Research at Kyushu University

Research Projects and Initiatives

Recent Studies at Faculty of Design

Estimating the seismic response of building from earthquake damage photographs

Department of Environmental Design, Faculty of Design
Associate Professor, Tomokazu Yoshioka

Our laboratory researches and develops technologies to continue using buildings damaged by severe earthquakes. Using the results of our studies, we aim to improve the resilience of cities and regions against earthquake disasters. Here, we introduce one recent research result, an application of image classification [1] using convolutional neural networks to earthquake disaster prevention.

■Estimation of seismic response of damaged condominiums

Shear failure of reinforced concrete non-structural walls that aren’t detached from the structural frame properly is frequently observed as typical earthquake damage in condominiums. Even if such earthquake damage occurs, the damage to the structural frame supporting the building is less than small damage [2]. Therefore, in many cases, the damaged building can be restored and can continue to be used (Figure 1). However, on-site surveys and determination of the degree of damage classification by skilled engineers are required to determine whether the building can continue to be used or not.

Figure 1 Characteristics of earthquake damage to condominiums [2]

Therefore, as an alternative to field surveys by specialists lacking during earthquake disasters, we are attempting to estimate the maximum rotation angle using semantic segmentation to detect damage and image measurement from earthquake damage photographs of RC non-structural walls taken by non-specialists (Figure 2) [3]. In addition, we are developing a method to estimate the seismic response of the structural frame and automatically determine the degree of damage (very minor, small, medium, large, or collapse) based on the estimated maximum rotation angle of the RC non-structural wall.

Figure 2 Estimation of the rotation angle of RC non-structural wall based on image measurement

(1) Tomokazu YOSHIOKA, A GENERATION OF DAMAGE CLASSIFIER FOR RC PARTIAL WALL USING DAMAGE PHOTOGRAPH BY DEEP LEARNING, AIJ Journal of Technology and Design, Volume 26, Issue 64, Pages 1252-1257, 2020.10,
(2) Tomokazu Yoshioka, Chapter 10 Damage of Steel Reinforced Concrete Structure 10.3 Examples 10.3.2 Detailed survey example, Report on the Damage Investigation of the 2016 Kumamoto Earthquake, AIJ, p.237-242, 2018.10
(3) Tomokazu YOSHIOKA, Kousuke TAKEDA, AN ESTIMATION OF CHORD ROTATION OF RC PARTIAL WALL BASED ON IMAGEMEASURED CONCRETE SPALLING AREA, AIJ Journal of Technology and Design, Volume 28, Issue 70, Pages 1224-1229, 2022.10 (printing)

Department of Environmental Design, Faculty of Design
Associate Professor, Tomokazu Yoshioka