Research On Spatial Variability And Uncertainties Within Pile Capacity Of Soft Soils Based On CPTU

Soft soils with low strength and strong spatial variability are widely distributed in the coastal areas of China.Geotechnical designs in these areas have been suffering the lack of proper determination of soil properties and have caused failure and great damage in practice.A main reason of the improper determination of soil properties is that they were generally selected according to the laboratory tests,which have been criticized for their high cost,sampling discontinuity and soil disturbance.The upshot is that the spatial variability characteristics of the soft soils were not representative of the in-situ conditions.Therefore,the geotechnical design may be either over-conservative or over-risky due to the improper assessment of the spatial variability of soil properties.The modern enhanced piezocone penetration test?CPTU?has been demonstrated to be a promising tool to accurately characterize the behavior of soft soils over the decades.It simultaneously provides three independent measurements including the cone tip resistance?q_t?,sleeve frictional resistance?f_s?and pore water pressure?u₂?,and can therefore be used to describe the failure strength,remolded strength,consolidation and permeability behavior of the soils.However,this advanced technique has gained insufficient application in China,especially for the probabilistic site characterization and reliability-based analysis.It has been widely acknowledged that geotechnical properties are site-specific and an understanding of local geology is crucial in the interpretation of site investigation data.This distinctive feature of geotechnical engineering contributes to the different criteria used in the actual design.To achieve a rational geotechnical design,the spatial variability and the uncertainties associated within soil properties shall be addressed accurately.In this study,the spatial variability of soil parameters and uncertainties within pile reliability are systematically analyzed based on the field CPTU technique with the funding from the"Twelfth Five-Year"National Science and Technology Support Plan and the National Key Research and Development Program of China.The main conclusions are summarized as follows:?1?The random field theory is used to evaluate the spatial variability characteristics of the CPTU data obtained from the soft soils within different geologic formations.The random field model?RFM?parameters of the investigated CPTU indices are found to be a function of the corresponding geologic formations of soils.The spatial variability characteristics of the CPTU indices are most significant for the Yangtze River-alluvial soft soils,perhaps due to the complex geologic history and wide coverage of the alluvial plain.The variabilities of the RFM parameters of the CPTU indices for the marine soft soils in Northern Jiangsu are also notable,and they are attributed to the frequent transgression of the Yellow Sea during the soil sedimentation.The RFM parameters of the CPTU indices of the Lixia-Lagoon and the Taihu-lacustrine soft soils are well concentrated within narrow ranges,which are consistent with the relatively uniform geological depositional environment in these areas.Besides,it seems that the evaluated spatial variability characteristics also depend on the scale of the observation.The large estimates of the scale of fluctuation seem to occur when the observation scale is large.This scale-effect shall not be neglected in the assessment of the spatial variability of soil parameters.?2?The application of geostatistics in predicting the soil parameters at unsampled locations are investigated using theoretical analysis,numerical analysis and field CPTU study.Moreover,it is theoretically proven that the simple Kriging is the same as the conditional random field.The main difference between the simple Kriging and the ordinary Kriging is that the sample data are considered to be sufficient to describe the uncertainty within the soil parameter at unsampled locations in the former,whereas this assumption is not taken in the latter.This assumption produces different results of the Kriging interpolations.Three simple examples are analyzed to illustrate the general rule of the Kriging interpolations,indicating that the geostatistics is capable of both the deterministic and uncertain nature of the geotechnical parameters.That is to say,the values of a soil parameter in space are deterministic in nature,whereas uncertainties manifest from lack of information,and these uncertainties tend to increase with the increasing of spatial distance between the prediction and sample points.Besides,it seems to be necessary to achieve the normality of the sample data to ensure that the probability density distribution of the Kriging prediction is traceable.The back Box-Cox transformation formulas are then derived to report the predictions when the transformation of a non-Gaussian soil parameter to a Gaussian variable is conducted.Finally,as an illustrative example,the probabilistic analysis and numerical simulation are performed on the results of the geostatistics to evaluate the spatial distribution of the soft clay and key design parameters using the CPTU data collected from the Chongqi Bridge project.?3?The multivariate distribution model is used to evaluate the multiple dependencies of design parameters and CPTU indices for the Jiangsu soft soils.The design parameters include the undrained shear strength,overconsolidation ratio,sensitivity,horizontal coefficient of consolidation and horizontal coefficient of permeability.Multivariate correlations among the design parameters and CPTU indices are derived based on the analysis of compiled databases for the Jiangsu soft soils.It is shown that the multivariate distribution model is capable of the multivariate correlations among the soils parameters for the Jiangsu soft soils.Moreover,the accuracy of a predicted value of a design parameter can be consistently improved when more CPTU indices are involved.However,the site-specific effect may play an important role in the application of the multivariate distribution models.It is advisable to construct a local multivariate distribution model based on local database with sufficient sample data to achieve an enhanced prediction.The Box-Cox transformation provides a convenient way to simplify the construction and application of the multivariate distribution model.However,the soil parameter after the Box-Cox transformation can only be viewed to approximate a normal distribution.Meaningless statistics,especially the upper bound,of the predicted soil parameter may be arrived if the Box-Cox method is indeed applied.?4?The reliability-based analysis is performed to obtain the correlation between the probability of pile failure?p_f?and the factor of safety?FS?used in the WSD with the incorporation of the geostatistical analysis.It is shown that the reliability-based analysis can fundamentally be viewed as an objective and quantitative representation of the general subjective rules followed in the WSD to determine the FS.By incorporating the predictions of soil properties at unsampled locations provided by the geostatistics,the reliability of pile design can be improved and therefore the minimal FS value required to maintain the stability of pile foundation can be reduced to achieve a more rational design.Using this approach,the minimal FS values for a single pile are reduced from2.51–3.15 to 2.38–2.96.Besides,the methods of reliability-based analysis and pile capacity prediction model also play an important role in the determination of FS.The mean-value first-order reliability method is less accurate than the first-order reliability method and the Monte Carlo simulation method,especially at the high FS levels.