Research And Application On The Algorithm Of Back Analysis Of Soil Parameter Based On The Non-parametric Regression

Soil parameters are important indicators in the geotechnical engineering design process, which are obtained generally through indoor test. Due to the reasons such as test condition, test method and so on, test results can often be discreteness, which have some difference with the real value. Moreover, if the construction method is different, the soil parameters of the same kind of soil will be different. Therefore, the appropriate method reasonable methods need to be adopted to determine the soil parameters. Combined with the existing monitoring data, the soil parameters are obtained through the back analysis method. And then the parameters obtained by the back analysis are used to predict the corresponding results. It can update data, guide the design and construction, or guide the similar project.In this paper, the back analysis method and its application are researched as follows:(1) Aiming at the problems existing in the back analysis method, a soil parameter back analysis algorithm based on the nonparametric regression is put forward. The same measured target in the different condition is regard as the constraint conditions. Considering the error of the measured target, the tolerance is selected to obtain the tolerance range of target values. And then the initial scopes of soil parameters are determined for back analysis. Combined with uniform testing and nonparametric regression method, the response surfaces between soil parameters and the measured target are established. A large number of parameters samples are enumerated according to the working condition. Then a large number of targets’ calculation values are obtained by response surface interpolation. By comparing the values and measured values of the target, the soil parameter data samples are screened out, which are responding to the calculated value of target. They are close to the target measurement values within the tolerance range. And the statistical range of the sample data is the corresponding actual value range of soil parameters corresponding to the target measured value within the certain tolerance range.(2) Back analysis algorithm applies to the foundation pit engineering. The force of steel support and horizontal displacements of pile body is chosen as the measured target. The back analysis algorithm is used to analyze the shear strength index of soils that exsit within the scope of supporting pile depth of foundation pit. The range value of the soil parameters are obtained and verified.(3) Back analysis algorithm is used in the shield tunnel project. The ground settlement is regarded as the measured target. The back analysis algorithm is used to analyze the shear strength index of soils that is in the influence range of shield tunnel excavation. The range values of the soil parameters are obtained. Finally, back analysis results are used to predict the ground surface settlement of the other section. The result is close to the measured values.Through the application of two engineering examples, it shows that the proposed back analysis algorithm is feasible. And it is a common algorithm. The algorithm considers the measurement uncertainty of the target value and the influence of construction condition. And it can select multiple measured targets as constraint conditions in order to analyze multiple parameters comprehensively and improve the analysis efficiency. It not only can be used to guide the design and construction but also provides an effective way for the soil parameter back analysis in geotechnical engineering.