Data-driven assessment and design recommendations for piloti-type RC buildings based on machine learning techniques

Abstract

This study aims to evaluate the seismic performance of piloti-type reinforced concrete (RC) buildings and to develop an intelligent framework for predicting inter-story drift ratio (IDR) using machine learning techniques. A dataset comprising 111 structural configurations from South Korean piloti structures was analyzed based on key input parameters, including column number (CN), column shape ratio (CSR), wall shape ratio (WSR), concrete compressive strength (fc′), and transverse spacing (TS). Two machine learning models-Support Vector Machine (SVM) and Adaptive Neuro-Fuzzy Inference System (ANFIS) were developed and compared. Among the models, ANFIS demonstrated superior accuracy, with R2 values exceeding 0.95 and error metrics (MAE, MSE, RMSE) within acceptable limits. The model's predictions were further validated through finite element (FE) analysis using LS-DYNA pushover simulations. The predicted IDR values closely matched the FE results, with discrepancies below 10%. Parametric analysis confirmed that optimal seismic performance was achieved when CSR ranged from 1.0 to 1.5 and WSR from 1.5 to 2.78. Based on these findings, design recommendations were proposed for both new construction and retrofit applications, tailored to moderate and high seismic hazard zones.