Research

Research projects

On-going

PI: Multimodal Deep Learning for Early Warning of Hazardous Vibrations, 2023-2026, National Research Foundation (우수신진: KRW 520M)
Co-PI: Multipurpose Coastal Floating Infrastructure Technology, 2023-2027, Korea Agency for Infrastructure Technology Advancement (KRW 500M)

Completed

Co-PI: Big Data Lab for Building Structures Retrofitted with Carbon Composite Materials in Multi-Hazard based on Natural Fire, 2021-2023, National Research Foundation (Basic Resarch Laboratory: KRW 1,375M)
PI: VibNET: Integrated framework for classification and assessment of multiple types of hazardous vibrations, 2021-2022, National Research Foundation (생애첫연구: KRW 60M)
PI: Development of real-time hazardous vibration detection algorithm based on CCTV images, 2021, Univeristy of Seoul (KRW 24M)

Multimodal deep learning for early warning of hazardous vibrations

Our current research mainly focuses on the integration of multi-modal data for improved condition assessment of cable-supported bridges. Our approach involves the extraction of comprehensive information from a variety of sources (modals), including traditional vibration and environmental data, images, sounds, videos, and results from on-site inspections, visual inspections, and CCTV image analysis. The concatenation of multiple modalities may allow the enhanced performance of deep learning model and enable preemptive prediction and early warning of various harmful vibrations in cable-supported bridges. We believe that to assess the condition of these structures accurately, information must be collected from multiple sources; and therefore, this research represents a holistic and data-driven approach towards ensuring the safety and longevity of our civil infrastructures.

Selected publications

Park S, Kim S*, Enhancing Field Applicability of Unsupervised Vibration-Based Dam-age Assessment via Time-Frequency Pattern Differentiation, Journal of Civil Structural Health Monitoring. 2025 (Accepted).
Park S, Kim S*, Hybrid CAE–DSVDD for Unsupervised Vibration-Based Damage Detection in In-Situ Steel Truss Bridge, Structural Health Monitoring. 2025 (Accepted).
Park S, Kim S*, Enhancing vibration-based damage assessment with one dimensional CNN: parametric studies and field applications, KSCE Journal of Civil Engineering. 2024 Mar 18:1-8.
Lee H, Yoon H, Kim S*. Vibration detection of stay-cable from low-quality CCTV images using deep-learning-based dehazing and semantic segmentation algorithms, Engineering Structures. 2023 Oct; 292: 116567.

Data-driven vibrational serviceability assessment

With a focus on addressing the challenge of vibrational serviceability issues in our civil infrastructures, this research aims to uncover the root causes of vibrations and their impact. Excessive vibrations can result in discomfort for users and threaten the longevity of these structures, and thus a multi-faceted approach has been taken to tackle this problem. This approach encompasses wind environmental analysis, wind tunnel testing, big data analysis, computer-vision algorithm, and output-only modal analysis. Our goal is to not only enhance the understanding of vibrations in these large-scale infrastructures but also provide practical solutions that ensure their continued safe and sustainable operation.

Selected publications

Kim S, Kim HM, Hwang Y*, Data-Driven Dynamic Response Forecasting and Anomaly Detection in Long-Span Bridges, Journal of Civil Structural Health Monitoring. 2025 Apr
Lee H, Yoon H, Kim S*. Vibration detection of stay-cable from low-quality CCTV images using deep-learning-based dehazing and semantic segmentation algorithms, Engineering Structures. 2023 Oct; 292: 116567.
Kim S, Kim SJ*, Kim HK, High-mode vortex-induced vibration of stay cable: monitoring, cause investigation, and mitigation, Journal of Sound and Vibration. 2022 Apr; 524: 116758
Hwang YC, Kim S, Kim HK*. Cause investigation of high-mode vortex-induced vibration in a long-span suspension bridge. Structure and Infrastructure Engineering. 2019 Apr 29:1-0.
Park J, Kim S, Kim HK*. Effect of gap distance on vortex-induced vibration in two parallel cable-stayed bridges. Journal of Wind Engineering and Industrial Aerodynamics. 2017 Mar 31;162:35-44.

Imperfect-data-driven VIV identification

Our research focus is on the development of data-driven methods for accurate identification of vortex-induced vibrations (VIVs) in the vibrational serviceability assessments of long-span bridges. Recognizing the challenges posed by imbalanced datasets, we aim to improve predictability by exploring one-class regression models, minority data augmentation, and semi-supervised labeling techniques. Currently, we are investigating ways to overcome the laborious task of manual labeling in VIV detection, by exploring synthetic data augmentation and unsupervised learning methods. Our ultimate goal is to contribute towards a label-free, robust framework for VIV identification.

Selected publications

Lee S, Kim S*, Pointwise Vortex-induced Vibration Detection: Learning from Time-Series Synthesis Data, Engineering Structures. 2025 Mar; 326: 119525.
Kim S, Lee SH, Kim S*. Pointwise multiclass vibration classification for cable-supported bridges using a signal- segmentation deep network, Engineering Structures. 2023 Mar; 279: 115599.
Lee SH, Kim S*. Unsupervised Vortex-induced Vibration Detection Using Data Synthesis. Journal of the Computational Structural Engineering Institute of Korea. 2023; 36(5): 315-321.
Kim S, Kim T*, Machine-learning-based prediction of vortex-induced vibration in long-span bridges using limited information, Engineering Structures. 2022 Sep; 266(1): 114551
Lim JY, Kim S*, Kim HK, Using supervised learning techniques to automatically classify vortex-induced vibration in long-span bridges, Journal of Wind Engineering and Industrial Aerodynamics. 2022 Feb; 221: 104904

Fault-tolerant output-only damping identification

The frequency of sensor faults found in the long-term monitored data of large-scale structures is a potential cause of errors in the damping estimates. We proposed a machine-learning-based fault-data management approach whereby erroneous data are identified and removed automatically. ML algorithm is used to automatically detect and recover/isolate multiple types of sensor faults from measured accelerations. The labeled training samples are artificially augmented using digital simulation of a random process with an envelope function. The recovered data provided a more robust and consistent damping estimate, and demonstrated the efficacy of the proposed fault-data management strategy. Currently, we are pursuing the study on underdetermined operational modal analysis for limited sensor data.

Selected publications

Kim S, Kim HK*, Spencer BF, Automated damping identification of long-span bridge using long-term wireless monitoring data with multiple sensor faults, Journal of Civil Structural Health Monitoring. 2022 Feb; 12:465-479.
Kim S, Kim HK*, Hwang YC. Enhanced Damping Estimation for Cable-Stayed Bridges Based on Operational Monitoring Data. Structural Engineering International, 2018 Jul 3;28(3): 308-317.
Kim S, Kim HK*. Damping Identification of Bridges under Nonstationary Ambient Vibration. Engineering. 2017;3(6):839-44.

Probability modeling of modal parameters of long-span bridges under environmental and operational variations

Our research aims to investigate on the long-term damping characteristics of cable-stayed bridges under environmental and operational variations. To achieve this, we have applied an optimized OMA-based damping estimation algorithm to 2.5 years of monitoring data collected from a twin cable-stayed bridge. We also proposed a novel framework to estimate modal damping ratios using a dynamic displacement reconstruction technique based on operational modal analysis of measured acceleration data. Our analysis has uncovered several key findings, including: (1) fluctuations in the long-term damping ratios over different seasons; (2) the impact of aerodynamic interference on the dynamic behavior of the bridge; (3) the dependency of the damping ratio on amplitude; and (4) log-normal models that describe the damping characteristics. We believe that this research will contribute to the understanding of damping characteristics of long-span bridges and inform future designs that can better withstand the demands of real-world conditions.

Selected publications

Hwang D, Kim S*, Kim HK, Deep Gaussian process regression for damping of a long-span bridge under varying environmental and operational conditions, Journal of Civil Structural Health Monitoring. 2023 Jun 18:1-5.
Hwang D, Kim S*, Kim HK. Long-Term Damping Characteristics of Twin Cable-Stayed Bridge under Environmental and Operational Variations. Journal of Bridge Engineering. 2021 Sep 1;26(9):04021062.
Kim S, Park J, Kim HK*. Damping identification and serviceability assessment of a cable-stayed bridge based on operational monitoring data. Journal of Bridge Engineering. 2016 Oct 20;22(3):04016123.
Kim SJ, Kim HK*, Calmer R, Park J, Kim GS, Lee DK. Operational field monitoring of interactive vortex-induced vibrations between two parallel cable-stayed bridges. Journal of wind engineering and industrial aerodynamics. 2013 Dec 31;123:143-54.

Page updated

Google Sites

Report abuse