| Karst collapse refers to the process of sudden collapse of the surface due to various factors in areas where soluble rock is distributed.As one of the most dangerous natural disasters,karst collapse has become the most common geological disaster in karst areas.The hazards of karst collapse are mainly consist of the destruction of roads,railways,buildings and surface water bodies,affecting agricultural water and land use,causing casualties and property damage.Due to the characteristics of concealment,suddenness,uncertainty,multi-factor and harmfulness,karst collapse has become an important factor affecting and restricting the social and economic development.At present,scholars have analyzed the incubation process and collapse mechanism of karst collapse in a variety of ways.However,because the incubation process of the collapse occurs underground,it is difficult to locate and observe.In addition,the damage occurs suddenly,which makes it difficult to summarize and verify the collapse mechanism.There are still some controversies about the generation mechanism of karst collapse,and the systematic and unified collapse mechanism is still in the process of gradual improvement.At present,one of the important ways to reduce the hazards of karst collapse is to predict the spatial distribution of karst collapse and to evaluate the susceptibility of karst collapse in appropriate ways.In the present study,Zhongliangshan area of Chongqing was selected as the study area.Based on the analysis of geological background and collapse characteristics in the area,the karst collapse mechanism was studied by numerical simulation method.Combined with the geospatial database in the area,the index system of collapse evaluation was established.Analytic hierarchy process,random forest and parameter optimization methods were used to study the regionalization of karst collapse susceptibility.The main research contents and achievements of this paper are as follows:1 On the basis of analyzing the characteristics of karst collapse in the study area,a generalized model of karst collapse was proposed.According to the existing survey data,by analyzing the hydrogeology,stratigraphic lithology,topography,karst morphology,human engineering activities and other conditions related to the formation of collapses in the study area,it was found that most of the collapses in the area are covered karst collapses,mainly distributed in groups,mainly distributed in karst depressions area,and mostly developed along the direction of structural fissures.Combined with the analysis of the geological structure background and collapse cases in the study area,the geological structure modes and types of karst collapses were summarized and discussed,and three typical failure types of karst collapses were proposed,namely straight-wall failure type,channel failure type and disintegration-creep failure type karst collapse.Combining with the results of field investigations,the correctness is confirmed.2 The study of the mechanism of karst collapse based on the coupling of seepage and stress revealed the influence of the thickness of the overburden on the karst collapse.Based on the theory of seepage-stress coupling,using ABAQUS software and secondary development based on Python language,the process of karst collapse with different overburden thicknesses under the action of rainfall was simulated and analyzed.It revealed the seepage law of the overburden soil under the action of rainfall.By analyzing the influence of the overburden thickness on the collapse process,the results showed that the failure mode of the overburden karst collapse in the area is related to the overburden thickness.In the thinner covering layer,the collapse is dominated by the straight-wall failure type.In the medium thickness covering layer,it has experienced the transition from disintegration-creep failure type I to disintegration-creep failure type II.In the thicker covering layer,mainly disintegration-creep failure type III collapse.3 The frequency ratio method was used to analyze the correlation between different influencing factors and karst collapse,and an evaluation index system for karst collapse in the study area was established.According to the characteristics of the study area,selected the topography(height,slope,slope direction,surface curvature,section curvature,slope,surface roughness),geological structure(distance from fault),lithology,hydrogeological conditions(formation rich in water,terrain humidity index),overburden characteristics(soil thickness)and underengineering activities(distance from tunnel)and so on six class 13 potential impact factor for karst collapse factor,and a geospatial database of the study area was constructed.The correlation between each influencing factor and karst collapse was analyzed.The results showed that the distribution of the collapse in the area is closely related to the location of the tunnel,and it occurs mainly in the karst trough valley area,and the farther away from the underengineering activities,the fewer the number of collapse.4 A method for evaluating the susceptibility of karst collapse combining geographic detectors and analytic hierarchy process was established,and the method was used to evaluate the susceptibility of karst collapse in the study area.When using the analytic hierarchy process,the geographic detector was used to select the influence factors and provide a reference for constructing the pair comparison matrix,reducing the influence of subjective factors and the experience,and the prediction accuracy of karst collapse susceptibility evaluation was improved obviously.In the susceptible zoning,89.6% of the collapse occurr in the highly susceptible zoning,while in the susceptible zoning without AHP optimization,only 64.5% of the collapse occurr in the highly susceptible zoning.5 The evaluation methods of karst collapse susceptibility based on grid search method and Bayesian algorithm optimization random forest were constructed respectively,and compared and analyzed with the analytic hierarchy process.The research results show that the AUC values of using grid search method and Bayesian algorithm to optimize random forest are 0.936 and 0.946,respectively.92.0% and93.9% of the collapse occurr in the high-prone areas with an area of 10.6% and 7.4%respectively.By comparing the ROC curve,AUC value,and susceptibility zone statistical results of four evaluation methods such as single analytic hierarchy method,analytic hierarchy method combined with geographic detector optimization,grid search method to optimize random forest,and Bayesian algorithm to optimize random forest,the research shows that the Bayesian optimization random forest model has the highest accuracy and is more consistent with the actual collapse distribution.Finally,the Bayesian optimization random forest model was used to construct the zoning map of karst collapse susceptibility in the study area. |
PDF Full Text Request