A highway filling area risk assessment and management method

Abstract

The present application belongs to the technical field of highway filling disease assessment and control, and provides a highway filling area risk assessment and control method, wherein the risk assessment method comprises the following steps: S1, establishing a database; the database contains data of several typical filling areas, and the data at least includes geometric characteristics, slope conditions and geomorphology hydrology of each filling area; S2, according to the database established in step S1, a statistical analysis method is used to determine main control factors related to filling area diseases, and part or all of the main control factors are selected as variable parameters to construct a risk assessment model; S3, using the risk assessment model constructed in step S2, the risk of the filling area to be evaluated is assessed, and the risk value of the filling area is output. The present application can assess the current risk level of the filling area, and provide a basis for the disease monitoring and control of the highway filling area.

Description

Technical Field

[0001] This invention belongs to the field of highway embankment disease assessment and control technology, specifically relating to a method for risk assessment and control of embankment areas on expressways. Background Technology

[0002] A fill area refers to a wide, flat area formed by the filling of a large amount of earth and stone due to road construction, terrain modification, or other needs. For example, when constructing highways in mountainous, hilly, and valley areas, it is a special artificial landform unit formed by filling and leveling adjacent valleys, gullies, or depressions to dispose of excavated or excess earth and stone generated during construction. Its formation is not due to proactive needs for terrain modification or route elevation design, but rather a passive accumulation of fill within partially or semi-enclosed terrain, constrained by engineering limitations such as large volumes of excavated material, insufficient storage space, and long transportation radii. Because of poor fill material properties, insufficient compaction, and the lack of pre-designed effective drainage measures, these fill areas are prone to water accumulation and groundwater level rise under extreme or continuous rainfall conditions. This shortens the seepage path to the opposite embankment, and water seepage to the opposite embankment can easily cause slope deformation or even instability, seriously threatening the safe operation of the highway.

[0003] From an engineering classification perspective, the fill area is essentially a type of artificial terrace for waste disposal, possessing the dual characteristics of "high fill" and "low-lying catchment area." Its foundation bearing layer is mostly soft soil or colluvial layer, making it difficult to control compaction uniformly, resulting in significant settlement differences. Furthermore, if the drainage system is poorly designed or maintained, it can easily create a "catchment pit effect," making this area a high-risk location for roadbed seepage and the stability of the opposite embankment slope.

[0004] As the service life of highways continues to increase, related defects in the fill areas are becoming increasingly apparent. However, a risk classification and grading method for highway fill areas has not yet been developed, making it difficult to conduct quantitative assessments and scientific classifications. Summary of the Invention

[0005] In order to overcome the above-mentioned shortcomings of the prior art, the purpose of this invention is to provide a risk assessment method for highway fill areas, so as to assess the current risk level of fill areas and provide a basis for the monitoring and management of defects in fill areas.

[0006] The technical solution adopted by this invention to solve its technical problem is: A risk assessment method for highway fill zones includes the following steps: S1. Establish a database; the database contains data on several typical fill areas, including at least the geometric features, slope conditions, and geomorphological and hydrological characteristics of each fill area. S2. Based on the database established in step S1, statistical analysis methods are used to determine the main controlling factors related to diseases in the filled area, and some or all of the main controlling factors are selected as variable parameters to construct a risk assessment model. S3. Using the risk assessment model constructed in step S2, perform a risk assessment on the fill area to be assessed and output the risk value of the fill area.

[0007] In a preferred embodiment of the present invention, in step S1, the data sources of the database include on-site investigation and verification data, design and survey data, and construction and treatment records; The on-site investigation and verification data include information on the topographic features, project scale, and distribution of defects in the filled area obtained through on-site surveys of typical road sections; The design and survey data, including the extraction of fill area, thickness, slope type, and terrain type, are used to establish engineering and terrain attributes; The construction and treatment archives include filling parameters, slope treatment and protection measures, construction period, and damage repair records.

[0008] In a preferred embodiment of the present invention, step S2, determining the controlling factors includes the following steps: A1. Based on the database established in step S1, conduct statistical analysis on the types of diseases in the filled area; A2. Based on the analysis results of step A1, establish a system of influencing factors for diseases in the filled-in area, analyze multiple factors, and determine the main controlling factors related to diseases in the filled-in area; among them, Step A1 includes: A101. Disease Type and Frequency Distribution Analysis: This step classifies the disease types in the filled area and statistically analyzes the distribution of each disease type.

[0009] A102. Risk Level Distribution Analysis: This step involves classifying the severity of damage in each fill area into risk levels and statistically analyzing the distribution of fill areas at each risk level. A103. Regional Comparative Analysis: This step involves statistically analyzing the regions where each leveled area is located, and combining this with the analysis of disease types to obtain the disease development patterns in the leveled areas. A104. Analysis of the relationship between disease and engineering features: This step involves statistical analysis and comparative study of engineering features to obtain the engineering features that have a significant impact on the disease in the fill area; the engineering features include fill thickness, fill area, opposite slope height, comprehensive slope ratio and original slope. A105. Analysis of Spatial and Environmental Characteristics of Diseases: This step involves statistically analyzing the combined effects of external characteristics on diseases in the filled area; these external characteristics include landform type, upstream and downstream location, and external water bodies. Step A2 includes: A201. Establish a system of influencing factors for diseases in the leveled area; based on the statistical analysis of steps A104 and A105, select the characteristics that have a greater impact on diseases in the leveled area as influencing factors, classify these influencing factors, and construct a system of influencing factors for diseases in the leveled area. The system of influencing factors for diseases in the fill-in area includes geometric features, slope conditions, and geomorphological hydrology; the geometric features include the following influencing factors: fill thickness and fill area; the slope conditions include the following influencing factors: opposite slope height, overall slope ratio, and original slope; the geomorphological hydrology includes the following influencing factors: landform type and upstream / downstream location. A202. Correlation Analysis: This step obtains the correlation between the influencing factors selected in step A201 and the diseases in the filled area. A203. Identification of controlling factors: Combining the correlation analysis in step A202, principal component analysis is introduced to comprehensively determine the influencing factors selected in step A201, and to identify the controlling factors related to the diseases in the filled area.

[0010] Preferably, step A2 further includes: A204. Multi-factor interaction analysis: Select different controlling factors to form combined factors and conduct interaction analysis; identify whether the disease risk in the fill area increases exponentially when different controlling factors are simultaneously in the unfavorable range. In step S2, a combination of factors that contribute to the additive increase of disease risk in the filled area is introduced as interaction terms into the risk assessment model.

[0011] In a preferred embodiment of the present invention, in step S2, the risk assessment model adopts a probabilistic model, and the construction process includes the following steps: B1. Select some or all of the main controlling factors as variable parameters, and introduce some or all of the combination factors with superadditive growth as interaction terms; B2. Logistic regression and conditional probability analysis are used to characterize the contribution relationship of each parameter to the disease probability P and determine the risk assessment model. B3. Validate and classify the risk assessment model determined in step B2.

[0012] Preferably, in step B1, the main controlling factors selected include fill thickness H, fill area A, comprehensive slope i, and landform type T; the introduced combination factors include H×i; Based on the selected primary control factors and combined factors, the risk assessment model constructed through step B2 is as follows:

[0013] Where P represents the high-risk probability, with a value range of [0,1]; β0 is the intercept term; β1, β2, β3, β4, and β5 are regression coefficients, the specific values ​​of which are obtained by fitting data from the database; continuous variables H, A, and i are standardized using Z-scores; categorical variable T is converted into a dummy variable. The unit of the continuous variable fill thickness H is meters (m), and the unit of the fill area A is meters (m²). 2 The unit of the overall slope ratio i is %.

[0014] Preferably, in steps B1 and B2, multiple risk assessment models are constructed by selecting different controlling factors, introducing different combination factors, or not introducing combination factors. In step B3, based on the data in the database, ROC curve analysis and confusion matrix are used to verify the multiple risk assessment models determined in step B2, and the discrimination ability of the multiple risk assessment models is compared. The risk assessment model with the strongest discrimination ability is selected as the final risk assessment model.

[0015] Preferably, in step B3, the ROC curve of the final risk assessment model is analyzed to determine the optimal classification threshold, and the predicted probability distribution map of the final risk assessment model is plotted. The optimal classification threshold is verified by combining the risk level of the disease severity in the filled area as divided in step A102.

[0016] The second objective of this invention is to provide a method for managing highway fill zones based on the aforementioned risk assessment method.

[0017] A method for managing fill zones along highways includes the following steps: (1) Establish a landform classification system for filled areas; The landform classification system for the filled area includes landform categories, typical landforms, dominant disease types, engineering characteristics, and control points. (2) Based on the risk assessment model finally determined above, establish risk grading standards; The risk classification criteria include probability range, risk level, feature description, and engineering recommendations; (3) Couple the landform classification system of the fill area in step (1) and the risk classification standard in step (2) to form a control decision matrix; (4) Based on the control decision matrix in step (3), control each fill area in the database.

[0018] Compared with the prior art, the beneficial effects of the present invention are: The method for risk assessment of highway fill areas of the present invention can assess the current risk level of fill areas, which is beneficial for providing a basis for the monitoring and control of defects in fill areas.

[0019] Furthermore, in conjunction with the preferred scheme of the highway fill area risk assessment method of the present invention, a data statistical analysis method is innovatively adopted. Based on the established database, the main control factors for model construction are obtained by analyzing the types of defects, risk levels, regional distribution, and influencing factors of the fill area, and finally a probabilistic model that can assess the risk of the fill area is obtained, which is conducive to providing a quantitative basis for refined risk management.

[0020] The present invention provides a risk management method for highway fill areas, establishing a "classification-grading-measures" management decision matrix for fill areas. This matrix transforms model output into engineering governance standards, directly guiding on-site management and enabling the engineering application of risk grading and treatment decisions. Furthermore, during the management process, risk assessment model parameters can be combined with patrol data to achieve dynamic correction and long-term risk monitoring, providing a quantitative basis for refined risk management of fill areas. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a distribution map of disease types.

[0023] Figure 2 This is a risk level distribution map.

[0024] Figure 3 A comparative diagram of disease types in the filled-in area.

[0025] Figure 4 The histogram of frequency distribution of fill thickness and the curve fitted with probability density are shown.

[0026] Figure 5 This is a graph showing the cumulative distribution function of the fill thickness.

[0027] Figure 6 The histogram of the area frequency distribution of the filled area and the fitted curve of the probability density are shown.

[0028] Figure 7 This is a graph showing the cumulative distribution function of the filled area.

[0029] Figure 8 This is a scatter plot showing the overall slope ratio, side slope height, and risk level.

[0030] Figure 9 This is a statistical chart showing the proportion of high-risk sections within the slope ratio and slope height zones.

[0031] Figure 10 This is a two-dimensional risk heat map of "fill thickness - comprehensive slope".

[0032] Figure 11 This is a two-dimensional risk heat map of "area of ​​filled area - landform type".

[0033] Figure 12 This is a comparison chart of the ROC curves for the three models. Detailed Implementation

[0034] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. Many specific details are set forth in the following description to provide a thorough understanding of the present invention; the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0035] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0036] Example 1 This embodiment discloses a risk assessment method for highway fill zones, including the following steps: S1. Establish a database to provide basic data support for subsequent statistical analysis of diseases and research on influencing factors. The database contains data on several typical fill areas, including at least the geometric characteristics, slope conditions, and geomorphological and hydrological data of each fill area.

[0037] S2. Based on the database established in step S1, statistical analysis methods are used to determine the main controlling factors related to diseases in the filled area, and some or all of the main controlling factors are selected as variable parameters to construct a risk assessment model.

[0038] The determination of the controlling factors can be divided into the following steps: A1. Based on the database established in step S1, conduct statistical analysis on the types of diseases in the filled area; A2. Based on the analysis results of step A1, establish a system of influencing factors of diseases in the filled area, analyze multiple factors, and determine the main controlling factors related to diseases in the filled area.

[0039] The construction of a risk assessment model can be divided into the following steps: B1. Select some or all of the main controlling factors as variable parameters, and introduce some or all of the combination factors with superadditive growth as interaction terms or do not introduce interaction terms. B2. Logistic regression and conditional probability analysis are used to characterize the contribution relationship of each parameter to the disease probability P and determine the risk assessment model. B3. Validate and classify the risk assessment model determined in step B2.

[0040] S3. Using the risk assessment model constructed in step S2, conduct a risk assessment on the fill area to be assessed, output the risk value of the fill area, quantify the probability of high risk occurrence under different engineering conditions, and thus provide support for operation and maintenance management and governance decisions.

[0041] This embodiment uses field survey data, survey and design documents, construction and maintenance archives, and related statistical ledgers from highway construction and operation projects in Guangdong Province as the data sources for the database. Through the integration of data from multiple channels, a comprehensive database covering typical fill areas throughout the province was established, providing basic data support for subsequent statistical analysis of defects and research on influencing factors.

[0042] Specifically, in step S1, data can be integrated from the following three aspects: on-site investigation and verification data, design and survey data, and construction and treatment archives. Specifically: the on-site investigation and verification data includes obtaining information on the topographic features, project scale, and disease distribution of the fill area through field surveys of typical road sections. The design and survey data includes extracting the fill area, thickness, slope type, and terrain category to establish project and terrain attributes. The construction and treatment archives include fill parameters, slope treatment and protection measures, construction period, and disease repair records.

[0043] The database was constructed following a unified standard: first, the original data was parsed and fields were extracted item by item; second, parameters such as area, thickness, and risk level were converted and standardized; then, a "highway-region" lookup table was used to categorize work sites in four major regions: northern Guangdong, eastern Guangdong, areas adjacent to or on the edge of the Pearl River Delta, and western Guangdong; finally, a comprehensive database covering multiple dimensions, including work site information, project scale, disease characteristics, and treatment measures, was formed. The data scale and composition are shown in the table below.

[0044]

[0045] In step S2, based on the database established in step S1, the statistical regularity of the types of diseases in the fill area is first analyzed. Through systematic statistical analysis of the characteristics of typical fill area diseases, it was found that fill area diseases not only include structural, seepage, and stability diseases occurring within the fill area itself, but also secondary diseases such as lateral or opposite slope slippage, seepage, and deformation indirectly induced by the formation of the fill area and changes in hydrogeological conditions. These two types of diseases have obvious spatial correlation and coupling, reflecting the dual role of the fill area in the regional hydrological and mechanical system. Through comprehensive statistical and comparative analysis of disease types, risk levels, and spatial distribution, the inherent regularity and evolution trend of fill area diseases can be revealed, providing data support for subsequent disease cause analysis and risk modeling. Specifically, the statistical regularity analysis of fill area disease types includes: A101. Disease Type and Frequency Distribution Analysis: This step classifies the disease types in the filled area and statistically analyzes the distribution of each disease type.

[0046] Based on the database, a total of 326 engineering samples were collected, containing 225 disease records. Statistical results show that the diseases in the fill areas in the database have significant complexity and coupling, and can be divided into two main types: one is the disease of the fill area itself, and the other is the disease induced by the fill area on the side or opposite slope.

[0047] The main defects in the filled areas are foundation softening, seepage accumulation, water retention, and uneven settlement, characterized primarily by structural deterioration and drainage system failure. Among these, water accumulation and poor drainage are the most common, accounting for about 40%, often manifesting as surface water accumulation, blocked drainage ditches, or missing interception facilities. In gully-enclosed filled areas, water retention accounts for about 20%, mostly seen in semi-enclosed valley areas in hilly regions. Seepage and mudslides caused by rising groundwater levels account for about 15%, closely related to low-lying terrain and high water content. Defects or damage to drainage ditches account for about 10%. Settlement and cracking of the fill body account for about 5%, although the proportion is low, it often leads to structural stress concentration and localized abnormal seepage.

[0048] Meanwhile, the hydrodynamics and seepage processes in the fill area often induce secondary diseases on adjacent slopes, accounting for approximately 40% of the total diseases. These diseases mainly include seepage at the slope toe, slope slippage, and deformation of the opposing slope. Especially in gully-type or low-lying fill areas, when the two slopes form a relatively closed water catchment system, the seepage channels under heavy rainfall conditions during the rainy season may form a continuous chain of failure: "water storage in the fill area—seepage on the opposing slope—slope instability." This process reflects the hydraulic-mechanical coupling relationship between seepage activity within the fill area and the stability of the external slope.

[0049] In summary, the disease system in the fill-in area of ​​the database exhibits a complex characteristic of "coexistence of internal deterioration and external response." Internal diseases within the fill-in area itself, such as seepage and settlement, are the root cause of system instability, while the responses of the opposite slopes triggered by seepage or water level changes further amplify the risk of disaster. These two factors are coupled in spatial structure and hydrodynamic processes, forming a "source-transmission-cause" chain evolution pattern of disasters. This characteristic reveals the multi-level impact mechanism of diseases in fill-in areas and has significant theoretical and engineering implications for subsequent risk model construction and the formulation of comprehensive management strategies.

[0050] Based on the statistical analysis of the above disease types, the following diagram was drawn: Figure 1 The pie chart showing the distribution of disease types in the filled area is shown.

[0051] A102. Risk Level Distribution Analysis: This step involves classifying the severity of damage in each fill area into risk levels and statistically analyzing the distribution of fill areas at each risk level.

[0052] Based on the data recorded in the database and the drawings, disease severity, and treatment information from the data sources, the fill areas are divided into: (1) areas with obvious disease defects, large fill scale, strong water catchment, or close contact with external water bodies; (2) areas with certain drainage defects, diseases, or groundwater activity, but which have not yet experienced large-scale damage; and (3) areas with well-developed drainage measures or relatively favorable conditions. Furthermore, the risk levels are simply divided into high risk, medium risk, and low risk, corresponding to the three types of areas mentioned above.

[0053] Based on the above risk level classification method, the risk level of defects in each fill area was marked. Statistics showed that approximately one-third of the work sites were classified as high-risk; medium-risk work sites accounted for about 45%; and low-risk work sites accounted for less than 25%. Simultaneously, a risk level distribution map was drawn based on the statistical data, as shown below. Figure 2 As shown.

[0054] A103. Regional Comparative Analysis: This step involves statistically analyzing the regions where each leveled area is located, and combining this with the analysis of disease types to obtain the disease development patterns in the leveled areas.

[0055] The hazards in fill-in areas vary significantly across different regions. This embodiment uses data from fill-in areas of highways in Guangdong Province. Therefore, taking Guangdong highways as an example, the northern mountainous region, due to its developed mountain valleys, suffers from poor drainage and groundwater seepage, resulting in the highest proportion of high-risk work sites. The eastern hilly region is primarily characterized by enclosed water accumulation, commonly found in small to medium-sized work sites enclosed by gullies. Fill-in areas near or on the edge of the Pearl River Delta are mostly located close to external water bodies, where external water seepage is a major concern. Data for the western mountainous region is relatively limited, but preliminary statistics show that drainage ditch defects and settlement problems are prominent. These statistical results indicate that the hazards in fill-in areas are not only related to the scale of the project but also exhibit a significant regional coupling relationship with the geomorphological environment and hydrological conditions.

[0056] Combining the disease type statistics from step A101 with the regions in the database, plot as follows: Figure 3 The diagram shows a comparison of disease types in the filled area.

[0057] In summary, the analysis reveals a clear evolutionary pattern in the development of soil defects in the filled areas of the database: the initial stage is characterized by poor surface drainage and shallow water accumulation. If not addressed promptly, rainwater will gradually infiltrate deeper layers, softening the soil and increasing hydraulic forces, potentially leading to seepage, mudslides, or even structural instability of the slope. This "from shallow to deep, gradually worsening" evolutionary process highlights the importance of early intervention; that is, when the defects are still in the surface stage, risks can be controlled by strengthening drainage measures.

[0058] A104. Analysis of the relationship between diseases and engineering features: This step involves statistical analysis and comparative study of engineering features to obtain the engineering features that have a significant impact on diseases in the fill area; the engineering features include fill thickness, fill area, opposite slope height, comprehensive slope ratio and original slope, etc.

[0059] 1) Analysis of fill thickness Statistical analysis was performed on the fill thickness data of complete records in the database. The thickness ranged from 1.5 to 15.0 m, with a mean of approximately 7.1 m, a median of 8.0 m, a standard deviation of 3.50 m, and a coefficient of variation of approximately 0.49, indicating strong dispersion in the fill thickness distribution. The histogram and distribution fitting results (e.g.) Figure 4 As shown in the figure, the fill thickness data generally shows a right-skewed distribution, with the lowest thickness section (3~6m) and the medium thickness section (6~10m) accounting for the highest proportion.

[0060] The lognormal distribution can fit this characteristic well, with a logarithmic mean μ≈1.88 and a standard deviation σ≈0.42. The corresponding fill thickness distribution is concentrated in the range of 7~9m, which is consistent with the common design range of high fill subgrade.

[0061] Further analysis using the cumulative distribution function (CDF) (e.g.) Figure 5 As shown in the figure, when the thickness is >10m, the cumulative proportion is less than 25%, indicating that the number of construction sites with large thickness is relatively limited, but they have a higher risk significance in structural stability and the occurrence of defects.

[0062] 2) Analysis of the area of ​​the fill zone Valid samples of fill area parameters were obtained from the database, ranging from 8 to 22450 m², with a mean of 3783 m², a median of 2100 m², a standard deviation as high as 4315 m², and a coefficient of variation of approximately 1.14. Compared with the fill thickness, the area distribution of the fill area is more discrete.

[0063] See Figure 6 The histogram shows that the area of ​​the fill-in zone exhibits a strong right-skewed distribution, with the vast majority of work sites having an area of ​​less than 5000 m², and only a small number exceeding 10000 m². The log-normal distribution fitting results indicate that the model can effectively characterize the right-skewed nature and long-tailed distribution of the area data. Its logarithmic mean μ≈7.7 and standard deviation σ≈1.4 correspond to areas mainly concentrated in the 1000~5000 m² range, consistent with the typical size distribution characteristics of fill-in zone work sites.

[0064] See Figure 7 Further analysis using the cumulative distribution function (CDF) revealed that areas > 5000 m² accounted for only about 20% of the sample, while extremely large areas (> 15000 m²) accounted for less than 5%. However, these areas are often the focus of attention in disease control and risk management. The log-normal model showed a smooth and good fit in the tails, with a coefficient of determination R² ≈ 0.986, indicating that the distribution accurately reflects the statistical characteristics and engineering significance of the area parameter.

[0065] 3) Analysis of the characteristics and defects of the opposite slope To identify the relationship between the geometric conditions of the opposite slope and the risks and defects of the filled area and its opposite slope, this embodiment uses the comprehensive slope ratio i (%) and the opposite slope height Hs (m) as independent variables to conduct two types of analysis: one is the comparison of scatter distribution and risk level (e.g., Figure 8 As shown); the second is the zoning statistics of "slope ratio × slope height" (e.g. Figure 9 As shown, i: <10%, 10~30%, >30%; Hs: <10m, 10~15m, >15m).

[0066] See Figure 8The risk increases nonlinearly with the synergistic enhancement of "slope ratio and slope height". When the comprehensive slope ratio reaches more than 30% and the height is greater than 10m, the cases of fill areas in the database are classified as medium to high risk. This shows that the degree of damage and risk of fill areas are positively correlated with slope ratio and slope height.

[0067] See Figure 9 High-risk samples were clearly clustered in the quadrant where i ≥ 30% and Hs ≥ 10m, while when i < 30% or Hs < 10m, they were mostly low-risk or medium-risk. Zonal statistics further quantified this pattern: when i < 10%, the proportion of high-risk sections in different slope height groups was close to 0%, indicating overall controllability; when 10% ≤ i < 30%, significant high risk only occurred in Hs = 10~15m (approximately 25%), with other slope heights generally not triggering high risk; when i ≥ 30%, the risk increased dramatically, with the highest proportion of high-risk sections in Hs = 10~15m (approximately 41%), while Hs > 15m remained at a relatively high level (approximately 18%), and Hs < 10m generally did not experience high risk. Therefore, i=30% can be identified as the "risk initiation threshold", and Hs=10~15m can be regarded as the most sensitive slope height window; when i≥30% and Hs>15m, although it is lower than the peak value, it is still in the high-risk range and should be managed as a high-risk segment.

[0068] The above analysis shows that the characteristics of the side slope and the degree of damage in the filled area and the whole area exhibit certain patterns. Therefore, the influence of side slope factors needs to be considered when modeling.

[0069] A105. Analysis of Spatial and Environmental Characteristics of Diseases: This step involves statistically analyzing the combined effects of external characteristics on diseases in the filled area; the external characteristics include landform type, upstream and downstream location, and external water bodies.

[0070] 1) Analysis of topographic and geomorphological types Filled areas can be mainly categorized into low-lying / gully types, semi-filled / semi-excavated types, and adjacent-water types. Based on database data, low-lying and gully types show the highest concentration of problems, accounting for 45% of the total, primarily characterized by water accumulation and seepage. Semi-filled / semi-excavated types often exhibit combined problems such as water accumulation at the slope toe and slope collapse. Adjacent-water types are affected by changes in external water levels, commonly exhibiting reverse seepage and mudslides. Therefore, the degree of topographic enclosure and the intensity of external water influence determine the basic spatial distribution pattern of these problems.

[0071] 2) Analysis of upstream and downstream locations The probability of structural damage in the fill areas located downstream or in depressions along the water flow direction was approximately 1.6 times that of the upstream samples. When the road runs perpendicular to the water flow direction, localized water accumulation is likely to occur during the rainy season, making it difficult to dissipate underground seepage pressure and thus accelerating structural deterioration.

[0072] 3) Analysis of external water bodies According to statistics from the "Geomorphological Type" field in the database, approximately 12% of the fill-in road sections belong to the "adjacent to water body type," mainly distributed along rivers, lakes, or near reservoirs. These sample points generally have a high rate of damage, with seepage, mudslides, and slope toe erosion being the primary types of damage. The causes are mainly related to the reversal of seepage direction due to changes in the external water level and the infiltration of the slope toe. During the flood season or typhoon season, when the external water level rises above the slope toe elevation, the seepage pressure increases and drainage channels are blocked, easily leading to softening and mudslide damage in the lower part of the fill slope. This indicates that dynamic changes in the external water environment have a significant impact on the structural safety of fill-in areas adjacent to water bodies.

[0073] Statistical analysis of steps A101-A105 indicates that geometric scale, slope stability, and catchment environment have a coupled amplifying effect on disease severity: when the fill thickness H > 10m or the fill area A > 5000m², the severity of disease increases significantly, and when both are large, the probability of high-level disease occurrence can reach 25-30%; the risk further increases under conditions where the overall slope ratio is steeper than 1:1.5 and the original slope is large; high-frequency diseases are prone to form in low-lying / gully terrain with high enclosure and located downstream / at the end of the catchment area. Therefore, the above statistical facts collectively point to three dominant pathways: "geometric scale—slope stability—catchment environment".

[0074] Based on the above statistical analysis results, the main controlling factors related to diseases in the leveled areas were identified, including: A201. Establish a system of influencing factors for diseases in the leveled area; based on the statistical analysis of steps A104 and A105, select the characteristics that have a greater impact on diseases in the leveled area as influencing factors, classify these influencing factors, and construct a system of influencing factors for diseases in the leveled area. The system of influencing factors for diseases in the filled area includes geometric characteristics, slope conditions, and geomorphological and hydrological factors. Among them: 1) Geometric characteristics include fill thickness and fill area. These indicators reflect the geometric scale and volume characteristics of the fill. The greater the thickness and the larger the area, the more stress accumulates and pore water pressure increases in the fill, making it prone to settlement and seepage-type defects.

[0075] 2) Slope conditions include the height of the opposite slope, the overall slope ratio, and the original slope gradient. These indicators reflect the slope characteristics between the filled area and the original topography. A larger overall slope ratio and original slope gradient will increase the gravity component and surface runoff velocity, causing stability problems such as slope toe erosion, landslides, and uneven drainage; the greater the slope elevation difference, the more uneven the stress on the slope and the more concentrated the strain, thus increasing the risk of disease.

[0076] 3) Geomorphological hydrology includes geomorphological type and upstream / downstream location. Geomorphological type represents the surface morphology and water catchment conditions. Valleys and gullies are prone to rainwater accumulation and poor drainage. Based on topographic enclosure, water catchment conditions, and the characteristics of external water influence, the filled areas are divided into five categories: low-lying waterlogged type, mountain gully water catchment type, semi-filled / semi-excavated type, adjacent water seepage type, and terrace / hill type. The first four categories account for 88% of the total sample, are highly representative, and can cover the main geomorphological units in the database. Upstream / downstream location reflects the confluence pattern of the water system along the route. Road sections located downstream or in the catchment area are more prone to seepage and water accumulation.

[0077] The established system of influencing factors and mechanisms of action for diseases in the filled-in area can be found in the table below:

[0078] A202. Correlation Analysis: This step obtains the correlation between the influencing factors selected in step A201 and the defects in the fill area. Specifically, this embodiment uses two non-parametric methods, Spearman's rank correlation method and Kendall's rank correlation coefficient, for correlation analysis. Spearman's method is suitable for correlation analysis of continuous variables, including fill thickness, fill area, overall slope, and opposite slope height. Kendall's method is suitable for correlation analysis of categorical variables, including landform type and upstream / downstream location. The analysis results are shown in the table below:

[0079] Comprehensive analysis shows that the formation of hazards in the filled areas of the database is jointly controlled by geometric scale effects (geometric characteristics), topographic slope effects (slope conditions), and water catchment concentration effects (geomorphological and hydrological factors). Specifically, fill thickness and filled area determine the volume and internal stress level of the fill; the overall slope and original slope dominate structural stability and drainage capacity; while geomorphological type and upstream / downstream location reflect external water catchment and drainage conditions. These three types of factors exhibit significant spatial coupling, jointly constituting a "geometric-slope-hydrological" ternary coupled disaster-causing mechanism. Furthermore, correlation results indicate that fill thickness, filled area, overall slope, geomorphological type, and upstream / downstream location are significantly positively correlated factors and are the preferred independent variables for subsequent model fitting. A203. Identification of controlling factors: Combining the correlation analysis in step A202, principal component analysis (PCA) is introduced to comprehensively determine the influencing factors selected in step A201 and identify the controlling factors related to the disease in the filled area.

[0080] The results of principal component analysis can be found in the table below:

[0081] According to the principal component analysis results in the table above, the cumulative variance contribution rate of the first two principal components reaches 74.2%, with the first principal component contributing 52.6% and the second principal component contributing 21.6%. The first two principal components can comprehensively reflect the changes in the overall characteristics of the sample, and therefore can be used as the main representatives of the disease control factors in the filled area.

[0082] The following principal component loading matrix table reflects the contribution of each variable to the principal components.

[0083]

[0084] The first principal component (PC1) is dominated by fill thickness and fill area, reflecting the "geometric scale effect"; the second principal component (PC2) is dominated by landform type and upstream / downstream location, reflecting the "catchment environment effect". Both the overall slope and original slope have moderate loads on PC1 and PC2, indicating their transitional and regulatory role between geometry and hydrology. According to the principal component load matrix, thickness (0.82) and area (0.79) have the highest load values, indicating that the larger the geometric volume, the more significant the internal stress concentration and seepage path extension effects, which are the internal driving forces for disease development. Landform type (0.73) and upstream / downstream location (0.69) have the highest loads in the second principal component, indicating that the risk of disease increases significantly in catchment areas and valleys due to external water flow convergence and drainage obstruction. The moderate loads (approximately 0.3~0.4) on slope and original slope reflect their minor but not negligible role in stability control.

[0085] In summary, the main controlling factors of diseases in filled areas can be categorized into three types: 1) Geometric characteristics: The thickness of the fill and the area of ​​the fill zone control the internal stress distribution and seepage accumulation, and are the dominant factors in the formation of the disease; 2) Slope conditions: The combined slope ratio and original slope affect slope stability and drainage path, and are regulating factors for structural stability; 3) Geomorphology and hydrology: The geomorphological type and upstream and downstream location determine the external hydrodynamics and seepage intensity, which are the main drivers of disease expansion and amplification.

[0086] The correlation analysis in step A202 and the results of the identification of the main controlling factors in step A203 corroborate each other, indicating that the risk of disease in the filled area is jointly controlled by three mechanisms: geometry, slope, and water catchment. Among these, the geometric scale effect reflected by the fill thickness and the area of ​​the filled area, and the water catchment environment effect reflected by the landform type, are the most significant. The main controlling factors identified through the above statistical analysis provide quantitative support and core parameter basis for the construction of the risk assessment model.

[0087] A204. Multi-factor interaction analysis: Different controlling factors are selected to form combined factors, and interaction analysis is carried out. This embodiment selects two sets of the most engineering-significant combined factors for interaction analysis: the first combination formed by mechanical coupling and the second combination formed by hydrological coupling. First combination: Fill thickness H × Overall slope ratio i; Second combination: Area A × Landform type T.

[0088] The analysis identifies whether the disease risk in the filled area exhibits additive growth when different controlling factors are simultaneously in unfavorable ranges. In subsequent step S2, the risk assessment model considers incorporating some or all of the combined factors that contribute to the additive growth of disease risk in the filled area as interaction terms. Specifically: 1) Analyze the first combination. Using fill thickness H (m) and overall slope i (%) as independent variables, a two-dimensional partitioning matrix of "fill thickness - overall slope" was constructed. The results are shown in the table below. Figure 10 .

[0089]

[0090] As shown in the table above, the interaction between fill thickness and overall slope ratio has a significant impact on the risk of embankment erosion. When the fill thickness of the embankment slope increases from thin to a certain thickness, combined with a steeper slope, morphologically, overall sliding is more likely to occur at the fill-cut contact surface. When the thickness is too thick, the maximum shear stress surface may not appear near the weak fill-cut contact surface, potentially reducing the proportion of high-risk areas. Overall, the risk increases significantly with the combined increase of fill thickness and overall slope ratio, indicating a significant mechanical coupling effect between geometric volume and slope steepness. When the fill thickness exceeds 6m and the overall slope ratio is greater than 30%, the proportion of high-risk areas reaches approximately 55%, significantly higher than the overall average, indicating that under thick fill and steep slope conditions, foundation stress concentration and insufficient overall stability are the main contributing factors. In the combination range of low thickness (<6m) and medium slope ratio (10~30%), the risk level is in a transitional stage, with the proportion of high-risk areas being only 6%. The risk is lowest in the low slope ratio range (<10%), with virtually no risk observed, reflecting the characteristics of small geometric volume and structural stability.

[0091] See Figure 10 The results showed that the proportion of high-risk areas increased significantly with the simultaneous increase of fill thickness and overall slope. When the fill thickness was greater than 6m and the overall slope was greater than 30%, the proportion of high-risk areas exceeded 50%, showing a typical mechanical amplification effect.

[0092] Overall, the interaction effect between fill thickness and overall slope ratio exhibits a clear superadditive characteristic: when both indicators are simultaneously in the high value range, the probability of disease occurrence far exceeds the single-factor superposition effect, verifying that the mechanical coupling of "fill thickness-overall slope ratio" is one of the key control mechanisms for the formation of diseases in filled areas. Therefore, the first combination can be considered as an interaction term in the risk assessment model.

[0093] 2) Analyze the second combination. The filled area was divided into three groups: ≤2000m², 2000~5000m², and >5000m². A two-dimensional partition matrix was constructed by cross-referencing these groups with the five landform types. The results are shown in the table below. Figure 11 .

[0094]

[0095] The two-dimensional area-topography zoning matrix reveals significant differences in disease risk across different landform types within filled areas. Overall, the proportion of high-risk diseases increases with the size of the filled area. In low-lying, semi-filled / semi-cut, and gully landforms, the combined high-risk proportion reaches 75% when the area exceeds 2000 m², 36% higher than the proportion for areas ≤2000 m², and significantly higher than the overall average. This indicates that under conditions of concentrated runoff and closed topography, the adverse effects of surface runoff and groundwater activity on structural stability are amplified. Water-adjacent landforms also exhibit high risks in small areas (≤2000 m²), suggesting that lateral water infiltration and bank slope weakening may be the main pathogenic mechanisms. In contrast, semi-filled / semi-cut and terrace / hilly landforms have lower overall risks. The main risk is significantly influenced by differential settlement at the cut-fill interface, which is independent of the filled area, and large filled areas are rarely formed in these landform types.

[0096] In summary, a clear hydrological coupling effect exists between the area of ​​the fill zone and the topography: when both indicators are simultaneously in the unfavorable range, the risk increases additively, verifying the dual control effect of "geometric volume-catchment environment" on the evolution of diseases in the fill zone. Therefore, a second combination could be considered as an interaction term in the risk assessment model.

[0097] Based on the statistical analysis results of steps A1-A2, the main controlling factors related to the defects in the filled area are identified as follows: fill thickness H, filled area A, overall slope i, original slope S, landform type T, and upstream / downstream location W. When constructing the risk assessment model, some or all of these main controlling factors can be selected. Furthermore, interaction terms consisting of combined factors can be introduced when constructing the risk assessment model, including: "fill thickness H × overall slope i" and "fill area A × landform type T".

[0098] In step S2, after determining the controlling factors, a risk assessment model is then constructed. This embodiment uses a probabilistic risk assessment model. Specifically, multiple versions of the risk assessment model can be constructed by selecting different controlling factors, introducing different combinations of factors, or not introducing any combinations of factors. Based on data from the database, ROC curve analysis and confusion matrices are used to validate these multiple versions of the risk assessment model. The discriminative power of the multiple versions of the risk assessment model is then compared, and the risk assessment model with the strongest discriminative power is selected as the final risk assessment model.

[0099] For example, this embodiment sets up three risk assessment models, as detailed in the table below.

[0100]

[0101] Where P represents the probability that a sample enters the high-risk category, with a value ranging from [0,1]. The dependent variable is the risk level (high = 1; medium-low = 0), and the independent variables are the standardized values ​​of each parameter. In terms of method selection, logistic regression offers advantages in significance testing and interpretability, while conditional probability models can intuitively reflect the combined effects of multiple factors. Considering both data scale and interpretability, this embodiment uses logistic regression as the core model and conditional probability as a supplementary verification.

[0102] Parameters were estimated using the maximum likelihood method, and significance was determined by the Wald test and p-value. The test results for the three versions of the model are shown in the table below.

[0103]

[0104] The results show that Model 1 can reflect the basic geometric control effect, but ignores the topographic and mechanical effects; Model 2 adds slope and landform, which improves the goodness of fit, but still cannot characterize the interaction between thickness and slope; Model 3 introduces the interaction term H×i on the basis of the enhanced model, which makes the prediction accuracy the highest (AUC=0.85), and the results are more consistent with the physical mechanism of the high risk of "thick fill in flat areas and steep slopes" in the field. Therefore, Model 3 is selected as the final risk assessment model in this embodiment.

[0105] That is, in step B1, the selected main control factors include fill thickness H, fill area A, overall slope i, and landform type T; and the introduced combination factor H×i is used as an interaction term. The final risk assessment model is as follows:

[0106] Where P is the high-risk probability, ranging from [0,1]; β0 is the intercept term; β1, β2, β3, β4, and β5 are regression coefficients, the specific values ​​of which are obtained by fitting data from the database; continuous variables H(m) and A(m) are also included.2 ), i (%) are standardized using Z-score; categorical variable T is converted into a dummy variable.

[0107] Furthermore, by fitting the data in the database, the parameter estimation results can be obtained as follows: β0=-1.392963, β1=+0.333457, β2=-0.079936, β3=+0.267971, β4=+0.2546, β5=+4.121596.

[0108] Therefore, the final risk assessment model can be expressed as:

[0109] Based on the established database, the variables were calculated using the sample mean as the benchmark condition. The results show that: When H > 10m and i > 30%, P ≈ 0.70; When H < 6m and i < 10%, P ≈ 0.12; It can be seen that the thickness of the fill and the overall slope jointly determine the main trend of the probability of disease occurrence, and the interaction term β5=4.12 indicates that there is a significant nonlinear amplification effect.

[0110] Meanwhile, if the landform type is "near water" or "gully," then under the same fill thickness and overall slope ratio, P will increase by approximately 0.10 to 0.15, meaning the risk level can change from medium to high. This indicates that landform type has a significant amplification effect in thick-fill, steep-slope environments.

[0111] Further, in step B3, the ROC curve of the final risk assessment model is analyzed to determine the optimal classification threshold. Simultaneously, the predicted probability distribution map of the final risk assessment model is plotted. Combined with the risk levels of disease severity in the filled area as defined in step A102, the optimal classification threshold is verified. Specifically, in this embodiment, the ROC curves of the three versions mentioned above are plotted and compared again, such as... Figure 12 As shown, the AUC values ​​of the three models are 0.79, 0.82 and 0.85 respectively. The curves gradually approach the upper left, and model 3 has the strongest discriminative ability, indicating that introducing the interaction term H×i can significantly improve the model performance.

[0112] Furthermore, ROC curve analysis was used to determine the optimal classification threshold for Model 3 (i.e., the final risk assessment model). It was found that when P=0.6, the sensitivity was 0.81 and the specificity was 0.77, achieving a relatively good balance. Based on this threshold, the risks in the filled-in area were divided into three categories, as shown in the table below.

[0113]

[0114] In actual engineering, the fill thickness H and slope height have a high correlation (generally 0.8 to 1.2 times). However, to maintain the consistency of the model input variables, this embodiment uses the fill thickness H as the classification basis. If the risk zoning is reclassified based on the slope height, the results are basically consistent.

[0115] Furthermore, based on the final risk assessment model, combined with typical landform types and engineering examples, the evolutionary characteristics and disaster-causing mechanisms of hazards in filled areas can be analyzed from both physical processes and measured data perspectives. Through internal validation and case comparison, the engineering interpretability and applicability of the final risk assessment model's prediction results can be further explored, providing a basis for subsequent prevention and control design.

[0116] Example 2 This embodiment provides a method for managing fill-in areas along highways, which includes the following steps: (1) Establish a landform classification system for filled areas.

[0117] Based on steps A1 and A2 in Example 1, and combining geomorphic attributes such as topographic enclosure, catchment conditions, and proximity to water, the filled areas in the database are divided into five typical types, establishing a classification system. This classification system comprehensively considers three major factors: causal characteristics, geometric scale, and hydrological environment, providing a geomorphic basis for risk classification and differentiated management. The geomorphic classification system for filled areas is shown in the table below:

[0118] (2) Based on the risk assessment model finally determined in Example 1, a risk grading standard is established. The risk probability P output by the model is used as a quantitative indicator to classify the fill area, and P=0.60 is selected as the threshold. The risk grading standard includes probability interval, risk level, feature description, and engineering recommendations, as detailed in the table below;

[0119] (3) Couple the landform classification system of the fill area in step (1) with the risk classification standard in step (2) to form a control decision matrix, as detailed in the table below:

[0120] The aforementioned control and decision matrix achieves a closed-loop connection between "model calculation - risk level - governance measures", enabling the classification system and quantitative model to form a unified management logic.

[0121] (4) Based on the control decision matrix in step (3), control each fill area in the database.

[0122] The implementation path can be referenced as follows: (4.1) Input layer: Collects basic parameters such as H, A, i, T; (4.2) Calculation layer: Calculate the risk probability P using the risk assessment model; (4.3) Tier: Risk is determined according to the risk grading standard; (4.4) Decision-making level: Match governance strategies according to control decision-making rules; (4.5) Feedback layer: Combine the on-site monitoring results to correct parameters and realize the dynamic updating of the risk assessment model.

[0123] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Therefore, any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A risk assessment method for highway fill areas, characterized in that, Includes the following steps: S1. Establish a database; the database contains data on several typical fill areas, including at least the geometric features, slope conditions, and geomorphological and hydrological characteristics of each fill area. S2. Based on the database established in step S1, statistical analysis methods are used to determine the main controlling factors related to diseases in the filled area, and some or all of the main controlling factors are selected as variable parameters to construct a risk assessment model. S3. Using the risk assessment model constructed in step S2, perform a risk assessment on the fill area to be assessed and output the risk value of the fill area.

2. The risk assessment method for highway fill areas according to claim 1, characterized in that, In step S1, the data sources for the database include on-site investigation and verification data, design and survey data, and construction and treatment records; The on-site investigation and verification data include information on the topographic features, project scale, and distribution of defects in the filled area obtained through on-site surveys of typical road sections; The design and survey data, including the extraction of fill area, thickness, slope type, and terrain type, are used to establish engineering and terrain attributes; The construction and treatment archives include filling parameters, slope treatment and protection measures, construction period, and damage repair records.

3. The risk assessment method for highway fill areas according to claim 1, characterized in that, In step S2, determining the controlling factors includes the following steps: A1. Based on the database established in step S1, conduct statistical analysis on the types of diseases in the filled area; A2. Based on the analysis results of step A1, establish a system of influencing factors for diseases in the filled-in area, analyze multiple factors, and determine the main controlling factors related to diseases in the filled-in area; among them, Step A1 includes: A101. Disease type and frequency distribution analysis; classify the disease types in the filled area and statistically analyze the distribution of each disease type; A102. Risk Level Distribution Analysis: The severity of damage in each fill-in area is classified into risk levels, and the distribution of fill-in areas at each risk level is statistically analyzed. A103. Regional comparative analysis: Statistics on the areas where each fill-in area is located, combined with the analysis of disease types, to obtain the disease development patterns in the fill-in areas; A104. Analysis of the relationship between disease and engineering characteristics; statistical analysis and comparative study of engineering characteristics to obtain engineering characteristics that have a significant impact on disease in the fill area; the engineering characteristics include fill thickness, fill area, opposite slope height, comprehensive slope ratio and original slope. A105. Analysis of Spatial and Environmental Characteristics of Diseases; Statistical analysis of the comprehensive effects of external characteristics on diseases in the filled area; The external characteristics include landform type, upstream and downstream location, and external water bodies; Step A2 includes: A201. Establish a system of influencing factors for diseases in the leveled area; based on the statistical analysis of steps A104 and A105, select the characteristics that have a greater impact on diseases in the leveled area as influencing factors, classify these influencing factors, and construct a system of influencing factors for diseases in the leveled area. The system of influencing factors for diseases in the fill-in area includes geometric features, slope conditions, and geomorphological hydrology; the geometric features include the following influencing factors: fill thickness and fill area; the slope conditions include the following influencing factors: opposite slope height, overall slope ratio, and original slope; the geomorphological hydrology includes the following influencing factors: landform type and upstream / downstream location. A202. Correlation analysis; Obtain the correlation between the influencing factors selected in step A201 and the diseases in the filled area; A203. Identification of controlling factors: Combining the correlation analysis in step A202, principal component analysis is introduced to comprehensively determine the influencing factors selected in step A201, and to identify the controlling factors related to the diseases in the filled area.

4. The method for risk assessment of highway fill areas according to claim 3, characterized in that, Step A2 also includes: A204. Multi-factor interaction analysis: Select different controlling factors to form combined factors and conduct interaction analysis; identify whether the disease risk in the fill area increases exponentially when different controlling factors are simultaneously in the unfavorable range. In step S2, a combination of factors that contribute to the additive increase of disease risk in the filled area is introduced as interaction terms into the risk assessment model.

5. The method for risk assessment of highway fill areas according to claim 4, characterized in that, In step S2, the risk assessment model adopts a probabilistic model, and the construction process includes the following steps: B1. Select some or all of the main controlling factors as variable parameters, and introduce some or all of the combination factors with superadditive growth as interaction terms; B2. Logistic regression and conditional probability analysis are used to characterize the contribution relationship of each parameter to the disease probability P and determine the risk assessment model. B3. Validate and classify the risk assessment model determined in step B2.

6. The risk assessment method for highway fill areas according to claim 5, characterized in that, In step B1, the main controlling factors selected include fill thickness H, fill area A, overall slope i, and landform type T; the combined factors introduced include H×i; Based on the selected primary control factors and combined factors, the risk assessment model constructed through step B2 is as follows: Where P is the high-risk probability, with a value range of [0,1]; β0 is the intercept term; β1, β2, β3, β4, and β5 are regression coefficients, and the specific values ​​of β0, β1, β2, β3, β4, and β5 are obtained by fitting data from the database; continuous variables H, A, and i are standardized using Z-score; and categorical variable T is converted into a dummy variable.

7. The method for risk assessment of highway fill areas according to claim 6, characterized in that, In steps B1 and B2, multiple risk assessment models are constructed by selecting different controlling factors, introducing different combination factors, or not introducing combination factors. In step B3, based on the data in the database, ROC curve analysis and confusion matrix are used to verify the multiple risk assessment models determined in step B2, and the discrimination ability of the multiple risk assessment models is compared. The risk assessment model with the strongest discrimination ability is selected as the final risk assessment model.

8. The method for risk assessment of highway fill areas according to claim 7, characterized in that, In step B3, the ROC curve of the final risk assessment model is analyzed to determine the optimal classification threshold. At the same time, the predicted probability distribution map of the final risk assessment model is plotted. Combined with the risk level of the disease severity in the filled area as divided in step A102, the optimal classification threshold is verified.

9. A method for managing fill areas along highways, characterized in that, Includes the following steps: (1) Establish a landform classification system for filled areas; The landform classification system for the filled area includes landform categories, typical landforms, dominant disease types, engineering characteristics, and control points. (2) Based on the risk assessment model determined in the risk assessment method for highway fill areas as described in any one of claims 1-8, establish a risk classification standard; The risk classification criteria include probability range, risk level, feature description, and engineering recommendations; (3) Couple the landform classification system of the fill area in step (1) and the risk classification standard in step (2) to form a control decision matrix; (4) Based on the control decision matrix in step (3), control each fill area in the database.

Images