Development of novel artificial intelligence functions based on 3D finite element method using February 6 Kahramanmaraş Seismic Records for earthquake effects prediction in various soils

• Published in: Engineering geology Vol. 336; p. 107570
• Main Authors: Ekmen, Arda Burak, Avci, Yusuf
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.07.2024
• Subjects: artificial intelligence; Artificial intelligence based on three-dimensional finite element analysis; Earthquake effects prediction; earthquakes; February 6 Kahramanmaraş Earthquakes; finite element analysis; geophysics; infrastructure; model validation; neural networks; prediction; Soil effect; soil profiles; Soil-pile-structure interaction; Artificial intelligence based on three-dimensional finite element analysis; Earthquake effects prediction; February 6 Kahramanmaraş Earthquakes; Soil effect; Soil-pile-structure interaction
• ISSN: 0013-7952; 1872-6917
• DOI: 10.1016/j.enggeo.2024.107570

The seismic events on February 6, 2023, in the province of Kahramanmaraş/Türkiye, caused severe damage and the collapse of numerous structures due to underlying soil issues. This catastrophe revealed the inevitable requirement to evaluate the effect of soil profile on structural safety. In the present study, novel artificial intelligence (AI) functions based on the three-dimensional finite element (3D FE) method considering various soil parameters were developed to predict the effects of earthquakes. A 3D FE model of the ten-story building with a known soil profile and structural elements was created in the first stage, accounting for the soil-pile-structure interaction. After model validation, numerous parametric time history earthquake analyses were performed using the February 6 Pazarcık/Kahramanmaraş (Mw = 7.7) earthquake records. Therefore, the effects of soil parameters on acceleration, settlement, and lateral deformations were investigated. An innovative coding infrastructure, leveraging the power of AI, was developed to generate optimal network solutions automatically for creating high-order regression prediction functions. The 3D FE data was integrated into the code, and subsequently, an artificial neural network was utilized to formulate a function that yielded statistically significant outcomes. The created function accurately predicted the accelerations, settlements, and deformations. A novel method for indicating the potential deformations and accelerations inflicted by earthquakes based on soil parameters was introduced. This methodology can serve as a practical guide for researchers and project implementers in the initial design phases. •Earthquake effects for various soil parameters were investigated.•Numerous parametric 3D finite element analyses were conducted to provide data for AI.•Novel AI functions were developed to guide researchers in earthquakes.