Prediction of seismic-induced bending moment and lateral displacement in closed and open-ended pipe piles:A genetic programming approach  

作　　者：Laith Sadik Duaa Al-Jeznawi Saif Alzabeebee Musab A.Q.Al-Janabi Suraparb Keawsawasvong 

机构地区：[1]Department of Civil and Architectural Engineering and Construction Management,University of Cincinnati,United States [2]Department of Civil Engineering,College of Engineering,Al-Nahrain University,Baghdad,Iraq [3]Department of Roads and Transport Engineering,University of Al-Qadisiyah,Al-Qadisiyah,Iraq [4]Department of Civil Engineering,Thammasat School of Engineering,Thammasat University,Pathumthani,12120,Thailand

出　　处：《Artificial Intelligence in Geosciences》2024年第1期82-95,共14页地学人工智能（英文）

摘　　要：Ensuring the reliability of pipe pile designs under earthquake loading necessitates an accurate determination of lateral displacement and bending moment,typically achieved through complex numerical modeling to address the intricacies of soil-pile interaction.Despite recent advancements in machine learning techniques,there is a persistent need to establish data-driven models that can predict these parameters without using numerical simulations due to the difficulties in conducting correct numerical simulations and the need for constitutive modelling parameters that are not readily available.This research presents novel lateral displacement and bending moment predictive models for closed and open-ended pipe piles,employing a Genetic Programming(GP)approach.Utilizing a soil dataset extracted from existing literature,comprising 392 data points for both pile types embedded in cohesionless soil and subjected to earthquake loading,the study intentionally limited input parameters to three features to enhance model simplicity:Standard Penetration Test(SPT)corrected blow count(N60),Peak Ground Acceleration(PGA),and pile slenderness ratio(L/D).Model performance was assessed via coefficient of determination(R^(2)),Root Mean Squared Error(RMSE),and Mean Absolute Error(MAE),with R^(2) values ranging from 0.95 to 0.99 for the training set,and from 0.92 to 0.98 for the testing set,which indicate of high accuracy of prediction.Finally,the study concludes with a sensitivity analysis,evaluating the influence of each input parameter across different pile types.

关 键 词：Genetic programming Pipe piles Lateral response Bending moment Earthquake loading Standard penetration test Machine learning 

分 类 号：TU4[建筑科学—土工工程]