| The development of fractures and weak layers,muddied interlayers,loose structure,and high porosity are properties of the red-bed soft rocks.Landslide disasters on red-bed soft rock slopes are frequent due to the infiltration of rainwater in the subtropical humid and rainy climate,thus threatening directly regional engineering activities and the safety of people’s lives.Nevertheless,many studies on the red-bed soft rock slopes focus on the lithological characteristics of the red-bed and the analysis methods of slope stability.The mechanism of instability and susceptibility prediction of the red-bed soft rock slopes under the effects of engineering excavation and rainwater infiltration are still not fully understood,such as the failure model,the contribution of rainfall infiltration to the damage of red-bed soft rock slopes,the pattern of temporal evolution and the prediction of regional red-bed soft rock landslide susceptibility.These resulted in invalid slope reinforcement measures and even serious slope safety accidents.In summary,the instability mechanism of red-bed soft rock slopes caused by rainfall infiltration requires to be clarified.Further,the prediction method of red-bed soft rock landslide susceptibility needs to be improved and a scientific basis could be formed for the route planning and slope reinforcement design technical measures of highway and railway projects in subtropical regions.In this paper,field investigation,finite element numerical simulation analysis,and machine learning were performed to examine the failure mechanism and susceptibility assessment of these red-bed soft rock slopes in the subtropical climate.The main research work and achievements of this paper are as follows:(1)Based on the field test of a typical engineering cutting slope,the failure mode,landslide formation condition,influencing factors,and failure evolution pattern of red-bed soft rock slope are found.The results show that during the excavation process,large horizontal displacement occurs at the front edge of the slope,and the initial plastic zone develops,resulting in a shallow landslide.During 20 days of continuous rainfall,the water content in the shallow layer of the slope increases continuously,and a transient saturated area forms at the surface of the slope.In addition,rainwater seeps down along the cracks to form a penetrating zone,thus accelerating the process of rock and soil mass softening,which further reduces the factor of safety of the slope.The combined effects of the excavation and rainfall ultimately lead to the failure of the siltstone slope;however,continuous rainfall is the key factor triggering deep sliding.The deformation and failure of the slope mainly undergo four stages:the local collapse of the slope surface,formation of the plastic zone at the foot of the slope,bulging at the toe,and formation of tension cracks in the crown of the landslide.The failure mode of the siltstone slope belongs to be a retrogressive-type of the front edge bulging and trailing edge tension cracking.Based on the deformation characteristics and the failure mechanism of the landslide,comprehensive control measures including interim emergency mitigation measures and long-term mitigation measures are proposed.(2)A physics-based slope model coupled with a hydrological model is used to simulate the factor of safety and porewater pressure development of unsaturated slopes with various rock masses under diverse rainfall scenarios.The change and water migration pattern of pore water pressure,safety factor of argillaceous sandstone,siltstone,and mudstone under regional rainfall conditions were studied:rainfall pattern affects infiltration characteristics,pore water pressure,and infiltration depth,the safety factor is also controlled by the rainfall pattern.Among all the rainfall patterns,the fastest decreasing rate of the safety factor with time is the advanced type,followed by uniform,normal,and delayed type.Both siltstone and mudstone slopes are more sensitive to long-term rainfall than short-term rainstorms.The infiltration depth of rainwater on siltstone slopes is in the range of 16-19 m,and the infiltration capacity is much greater than that of mudstone and argillaceous sandstone slopes,the infiltration depth is 3 to 5 times that of mudstone and argillaceous sandstone slopes.(3)A progressive failure model of red-bed soft rock slopes under rainfall conditions was established,and the progressive failure mode and the depth of the sliding surface were identified during the effects of rainfall pattern and duration:there is a saturation zone on the slope surface of the argillaceous sandstone slope,which provides a seepage channel for the infiltration of rainwater along the parallel slope surface.Translational shallow landslides occurred in the transient saturation zone of the sandstone slope,and the depth of the failure plane was controlled by rock mass properties,independent of rainfall patterns and duration,the depth of the sliding surface is between 1 and 5 m.The instability mode of the siltstone slope exhibits a retrogressive type in which the sliding surface of the slope increases with an increase in rainfall duration.(4)For several forms of red-bed soft rock progressive slip,11 evaluation index factors in 4 categories were selected to study the spatial distribution on the red-bed basin in the Sanshui area,including topography,stratigraphic lithology,hydrological environment,and human engineering activities.Herein,the multi-factor-based red-bed landslide catalog database was established.Based on a naive bayesian model,random forest and logistic regression,a coupled model has been constructed,thus proposing a regional red-bed landslide susceptibility prediction method.According to the comparative analysis of the susceptibility prediction accuracy,it can be seen that the area under the curve of the success rate is relatively high using the logistic regression coupled with the naive bayesian model and the random forest model.It means that the coupled model has accurate prediction performance for the spatial distribution of landslide susceptibility.Compared with the frequency ratio of the traditional coupled model,the random forest models and logistic regression models coupled with the naive bayesian model without landslide classification has limited evaluation accuracy.It is noted that the prediction accuracy can be greatly improved with the landslide classification.The random forest model coupled with the naive bayesian model can process discrete data and continuous data.Moreover,it is more suitable for the prediction of red-bed landslide susceptibility.(5)Based on the big data of regional red-bed landslides and maximum rolling rainfall in Guangdong Province during the past 20 years,two types of typical red-bed soft rock landslides(argillaceous clastic and argillaceous rock)were analyzed,including the distribution,scale,precipitation,topography,and geological factors.The number and volume of argillaceous clastic,argillaceous rock landslides and the maximum rolling rainfall intensity are combined to derive correlations.To achieve a relatively fast and accurate prediction for argillaceous clastic and argillaceous rock landslide activities under rainfall conditions,thresholds for the volumelevel of the red-bed soft rock landslide scale are derived by identifying the maximum rolling rainfall intensity.The magnitudes of argillaceous clastic and argillaceous rock landslide occurrences are closely related to the 1 to 144 hours MRR intensity.The prediction accuracy of argillaceous clastic landslide volume-level threshold equations is increased as small-sized,medium-sized,large-sized,super-large-sized,and giant-sized landslides.In summary,the overall prediction accuracy of argillaceous landslide volume-level rainfall threshold equations is higher and acceptable. |
PDF Full Text Request