关键词: 岩土工程, 差分进化算法, 支持向量机, 黄土滑坡体, 参数反演

Abstract: According to the characteristics of loess landslide and the advantages of differential evolution algorithm and support vector machine, the inversion model of DE-SVM was established. The inversion software of landslide strength parameters was developed based on three key techniques: object-oriented, self-adaptability and visualization. It was programmed by the GUI module of MATLAB software. Two factors and twenty levels of calculation programs were constructed according to the uniform design method. The safety coefficients of each section with different c and φ calculated by the simplified Bishop method were taken as the learning samples. And the strength parameters of No.3 loess landslide of Du-jia gully in Yan’an new district were inverted. The results show that: the cohesion and internal friction angle of loess landslide obtained by inversion are 15.78kPa and 19.6° respectively. The error between the safety coefficients of landslide calculated by feedback and the stability evaluation index is small, which verifies the rationality of parameter inversion and also shows that the established DE-SVM method is suitable for parameter inversion of loess landslide. The developed software is stable and reliable for the strength parameters inversion of landslide.

Key words: geotechnical engineering, differential evolution algorithm, support vector machine, loess landslide, parameter inversion