Zhongjiawan Slope Reliability Analysis Based On Support Vector Machine

Slope stability analysis is an important research topic in geotechnical engineering,and reliability is an important indicator of its evaluation. There are a lot of random uncertainty and fuzzy uncertainty in slope engineering, and the existing slope reliability analysis methods mostly consider only one of them. There are a few methods taking into account two uncertainty, but usually with computationally intensive and complex principles, and difficult to use in practical engineering. Therefore, there is important practical significance to study a simple slope random-fuzzy reliability analysis method with high precision.Combining fuzzy mathematics-related methods and random fields method, this thesis uses support vector machine(SVM) method to analyze slope fuzzy-random reliability. First of all, based on the slope safety coefficient of the SVM prediction model,and use genetic algorithm to optimize the SVM kernel function parameter, to fit and classify the slope safety factor. Finally, the slope reliability analysis method based on SVM is proposed and the parameter sensitivity is analyzed. With examples the thesis shows the following conclusions:The safety factor of slope prediction model based on SVM is high accuracy, and the training sample of slope reliability analysis method based on SVM needn't high representativeness and completeness, which reducing the amount of calculation, and the result with high accuracy mainly depends on the number and distribution of the sample.Reliability of slope associated with internal friction angle and cohesion is positive, and associated with cross-correlation is negative; when the variability of parameters is large variability negatively correlated with the reliability, when the variability is small variability positive ly correlated with the reliability.