Postearthquake Damage Classification of Buildings Using a Recording of Low-Cost IoT Devices

People live in a temporary shelter after an earthquake due to the unknown severity of the damage level of buildings. Recently, machine-learning-based rapid building damage classification after the earthquake has increased considering the structural and seismic characteristics. The structural properties can be obtained using Google Earth, Open Street Map, unmanned aerial vehicles, etc. However, it is challenging to obtain the ground motion characteristics. The dense seismic networks developed in Japan and California are 20 and 10 km, respectively. This sparse distribution limits the precision of seismic parameter estimation in specific areas; for instance, those at Saitama University are unknown as the nearest recording station is Omiya, which is 13 km away. Moreover, the dense installation of sensors requires high installation and maintenance costs and cannot be affordable for developing and underdeveloped countries that are prone to earthquakes. Furthermore, the regression analysis and machine-learning approach used to predict the ground motion parameters require a larger database of past earthquakes, which is not feasible for countries where there is a lack of seismic recording stations. Thus, this study presents a novel machine-learning approach to predict the ground motion parameters using a recording from a low-cost Internet of things (IoT) device after an earthquake. These predicted seismic properties along with the structural features are utilized by the damage classification model to classify the damage to RC buildings at Saitama University after an earthquake.