangzhou Metro Group Co., LTD., Tsinghua SIGS, Tsinghua University, Guangzhou Metro Construction Management Co., LTD., China Railway Construction South China Construction Co., LTD., Guangzhou Rail Transit Construction Supervision Co., LTD., Beijing Cloud Construction Technology Co., LTD. 

2024-02-27

Inertia load reduction for loadoff during floating offshore wind turbine installation: Release decision and ballast control has been published in the journal Sustainable Horizons.

High offshore installation costs are a significant factor limiting the competitiveness of offshore wind energy. One efficient installation approach for floating offshore wind turbines is to preassemble the tower, nacelle, and rotor onshore and perform a single lifting operation to mate the superstructure with the floating foundation at the installation site. It is heavy lifting, due to the weighty payload. At the end of the mating process, a loadoff operation is conducted to transfer the preassembly to the floating foundation. It results in a sudden change in total force acting on the vessel and causes substantial acceleration and potential damage to the mechanism in the onboard nacelles. The magnitude of acceleration of the onboard nacelles can vary greatly at different release instants. In this research, a simplified two-degrees-of-freedom (DOF) (heave and pitch) model is also proposed to account for the heavy lifting process and variable ballast tanks. The sudden payload transfer is approximated using a hyperbolic tangent function to guarantee continuity and differentiability. The loadoff operation consists of the decision-making and vessel-stabilizing phases. Based on the nonlinear model predictive control method, a payload-transfer time selector and anti-pitch ballast controller have been developed to achieve optimal release time decisions and stabilize the vessel after payload release, respectively. Six-DOF simulation results show that the proposed algorithms are capable to a satisfying level of robustness of deciding the optimal payload release time instant, as well as limiting the peak and mean acceleration magnitudes of the onboard nacelles after payload release. The decision-making and control strategies may promote the sustainable energy transformation by extending the operation window and reduce the installation costs.

Note: The first author of this paper is Ma Can, a master's student in 2021, and Assistant Professor Ren Zhengru is the corresponding author. The research results were supported by Shen-zhen Science and Technology Program, Shen-zhen Science & Technology Commission, and Shenzhen Science and  the Guangdong Basic and Applied Basic Research Foundation.