CN113240357A

CN113240357A - Rapid evaluation method for stability of highway side slope in red layer area

Info

Publication number: CN113240357A
Application number: CN202110784548.XA
Authority: CN
Inventors: 郑达; 陈强; 张文; 张硕; 吴鑫泷
Original assignee: China Huaxi Engineering Design & Construction Co ltd; Chengdu Univeristy of Technology; Institute of Exploration Technology Chinese Academy of Geological Sciences
Current assignee: China Huaxi Engineering Design & Construction Co ltd; Chengdu Univeristy of Technology; Institute of Exploration Technology Chinese Academy of Geological Sciences
Priority date: 2021-07-12
Filing date: 2021-07-12
Publication date: 2021-08-10

Abstract

The application provides a method for quickly evaluating the stability of a highway slope in a red layer area, which comprises the following steps: determining various types of evaluation indexes influencing the stability of the red layer slope; rapidly acquiring a sample data set, wherein the sample data set is data of multiple groups of multi-type evaluation indexes of the red layer slope of the sample, and the data of the multi-type evaluation indexes of the red layer slope of each group of sample carries pre-labeled actual stability grade of the corresponding red layer slope of the sample; training based on an SVM algorithm according to the sample data set, and establishing a rapid evaluation model; rapidly acquiring data of various evaluation indexes of the red layer slope to be evaluated; and inputting the multi-type evaluation index data of the red layer slope to be evaluated into the rapid evaluation model to obtain the stability grade of the red layer slope to be evaluated. By the rapid evaluation method, the stability grade of the red layer side slope can be rapidly evaluated, timely and effective support of the side slope is facilitated, and the construction time and the support magnitude are saved.

Description

Rapid evaluation method for stability of highway side slope in red layer area

Technical Field

The embodiment of the application relates to the technical field of rock and soil, in particular to a method for quickly evaluating the stability of a road slope in a red layer area.

Background

The redzone is a land debris sedimentary rock stratum mainly comprising a mountain foot Hongguan phase, a river phase, a lake phase or a river-lake alternate phase formed in the Jurassic period, the chalk phase, the ancient period and the recent period, the appearance usually takes red as a main color tone, lithology mainly comprises soft rock (claystone, mudstone, shale and the like) and hard rock (sandstone, siltstone and the like), the sedimentary environment is special, internal cementation is not tight, the unique properties of redzone rock mass, particularly redzone soft rock are formed, the mudstone and the siltstone have water sensitivity and expansibility, and are easy to disintegrate and weather or even argillize when meeting water, and as a special stratum, the redzone slope deformation instability mode is complex, and landslide disasters are easy to occur.

Construction of engineering inevitably needs to cross red layer areas, and highway construction usually adopts simultaneous construction of a plurality of sections on the whole line, so that numerous excavation side slopes appear in a short period, and disasters such as bedding landslide caused by rock deformation, local collapse and exposure of weak interlayers, mass-generated near-horizontal landslide caused by strong rainfall, and accumulation landslide can happen to a plurality of red layer side slopes in the excavation process or shortly after excavation. Based on this, the related technical staff needs to perform stability evaluation on the side slope, determine the support magnitude of the side slope according to the stability grade of the side slope, and perform effective support on the side slope.

Based on the characteristics that the deformation instability mode of the red layer slope is complex and landslide and collapse disasters easily occur, aiming at numerous red layer slopes excavated in short time in road construction, if a traditional slope stability evaluation method in the related technology is adopted, namely stability evaluation is performed on the red layer slope according to manual working experience, the workload is huge, the stability evaluation of numerous red layer slopes is difficult to complete in short time, and the situation that landslide disasters occur due to untimely stability evaluation may occur. Moreover, since the manual evaluation mode is greatly influenced by subjective factors such as the working experience of technicians, objective basis is lacked for slope stability evaluation, and the evaluation effect of slope stability is poor.

Disclosure of Invention

Based on the technical problems, the embodiment of the application provides a method for quickly evaluating the stability of the side slope of the highway in the red-layer area, which can comprehensively, quickly and objectively evaluate the stability of the side slope in the construction process of the highway in the red-layer area, is beneficial to timely and effectively supporting the side slope, and saves the construction time and the supporting magnitude.

The embodiment of the application provides a method for quickly evaluating the stability of a highway slope in a red layer area, which comprises the following steps:

determining various evaluation indexes of influencing factors influencing the stability of the red layer slope; the multi-class evaluation indexes comprise: the method comprises the following steps of (1) slope height, slope width, lithology combination characteristics, structural surface filling characteristics, rock mass structure type, slope excavation series, slope excavation height, average excavation slope and maximum single-day rainfall;

rapidly acquiring a sample data set, wherein the sample data set is data of multiple groups of multi-type evaluation indexes of the red layer slope of the sample, and the data of the multi-type evaluation indexes of the red layer slope of each group of sample carries pre-labeled actual stability grade of the red layer slope of the corresponding sample;

training based on an SVM algorithm according to the sample data set, and establishing a rapid evaluation model;

rapidly acquiring data of various evaluation indexes of the red layer slope to be evaluated;

and inputting the data of the various evaluation indexes of the red layer slope to be evaluated into the rapid evaluation model to obtain the stability grade of the red layer slope to be evaluated.

According to the method, a rapid evaluation model is established by adopting various evaluation indexes including side slope height, side slope width, lithologic combination characteristics, structural surface filling characteristics, rock mass structure types, slope body structure types, side slope excavation series, side slope excavation height, average excavation slope and maximum single-day rainfall, and the side slope stability in the highway construction process in the red layer area can be comprehensively, rapidly and objectively evaluated based on the rapid evaluation model.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the description of the embodiments of the present application will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without inventive exercise.

Fig. 1 is a flowchart illustrating a method for rapidly evaluating the stability of a road slope in a red layer region according to an embodiment of the present application;

FIG. 2 is a chi-square test result diagram of each evaluation index in a sample data set according to an embodiment of the present application;

fig. 3 is a flowchart illustrating a model building method according to an embodiment of the present application.

Detailed Description

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are some, but not all, embodiments of the present application. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.

Based on the method, the method can be used for comprehensively, quickly and objectively evaluating the stability of the side slope in the construction process of the highway in the red layer, is beneficial to timely and effectively supporting the side slope, and saves the construction time and the magnitude of the support.

Referring to fig. 1, fig. 1 is a flowchart illustrating a method for rapidly evaluating stability of a road slope in a red layer region according to an embodiment of the present application. As shown in fig. 1, the method for rapidly evaluating the stability of a red layer road slope provided by this embodiment is applied to a red layer slope, and is particularly suitable for evaluating the stability of a red layer slope in a red layer road construction process, and the method includes the following steps:

step S11: determining various types of evaluation indexes influencing the stability of the red layer slope; the multi-class evaluation indexes comprise: the method comprises the following steps of slope height of the side slope, slope width of the side slope, lithology combination characteristics, structural surface filling characteristics, rock mass structure type, slope excavation series, slope excavation height, average excavation slope and maximum single-day rainfall.

In this embodiment, the influence factors of the red layer slope collapsing disaster can be obtained by analyzing the common deformation and damage mode of the red layer area highway slope (hereinafter referred to as red layer slope). Wherein, the deformation failure mode of red layer side slope can be divided into: flat push type slip damage, plastic flow-pull crack type damage, bedding slip pull crack type damage, wedge damage, compression pull crack type collapse damage, slip-bend-shear type damage, basal overlay interface slip-pull crack type slip, and internal creep-pull crack type slip.

In this embodiment, through analyzing the common deformation failure modes of the red layer slope, it is found that the occurrence of the red layer slope collapse disaster is a result of the combined action of a plurality of influence factors, and the red layer slope collapse disaster can be generally divided into two types: the method comprises the following steps of firstly, geological factors, namely geological environment conditions of the side slope, which are basic conditions affecting the stability degree of the side slope under natural conditions, such as topographic and geomorphic conditions, stratigraphic and lithological characteristics, rock structural plane characteristics, slope structural characteristics, hydrogeological conditions and the like; the second is the inducing factors such as engineering excavation, rainfall, earthquake and the like.

Wherein, the topographic conditions comprise the slope height, scale, slope shape, temporary condition and the like of the side slope. Under the condition that other conditions are not changed, the larger the slope height is, the more easy the side slope is to be unstable; the slope and the slope stability are not simple linear relations but act together with other geological condition factors, and when the combination of the strata lithology of the slope is combined into a whole, the slope is easy to destabilize when the slope is larger; when the rock property combination of the side slope stratum is combined, the larger the gradient is, the more easy the side slope is to be unstable; the same slope design, the larger the gradient, the lower the height of the side slope which can be excavated in construction; the larger the slope height and the slope width are, the larger the scale of the slope body is, the influence range of the slope during deformation is increased, and the damage is more serious; the larger the face range is, the probability and scale of uncovering the unstable rock mass are increased.

2. The lithology characteristics of the stratum are the material basis for forming the side slope, determine the physical and mechanical properties of the side slope, influence the position and range of the instability of the side slope and play a key role in the instability mechanism of the side slope. The red layer side slope has its special lithology constitution and inner structure, according to reconnaissance and research on the spot, the red layer rock mainly divide into two main types: one is hard rock (hard and harder), such as sandstone, siltstone, etc.; one is soft rock such as mudstone, shale, claystone, etc.

3. The structural surface characteristics of the rock mass are important factors influencing the stability of the rock mass, generally, the structural surface characteristics refer to an interface formed inside the rock mass in the diagenesis process or the later-stage transformation effect, the rock mass is divided and layered, the interior of the rock mass is more complex and is full of uncertainty, the structural surface often forms a potential sliding surface of a side slope and a rock mass cutting surface, and the overall and local stability of the side slope is controlled.

4. The slope body structure which is easy to destabilize after the excavation of the road side slope comprises a nearly horizontal rock stratum structure, an inclined rock stratum structure and a stacking layer structure; the slope structural characteristics play a control role in the stability of the slope, mainly refer to the relationship between the occurrence, properties, spatial positions and the like of a structural surface and the slope, a rock body generates a large number of joint cracks in the rock body under the influence of the structural action or external stress, the joint cracks in different directions and types and the slope form various combined structures, and the combined structures are discussed to influence the stability of the slope, so that the type, the range and the scale of the instability of the slope are analyzed.

5. The hydrogeological conditions mainly refer to red layer slope groundwater, and the main types of the red layer slope groundwater comprise loose rock pore water, clastic rock pore water, erosion pore water and bedrock fracture water. The bed rock fracture system is a main space for infiltration, migration and storage of underground water, and factors such as the scale, density, opening degree, connectivity, water permeability and the like of the fracture determine the water supply and water conductivity of the red layer slope and are of great importance to the stability of the slope. Rock permeability has been considered in both lithological and structural surface characteristics, and excavation of highway slopes often creates favorable conditions for groundwater to be drained to the outside.

6. Most of slope deformation and instability are related to rainfall, and the rainfall mainly has the following contribution effects on red layer landslide: the weathering and disintegration speed of rock mass is accelerated. ② the self weight of the slope is increased. Softening effect: rainfall infiltrates into the slope and is converted into underground water, and the underground water can soften the rock-soil body, which is mainly reflected in lubrication of the structural surface and softening of the slippery soil. Mechanical action: when water is filled in the fractured vertical fractured rock mass, hydrostatic pressure which is nearly vertical to the fracture wall can be generated, so that the fracture deformation is increased.

7. The engineering excavation changes the geometrical characteristics and the geological environment of a natural side slope, is an important inducing factor of slope instability, and large-area unloading can occur in the direction of the free face after excavation, so that large-scale disturbance deformation is caused.

In this embodiment, based on the analysis of the red layer slope instability mode (i.e., deformation failure mode) and the slope stability influence factors, a plurality of types of suitable evaluation indexes are selected according to systematic, scientific, easy-to-obtain, representative screening principles and the role of each influence factor in the red layer slope stability.

The multiple types of evaluation indexes selected in this embodiment include: the method comprises the following steps of slope height of the side slope, slope width of the side slope, lithology combination characteristics, structural surface filling characteristics, rock mass structure type, slope excavation series, slope excavation height, average excavation slope and maximum single-day rainfall.

Wherein, (1) the width of the side slope

In this embodiment, the width of the side slope refers to the length of the excavation surface of the side slope along the line, and may often reflect the extent of the artificial disturbance of the side slope, and is in positive correlation with the deformation scale of the slope body. This evaluation index can be quantitatively expressed in the unit "meter (m)".

(2) Side slope height

In this embodiment, the slope height of the side slope refers to the height between the bottom of the side slope and the top of the side slope, which reflects the scale of the side slope. This index can be expressed quantitatively in units of "meters (m)".

(3) Lithological combination of features

Different lithologic combination characteristics determine different stable slope angles of the side slope, and further influence the stability of the side slope, for example, in other side slopes, a stable near-horizontal rock layer side slope is usually formed, and in a red layer side slope, if rock layer combination is a soft-hard interbed, and under the conditions that cracks and pulling grooves exist on the rear edge, flat-pushing type landslide damage is easily generated. Therefore, important consideration is needed when evaluating the stability of the side slope, and the lithology combination characteristic indexes are divided into the following parts by combining the characteristics of the red layer: hard rock, soft rock, hard rock and soft rock, soft rock and hard rock, sand-mud rock interbed and fourth series accumulation layer, and the indexes are expressed in qualitative form.

(4) Average excavation slope

The main body evaluated in the embodiment is an excavation slope, and the slope rate of the excavation slope commonly seen in engineering is 1:0.25, 1: 0.5, 1:0.75, 1:1, 1:1.25 and 1:1.5, and adopting various slope-releasing rates for some side slopes, so that an average excavation slope concept is introduced, the average excavation slope refers to an included angle between a connecting line between a slope foot excavation position and an opening line and a horizontal plane, the steepness degree of the excavation surface of the side slope can be visually represented, and the index can be quantitatively represented in unit degree (degree).

(5) Height of slope excavation

In this embodiment, the slope excavation height refers to a height difference between an excavation position of a slope toe and an opening line, and may reflect the size of an excavation section and the degree of the slope being artificially disturbed to a certain extent, and generally form a positive correlation with the deformation scale of a slope body, and this index may be quantitatively expressed in a unit of "meter (m)".

(6) Structural surface filling features

The mechanical property and the space position of the structural surface play a role in controlling the stability of the side slope, and the filling characteristics of the red-layer structural surface are divided into the following four types according to the shear strength: closed no-filling, open less filling, thin layer filling, thick layer filling. This index is expressed in qualitative form.

Closed without filling: the structural surface features are mainly presented in a deposition structural surface and a closed structural joint surface, and the cementation property between layers is better without filling.

Opening and less filling: the structural surface is characterized by mainly existing in a structural joint surface, the structural surface is opened by 3-50 mm and is filled with a small amount of broken stone and argillaceous fillers, the structural surface is usually found in sandstone rock bodies of red bed nearly horizontal and gentle dip rock slope, the structural surface is in rough and rough state, the joint surface is usually used as a boundary surface of rock body deformation instead of a shear surface, and most of the joint surface is a channel for rainwater infiltration, gathering and migration.

Thin layer filling: the structural surface is characterized in that a structural joint surface and an interlayer are dislocated, rainfall infiltration is accompanied by argillaceous and clay filling, the thickness is about 2-5 mm, and the interlamination is smooth.

Filling a thick layer: the structural surface features mainly exist between adjacent rock layers, and thick-layer fillers of about 5-30 cm are filled between the rock layers, and generally are products of soft rock which is transformed into cement and softened when meeting water.

(7) Slope excavation progression

In this embodiment, the grade of the side slope excavation is designed during the side slope excavation, a reasonable construction scheme is designed according to the geological conditions and the construction conditions of the side slope engineering, the grade is usually 1-8, the influence of the height of the side slope excavation is large, a positive correlation relationship is formed with the deformation scale of the side slope under general conditions, and the index can be quantitatively expressed in unit grade.

(8) Structural type of slope

The slope body structure reflects the combination relation between the attitude of the structural plane and the trend of the slope, and for the highway side slope in the red layer area, the difference of the slope body structure determines the difference of the deformation position, the instability mode and the cause type of the side slope, and the difference is a key factor for controlling the stability of the red layer side slope, and the slope body structure type of the red layer side slope is divided into: the near-horizontal side slope, the forward slope, the reverse slope and the accumulation side slope are four categories. This index is expressed in qualitative form.

(9) Structural type of rock mass

The red horizon rock mass constitutes and uses sandstone, shale, mudstone, conglomerate etc. as the main, mostly is the sedimentary rock, lamellar structure, consequently, this embodiment divides red horizon slope rock mass structure type into according to rock stratum thickness, lithology combination: thick-layer structure, medium-thickness layer structure, thin-layer structure and (stacked body and bedrock) binary structure. This index is expressed in qualitative form.

(10) Maximum single-day rainfall

Rainfall is a main inducing factor of red layer landslide, the density of landslide points in an area and single-day rainfall form a positive correlation relationship, the maximum single-day rainfall is an important index for measuring the rainfall intensity of a certain area, the range is usually 0-250 mm, and in order to facilitate collection and modeling of related technicians, the evaluation index is expressed in an interval form and the unit is 'mm'.

In a preferred embodiment, step S11 specifically includes the following sub-steps:

step S111: analyzing and determining influence factors influencing the stability of the red layer slope; the influencing factors at least comprise: geometric, lithological and structural, and external triggering features.

In this embodiment, for the engineering side slope in the red zone, the stability mainly focuses on the geometric characteristics, lithology and structural characteristics, excavation rainfall and other external factors.

Step S112: and determining various types of evaluation indexes representing the influence factors according to the influence factors.

In this embodiment, step S112 specifically includes the following sub-steps:

sub-step S1121: and determining the slope height and the slope width as evaluation indexes representing the geometric characteristics based on the slope height and the slope width in the geometric characteristics and the susceptibility degree and the correlation degree of slope instability.

In this embodiment, the geometric characteristics of the side slope influence the susceptibility and the instability scale of the side slope instability, wherein the factors such as the side slope height, the side slope width, and the slope in the geometric characteristics are closely related to the susceptibility and the scale of the side slope instability, and therefore, the side slope height and the side slope width in the multiple types of evaluation indexes are determined as the evaluation indexes representing the geometric characteristics.

Substep S1122: and determining the lithologic combination characteristics, the structural surface filling characteristics, the rock mass structure type and the slope body structure type as evaluation indexes representing lithologic and structural characteristics based on geological factors forming the red bed slope.

In this embodiment, the lithological combination of the side slope is a material foundation constituting the side slope, determines the physical and mechanical properties of the side slope, and determines the stable slope angle and the instability mechanism of the excavated side slope to a great extent, and it is because the red layer argillaceous rock contains a large amount of loose or consolidated argillaceous minerals, such as montmorillonite and illite, which have strong hydrophilicity, soften and expand when meeting water, so that the argillaceous rock is relatively easy to generate plastic deformation under the action of internal and external forces, and the red layer specific sand-shale interbedded combination is a foundation for the occurrence of gentle rock layer landslide and large rock landslide in the red layer region. The combination between the red layer surfaces is poor, and is influenced by the structural action and the weathering unloading action, a frequently developed argillized interlayer, an interlaminar dislocation zone, a large number of structural joints and cracks in a slope body, the fluctuation degree, the development density, the contact state and the filler characteristics play a key control role in the shear strength of the structural surfaces.

Therefore, in the present embodiment, based on the above-described geological factors constituting the red bed slope, the lithologic combination characteristic, the structural plane filling characteristic, the rock mass structure type, and the slope structure type of the multiple types of evaluation indexes are determined as the evaluation indexes representing the lithologic and structural characteristics.

Substep S1123: and determining the grade number of the slope excavation, the height of the slope excavation, the average excavation gradient and the maximum single-day rainfall as evaluation indexes representing the external triggering characteristics based on the triggering factors of red layer slope instability.

In the embodiment, through research on different types of slope instability modes, it is found that the instability of a near-horizontal slope is mainly influenced by factors such as rainfall, excavation unloading, structural cracks, lithologic composition and the like, and in the excavation unloading process of the near-horizontal sand-shale interbedded slope, due to inconsistent layer strain, an interlayer shear crack surface is generated, under the action of rainfall and underground water, a argillized interlayer is softened, the shear strength is reduced, and the instability occurs under the combined action of hydrostatic pressure and bottom uplift pressure in the cracks. The position of a soft rock stratum in a bedding side slope is very important, if a soft interlayer is exposed during excavation, the side slope can creep and slide under the action of only self-weight stress, the side slope has the risk of large-scale instability during excavation under the promotion of rainfall, the stability of the reversely inclined side slope is influenced by the combination of a main control structural surface and lithology, and collapse disasters frequently occur on the side slope under the conditions that soft and hard rocks are differentially weathered and joints penetrate through the stratum. The slope of the accumulation body is closely related to the hydrogeological conditions and excavation degree of the slope no matter whether the base-cover interface slides, pulls, cracks and is unstable or the internal creep-pull crack of the accumulation body is unstable.

Therefore, in this embodiment, based on the triggering factor of red slope instability, the slope excavation order, the slope excavation height, the average excavation slope and the maximum single-day rainfall in the multiple types of evaluation indexes are determined as the evaluation indexes representing the external triggering characteristics.

In this embodiment, 10 rapid stability evaluation indexes reflecting side slope stability influence factors are selected, where the geometric characteristic indexes of the side slope include side slope height and slope width, the lithological and structural characteristic indexes include lithological combination characteristics, structural surface filling characteristics, rock mass structure types, and slope body structure types, the external trigger factor indexes include side slope excavation progression, side slope excavation height, average excavation slope, and maximum single-day rainfall, and these indexes are all easily and rapidly acquired, and they interact to form a rapid red layer highway side slope stability evaluation index system, and the rapid red layer highway side slope stability evaluation index system can relatively and comprehensively evaluate the red layer side slope stability.

Step S12: and rapidly acquiring a sample data set, wherein the sample data set is data of multiple groups of multi-type evaluation indexes of the red layer slope of the sample, and the data of the multi-type evaluation indexes of the red layer slope of each group of sample carries the pre-labeled actual stability grade of the corresponding red layer slope of the sample.

In this embodiment, after determining multiple types of evaluation indexes representing red slope stability influence factors, a sample data set needs to be obtained quickly. The sample data set is data of multiple types of evaluation indexes of multiple groups of sample red layer slopes, and the sample red layer slopes are red layer slopes which are subjected to stability evaluation and known to have actual stability grades. For a sample red layer side slope, data of multiple types of evaluation indexes corresponding to each sample red layer side slope in multiple sample red layer side slopes needs to be quickly acquired, that is, each sample red layer side slope in a sample data set needs to quickly acquire the slope height, the slope width, the lithological combination characteristics, the structural surface filling characteristics, the rock mass structure type, the slope body structure type, the side slope excavation series, the side slope excavation height, the average excavation slope and the maximum single-day rainfall of the sample red layer side slope. And the data of the slope height, the slope width, the lithologic combination characteristic, the structural surface filling characteristic, the rock mass structure type, the slope body structure type, the slope excavation series, the slope excavation height, the average excavation slope and the maximum single-day rainfall of one sample red layer slope are the data of multiple types of evaluation indexes of one group of sample red layer slopes, and the data of the multiple types of evaluation indexes of each group of sample red layer slopes carries the actual stability grade of the corresponding sample red layer slope marked in advance, so that a sample data set is formed by the data of the multiple types of evaluation indexes of the multiple groups of sample red layer slopes. It should be noted that the number of the red layer slopes of the samples in the sample data set in this embodiment may be set according to actual requirements, for example, the sample data set may have 50 sets of data of multiple types of evaluation indexes of the red layer slopes of the samples, and the number of the red layer slopes of the samples in the sample data set is not specifically limited in this application.

In a preferred embodiment, the present application further provides a method for quickly acquiring a sample data set, and specifically, the method may include:

step S21: and rapidly acquiring the data of the slope width, the slope height, the slope excavation progression and the average excavation slope index of each red layer slope of the sample data set by using an unmanned aerial vehicle three-dimensional live-action modeling technology.

In actual engineering, once a slope is disturbed by excavation, large-area unloading can occur in the direction of an empty surface, and large-scale disturbance deformation is likely to be caused, so that researchers are required to rapidly acquire evaluation index data of the slope. In the related technology, data acquisition is carried out by means of a tape measure, a compass and the like under most conditions of geometric characteristics (slope width and slope height) of a side slope and excavation characteristics (excavation height, excavation series and excavation slope), in order to record accurate engineering geological condition information of a work point, an investigator always goes to a site, investigation work is hard and dangerous, and meanwhile, the problems of investigation data loss and insufficient accuracy caused by the fact that limited personnel of the terrain cannot reach are often caused,

in order to solve the above problem, in this embodiment, data of the side slope width, the side slope height, the side slope excavation progression and the average excavation slope index of each red layer side slope of the sample data set is quickly obtained by using an unmanned aerial vehicle three-dimensional live-action modeling technology.

For example, Earth Survey software integrating data processing, display and report generation and integrating geological disaster prevention and control of university of Country and geological environment protection national key laboratories can be used for data interpretation of a geological three-dimensional model, the processed three-dimensional model is opened in Earth Survey, and geometric feature data of a side slope, such as the stage number of side slope excavation, the height of the side slope, the excavation height of the side slope, the width of the side slope, the area of an excavation section and the like, can be quickly and accurately extracted within minutes by using a measurement module in the software, so that the average slope excavation data can be calculated by the difference between the horizontal distance and the difference between the altitude of two points from the excavation position of the slope toe to the opening of the slope top.

In this embodiment, acquire corresponding evaluation index data fast through unmanned aerial vehicle three-dimensional live-action modeling technique, compare in traditional survey means, unmanned aerial vehicle three-dimensional live-action modeling technique has advantages such as simple and fast, not restricted by terrain factor in the investigation of side slope rock mass geological conditions, have higher discernment degree to information such as side slope geometric characteristics data, excavation characteristic data, rock mass structure parameter, can solve traditional manual investigation and waste time and energy and precision not high, even because the complicated geology personnel of topography can't the condition of investigation, realized the scalability, the storability, the digitization of geological information, be the crucial ring in the quick evaluation work of side slope stability.

Step S22: and inquiring observation station data closest to each sample red layer side slope in the sample data set from a meteorological database to obtain the maximum single-day rainfall capacity of each sample red layer side slope in the sample data set.

In this embodiment, the data of the maximum single-day rainfall index of the observation station closest to the red layer slope of the sample can be queried through a meteorological database (such as a chinese meteorological data network), and this work can be acquired through the specific coordinates of the red layer slope of the sample before construction. For example, if data of the maximum single-day rainfall of the sample red layer slope a is to be acquired, historical data of the maximum single-day rainfall of the observation station B closest to the sample red layer slope a can be queried through the chinese meteorological data network before construction. Wherein the maximum single-day rainfall is the maximum total rainfall all day. For example, the maximum total rainfall all day in nearly ten years may be relatively representative, or the maximum total rainfall all day in nearly five years may be representative, in this embodiment, the maximum total rainfall all day in which time period the maximum rainfall all day is may be set according to an actual requirement, and the time period of the maximum rainfall all day is not specifically limited in this application.

Step S23: and inquiring survey data to obtain data of lithologic combination characteristics, structural plane filling characteristics, rock mass structure types and slope structure type indexes of each red layer slope of the sample data set.

In the embodiment, the data of the lithologic combination characteristic, the structural surface filling characteristic, the rock structure type and the slope structure type index of each red layer slope of the sample data set can be quickly obtained through the early-stage survey data.

It should be noted that, the steps S21 to S23 may be executed sequentially, may be executed simultaneously, or may be executed sequentially after being combined in pairs, and this embodiment does not set any specific limitation on the execution order of the steps S21 to S23.

Step S13: and training based on an SVM algorithm according to the sample data set, and establishing a rapid evaluation model.

Based on the analysis, the red layer slope stability influence factors are numerous, and the slope stability and the influence factors are in a very complex nonlinear relation, so that a better relational expression is difficult to find to describe the nonlinear mapping relation. Therefore, in this embodiment, after the sample data set is quickly obtained, the sample data set may be trained according to a supervised learning method according to the sample data set, each evaluation index data of the known sample red layer slope and the actual stability level are analyzed, and a nonlinear relationship between the two is found, so as to establish a red layer slope stability quick evaluation model, so that only the evaluation index data of the red layer slope is input later, and the stability level of the red layer slope can be predicted through the quick evaluation model.

In this embodiment, the specific training may be performed on the sample data set by using an SVM algorithm. The SVM is a short for Support Vector Machine, and the Chinese name of the SVM is a Support Vector Machine, belongs to a Machine learning algorithm with supervision, and can be used for classification of discrete dependent variables and prediction of continuous dependent variables. The SVM algorithm is used for finding a classification plane of data, and the idea of the SVM algorithm is to divide sample points of different classes by using a hyperplane formed by certain support vectors.

Step S14: and rapidly acquiring data of various evaluation indexes of the red layer slope to be evaluated.

After the rapid evaluation model is established, stability evaluation can be performed on the red layer slope to be evaluated through the rapid evaluation model, and data of various evaluation indexes of the red layer slope to be evaluated needs to be rapidly acquired at the moment. In this embodiment, the multiple types of evaluation indexes of the red layer slope to be evaluated correspond to the multiple types of evaluation indexes of the sample red layer slope in the sample data set for training the rapid evaluation model one by one, and the method for rapidly obtaining the data of the multiple types of evaluation indexes of the red layer slope to be evaluated is the same as the method for rapidly obtaining the sample data set in this application, and therefore, redundant description is not repeated.

Step S15: and inputting the data of the various evaluation indexes of the red layer slope to be evaluated into the rapid evaluation model to obtain the stability grade of the red layer slope to be evaluated.

In this embodiment, after obtaining the data of the multiple types of evaluation indexes of the red layer side slope to be evaluated, the data of the multiple types of evaluation indexes of the red layer side slope to be evaluated (including the data of the side slope height, the side slope width, the lithology combination characteristic, the structural plane filling characteristic, the rock mass structure type, the slope body structure type, the side slope excavation stage number, the side slope excavation height, the average excavation gradient, and the maximum single-day rainfall amount of the red layer side slope to be evaluated) may be input to the trained rapid evaluation model, so as to obtain the output stability grade of the red layer side slope to be evaluated, thereby predicting the stability grade of the red layer side slope to be evaluated.

In the embodiment, various types of evaluation indexes influencing the stability of the red layer slope are determined; rapidly acquiring a sample data set, wherein the sample data set is multi-class evaluation index data of a plurality of groups of sample red-layer slopes; training based on an SVM algorithm according to the sample data set, and establishing a rapid evaluation model; rapidly acquiring various types of evaluation index data of the red layer slope to be evaluated; and inputting the multi-type evaluation index data of the red layer slope to be evaluated into the rapid evaluation model to obtain the stability grade of the red layer slope to be evaluated. According to the method, a rapid evaluation model is established by adopting various evaluation indexes including side slope height, side slope width, lithologic combination characteristics, structural surface filling characteristics, rock mass structure types, slope body structure types, side slope excavation series, side slope excavation height, average excavation gradient and maximum single-day rainfall, based on the rapid evaluation model, comprehensive, rapid and objective evaluation can be performed on the side slope stability in the highway construction process in the red layer area, the evaluation method is not influenced by subjective factors such as the working experience of technical personnel, timely and effective support can be performed on the side slope, the construction time and the magnitude of the support are saved, and the condition that red layer slope collapse disasters are caused when the red layer side slope stability evaluation is not in time is avoided.

With reference to the foregoing embodiment, in an implementation manner, the present application further provides a method for establishing a rapid evaluation model, and specifically, the method may include:

step S31: analyzing the acquired sample data set, including: and analyzing the sample data set through univariate analysis and bivariate analysis to obtain an analysis result of the sample data set. Wherein the analysis result at least comprises one of the following: correlation among evaluation indexes in the sample data set and dirty data in the sample data set; and the correlation among the evaluation indexes in the sample data set is obtained by the bivariate analysis method.

In this embodiment, after obtaining a sample data set, the obtained sample data set is analyzed to reveal relationships between index data in multiple types of evaluation index data and determine which aspects can optimize the quality of the data set, so as to obtain an analysis result of the sample data set, where the obtained analysis result at least includes one of: correlation among evaluation indexes in the sample data set and dirty data in the sample data set; the dirty data may be duplicate data, invalid data, outliers, inconsistent values, and the like.

In this embodiment, the analysis work on the sample data set is performed by starting with single-variable analysis and double-variable analysis. During univariate analysis, the data can be divided into continuous variables (continuous numerical values, such as slope width, slope height, average excavation gradient, slope excavation height and slope excavation progression) and classified variables (discrete classification values, lithological combination characteristics, structural plane filling characteristics, rock mass structure types, slope mass structure types and maximum single-day rainfall intervals) according to different types of sample data (data of multiple types of evaluation indexes). And different analyses are carried out on the data according to the variable types, and the data are displayed in a graph mode, so that the understanding of the data is deepened more visually and clearly, and some internal rules among the data are found. The maximum single-day rainfall interval in this embodiment may be a classification type variable or a continuity variable. The data type of the maximum single-day rainfall interval is related to whether the observation station points closest to the sample red layer slopes are consistent or not, if the sample red layer slopes correspond to the same observation station point, the maximum single-day rainfall of the sample red layer slopes is the same, the maximum single-day rainfall interval can be a type-divided variable, and the data type of the maximum single-day rainfall interval is not specifically limited in the application.

For the continuous variable, the statistic feature value can be calculated and collected by checking whether the data has an outlier or a missing value through a box and whisker diagram, a histogram and the like. For the branch type variable, whether the data has a missing value or not can be checked mainly by checking a data set, and then the frequency of each evaluation index category is counted to find the characteristic which is greatly different from other types from the frequency, and the characteristic is determined as an outlier.

And carrying out double-variable analysis on the data, analyzing the joint distribution relation between the two variables, and checking whether dirty data appear or not. When carrying out bivariate analysis on the data, the correlation between every two characteristic data can be obtained in various ways, and the application does not specifically limit the correlation. For example, the correlation between features can be calculated using a function of python, using mainly pearson's coefficient, resulting in a value of-1 to 1. -1 indicates a completely negative correlation between features, 1 indicates a completely positive correlation, and 0 indicates no correlation. For the features with strong correlation, the performance of the model is reduced, and dimension reduction processing can be performed in subsequent feature engineering. For example, the result of performing on the sample data set X is shown in table 1.

TABLE 1 correlation of two-variable evaluation indices

Through bivariate analysis, the joint distribution relation between two variables can be obtained. For example, after bivariate analysis is performed on the sample data set X (as shown in table 1), it can be seen that the excavation progression and the excavation height have strong positive correlation, that is, the excavation progression increases with the increase of the excavation height; secondly, the excavation height and the slope height, the excavation progression and the slope height, and the excavation progression and the slope width are in weak positive correlation. The slope height, the slope width, the excavation height and the average excavation gradient have little relation.

Step S32: and preprocessing the sample data set according to the analysis result of the sample data set.

In this embodiment, after the sample data set is analyzed to obtain the analysis result, the sample data set may be subjected to corresponding preprocessing operation according to the corresponding analysis result. For example, the sample data set may be preprocessed according to the characteristics of the data itself, the analysis result obtained by the analysis, and different data formats and different requirements, so as to improve the accuracy of the algorithm model.

Step S33: training based on SVM algorithm according to the preprocessed sample data set, and establishing the rapid evaluation model.

In this embodiment, in the model training process, the preprocessed sample data set may be trained based on a supervised learning method according to the preprocessed sample data set, and a fast evaluation model of the red layer slope stability is established, so that a technician only needs to input evaluation index data of the red layer slope into the fast evaluation model, and the fast evaluation model may directly preprocess the data inside the model, and predict the stability grade of the red layer slope through the fast evaluation model.

In the embodiment, in the process of establishing the model, the sample data set is analyzed, the sample data set is preprocessed based on the analysis result, and then the rapid evaluation model is established according to the preprocessed sample data set, so that the accuracy of predicting the stability level of the model is improved.

With reference to the foregoing embodiments, in one implementation manner, the present application further provides a data preprocessing method, where the preprocessing at least includes: data cleansing, data transformation and data feature selection, in particular, the method may comprise:

step S41: and performing the data cleaning on the dirty data in the sample data set.

In this embodiment, data cleaning is performed on dirty data in the sample data set in the analysis result, so as to improve the quality of the sample data set. For example, the repeated data or invalid data in the dirty data can be removed, the missing value can be filled, the inconsistent value can be further analyzed and processed, the outlier can be analyzed for the reason of the outlier and correspondingly processed, and the reason of the outlier and the processing method are shown in table 2.

TABLE 2 reason for discrete value and processing method

Step S42: and performing the data conversion on continuous data in the data of the multiple types of evaluation indexes of the multiple groups of sample red layer slopes in the sample data set, wherein the data conversion comprises the following steps: and carrying out data scale adjustment and normalization processing.

In the embodiment, continuous data in the sample data set, such as the range of slope height is 20-89 m, the range of slope width is 19-200 m, the range of average excavation slope is 28-70 °, and the continuous variable distribution span is too large, so that data scale adjustment, standardization and normalization processing are required.

In general, each attribute of data is measured according to different modes, and then the scales of all the attributes of the data are unified by adjusting the scales of the data, so that great convenience is brought to training of an algorithm model for machine learning. This approach will typically normalize all attributes of the data and convert the data to a value between 0 and 1. And carrying out data scale adjustment on continuous data in the data of the multi-class evaluation indexes of the multi-group sample red layer slope in the sample data set, gathering the data to be close to 0, and setting the variance to be 1. The same data scale can improve the accuracy of distance-dependent algorithms. After the data range adjustment, the eigenvalues of all data points are limited to the specified range, for example, the partial execution results are shown in table 3.

TABLE 3 scaling of data characteristic values

In this embodiment, in order to optimize the convergence effect of the established model, after data scale adjustment is performed on the data, normalization processing needs to be performed on the data, the distance of each line of data is processed into 1, the data with the vector distance of 1 in linear algebra is also called "normalization element" processing, and the calculation formula is as follows:

（1）

in the formula (1), x is processed data,

for the ith data in the original data,

the minimum and maximum values of the type variable in the raw data, respectively. After normalization processing, the data performance can be improved, numerical value sorting which is the same as that of original data variables is also reserved, performance improvement of an algorithm which partially uses weight input or distance input is effective, and partial processing results are shown in table 4.

TABLE 4 normalization of data feature values

Step S43: and selecting data characteristics of the classified data in the data of the multi-class evaluation indexes of the multi-group sample red layer slope in the sample data set by using a One-Hot coding method, and quantizing the classified data.

In this embodiment, "side slope height", "side slope width", "average excavation gradient", "side slope excavation height", and "side slope excavation number" in the sample data set are continuous variables, and the data processing methods described above are all used to process such characteristics. However, in the present embodiment, it is necessary to quantify the classification type data in the data of the multiple types of evaluation indexes of the multiple sets of sample red layer slopes in the sample data set.

In this embodiment, the classification variables are subjected to data feature selection by the One-Hot coding method, and the classification data are quantized. The principle of One-Hot encoding is to represent a classification variable as One or more sub-features, which take the values 0 and 1. For a linear two-class formula, two values of 0 and 1 are meaningful, and a sub-feature can be introduced for each class to represent any number of classes. For example, possible values of lithological combination features include "sandstone-shale interbed", "hard rock-soft rock", "soft rock-hard rock", "soft rock", "hard rock", and "fourth series of deposits". To encode these 6 possible values, 6 sub-features can be created, respectively "lithology feature _ sandstone interbed", "lithology feature _ hard rock-soft rock", "lithology feature _ soft rock-hard rock", "lithology feature _ soft rock", "lithology feature _ hard rock", and "lithology feature _ fourth series of accumulations". If the lithological combination characteristic of one slope takes a certain value, the corresponding characteristic value is 1, and the other characteristics all take values of 0. Thus, for each data point, only one of the 6 sub-features takes the value 1. Example results of quantification are shown in table 5.

TABLE 5 data characteristic value One-Hot processing partial data display

In the embodiment, the sample data set is preprocessed in multiple modes, and then the rapid evaluation model is established according to the preprocessed sample data set, so that great convenience can be brought to model training, and the accuracy of the model for predicting the stability grade can be improved.

With reference to the foregoing embodiment, in an implementation manner, the present application further provides a data preprocessing method, where the preprocessing further includes: performing dimension reduction processing on the sample data set, specifically, the method may include:

step S51: evaluating the correlation degree between each evaluation index in the multiple types of evaluation indexes and the stability of the red layer slope; wherein the higher the correlation degree between the evaluation index and the stability of the red layer slope is, the higher the importance of the evaluation index is.

For data in a sample data set, the more the types of the data are, the larger the dimensionality is, and the excessive dimensionality can cause a 'dimensional disaster' of machine learning, that is, the time and the learning cost required by model building and running are greatly increased, and the performance is reduced, so that the dimension reduction processing needs to be performed on a high-dimensionality data set.

In this embodiment, the data types of the sample data set for training the rapid evaluation model include 10 types, namely, a slope height, a slope width, a lithology combination characteristic, a structural plane filling characteristic, a rock structure type, a slope excavation stage number, a slope excavation height, an average excavation slope, and a maximum single-day rainfall. In order to further improve the performance of the model and reduce unnecessary workload during data acquisition in actual work, dimension reduction processing is performed, and feature selection is required for which features should be simplified.

In this embodiment, the relevance between each evaluation index of the multiple types of evaluation indexes and the stability of the red layer slope needs to be evaluated, in one mode, the relevance between each evaluation index and the stability of the red layer slope can be evaluated through a chi-square test method, and the higher the score obtained by the chi-square test is, the more important the evaluation index is. That is, the higher the degree of association between the evaluation index and the red slope stability, the higher the importance of the evaluation index. For example, the chi-squared test results obtained for a certain sample data set are shown in fig. 2 for 10 evaluation indexes in the present embodiment.

Step S52: and selecting a plurality of evaluation indexes to be eliminated with the lowest relevance.

In this embodiment, a plurality of evaluation indexes to be rejected with the lowest correlation degree may be selected when obtaining the correlation degree between each evaluation index in the plurality of types of evaluation indexes and the stability of the red layer slope. It should be noted that the number of evaluation indexes to be rejected with the lowest relevance degree selected in this embodiment may be set according to actual requirements, for example, 2 evaluation indexes to be rejected with the lowest relevance degree may be selected, and whether the evaluation indexes are rejected or not is discussed. For example, as can be seen from fig. 2, the importance of the features (evaluation indexes) is, in order from large to small: the method comprises the following steps of slope structure type, maximum single-day rainfall, lithologic combination characteristics, side slope excavation height, side slope height, structural surface filling characteristics, rock mass structure type, average excavation slope, side slope width and side slope excavation progression. The width of the side slope and the excavation grade number of the side slope are two features with the lowest importance, and whether the side slope is removed or not will be discussed.

Step S53: and comparing the importance of the plurality of evaluation indexes to be rejected with the importance of other evaluation indexes, and determining whether the evaluation indexes to be rejected are rejected.

In this embodiment, the importance of the selected multiple evaluation indexes to be rejected is compared with the importance of other evaluation indexes, so as to determine whether to reject the evaluation indexes to be rejected. In this embodiment, the other evaluation indexes are the other evaluation indexes except the evaluation index to be rejected in the multiple types of evaluation indexes.

With reference to the foregoing embodiments, in one implementation manner, the present application further provides a data dimension reduction method, where the method may include:

step S61: and obtaining the correlation between the plurality of evaluation indexes to be eliminated and other evaluation indexes according to the analysis result.

In this embodiment, the correlations between the selected multiple evaluation indexes to be rejected and other evaluation indexes can be obtained according to the correlations between the evaluation indexes in the sample data set obtained by the bivariate analysis method in the analysis result.

Step S62: and determining candidate evaluation indexes representing the evaluation indexes to be rejected to participate in stability evaluation according to the correlation between the evaluation indexes to be rejected and the other evaluation indexes.

In this embodiment, according to the correlation between the evaluation index to be rejected and the other evaluation indexes, a candidate evaluation index having a strong correlation with the evaluation index to be rejected may be determined, and whether the candidate evaluation index can represent the evaluation index to be rejected to participate in the stability evaluation is discussed, so that the candidate evaluation index that can represent the evaluation index to be rejected to participate in the stability evaluation is determined.

Step S63: and determining to reject the evaluation index to be rejected under the condition that the importance of the candidate evaluation index is higher than that of the evaluation index to be rejected.

In the embodiment, after the candidate evaluation index which can represent the evaluation index to be rejected to participate in the stability evaluation is determined, the importance of the candidate evaluation index is compared with the importance of the evaluation index to be rejected, and if the importance of the candidate evaluation index is higher than that of the evaluation index to be rejected, the evaluation index to be rejected is determined to be rejected, so that the evaluation index to be rejected is rejected.

The following description is given by taking the selected slope excavation stage number as the evaluation index to be rejected by combining table 1 and fig. 2 as an example: as can be seen from table 1, there is a strong positive correlation between the side slope excavation number of steps and the side slope excavation height, that is, the side slope excavation number of steps increases with the increase of the side slope excavation height, which also indicates that the side slope excavation number of steps can be represented by the side slope excavation height to participate in the stability evaluation, and the importance of the side slope excavation height is much higher than that of the side slope excavation number of steps, so that the side slope excavation number of steps is determined to be rejected.

step S71: and determining the importance gap between the importance of the evaluation index to be eliminated and the other evaluation indexes.

In this embodiment, the importance difference between the importance of the evaluation index to be rejected and the importance of the other evaluation indexes may also be determined, that is, the difference between the importance of the evaluation index to be rejected and the importance of the other evaluation indexes is calculated, for example, as shown in fig. 2, the chi-square test score of each evaluation index is obtained by using a chi-square test method, and the difference between the chi-square test score of the evaluation index to be rejected and the chi-square test score of the other evaluation indexes is calculated.

Step S72: and comparing the data acquisition difficulty for acquiring the evaluation index to be eliminated with the data acquisition difficulty of other evaluation indexes.

In this embodiment, the data acquisition difficulty of the evaluation index to be rejected and the data acquisition difficulty of other evaluation indexes are respectively evaluated, and the acquired data acquisition difficulty of the evaluation index to be rejected is compared with the data acquisition difficulty of other evaluation indexes.

Step S73: and determining to retain the to-be-rejected index under the condition that the importance difference is smaller than a second preset threshold and the data acquisition difficulty of the to-be-rejected evaluation index is lower than that of the other evaluation indexes.

In this embodiment, a second preset threshold is preset, and the second preset threshold is set empirically, and when the importance difference is smaller than the second preset threshold and the data acquisition difficulty of the evaluation index to be rejected is lower than the data acquisition difficulty of other evaluation indexes, it is determined that the evaluation index to be rejected is not to be rejected, and the evaluation index to be rejected is retained.

The following description will be given by taking the selection of the slope width as the evaluation index to be rejected with reference to fig. 2 as an example: the importance ranking of the side slope width is the second last, meanwhile, the side slope width is 1.5 times of the importance score of the excavation series of the first last side slope, the side slope width is not separated from the importance scores of the average excavation slope of the third last and the lithological structure type characteristics of the fourth last, the side slope width can be used as one of the characteristics representing the side slope scale, the field data of the side slope width is easy to obtain as with other evaluation indexes, and therefore the side slope width characteristic is considered to be reserved.

In this embodiment, the evaluation index types in the sample data set can be processed according to two methods, wherein an appropriate method can be selected according to actual conditions to determine whether to reject the evaluation index to be rejected so as to perform dimension reduction processing on the sample data set, thereby further improving the performance of the model and reducing unnecessary workload during data acquisition in actual work. It should be noted that, in the present application, for the plurality of evaluation indexes to be rejected in steps S52-S53, the above two methods may be used for combination processing, or only one of the methods may be used for dimension reduction processing, which is not specifically limited in the embodiment of the present application.

In a preferred embodiment, when the dimension reduction processing is performed on the sample data set during the establishment of the fast evaluation model, the following fast acquisition of data of multiple types of evaluation indexes of the red layer slope to be evaluated includes:

step S81: and determining various evaluation indexes adopted after the dimension reduction processing is carried out on the sample data set.

In this embodiment, dimension reduction processing is further performed on the sample data set when the rapid evaluation model is established, and after the dimension reduction processing is performed, various adopted evaluation indexes are determined, where the various adopted evaluation indexes are various evaluation indexes left after the evaluation indexes to be eliminated are eliminated from the sample data set. For example, in this embodiment, the evaluation indexes to be rejected are the slope excavation levels, and then the adopted various evaluation indexes are: the method comprises the following steps of slope height of the side slope, slope width of the side slope, lithology combination characteristics, structural plane filling characteristics, rock mass structure type, slope excavation height, average excavation slope and maximum single-day rainfall.

Step S82: and rapidly acquiring data in the red layer slope to be evaluated according to the various adopted evaluation indexes.

In this embodiment, when data of multiple types of evaluation indexes of the red layer slope to be evaluated is rapidly acquired, the data of the red layer slope to be evaluated for the various types of evaluation indexes adopted by the red layer slope to be evaluated is rapidly acquired according to the various types of evaluation indexes adopted after the dimension reduction processing is performed when the model is established. For example, the method for rapidly obtaining the data of the slope height, the slope width, the lithology combination characteristic, the structural plane filling characteristic, the rock mass structure type, the slope body structure type, the slope excavation height, the average excavation slope and the maximum single-day rainfall in the red layer slope to be evaluated is the same as the method for rapidly obtaining the sample data set in the application, and therefore, redundant description is not repeated.

In the embodiment, the dimension reduction processing is performed on the sample data set, so that the time and the learning cost required by model establishment and operation are reduced, the performance of the model is improved, and unnecessary workload during data acquisition in actual work is reduced.

In an implementation manner, with reference to the above embodiment, the present application further provides a model building method, as shown in fig. 3. Fig. 3 is a flowchart illustrating a model building method according to an embodiment of the present application. The method can comprise the following steps:

step S91: and dividing the preprocessed sample data set into at least a first training sample set and a first testing sample set.

In this embodiment, after the sample data set is preprocessed, the sample data set may be divided into at least two parts, one part being a first training sample set for training an algorithm to establish a model; the other part is a first test sample set which is used for judging (evaluating and predicting) the trained model and comparing with an expert evaluation result (namely the actual stability grade of the red layer slope) to evaluate the performance and the accuracy of the model.

The division ratios adopted for different sample data set scales are different, and the division ratios of the sample data sets in this embodiment may be set according to actual requirements, for example, 70% of data may be randomly divided into a training set, and the remaining 30% of data may be divided into a test set. The present application does not specifically limit the division ratio of the sample data set.

Step S92: and training based on an SVM algorithm and a Gaussian kernel function with preset hyper-parameters according to the first training sample set to obtain a rapid evaluation initial model.

In this embodiment, after the first training sample set is obtained, training may be performed based on an SVM algorithm according to the first training sample set. The SVM algorithm model is realized by adopting an SVM function package in a Sciket-learn machine learning toolkit based on a Python language platform, and the GridSearchCV package in the Sciket-learn machine learning toolkit is also adopted for realizing the hyper-parameter search and cross validation. The SVM algorithm is used for finding a classification plane of data, and the idea of the SVM algorithm is that a hyperplane formed by certain support vectors is not usedThe sample points of the same category are divided, so that when the classification is carried out in the two-dimensional space, different categories can be separated by using a straight line, and the straight line can be used

Expressed by expressions, in multidimensional space, planes are needed to separate different classes of data, and the expressions are

Solving the function involves the inner product of the data points

However, this method can only solve the linear separable problem, and when a nonlinear inseparable problem like slope stability prediction is encountered, a data point with a lower dimension needs to be mapped to a field with a higher dimension and converted into a linear sample for solving, and this method applies a kernel function

The method (1).

The characteristics of different kernel functions are different from the application scenes, the kernel function is selected as an important factor influencing the quality of the SVM algorithm, the Gaussian kernel function has fewer parameters, the anti-interference characteristic aiming at the strong noise in sample data is realized, the local fitting effect is strong, the nonlinear sample data samples can be mapped to a high-dimensional space to be separable, and the kernel function is the kernel function with the widest application range.

In this embodiment, the problem of deformation stability of the red layer slope is a complex and nonlinear problem, many influencing factors influencing the stability of the red layer slope are provided, and based on the model calculation amount and complex consideration, a gaussian kernel function is selected as the kernel function of the rapid evaluation model. Preferably, the SVM packet relates to important hyper-parameters: the "C" and "gamma" are shown in Table 6.

Hyper-parameter	Brief introduction to the drawings
		C	C is a penalty factor, understood as the weight of the preference of adjusting two indices (size of separation, accuracy of classification) in the optimization direction, i.e. the tolerance to errors, the higher C is, the more intolerable the error is, the more overfitting is easy to occur, the smaller C is, the less overfitting is easy, the larger or smaller C is, and the generalization capability is poor.
gamma	gamma is a parameter of the gaussian function itself after the function is selected as the kernel function. Implicitly deciding after mapping the data to a new feature space The larger the gamma is, the fewer the support vectors are, and the smaller the gamma value is, the more the support vectors are. The number of support vectors affects the speed of training and prediction And (4) degree.

TABLE 6 introduction of important hyper-parameters in SVM bag

In this embodiment, a grid search method is used to perform parameter optimization, and C and gamma are selected as initial search limits (10)^-3，10³) The step size is equal power of 10, the step size is reduced near the optimal parameter of the search, and the search is continued, and the penalty factor C cannot be too large or too small, and usually cannot exceed the set search range. Finally, the optimal parameter combination established by the model is obtained by using a grid search method: c =10 and gamma = 100.

That is, in a preferred embodiment, the training is performed based on the SVM algorithm and the gaussian kernel function (C =10, gamma = 100) with the predetermined hyper-parameter according to the first training sample set, and the sample data points are projected into the sample feature space of the high dimension. Finding an optimal classification plane between each classification characteristic data and other characteristic data through an SVM algorithm, obtaining a support vector set representing each classification characteristic and VC (vitamin C) credibility corresponding to the support vector set, and finally generating a rapid evaluation initial model capable of evaluating the stability of each red layer slope.

Step S93: and judging the rapid evaluation initial model through the first test sample set to obtain the rapid evaluation model.

In this embodiment, after the fast evaluation initial model is obtained, the first test sample set is input into the fast evaluation initial model, the first test sample set is used as an input source of the trained fast evaluation initial model, and the feature information of the first test sample set is mapped into the feature space by the kernel function, so as to obtain a predicted stability level result. And judging the rapid evaluation initial model according to the predicted stability grade result, thereby obtaining the rapid evaluation model.

In addition, it should be noted that, in the rapid evaluation method provided by the present application, the sample data set may be preprocessed, and then the preprocessed sample data set may be divided into the test sample set and the training sample set; the method may also divide the test sample set and the training sample set into sample data sets, and then perform corresponding preprocessing on the data in the test sample set and the training sample set, which is not limited in the present application.

In the embodiment, in the model training process, the model is trained based on the number of training samples to obtain the initial model, and the initial model is further verified by the test sample set to obtain the rapid evaluation model with high accuracy, good evaluation effect and strong generalization capability.

With reference to the foregoing embodiments, in one implementation manner, the present application further provides a model building method, where the method may include:

step S101: and taking the first test sample set as the input of the rapid evaluation initial model, judging whether the actual stability grade of each sample red layer side slope marked in the first test sample set is consistent with the corresponding stability grade of each sample red layer side slope output by the rapid evaluation initial model, and determining a first number of consistent stability grades.

In this embodiment, when the fast evaluation initial model is determined according to the first test sample set, the first test sample set may be used as an input of the fast evaluation initial model, and whether the actual stability level of each red layer slope of the samples marked in the first test sample set is consistent with the stability level of each red layer slope of the corresponding samples output by the fast evaluation initial model is determined, so as to determine a first number of consistent stability levels. For example, the first test sample set contains data of multiple types of evaluation indexes of the sample red layer slope F, the actual stability grade of the sample red layer slope F identifier is "poor", and after the data of the multiple types of evaluation indexes of the sample red layer slope F are input into the rapid evaluation initial model, if the obtained output result is that the stability grade is "poor", the stability grades are consistent; if the obtained output result indicates that the stability grade is poor, the stability grade is inconsistent. In this embodiment, a first number of consistent stability levels needs to be determined; wherein, first quantity can be for the unanimous sample red layer side slope number of the stability grade of test sample concentration, also can be for the unanimous rate of accuracy of test sample concentration sample red layer side slope stability grade, and first quantity in this embodiment can set up according to actual demand, and this application does not do specific restriction to first quantity.

Step S102: when the first number is larger than or equal to a first preset threshold value, the characterization model is judged to be passed, and the fast evaluation initial model is determined as the fast evaluation model.

In this embodiment, a first preset threshold may be set in advance according to experience, where the property of the first preset threshold corresponds to the first number, and for example, when the first number is the number of sample red-bed slopes with consistent stability levels, the first preset threshold is the number of sample red-bed slopes with consistent stability levels that meet the model requirements; when the first quantity is the accuracy rate of consistency of the stability grade of the red-layer slopes of the samples, the first preset threshold value is the accuracy rate of consistency of the stability grade of the red-layer slopes of the samples, which meets the requirements of the model, of the minimum red-layer slopes of the samples.

When the first quantity is larger than or equal to a first preset threshold value, the characterization model is judged to be passed, and the trained rapid evaluation initial model meets the requirements of high accuracy, good evaluation effect and strong generalization capability, so that the rapid evaluation initial model is determined as the rapid evaluation model.

Step S103: when the first quantity is smaller than the first preset threshold value, the characterization model fails to be judged, the sample data set is divided into at least a second training sample set and a second testing sample set again, training is carried out based on an SVM algorithm and a Gaussian kernel function with preset hyper-parameters according to the second training sample set to obtain a second quick evaluation initial model, and the second quick evaluation initial model is judged through the second testing sample set to obtain the quick evaluation model.

In this embodiment, when the first number is smaller than the first preset threshold, the characterization model fails to pass judgment, and the fast evaluation initial model trained by the characterization at this time cannot meet the requirements of high accuracy, good evaluation effect, and strong generalization capability, so that the fast evaluation initial model needs to be overturned to reestablish the model.

In this embodiment, the step of obtaining the fast evaluation model by determining the second fast evaluation initial model through the second test sample set corresponds to the step of obtaining the fast evaluation model by determining the fast evaluation initial model through the first test sample set, and may be: dividing the sample data set into at least a second training sample set and a second testing sample set, training based on an SVM algorithm and a Gaussian kernel function with preset hyper-parameters according to the second training sample set to obtain a second quick evaluation initial model, and judging the second quick evaluation initial model through the second testing sample set to obtain the quick evaluation model. The steps are repeatedly executed until a rapid evaluation model which meets a first preset threshold value, is high in accuracy rate, good in evaluation effect and strong in generalization ability is obtained.

In the embodiment, the model is trained through the training sample set to obtain the initial model, the initial model is further verified through the testing sample set, if the condition is not met, the sample data set is randomly divided again, the initial model is established again, and the initial model is judged until the rapid evaluation model with high accuracy, good evaluation effect and strong generalization capability is obtained.

The embodiments in the present specification are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.

It should be noted that, for simplicity of description, the method embodiments are described as a series of acts or combination of acts, but those skilled in the art will recognize that the present invention is not limited by the illustrated order of acts, as some steps may occur in other orders or concurrently in accordance with the embodiments of the present invention. Further, those skilled in the art will appreciate that the embodiments described in the specification are presently preferred and that no particular act is required to implement the invention.

While preferred embodiments of the present application have been described, additional variations and modifications of these embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. Therefore, it is intended that the appended claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the true scope of the embodiments of the application.

Finally, it should also be noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or terminal that comprises the element.

The method for rapidly evaluating the stability of the road slope in the red zone is described in detail, a specific example is applied in the method for explaining the principle and the implementation mode of the method, and the description of the example is only used for helping to understand the method and the core idea of the method; meanwhile, for a person skilled in the art, according to the idea of the present application, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present application.

Claims

1. A method for quickly evaluating the stability of a road slope in a red layer area is characterized by comprising the following steps:

determining various types of evaluation indexes influencing the stability of the red layer slope; the multi-class evaluation indexes comprise: the method comprises the following steps of (1) slope height, slope width, lithology combination characteristics, structural surface filling characteristics, rock mass structure type, slope excavation series, slope excavation height, average excavation slope and maximum single-day rainfall;

2. The method of claim 1, wherein the fast acquiring of the sample data set comprises:

rapidly acquiring data of the slope width, the slope height, the slope excavation progression and the average excavation slope index of each red layer slope of the sample data set by using an unmanned aerial vehicle three-dimensional live-action modeling technology;

inquiring observation station data closest to each sample red layer side slope in the sample data set from a meteorological database to obtain the maximum single-day rainfall capacity of each sample red layer side slope in the sample data set;

and inquiring survey data to obtain data of lithologic combination characteristics, structural plane filling characteristics, rock mass structure types and slope structure type indexes of each red layer slope of the sample data set.

3. The method of claim 1, wherein training based on an SVM algorithm according to the sample data set to establish a fast evaluation model comprises:

analyzing the acquired sample data set, including: analyzing the sample data set through univariate analysis and bivariate analysis to obtain an analysis result of the sample data set; wherein the analysis result at least comprises one of the following: correlation among evaluation indexes in the sample data set and dirty data in the sample data set; the correlation among all evaluation indexes in the sample data set is obtained by the bivariate analysis method;

preprocessing the sample data set according to the analysis result of the sample data set;

training based on SVM algorithm according to the preprocessed sample data set, and establishing the rapid evaluation model.

4. The method of claim 3, wherein the pre-processing comprises: performing dimensionality reduction on the sample data set, specifically comprising:

evaluating the correlation degree between each evaluation index in the multiple types of evaluation indexes and the stability of the red layer slope; the higher the correlation degree between the evaluation index and the stability of the red layer slope is, the higher the importance of the evaluation index is;

selecting a plurality of evaluation indexes to be eliminated with the lowest relevance;

and comparing the importance of the plurality of evaluation indexes to be rejected with the importance of other evaluation indexes, and determining whether the evaluation indexes to be rejected are rejected.

5. The method according to claim 4, wherein the comparing the importance of the plurality of evaluation indexes to be removed with the importance of other evaluation indexes to determine whether to remove the evaluation indexes to be removed comprises:

obtaining the correlation between the plurality of evaluation indexes to be rejected and other evaluation indexes according to the analysis result;

determining candidate evaluation indexes representing the evaluation indexes to be rejected to participate in stability evaluation according to the correlation between the evaluation indexes to be rejected and the other evaluation indexes;

and determining to reject the evaluation index to be rejected under the condition that the importance of the candidate evaluation index is higher than that of the evaluation index to be rejected.

6. The method according to claim 4, wherein the comparing the importance of the plurality of evaluation indexes to be removed with the importance of other evaluation indexes to determine whether to remove the evaluation indexes to be removed comprises:

determining an importance gap between the importance of the evaluation index to be rejected and the importance of the other evaluation indexes;

comparing the data acquisition difficulty for acquiring the evaluation index to be rejected with the data acquisition difficulty of other evaluation indexes;

and determining to retain the to-be-rejected index under the condition that the importance difference is smaller than a second preset threshold and the data acquisition difficulty of the to-be-rejected evaluation index is lower than that of the other evaluation indexes.

7. The method according to any one of claims 4 to 6, wherein the fast obtaining of the data of multiple types of evaluation indexes of the red layer slope to be evaluated comprises:

determining various evaluation indexes adopted after the dimension reduction processing is carried out on the sample data set;

and rapidly acquiring data in the red layer slope to be evaluated according to the various adopted evaluation indexes.

8. The method according to any one of claims 3 to 6, wherein the training based on SVM algorithm according to the preprocessed sample data set, and establishing the fast evaluation model, comprises:

dividing the preprocessed sample data set into at least a first training sample set and a first testing sample set;

training based on an SVM algorithm and a Gaussian kernel function with preset hyper-parameters according to the first training sample set to obtain a rapid evaluation initial model;

and judging the rapid evaluation initial model through the first test sample set to obtain the rapid evaluation model.

9. The method of claim 8, wherein said deciding the fast evaluation initial model through the first set of test samples, resulting in the fast evaluation model, comprises:

taking the first test sample set as the input of the rapid evaluation initial model, judging whether the actual stability grade of each red layer slope of the samples marked in the first test sample set is consistent with the stability grade of each corresponding red layer slope of the samples output by the rapid evaluation initial model, and determining a first number of consistent stability grades;

when the first number is larger than or equal to a first preset threshold value, judging that the representation model passes, and determining the rapid evaluation initial model as the rapid evaluation model;

when the first quantity is smaller than the first preset threshold value, the characterization model fails to be judged, the sample data set is divided into at least a second training sample set and a second test sample set again, training is carried out based on an SVM algorithm and a Gaussian kernel function with preset hyper-parameters according to the second training sample set to obtain a second quick evaluation initial model, and the second quick evaluation initial model is judged through the second test sample set to obtain the quick evaluation model;

the step of obtaining the rapid evaluation model by judging the second rapid evaluation initial model through the second test sample set corresponds to the step of obtaining the rapid evaluation model by judging the rapid evaluation initial model through the first test sample set until obtaining the rapid evaluation model.

10. The method of claim 1, wherein the determining of the plurality of types of evaluation indicators that affect the red slope stability comprises:

analyzing and determining influence factors influencing the stability of the red layer slope, wherein the influence factors comprise: geometric, lithological and structural, and external triggering features;

determining various types of evaluation indexes representing the influence factors according to the influence factors;

wherein, according to the influence factors, determining multiple types of evaluation indexes representing the influence factors specifically comprises:

determining the slope height and the slope width as evaluation indexes representing the geometric characteristics based on the slope height and the slope width in the geometric characteristics and the susceptibility and the correlation degree of slope instability;

determining the lithologic combination characteristics, the structural surface filling characteristics, the rock mass structure type and the slope body structure type as evaluation indexes representing the lithologic and structural characteristics based on geological factors forming the red bed slope;

and determining the grade number of the slope excavation, the height of the slope excavation, the average excavation gradient and the maximum single-day rainfall as evaluation indexes representing the external triggering characteristics based on the triggering factors of red layer slope instability.