Research Projects
On-going
PI: AU (Startup fund), Core technologies for disaster resilience of complex structural systems, 2023.03.01 ~ 2025.02.28, KRW 40,000,000
PI: NRF (Basic research project), Development of resilience assessment technologies for structural systems under multi-hazards, 2023.06.01 ~ 2025.12.31, KRW 154,000,000
Completed
–
Research Topics
1. Deep learning-based seismic response prediction of structures
The aim of this research topic is to develop a practical and computationally efficient method for predicting the seismic responses of structural systems using deep learning. To achieve this goal, our group first develops a seismic demand database by conducting a large number of dynamic analyses. Then, we introduces an appropriate deep neural network model to predict the peak responses of structural systems. The developedmethods can be used in regional seismic loss assessment that requires seismic responses of a set of structural systems and can replace nonlinear static procedures such as the coefficient method in ASCE 41-17 or the capacity spectrum method in FEMA 440. Moreover, we have made the supporting source codes and interactive visualization web service available at http://ERD2.snu.ac.kr.
2. Assessment of disaster resilience performance of civil infrastructures
The research group focuses on developing a framework to assess the resilience performance of civil infrastructure systems, including individual structures, lifeline networks, and urban communities. Compared to the conventional risk management framework, the resilience analysis considers not only the initial disruptions but also the recovery phase to account for short- and long-term impacts. The research group has made two main contributions regarding this research topic. First, we have developed indices that assess the resilience performance of systems based on probability and reliability theories. Second, we have proposed efficient algorithms to reduce computational costs in the resilience analysis. The research group is actively collaborating with distinguished researchers, including Dr. Sang-ri Yi, Dr. Jihwan Kim, and Dr. Chulyoung Kang, to advance this research topic further.
3. Manufacture of differential load cells
A hybrid simulation is an effective approach when it is difficult to represent either the structural components or external forces. In real-time aeroelastic hybrid simulation (RTAHS), the wind-induced force from a physical model is directly measured with sensors, while the structural systems are numerically modeled. However, due to the dynamic simulation’s characteristics, a commercial force sensor measures a resultant force of wind-induced force and the specimen’s inertia. The measure inertial hampers calculating the corresponding structural responses. To overcome this challenge, this research group developed a new force sensor that can accurately measure wind-induced force without the influence of the specimen’s inertia. The application of this new load cell is not limited to RTAHS but is also applicable to measure force in real-time when the systems are in motion.
4. Statistical/machine learning for assessing performance of structural systems under uncertainties
There exist numerous uncertainties in both load effects and structural systems, which present two primary issues. First, to obtain a comprehensive understanding of the performance of structural systems under natural and man-made hazards, a large number of analyses are often required. Second, identifying inherent patterns between loads and systems is challenging. To address these research needs, our group proposes various statistical and machine learning methods to reduce the computational costs of structural analyses. Furthermore, we investigate the developed models and the database for training these models to identify inherent patterns between input and output, and perform uncertainty quantification to determine essential random variables. This framework is applicable not only to individual structures but also to urban communities under different hazards.
IDRG 