In this study, a series of geotechnical centrifugal tests were conducted to investigate the effectiveness of settlement control of two types of rigid pile structure embankments (PRSE) in collapsible loess under high-speed railway embankments. The research results show that ground reinforcement is required to reduce the post-construction settlement and settlement rate of the embankments. The rigid pile structure embankments using rigid piles can substantially reduce the embankment settlement in the construction of embankments on collapsible loess, and the efficiency in settlement reduction is affected by the pile spacing. The pile-raft structure embankments ( PRSE) have much stronger ability in terms of the effectiveness of settlement control, while the pile-geogrid structure embankments (PGSE) provides rapid construction as well as economic benefits. Rational range of pile spacing of PRSE and PGSE are suggested based on the requirements of various railways design speeds. Furthermore, the time effectiveness of negative skin friction of piles and the action of pile-cap setting are also investigated. The relevant measures for improving the bearing capacity and two parts of transition zone forms as positive control mean have been suggested.

To research and analyze the differential settlements of foundations specifically, site investigations of existing railways and metro were firstly carried out. Then, the centrifugal test was used to observe differential settlements in different position between foundations on the basis of investigation. The theoretical model was established according to the stress diffusion method and Fourier method to establish an analytical solution of embankment differential settlement between different foundations. Finally, theoretical values and experimental values were analyzed comparatively. The research results show that both in horizontal and vertical directions, evident differential settlement exists in a limited area on both sides of the vertical interface between different foundations. The foundation with larger elastic modulus can transfer more additional stress and cause relatively less settlement. Differential settlement value decreases as the distance to vertical interface decreases. In the vertical direction of foundation, mass differential settlement also exists on both sides of the vertical interface and foundation with larger elastic modulus can transfer more additional stress. With the increase of relative modulus of different foundations, foundation with lower elastic modulus has larger settlement. Meanwhile, differential settlement is more obvious. The main error sources in theoretical and experimental values include: (a) different load form; (b) foundation characteristics differences; (c) modulus conversion; (d) effect of soil internal friction.

To research and analyze the additional stress distribution and change of granular materials, the model tests are used to observe vertical additional stress in different position and depth in different foundations. And the comparison between observed values both in different soils and single soils is conducted to analyze the transmission and attenuation of additional stress in granular materials. The research results show that the existing of boundary surface can lead to different vertical additional stress transmit obviously. And with the increasing of loading, the vertical additional stress differences between that of different soils get larger, meanwhile, the ratio of stress differences to smaller additional stress increases slightly. With the increasing of depth, the attenuation rate of vertical additional stress of silty soil changes much fast than that of fine sand. Wherever in horizontal direction or vertical direction, the vertical additional stress of fine sand which has higher compression modulus in different soils is slightly larger than that in single soil and transfers additional stress (loading) more under the same loading. To granular materials, inner friction structure effect is evident influence to additional stress transfer.

To research and analyze the additional stress distribution and change of granular materials, the model tests are used to observe vertical additional stress in different position and depth. And the comparison between observed values and theoretical values is conducted to analyze the transmission and attenuation of additional stress in granular materials. The research results show that calculated values are based on Boussinesq solution which ignores the property of soil layer (materials), the distribution of additional stress for fine sand which belongs to granular materials is largely deviated from theoretical value. For granular materials, inner friction structure effect is evident influence to additional stress transfer. And continue using calculation method which is based on continuum materials will have bigger difference and even wrong.

To research and analyze the neutral point distribution of rigid pile composite foundation on collapsible loess ground in high speed railway, the centrifuge model tests are used to imitate pile raft composite foundation and pile net composite foundation in different pile spacing based on the experimental study at home and abroad. The research results show that the movement of negative skin friction and the position of neutral point which is drop conspicuously with larger pile spacing are a continuous changing process in composite foundation. In the pile spacing 2D,4D,6D of pile raft composite foundation, the stable range of neutral point position are 0.25(the pile length ratio), 0.59,0.63 separately in post-construction stage. In the pile spacing 2D(no pile-cap),4D(no pile-cap),4D(pile-cap),6D(pi le-cap) of pile net composite foundation, the stable range of neutral point position are 0.37to0.43(the pile length ratio),0.50,0.80,0.81separately in post-construction stage. The pile-cap setting has more influence on Neutral point distribution. Suggestions are given that the waterproof and drainage measures in foundation must be considered when setting up composite foundation on collapsible loess.