Estimation of spectral acceleration based on neural networks

Abstract

This study presents an effective method based on artificial intelligence to predict spectral acceleration from the data of the 'Next generation attenuation' project. The proposed method uses the learning abilities of artificial neural networks to expand the knowledge of the mapping from earthquake parameters to spectral accelerations, which results in spectral accelerations for special frequencies. In this paper, the Levenberg-Marquardt algorithm is applied for training neural networks. For each type of faults, considered earthquake parameters included in the input of the neural network are moment magnitude, distance from the recording site to epicentre, hypocentre depth and average shear-wave velocity in the top 30 m. To test the adequacy of the trained neural networks, they were examined through their training set and new data. The obtained results demonstrate that neural networks can be effectively and reliably employed in the estimation of spectral acceleration.

The promise of implementing machine learning in earthquake engineering: A state-of-the-art review

“…The identified significant predictors are moment magnitude of the earthquake, source-to-site distance, the average shear-wave velocity of the site, faulting mechanism, and focal depth. The ML tools utilized in GMPEs include the ANN (the top row in Figure 3) (Bakhshi et al, 2014; Derras et al, 2014; Dhanya and Raghukanth, 2018; Güllü and Erçelebi, 2007; Kerh and Ting, 2005; Khosravikia et al, 2019), genetic programming (GP) (Cabalar and Cevik, 2009), multi-expression programming (MEP) (Alavi et al, 2011), SVR (Tezcan and Cheng, 2012; Thomas et al, 2017), GEP (Güllü, 2012; Javan-Emrooz et al, 2018), Lagrange equation discovery (ED) system (Markič and Stankovski, 2013), conic multivariate adaptive regression splines (CMARS) (Yerlikaya-Ozkurt et al, 2014), randomized adaptive neuro-fuzzy inference system (RANFIS) (Thomas et al, 2016), M5’ model tree and CART (Hamze-Ziabari and Bakhshpoori, 2018; Kaveh et al, 2016), DNN (Derakhshani and Foruzan, 2019), and hybrid methods such as the coupling of GP and orthogonal least squares (OLS) (Gandomi et al, 2011), the combination of ANN and simulated annealing (SA) (Alavi and Gandomi, 2011), the coupling of GP and SA (Mohammadnejad et al, 2012), and the coupling of GA, ANN, and regression analysis (RA) (Akhani et al, 2019).…”

Editorial

“…The numerical modelling of the structural connections that often represent the most vulnerable structural elements in steel buildings during progressive collapse is analysed in detail. The paper highlights the benefits and the drawbacks of the aforementioned methods with respect to the treatment of the connections.The second paper, by Bakhshi et al (2014), presents an effective method to estimate spectral acceleration based on artificial intelligence, which uses the learning abilities of artificial neural networks to expand knowledge of mapping from earthquake parameters to spectral accelerations, resulting in spectral accelerations for special frequencies. For the selected faults, earthquake characteristics are used as inputs for the trained neural networks, whereas the adequacy of the trained neural networks was examined through the authors' training set and new data.…”

“…The second paper, by Bakhshi et al (2014), presents an effective method to estimate spectral acceleration based on artificial intelligence, which uses the learning abilities of artificial neural networks to expand knowledge of mapping from earthquake parameters to spectral accelerations, resulting in spectral accelerations for special frequencies. For the selected faults, earthquake characteristics are used as inputs for the trained neural networks, whereas the adequacy of the trained neural networks was examined through the authors' training set and new data.…”