Study On Inclined Pile Behaviour Under Vertical And Horizontal Load Using Finite Difference Method And Machine Learning

The inclined piles are widely used in the practice.The inclined piles can be classified into categories including intended inclined pile and non-intended inclined pile.The former,for instance,the inclined pile in wharf,is mainly used to sustain the horizontal loads.The latter is formed by the artificial error and the equipment error in installation,or by the influence of other construction actives,such as the nearby excavations and piling.Due to the wide use,the characteristics of bearing capacity and deformation of inclined pile was systematically investigated in this thesis.The FLAC3 D was adopted to establish the finite difference model of inclined under vertical and horizontal load.This research focuses on the influencing factors on the bearing capacity and deformation,such as soil modulus,inclined angle of pile,length of pile and interface friction angle.The load-deformation curve of the different cases was systematically researched.The pile loading stiffness was defined to qualify the deformation behaviour of pile under working load.The results show that for the inclined piled under vertical load,the bearing capacity increases significantly as the pile length and interface friction angle increase.The pile loading stiffness increases with the increasing soil modulus,while the bearing capacity change slightly with the soil modulus.The larger inclined angle of pile can increase the pile bearing capacity.For the inclined pile under horizontal load,the axial force has a similar trend with that under vertical load.A different behaviour is that the stiffness of inclined pile under horizontal load can be increased by increasing the inclined angle.The radial force was not influence by the soil modulus,inclined angle of pile,length of pile and interface friction angle.The radial load-deformation curve of the radial force does not change sharply,while the axial load-deformation curve always has an inflexion point at which the deformation increases dramatically and the force keep nearly constant.Base on the results from the 432 finite difference models,a complete database including 388,800 data sets was established.The database was divided as training set and validation set and then the deep neural network(DNN)method was employed to form a prediction model,which can predict the load-defamation behaviour under any practical situation.The bearing capacity and loading stiffness can be easily obtained by the DNN model.Furthermore,a stochastic analysis was performed using the DNN model to investigate the distribution of bearing capacity of slightly inclined pile caused by installation error or impact of adjacent construction.It is found that the bearing capacity distribution of short inclined pile can be divided two parts,while the long pile shows a three-part distribution.The long inclined pile has a high percentage and a more concentrated distribution of maximum bearing capacity than the short pile.