Seismic Building Damage Prediction From GIS-Based Building Data Using Artificial Intelligence System

The estimation of seismic damage to buildings is complicated due to the many sources of uncertainties. This study aims to develop a new approach using an artificial intelligence system called adaptive neuro-fuzzy inference system (ANFIS) model to predict the damage of buildings at urban scale considering input uncertainties. First, the study performed seismic damage evaluation of buildings utilizing the capacity spectrum method (CSM) to obtain a set of 57,648 training data from a combination of three main parameters, i.e., 6 earthquake magnitudes, 8 structural types, and 1,201 distances. Next, the data was used to develop a practical ANFIS model for the seismic damage prediction. The variables of the fuzzy system are earthquake magnitudes, structural types, and distance between epicenter and building. To validate the applicability of the proposed model, analyses of spatial seismic building damage under five possible earthquakes in Chiang Mai Municipality were performed by using the proposed methodology. From the comparison of the damaged urban area, small discrepancies between the CSM and the ANFIS results could be observed. It should be noted that the proposed ANFIS model can predict the seismic building damage reasonably well compared with the CSM. Using the method proposed herein, it is possible to create damage scenarios for earthquake-prone areas where only a few seismic data are available, such as developing countries.