An Efficient Machine Learning Method for Estimating the Damping Ratio of RC Moment-Resisting Frames

The structural damping ratio, a measure of energy dissipation in structures subjected to dynamic loads, is a key factor in building seismic design and assessment. Despite its significance in seismic design and evaluation, it is commonly approached as an estimated value during the design process or handled as a flexible parameter depending on the structural type, as indicated in seismic design codes. This study determines the system equivalent damping ratio by equating the total dissipated energy, including hysteretic and elastic strain energies, by summating the energy dissipated in individual members. This calculation employs Jacobsen's approach, which provides a systematic framework for energy equivalence in dynamic systems. For this purpose, several tree ensemble-based machine learning algorithms estimate the members' hysteretic and strain energies and evaluate them using various statistical metrics. The PEER dataset, consisting of experimental results from ۳۵۳ rectangular reinforced concrete columns, is the basis for training and evaluating the models. The findings demonstrate ۹۶% and ۹۲% accuracy for predicting elastic strain energy and hysteretic energy, respectively, based on the R² metric achieved by the XGBoost algorithm.

Keywords:

Equivalent damping ratio , Hysteretic energy , Elastic strain energy , Machine learning , Tree ensemble model