Prediction of inelastic seismic performance of RC structures using machine learning algorithms

Abstract

Incorporation of Machine Learning (ML) techniques in determining dynamic properties for the structural systems that manifest non-linearity in behavior with respect to the geometry attributes under seismic response was the main scope of the current work. The fundamental period of the vibration (T-period) for the moment-resisting frame of reinforced concrete structures was selected as a studied parameter for validating applicability of utilizing ML approach in prediction of uncertainties earthquake engineering.  Artificial neural network (ANN) and Vector Machine (SVM) with devised embedded Kernel functions was based ML implementation for prediction of T-period. Radial basis function (RBF), Exponential RBF and Sigmoid were set up as kernel functions for supervising and enhancing accuracy of learning SVM model to the primitive dataset.   The findings attempted to generate intuitive with high accuracy relationships for the models with less discrepancies compared to the conventional that based linear regression.