Gaussian Process Regression for Seismic Fragility Assessment: Application to Non-Engineered Residential Buildings in Indonesia

Indonesia is located in a high-seismic-risk region with significant number of non-engineered houses, which typically have higher risk during earthquakes. Due to the wide variety differences even among parameters within one building typology, it difficult capture total population, as typical structural engineering approach understanding fragility involves tedious numerical modeling individual buildings—which computationally costly for large population buildings. This study uses statistical learning technique based on Gaussian Process Regression (GPR) build family curves. The current research takes column height and side length input variables, linear analysis used calculate failure probability. GPR then utilized predict curve probability collapse, given data evaluated at finite set experimental design. result shows that can collapse well, efficiently allowing rapid estimation an prediction single configuration. Most importantly, also provides uncertainty band associated curve, crucial information real-world analysis.