A Data-Driven Seismic Design Framework for RC Frames Using Wavelet Transforms and Machine Learning

Nowadays, reinforced concrete (RC) frames are one of the most well-developed lateral load-resisting systems used around the world. However, the complex nonlinear behavior of the composite material system, which combines steel and concrete, makes precise seismic analysis and design particularly challenging. Established methods, such as modal response history analysis and time history analysis, are often difficult to apply. This challenge is compounded when significant soil structure interaction alters the structural response through higher mode effects. Consequently, the development of innovative analytical methods that are both simplified and robust remains a primary challenge in advancing the design of these structural systems. This study presents a novel wavelet transform-based machine learning method (WTMLM) for the seismic design of RC frames. The method incorporates soil-structure interaction (SSI) effects and formulates the dynamic equilibrium of the system. The proposed WTMLM is used to design five RC frames with varying story heights on two different soil types. The performance is evaluated using nonlinear time-history analyses (NTHA) under real ground motions. Finally, a new equation for predicting displacement distribution is developed based on machine learning and median NTHA results. The findings indicate the WTMLM is an effective method for controlling the seismic performance of RC structures.

Keywords:

Reinforced concrete moment-resisting frame , Wavelet Transform , Machine learning-based method , Seismic behavior , Dynamic analysis