| As one of the most frequent,common,destructive and widespread geological disasters,landslides cause a large number of casualties and economic losses every year.Since the completion of the Three Gorges Project,the Three Gorges reservoir area has become a high incidence of landslide disasters in China under the dual influence of periodic changes in reservoir water level and rainfall.Therefore,how to make effective prediction of landslide has an extremely important role in the rescue and relief of landslide.In landslide prediction,its accuracy depends on the selection of influence factors and the construction of high-performance models.The traditional landslide displacement prediction model has less selection of displacement influence factors;and lacks consideration of the influence on the prediction results brought by the noise of monitoring data,the subjectivity of model structure selection and the randomness of parameter setting.In this thesis,an interval prediction method is proposed to address the shortcomings of the existing models.Firstly,the landslide displacement influence factors are studied quantitatively,and then the landslide displacement prediction model is constructed by Bootstrap method,genetic algorithm and least squares support vector machine to achieve accurate prediction of landslide displacement while characterizing the uncertainty of the prediction results by means of intervals,and the main research contents are as follows:(1)The Shuping landslide in the Three Gorges reservoir area is selected as the research object,and the relationship between factors such as rainfall,reservoir water level and landslide deformation state and the accumulated landslide displacement is quantitatively analyzed.The results show that the influence factors such as cumulative rainfall,maximum continuous rainfall,reservoir water level elevation,reservoir water level change rate,landslide displacement increment and displacement growth rate,which take into account the hysteresis effect,can well reflect the relationship between rainfall,reservoir water level and landslide deformation state and the cumulative displacement deformation characteristics of Shuping landslide.(2)In response to the lack of consideration of the influence of uncertainties in the model and data in the existing model,this thesis introduces Bootstrap method based on the displacement prediction model of genetic algorithm optimized least squares support vector machine,and the uncertainties caused by the noise of monitoring data,the subjectivity of model structure selection and the randomness of model parameter setting through Bootstrap residual resampling The uncertainties caused by noise of monitoring data,subjectivity of model structure selection and randomness of model parameter settings are quantified,and the corresponding prediction interval is constructed to characterize the uncertainty degree of the prediction results of Shuping landslide.(3)The interval prediction model constructed in this thesis was applied to the displacement prediction of two monitoring points ZG85 and ZG87 of the Shuping landslide.The application results confirm that the prediction model based on Bootstrap and genetic algorithm optimized least squares support vector machine can accurately predict the future displacement trends of two monitoring points ZG85 and ZG87 both in the form of point values and also evaluate the degree of uncertainty in the displacement prediction process of ZG85 and ZG87 monitoring points by means of intervals.Finally,the results are compared and analyzed with the point prediction and interval prediction results corresponding to prediction methods such as genetic algorithm optimized limit learning machine,and the results show that the prediction model proposed in this thesis obtains high interval coverage at two monitoring points ZG85 and ZG87 while the constructed intervals are narrower,and their width range comprehensive indexes are53.75 mm and 31.29 mm,respectively. |
PDF Full Text Request