Research On Prediction Model Of Loess Collapsibility Based On Microstructure Unit Theory

The loess collapsibility evaluation is one of the key fundamental issues that are indispensable in loess collapsibility theoretical study and engineering practice,but urgently need to be broken through.In this paper,Q₃Malan Loess in typical collapsible loess distribution areas of the Loess Plateau（Lanzhou,Yan’an,Yanchuan,Jingyang,Xi’an）is taken as the research object.By a sequence of laboratory tests and microscopic test,the types and properties of different scale particles and aggregates are systematically analyzed,and establish a clear conceptual model of the basic structural unit of loess collapsibility.Quantify different scale particles and Interparticle connections,and establish the relationship between the microstructure parameters（pore proportion,silt proportion,colloidal-clay proportion,inter-particle connection proportion,etc.）and collapsibility coefficient.On this basis,a new way to quantitatively characterize the collapsible deformation based on the microstructural unit theory is explored.The main research subjects and achievements include:1.The regional distribution law of basic physical and chemical properties of the loess samples in Lanzhou,Yanchuan,Yan’an,Jingyang,and Xi’an such as water content,specific gravity,void ratio,liquid plastic limit,particle distribution,mineral composition,chemical composition,microstructure,pore distribution,specific surface area were achieved,the results provide theoretical basis for further study the genesis and engineering properties of collapsible loess.2.By series of experiments,the influence of different pretreatment methods on the measurement of the grain size of Malan loess was revealed,and the suitable pretreatment methods and particle separation methods were found.3.Based on the analysis of a series of microscopic images,the type,material composition and properties of different scale particles and aggregates were studied in depth,and a clear conceptual model of the basic structural unit of loess collapsibility was established.The basic structural unit in loess in the form of single particles was mainly primitive mineral fragments.According to particle sizes,it can be divided into coarse debris（>75μm）,fine debris（5-75μm）,micro debris（<5μm）submicron（100 nm-1.0μm）and nano（1-100 nm）debris.In loess,the main structural units in the form of particle unit aggregates were coated particles and aggregates.According to the formation effect of the coated particles,they can be divided into types:filling effect,filling rounding effect,and thickening increasing effect.According to the composition and size of aggregates,they can be divided into four types:colloidal clay aggregates,single aggregates,multi-particle aggregates,and composite aggregates.The stability of each type of aggregates was as follows:colloidal clay aggregates>single aggregates>multi-particle aggregates>composite aggregates,and on this basis,a conceptual model of aggregates aggregation hierarchy was established.4.By analyzing the changes of pore area,quantity,total porosity and pore number of different sizes before and after collapsibility,the contribution of different pore sizes to the total collapsibility during the process of loess collapsibility was studied.Among them,the existence of mesopores was the spatial basis for the collapsible deformation.The humidification process was a necessary prerequisite for the occurrence of loess collapsible deformation.In the process of humidification,the water content change will inevitably cause the change of loess basic structural units,the water content change will inevitably cause the change of the connection between loess basic structural units.During the process of loess collapsibility,there were both strength changes caused by the change of water film during compression,and strength changes caused by a series of physical and chemical changes of clay minerals.5.In the process of loess microstructure image processing,an image segmentation method based on Gaussian mixture model was proposed,which can segment the image into three parts of silt,colloidal-clay and pores without manual threshold,and the obtained pore area ratio P_ARdecreases with the increase of collapsibility pressure,silt area ratio P_S,colloidal-clay area ratio P_Cincreases with the increase of pressure.Through qualitative analysis of microstructure images,the concept of contact area ratio among loess basic structural units was proposed and quantified.The results showed that with the increase of collapsible pressure,both direct contact area ratio P_CRand indirect contact area ratio C_CRshowed an increasing trend.Pore area ratio P_AR,colloidal-clay area ratio P_C,test resultant adhesive particle content P_W,direct contact area ratio P_CR,indirect contact area ratio C_CRand loess collapsibility coefficient were all significantly correlated,among which collapsibility coefficient and pore area ratio P_ARwas positively correlated,and colloidal-clay area ratio P_C,test resultant colloidal-clay content P_W,direct contact area ratio P_CR,indirect contact area ratio C_CRwere negatively correlated.Compared with colloidal-clay content P_Wobtained in the test,the correlation between the colloidal-clay area ratio P_Cand the collapsibility was more significant.6.Based on the above research results,the effective combination of the simple multiple linear regression model and the generalized accumulation model is explored to realize the prediction and evaluation of the collapsibility of loess.The model is based on the theory of microstructure units and comprehensively considers the macro to micro factors affecting the loess collapsibility,for rapid and accurate quantitative characterization of loess collapsibility coefficient provides a new way.At the same time,the generalized accumulation model extensively considers the spatial random effects in the collapsibility data,and the model can also be used for spatial prediction of loess collapsibility.In addition,the process of model building not only revealed the statistical significance of the explanatory variables such as water content,vertical pressure,porosity,dry density,colloidal-clay area ratio,and contact area ratio,but also the sensitivity of different explanatory variables to loess collapsibility is also quantitatively evaluated.