Mengmeng Sun
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mengmeng Sun has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Chaoxian Yang1 shared publication
- Chenzhi Cai
School of Civil Engineering, Central South University, Changsha, China
1 shared publication - Panping Liu1 shared publication
- Qiu-Sheng Li
City Univ. of Hong Kong
1 shared publication - Shangyu Hu
Department of Civil Engineering and Smart Cities, College of Engineering, Shantou University, Shantou 515063, P. R. China
1 shared publication - Shiji Huang
School of Civil Engineering, Central South University, Changsha, P. R. China
1 shared publication
Latest Publications
- Open Access
- 15 April 2026
- Published by Elsevier in SSRN Electronic Journal
Standardizing Road Traffic Safety Facility Operation and Maintenance: A Multi-Agent Large Language Model Approach
With the rapid expansion of road networks, there is an increasingly urgent demand for intelligent management of Traffic Safety Facilities (TSFs). Current operation and maintenance practices still rely heavily on manual expertise, resulting in low efficiency and a high risk of human error. Although large language models (LLMs) have demonstrated strong reasoning capabilities, their application to TSF-related tasks is constrained by limited domain knowledge and delayed knowledge updates. To address these challenges, this study proposes a knowledge-base-enhanced multi-agent collaborative reasoning approach for TSF-related tasks. A two-stage knowledge modeling strategy is developed to transform heterogeneous and regulation-intensive source knowledge into hierarchical knowledge units with semantic fingerprints, thereby enabling structured knowledge storage and efficient retrieval. On this basis, a multi-agent collaborative framework is further constructed to support intelligent task routing, hierarchical query augmentation, dense-sparse collaborative retrieval, multidimensional re-ranking, and iterative self-reflective generation. Meanwhile, a TSF question-answering (QA) benchmark is established, and systematic comparative experiments are conducted on closed-ended detail discrimination, open-ended comprehensive generation, and representative case analysis against original open-source LLMs, Naive Retrieval-Augmented Generation (RAG), Light RAG, and state-of-the-art closed-source models. Experimental results demonstrate that the proposed approach achieves superior performance in factual accuracy, semantic consistency, response completeness, and regulatory compliance, while effectively mitigating hallucinations without requiring large-scale fine-tuning. The proposed method provides a promising technical pathway for intelligent TSF operation and maintenance as well as engineering decision support.
- 15 April 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessStandardizing Road Traffic Safety Facility Operation and Maintenance: A Multi-Agent Large Language Model Approach
With the rapid expansion of road networks, there is an increasingly urgent demand for intelligent management of Traffic Safety Facilities (TSFs). Current operation and maintenance practices still rely heavily on manual expertise, resulting in low efficiency and a high risk of human error. Although large language models (LLMs) have demonstrated strong reasoning capabilities, their application to TSF-related tasks is constrained by limited domain knowledge and delayed knowledge updates. To address these challenges, this study proposes a knowledge-base-enhanced multi-agent collaborative reasoning approach for TSF-related tasks. A two-stage knowledge modeling strategy is developed to transform heterogeneous and regulation-intensive source knowledge into hierarchical knowledge units with semantic fingerprints, thereby enabling structured knowledge storage and efficient retrieval. On this basis, a multi-agent collaborative framework is further constructed to support intelligent task routing, hierarchical query augmentation, dense-sparse collaborative retrieval, multidimensional re-ranking, and iterative self-reflective generation. Meanwhile, a TSF question-answering (QA) benchmark is established, and systematic comparative experiments are conducted on closed-ended detail discrimination, open-ended comprehensive generation, and representative case analysis against original open-source LLMs, Naive Retrieval-Augmented Generation (RAG), Light RAG, and state-of-the-art closed-source models. Experimental results demonstrate that the proposed approach achieves superior performance in factual accuracy, semantic consistency, response completeness, and regulatory compliance, while effectively mitigating hallucinations without requiring large-scale fine-tuning. The proposed method provides a promising technical pathway for intelligent TSF operation and maintenance as well as engineering decision support.
- 13 March 2026
- Published by Elsevier in Measurement
- Vol. 273
Evaluation of dynamic characteristics of high-rise buildings with or without tuned liquid dampers under typhoon conditions
- 13 March 2026
- Published by Elsevier in Measurement
- Vol. 273
Evaluation of dynamic characteristics of high-rise buildings with or without tuned liquid dampers under typhoon conditions
- 11 November 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Structural Stability and Dynamics
Modal Identification of a Wind Turbine Via a Hybrid Operational Modal Analysis Scheme Considering Harmonic Interference
Operational Modal Analysis (OMA) is an essential tool for determining structural dynamic properties of wind turbines. However, the presence of harmonic excitation in wind turbines poses a significant challenge for the effective application of traditional white noise-based OMA techniques. When some harmonic frequencies are close to structural frequencies, harmonics may introduce bias in modal estimations or be misidentified as structural modes, thereby resulting in inaccurate determination of structural dynamic properties of wind turbines. To this end, a hybrid OMA scheme is developed in this paper to mitigate harmonic disturbances in vibration responses and thus enable performing precise modal estimation with uncertainty quantification for wind turbines. The noise robustness, stability, and accuracy of the hybrid OMA scheme are verified through a numerical simulation study. Furthermore, the hybrid OMA scheme is employed to evaluate the structural dynamic properties of a three-bladed 2.0 MW wind turbine for assessing its feasibility in practical applications. The primary objective of this study is to provide a hybrid OMA scheme that can accurately estimate modal parameters of wind turbines and other structures with rotating components.
- 11 November 2025
- Published by World Scientific Pub Co Pte Ltd in International Journal of Structural Stability and Dynamics
Modal Identification of a Wind Turbine Via a Hybrid Operational Modal Analysis Scheme Considering Harmonic Interference
Operational Modal Analysis (OMA) is an essential tool for determining structural dynamic properties of wind turbines. However, the presence of harmonic excitation in wind turbines poses a significant challenge for the effective application of traditional white noise-based OMA techniques. When some harmonic frequencies are close to structural frequencies, harmonics may introduce bias in modal estimations or be misidentified as structural modes, thereby resulting in inaccurate determination of structural dynamic properties of wind turbines. To this end, a hybrid OMA scheme is developed in this paper to mitigate harmonic disturbances in vibration responses and thus enable performing precise modal estimation with uncertainty quantification for wind turbines. The noise robustness, stability, and accuracy of the hybrid OMA scheme are verified through a numerical simulation study. Furthermore, the hybrid OMA scheme is employed to evaluate the structural dynamic properties of a three-bladed 2.0 MW wind turbine for assessing its feasibility in practical applications. The primary objective of this study is to provide a hybrid OMA scheme that can accurately estimate modal parameters of wind turbines and other structures with rotating components.