Desarrollo de una metodología de predicción de respuestas dinámicas usando Machine Learning (ML) para el edificio del CEC bajo cargas sísmicas.

Abstract

The purpose of this work is to develop an algorithm using Machine Learning to predict dynamic responses under the action of different seismic spectra. First, it is necessary to create a three-dimensional model of the structure in a finite element software, in this case SAP2000 was used, to simulate its behavior under the action of loads and the different seismic spectra, which need the acceleration values. spectral at 0.2 and 1 second of the seismic activity and the PGA (Peak Ground Acceleration). In this way, the usage factors of the structural elements are obtained, of which the most critical ones will be selected, which will form part of the database for the training of the automatic learning algorithm. Once the database has been generated, data cleaning and treatment is carried out, then they are normalized so that in this way the data belongs to the same scale and thus obtain more precise results, then with this database normalized data, we proceed to train and test the multivariate linear regression algorithm. Finally, we proceed to evaluate the accuracy of the model by calculating errors between real and predicted values.