AI-based system for quick seismic estimation of building structures on urban disaster-prevention in Taiwan

ABSTRACTAccording to a survey by the Ministry of the Interior (MOI) in Taiwan, around half of the 8.93 million buildings in the country, which are over 30 years old, have inadequate seismic capacity due to outdated design standards or aging materials. To evaluate seismic capacity, a preliminary seismic evaluation (PSE) system that involves site investigation and shop drawing review (if available) by professional engineers is typically used. However, given the significant financial and manpower resources required, performing PSE on all buildings in Taiwan is not practical. In order to overcome the challenge of evaluating the seismic capacity of buildings in a cost-effective and efficient manner, this study developed an enhanced PSE system called QSEBS, based on deep learning technology. By leveraging government property tax databases, QSEBS can rapidly estimate the seismic capacity of buildings, with results consistent with those of the PSERCB system. The key advantage of QSEBS is its ability to eliminate the need for human labors in PSE, saving significant amounts of money and manpower, particularly for a large number of buildings. Thus, QSEBS can serve as a valuable tool to support the government’s urban disaster-prevention strategy and can be widely implemented.CO EDITOR-IN-CHIEF: Ou, Yu-ChenASSOCIATE EDITOR: Ou, Yu-ChenKEYWORDS: Back-propagation neural network (BPNN)preliminary seismic evaluation of reinforced concrete building (PSERCB)quick seismic estimation of building structures (QSEBS)Kruskal-Wallis H testdata cleaning Nomenclature Ac2=seismic-capacity indexA2500=seismic demand for a 2500-year return period earthquakeAc2/IA2500=seismic capacity-demand ratio for seismic vulnerability assessmentC=ratio of spectral acceleration divided by ground acceleration for a specific structural period in elastic normalized response spectrum of accelerationD=diameter of the rebars and stirrupsE=convenient representation of 2μ−1E_TACW=equivalent total area of column-wallE_W/CW=equivalent width per column-wallE_D/CW=equivalent depth per column-wallH=value of Kruskal-Wallis H testH0=null hypotheses for correlation evaluationI=importance factorR=response reduction factorSa=parameter of elastic design spectral acceleration responseTn=structural periodVu, e=ultimate elastic base shear demandVy=yield base shear demandVS30=average shear wave velocity for a soil depth of 30 mW=sum of weight lumped at the ground floor’s ceiling levelμ=ductility level△u=ultimate or code-specified displacement△y=yield displacementχ2=Chi-square valueDisclosure statementNo potential conflict of interest was reported by the author(s).Additional informationFundingThis study was supported by the National Science and Technology Center for Disaster Reduction, Taiwan [Grant No. NCDR-S-111012].