Analysis And Evaluation Method Of Seabed Stability Based On Reliability Theory

Under the background of building an "ocean power".port.coastal and ocean engineering in China have seen tremendous development during recent years.With the exploration,development.and utilization of marine oil and gas resources and the construction of offshore wind turbines,platforms,and other offshore structures,seabed stability evaluation methods are becoming a research hotspot in geotechnical engineering.The seabed stability has become a prerequisite to ensure construction safety and guarantee the long-term service of offshore structures.Many uncertainties are involved in the evaluation of seabed stability,such as the spatial variability of sediment properties and the randomness of related loads.Traditional deterministic analysis methods cannot quantitatively characterize these uncertainties.In addition,there are many factors involved in seabed instability,and the failure mechanism is very complicated and not fully revealed.Therefore,reliability analysis and probabilistic methods are more reasonable means for submarine slope stability evaluation and serve as a supplement to deterministic analysis methods for the uncertainties quantification in the analysis.Some reliability methods have been applied to the field of geotechnical engineering.However,to date,application to the submarine slope has been relatively limited.In particular,analysis that considers the spatially variable parameters of marine sediments is rarely reported.In the present study,the randomness of seismic and ocean wave loadings and the spatial variability of marine sediment properties are considered.The probabilistic framework of seabed stability evaluation is constructed based on the reliability method to investigate the effects of load randomness and spatial variability of soil properties on seabed stability.The research is carried out on the following scientific issues:(?)primary sources and quantitative characterization of uncertainties in seabed stability analysis,(?)efficient surrogate model for low probability level reliability problems such as seabed instability,and(?)efficient simulation method for large-scale,high-precision,and multi-dimensional random fields.The main work of this paper is as follows:(1)Reliability methods combined with the random variable model are used to evaluate the stability of submarine slopes.A probabilistic analysis of the stability of a typical section of the northern slope in the South China Sea is carried out based on the second-order response surface method and the advanced first-order second-moment method.The effects of seismic loading on the stability of submarine slopes were preliminarily investigated using the pseudo-static method.The results show that although the overall fitting of the highly nonlinear performance function is less effective.the polynomial-based response surface method can approximate the response surface well local to the sample point.(2)The procedure of discretizing stationary.non-stationary.Gaussian,and non-Gaussian random fields based on the Karhunen-Loeve expansion method and the numerical method for solving the eigenpairs of the autocorrelation function are introduced in detail.The spatial variability of the undrained shear strength of marine sediments is simulated by stationary and non-stationary random fields.One-and two-dimensional random fields are embedded in the limit equilibrium method.to construct a random field model for submarine slopes.The stability of the submarine slope is evaluated using probabilistic analysis,and the effects of seismic loading and correlation length of the random field are investigated based on the quasi-static method.(3)A probabilistic method was proposed combining the enhanced Newmark method and random fields to evaluate the seismic stability of infinite submarine slopes.Compared with deterministic analysis and the traditional Newmark method,the probabilistic framework has advantages in the following aspects:(?)The strengths of the marine sediments were simulated by random fields discretized by the K-L expansion to accommodate the spatial variability.(?)The yield acceleration in the upslope direction is considered for gentle slopes in the submarine environment.(?)The potential failure positions are automatically identified in the probabilistic method rather than being predefined as in the deterministic analysis.(?)A series of artificial earthquake accelerations with the same spectral characteristics are simulated for consideration of the uncertainties in the input ground motions.(?)The pore water pressure generation and dissipation models are incorporated in the Newmark method to calculate the yield accelerations with consideration of the seismically-induced pore water pressure ratio changes.The results show the superiority of this probabilistic framework and demonstrate the significant effects of changes in pore water pressure and the spatial variability of soil strength on permanent displacements and failure probability of submarine slopes under seismic action.(4)A response surface method based on Gaussian process regression is proposed for building a surrogate model of the actual performance function in slope stability evaluation.The response surface is updated dynamically by adding new training points in the iterative algorithms.The surrogate model reduces the call to actual slope stability analysis in Monte Carlo simulations.The proposed method was applied to three random variable model cases and three random field model cases.The validity and efficiency of the proposed method were verified by comparing with the results of direct Monte Carlo simulations and other surrogate models.Comparisons with other surrogate models show that the response surface method based on Gaussian process regression requires fewer performance function calls and has a higher firting accuracy.For slope stability analysis problems with low probability levels,the proposed method is more effective,especially when the output of the training samples must be obtained using software analysis.Finally,the response surface method based on Gaussian process regression is applied to the reliability analysis of submarine slopes with embedded twodimensional random fields to achieve an accurate and efficient evaluation of submarine slope stability.(5)A probabilistic framework based on the stochastic finite element method is developed to study the wave-induced response in two-and three-dimensional spatially heterogeneous seabed.The framework couples the simulation of spatially heterogeneous soil in MATLAB and the finite element analysis of poroelasticity in COMSOL through the LiveLink program.Firstly,a 2D stochastic finite element model is used for the probabilistic analysis of the dynamic response of the sloping seabed under regular waves.The effects of cross-correlation of random fields and the slope angle on the results of probabilistic analysis were investigated,with the consideration of the spatial variability of permeability and shear modulus of marine sediment.In the analysis of 3D spatially heterogeneous seabed response,the effect of randomness of wave loads is considered,and the random wave generated by the JONSWAP spectrum is coupled to the stochastic finite element method through the LiveLink code for probabilistic analysis.A decomposed K-L expansion method was proposed,which took less computational time and memory space,making the generation of multi-dimensional random fields with high resolution and large scale more effective.Monte Carlo simulations were conducted for the statistics of the seabed response,and the effect of the correlation length of the random field on the depth of the potential liquefaction zone was investigated.