The present study explores the potential of simulating the seismic response of non-linear structural systems using Machine Learning (ML) methods. The main objective is to identify the seismic parameters of the input excitation that provide the most valuable and predictive insights, thereby selecting the most suitable ones for soft computing methods to predict dynamic response. Different signals and seismic parameters were used in a parametric numerical analysis to compute the dynamic responses of 1,000 distinct non-linear single-degree-of-freedom (SDOF) systems. The results were then analyzed using supervised learning techniques, employing 20 different ML algorithms, including linear regression, decision trees, support vector machines (SVM), boosted and bagged trees, and artificial neural networks (ANNs). A feature selection analysis was conducted to reduce the number of seismic parameters considered; those with the highest importance scores were used as input for each ML algorithm to determine which parameters yield the most effective predictions of structural response.
IDENTIFICATION OF EFFECTIVE SEISMIC PARAMETERS FOR SIMULATING NON-LINEAR STRUCTURAL RESPONSE USING MACHINE LEARNING APPROACH / De Iuliis, M.. - (2025). (10 th ECCOMAS Thematic Conference on Computational Methods in Structural Dynamics and Earthquake Engineering Rethymno 15-18 June).
IDENTIFICATION OF EFFECTIVE SEISMIC PARAMETERS FOR SIMULATING NON-LINEAR STRUCTURAL RESPONSE USING MACHINE LEARNING APPROACH
De Iuliis, Massimiliano
2025
Abstract
The present study explores the potential of simulating the seismic response of non-linear structural systems using Machine Learning (ML) methods. The main objective is to identify the seismic parameters of the input excitation that provide the most valuable and predictive insights, thereby selecting the most suitable ones for soft computing methods to predict dynamic response. Different signals and seismic parameters were used in a parametric numerical analysis to compute the dynamic responses of 1,000 distinct non-linear single-degree-of-freedom (SDOF) systems. The results were then analyzed using supervised learning techniques, employing 20 different ML algorithms, including linear regression, decision trees, support vector machines (SVM), boosted and bagged trees, and artificial neural networks (ANNs). A feature selection analysis was conducted to reduce the number of seismic parameters considered; those with the highest importance scores were used as input for each ML algorithm to determine which parameters yield the most effective predictions of structural response.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11583/3018085
Attenzione
Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo
