A method for predicting strength of a slip zone soil based on microstructure parameters of CT scanning

Abstract

The present application relates to the field of landslide soil strength prediction, and more particularly to a landslide soil strength prediction method based on CT scanning microstructure parameters. The content includes: CT scanning of landslide soil samples, obtaining CT images of landslide soil, and performing gray scale normalization processing; after gray scale normalization processing, a direction-scale joint structure response operator is introduced to construct a structure density field; a continuous scale disturbance calculation mechanism is introduced, the CT image of the landslide soil is divided into multi-scale voxel blocks according to the spatial scale, and based on the structure density field, the normalized structure disturbance strength index under each spatial scale is generated; based on the normalized structure disturbance strength index under each spatial scale, the scale response parameter is generated; a structure response prediction model is introduced, and based on the structure density field and the scale response parameter, the shear strength of the landslide soil is predicted. The problems of the existing landslide soil strength prediction method, such as lack of structure resolution capability, low prediction accuracy and difficulty in adapting to multi-scale structure characteristics, are solved.

Description

Technical Field

[0001] This invention relates to the field of slip zone soil strength prediction, and in particular to a method for predicting slip zone soil strength based on CT scan microstructural parameters. Background Technology

[0002] Slip zone soil is the most critical controlling material in a landslide, and its shear strength directly determines the stability and evolution trend of the landslide. Currently, obtaining the strength of slip zone soil in engineering mainly relies on indoor direct shear or triaxial tests, or on-site measurements using in-situ testing equipment. These methods are destructive and disturbing, and in practical applications, they suffer from many limitations, such as cumbersome operation, long data cycles, unstable test environments, and inability to reflect in-situ structures. Furthermore, due to the significant microscopic heterogeneity, structural directionality, and scale dependence of slip zone soil, traditional homogeneous material assumptions and strength models cannot accurately reflect its actual mechanical behavior. Although some studies have attempted to analyze microscopic pores and cracks using CT images, these are mostly limited to qualitative descriptions and have failed to establish a quantitative mapping mechanism from microscopic structural parameters to macroscopic strength indicators. Therefore, there is an urgent need for a novel strength prediction method for the multi-scale complex structure of slip zone soil that integrates CT image characterization and interpretable mathematical models to overcome the limitations of traditional tests in rapid landslide disaster assessment and real-time early warning. Summary of the Invention

[0003] This invention provides a method for predicting the strength of slip zone soil based on CT scan microstructure parameters, in order to solve the problems of existing slip zone soil strength prediction methods lacking structural resolution, having low prediction accuracy, and being difficult to adapt to multi-scale structural features.

[0004] The present invention provides a method for predicting the strength of slip zone soil based on CT scan microstructural parameters, specifically including the following technical solutions:

[0005] A method for predicting the strength of slip zone soil based on CT scan microstructural parameters includes the following steps:

[0006] S1. Perform CT scans on the slip zone soil samples to obtain CT images of the slip zone soil, and perform grayscale normalization processing; after grayscale normalization processing, introduce the direction-scale joint structural response operator to construct the structural density field.

[0007] S2. A continuous-scale perturbation calculation mechanism is introduced, and the CT image of the slip zone soil is divided into multi-scale voxel blocks according to spatial scale. Based on the structural density field, a normalized structural perturbation intensity index is generated at each spatial scale. Based on the normalized structural perturbation intensity index at each spatial scale, a scale response parameter is generated. A structural response prediction model is introduced, and the shear strength of the slip zone soil is predicted based on the structural density field and the scale response parameter.

[0008] Preferably, S1 specifically includes:

[0009] The CT image of the slip zone soil is a three-dimensional gray volume dataset composed of continuous two-dimensional slices. The three-dimensional gray volume dataset uses voxels as the smallest data unit. Each voxel corresponds to an original gray value. The original gray values ​​are normalized to obtain standardized gray values.

[0010] Preferably, S1 specifically includes:

[0011] After completing the grayscale normalization process, a local spatial analysis window is constructed with the voxel as the center. For any voxel position, a cubic neighborhood is constructed around it. Within the cubic neighborhood, the grayscale changes are weighted and accumulated through a spatial weight function with orientation modulation to generate an orientation-scale joint structural response operator.

[0012] Preferably, S1 specifically includes:

[0013] The structure density values ​​of each voxel are obtained by power-weighted integration of the direction-scale joint structure response operator; a structure density field is constructed based on the structure density values ​​of all voxels.

[0014] Preferably, S2 specifically includes:

[0015] Based on the structural density field, the average structural density value of each voxel block at each spatial scale is calculated; based on the average structural density value of each voxel block at each spatial scale, combined with the deviation enhancement term and the normalized harmonic term, the normalized structural perturbation intensity index at each spatial scale is calculated.

[0016] Preferably, S2 specifically includes:

[0017] The normalized structural perturbation intensity indices at various spatial scales are integrated into a single scale response parameter through nonlinear mapping.

[0018] Preferably, S2 specifically includes:

[0019] Based on the structural density field, a local structural perturbation term is constructed; based on the structural density field and a single scale response parameter, a power-mean term of structural density is constructed.

[0020] Preferably, S2 specifically includes:

[0021] A structural response prediction model is constructed based on the local structural disturbance term and the power mean term of the structural density.

[0022] The beneficial effects of the technical solution of the present invention are:

[0023] 1. Traditional image processing only identifies structural features in a certain direction or at a certain spatial scale, and cannot take into account multi-dimensional structural patterns such as pore orientation, microcrack orientation, and particle orientation. This invention constructs a direction-scale joint structural response operator to accurately capture the complexity of three-dimensional structures. Its physical meaning is clear, the calculation process is stable, and it can more realistically reflect the orientation and continuity of the "weak zones" and "slip surfaces" inside the slip zone soil.

[0024] 2. A continuous-scale perturbation calculation mechanism was introduced to establish a normalized structural perturbation intensity index and a single scale response parameter, which fully captured the degree of perturbation of the microstructure of slip zone soil at different scales and effectively characterized structural features such as weak structural zones that significantly affect the perturbation intensity.

[0025] 3. A structural response prediction model is constructed, establishing a mapping relationship between structure and strength. Local structural disturbance terms and power mean terms of structural density are introduced to capture the unstable behavior caused by micro-disturbances and the contribution of macro-compactness to the improvement of shear strength, respectively. The mathematical structure is clear, the physical meaning is clear, and it has good prediction accuracy, physical interpretability and wide applicability. Attached Figure Description

[0026] Figure 1 This is a flowchart of a method for predicting the strength of slip zone soil based on CT scan microstructure parameters, as described in this invention. Detailed Implementation

[0027] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] 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.

[0029] The following description, in conjunction with the accompanying drawings, details a specific scheme for a method for predicting the strength of slip zone soil based on CT scan microstructural parameters provided by this invention.

[0030] See attached document Figure 1 The diagram illustrates a flowchart of a method for predicting the strength of slip zone soil based on CT scan microstructural parameters, provided by an embodiment of the present invention. The method includes the following steps:

[0031] S1. Perform CT scans on the slip zone soil samples to obtain CT images of the slip zone soil, and perform grayscale normalization processing; after grayscale normalization processing, introduce the direction-scale joint structural response operator to construct the structural density field.

[0032] CT scans were performed on the slip zone soil samples to obtain CT images of the slip zone soil. These CT images are three-dimensional grayscale datasets composed of continuous two-dimensional slices. The three-dimensional grayscale dataset uses voxels as the smallest data unit, with each voxel corresponding to an original grayscale value. The original grayscale value reflects the density difference of the soil material at the corresponding spatial location. To ensure that subsequent algorithms are affected only by structural differences and not by equipment differences, the original grayscale values ​​need to be normalized using the min-max normalization method to obtain standardized grayscale values. The purpose of grayscale normalization is to eliminate the influence of different scanning batches and different CT energy parameters on the grayscale scale, ensuring that all subsequent spatial structure calculations are based on a unified dimensionless grayscale foundation.

[0033] After grayscale normalization, a local spatial analysis window is constructed centered on the voxel. For any voxel location... Build a side length of around it The cubic neighborhood containing Individual elements. Subsequently, a direction-scale joint structural response operator is introduced within this neighborhood to characterize the directionality and undulation of the microstructure in space. The direction-scale joint structural response operator does not directly calculate the traditional gradient, but rather uses a direction-modulated spatial weighting function to weight and accumulate grayscale changes. Its mathematical form is defined as:

[0034] in, It is the orientation-scale joint structural response operator, representing the voxel position. In direction The structural response intensity is the magnitude of the structural response intensity; the greater the structural response intensity, the more obvious the structural change. It is a neighborhood voxel position index; It is a local neighborhood window, representing the voxel position. Center point cubic neighborhood; It is the size of the cubic neighborhood window, used to control the spatial extent in which each voxel perceives structural information. , Indicates rounding up; The target structural scale, such as pore size, particle size, and microcrack width, is obtained from the existing slip zone soil sample database. This refers to the scan resolution, which is obtained from the parameters of the CT equipment. At the voxel position The gray value at the point after minimum-maximum normalization; It is a local spatial scale coefficient used to control the influence of neighborhood distance on the weighted response, similar to the Gaussian blur radius. ; It is a spatial Gaussian weight term, which represents the spatial distance weight of the voxel from the center point. The farther the distance, the smaller the weight. It is the spatial modulation frequency, which controls the periodicity of the cosine function, i.e., the spatial frequency density of the directional response. , The expected detection cycle is preset by the technicians. As a specific example, the value can be 4. It is a horizontal angle, controlling the angle perceived in the planar direction, used to align with the orientation of a specific structure, such as a crack. It is a variable with a value range of [missing value]. ; Voxel representation The relative center point in the two-dimensional plane direction Projection distance on; It is a directional periodic modulation response term. After inputting the directional projected coordinates into the cosine function, a positive and negative periodic response structure is constructed.

[0035] The above formula captures the spatial location, directional distribution, and scale characteristics of grayscale changes simultaneously, so that information such as pore arrangement direction and particle orientation are explicitly encoded into the calculation results.

[0036] Obtaining multiple horizontal angles After analyzing the directional structural response, the directional information is integrated into a scalar structural quantity that can directly participate in subsequent calculations. To this end, a power-weighted integral is performed on the directional structural response to construct the structural density field. any element The expression is:

[0037] in, This indicates the location of a voxel in a CT image of slip zone soil. The structure density value represents the structure perception intensity of the voxel in all directions; The directional enhancement index represents a control parameter for the fusion of directional responses. It is used to suppress weak directional responses, highlight the dominant structural direction, and reflect the nonlinearity of response integration. It is determined through curve fitting, such as the minimum mean square error (MSE) fitting method, based on existing CT images of slip zone soil and measured structural sensing intensities retrieved from an existing slip zone soil sample database. Specifically, the mean square error is calculated based on the structural sensing intensities obtained from existing CT images of slip zone soil and the measured structural sensing intensities. The index that minimizes the mean square error value is taken as the index. The reference value range is As a specific implementation, can be set to a value of 2; the above formula is based on the idea of ​​generalized mean, which aims to compress multi-directional structural information into a single scalar reflecting the complexity of the local structure. At this point, The magnitude of the value directly reflects the non-uniformity and directional concentration of the pore-particle structure in the neighborhood of the voxel.

[0038] Subsequently, a structure density field is constructed based on the structure density values ​​of all voxels in the entire three-dimensional space. The structural density field serves as the sole input for subsequent scale analysis.

[0039] S2. A continuous-scale perturbation calculation mechanism is introduced, and the CT image of the slip zone soil is divided into multi-scale voxel blocks according to spatial scale. Based on the structural density field, a normalized structural perturbation intensity index is generated at each spatial scale. Based on the normalized structural perturbation intensity index at each spatial scale, a scale response parameter is generated. A structural response prediction model is introduced, and the shear strength of the slip zone soil is predicted based on the structural density field and the scale response parameter.

[0040] To characterize the variation of the microstructure of slip zone soil at different spatial scales, a continuous-scale perturbation calculation mechanism is introduced. The specific operation is as follows:

[0041] By using a top-down voxel grouping method, the entire 3D volume is grouped according to its global spatial scale. (in The voxel side lengths in the 3D grayscale data set are obtained through CT equipment parameters; The spatial scale index is divided into multiple non-overlapping sub-blocks, i.e., multi-scale voxel blocks. At each spatial scale... Next, calculate the first... The average structural density value of each sub-block, i.e., the equivalent structural density value. :

[0042] in, In terms of spatial scale Next, the The average structural density value of individual primitive blocks; It is the current spatial scale The number of voxels in each voxel block; Indicates spatial scale Next, the Individual voxels, each representing a statistical unit of a local region. The above steps map microstructural information onto different physical scales, providing a foundation for subsequent analysis of structural fluctuations.

[0043] After obtaining the average structural density values ​​of each voxel block at various spatial scales, the spatial scale is then... The mathematical expression for calculating the disturbance intensity of the structural density within this spatial scale is:

[0044] in, Spatial scale The normalized structural perturbation intensity index under the given conditions characterizes the spatial scale. The overall deviation of the local structural density of the subsidence zone soil is an important quantitative indicator of structural complexity. The larger the value, the greater the difference in microstructure at that spatial scale, which may lead to uneven strength or weak structural areas. In terms of spatial scale The following is the number of spatial units (cubic sub-blocks) into which the entire CT image of the slip zone soil is divided; , Indicates rounding down; , , These are the number of voxels in the horizontal direction, the number of voxels in the vertical direction, and the number of slices in the scanning direction, respectively. Indicates spatial scale The central level of the global structure density distribution. , used as a benchmark for perturbation comparison; It is the standard deviation of structural density. Used to measure spatial scale The overall dispersion of the lower structure density distribution is used as the denominator term for perturbation standardization; It is a power-order response enhancement factor that controls the nonlinear amplification of the bias enhancement term. It is used to emphasize regions with high structural volatility and is determined through Bayesian optimization. The reference value range is [value range missing]. As a specific implementation, the value can be 2. Specifically, by using the mean squared error as the objective function, Considering these variables as input variables to the objective function, a Gaussian process regression model is established to probabilistically model the objective function, and then the optimal variable is selected in each iteration. The value minimizes the objective function and eventually converges to the optimal solution; It is a deviation enhancement term, used to amplify the nonlinearity of local structural differences; It is a normalized harmonic term that makes the disturbance value not only related to the absolute fluctuation, but also controlled by the number of sub-blocks and the fluctuation basis.

[0045] The above formula is derived from the idea of ​​higher-order central moments, and its purpose is to avoid the problem of insensitivity to extreme structural changes caused by using only variance.

[0046] Subsequently, the normalized structural disturbance intensity indices at each spatial scale were... Further integration into a single scale response parameter This refers to the multi-scale structural response intensity index, which serves as a direct input variable for intensity prediction. The scale response parameters are derived through a nonlinear mapping formula:

[0047] in, It is a single scale response parameter, namely a multi-scale structural response intensity index, used to quantify the microstructural complexity of slip zone soil. It can be interpreted as an index of the comprehensive complexity of the microstructure of slip zone soil at multiple spatial scales. This is the maximum spatial scale index, i.e., the maximum number of scales participating in the continuous scale perturbation calculation mechanism. It is calculated by dividing the minimum directional dimension (i.e., the number of voxels in the minimum direction) of the CT image of the slip zone soil acquired by the CT equipment in three-dimensional space by 8 (assuming at least 8 effective structural units), and then taking the logarithm to base 2. ; It is a scale weighting factor, representing the spatial scale. The contribution ratio of the microstructural complexity assessment of slip zone soil is determined through the Transformer attention mechanism. Specifically, the normalized structural perturbation intensity index at each spatial scale is constructed as a sequence vector. , To represent the transpose and, in order to perceive the spatial hierarchy between scales, the normalized structural perturbation intensity index at each spatial scale is positionally encoded to obtain the enhanced feature vector. Three learnable weight matrices are introduced. The enhanced feature vector is mapped to a high-dimensional space through a linear transformation: query vector key vector value vector Calculate spatial scale using the dot product attention formula Spatial Scale Association weight between In the structural analysis of slip zone soils, this reflects the explanatory power of characteristics at one spatial scale on characteristics at another spatial scale. ,in, Spatial scale The query vector below, Spatial scale The transpose of the query vector below, As a scaling factor to prevent gradient vanishing, its value is typically set to a constant such as 16, 32, 64, or 128, depending on the convergence requirements of Transformer attention mechanism training. In a preferred embodiment of the present invention, The value is set to 64; the association weights are normalized using Softmax to obtain the importance score of each spatial scale in the global structural representation, which is the final scale weight factor. Defined as the mean output of the attention head corresponding to this spatial scale, its value range is: ; It is a nonlinear enhancement index used to improve the response contribution of the normalized structural perturbation intensity index, calculated based on the coefficient of variation. ,in, It is the structure density field The coefficient of variation is obtained by dividing the mean of the structure density field by the standard deviation of the structure density field. It is a logarithmic compression term used to control the dominant effect of extreme perturbation values ​​and enhance stability.

[0048] After completing the structural density field With scale response parameters After calculation, a structural response prediction model is introduced to map the shear strength of the slip zone soil. The specific implementation formula is as follows:

[0049] in, This represents the predicted shear strength of the slip zone soil; This is an adjustment coefficient for the local structural disturbance term, used to adjust the weight of its influence on shear strength. The reference value range is... ; This is an adjustment coefficient for the power-mean term of the structural density, used to adjust the weight of its influence on shear strength. The reference value range is... ; , All parameters are obtained from the shear strength and corresponding structural parameters such as single scale response parameters and structural density values ​​obtained from the slip zone soil sample database. The least squares method is used for fitting. Specifically, slip zone soil samples with known measured shear strength values ​​are selected from the database, and the single scale response parameter and structural density value corresponding to each sample are calculated. The objective function is to minimize the sum of squared errors between the predicted and measured shear strengths. , ; It is the total number of voxels in the CT image of the slip zone soil, that is, the number of structural units involved in the statistics; It is the first The local perturbation response value of an individual element represents the intensity of its deviation from the mean, and is used to amplify local anomalous structures. It is defined as follows: ,in, It is in the Voxel position of a voxel ( The structural density value is taken from the structural density field. ; It is the average structural density value at all voxel locations; It is the offset penalty factor, determined by a grid search method, with a reference value range of [value missing]. ; It is a structural enhancement index used to control the nonlinear enhancement of structural density against shear strength. It is obtained using a grid search method based on shear strength data obtained from a slip zone soil sample database and corresponding structural parameters such as single scale response parameters and structural density values. The reference value range is [insert range here]. ; It is a local structural disturbance term, representing the amplification of the intensity of the local structural disturbance in the overall structure; It is the power-mean term of the structural density, which describes the nonlinear saturation value response of macroscopic compactness to the enhancement of shear strength.

[0050] In summary, a method for predicting the strength of slip zone soil based on CT scan microstructural parameters has been developed.

[0051] The order of the embodiments is for illustrative purposes only and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0052] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0053] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for predicting the strength of slip zone soil based on CT scan microstructural parameters, characterized in that, Includes the following steps: S1. Perform CT scans on the slip zone soil samples to obtain CT images of the slip zone soil, and perform grayscale normalization processing; after grayscale normalization processing, introduce the direction-scale joint structural response operator to construct the structural density field. S2. Introduce a continuous-scale perturbation calculation mechanism to divide the CT image of the slip zone soil into multi-scale voxel blocks according to spatial scale, and generate normalized structural perturbation intensity indices at each spatial scale based on the structural density field. Based on the normalized structural disturbance strength index at various spatial scales, scale response parameters are generated; a structural response prediction model is introduced, and the shear strength of the slip zone soil is predicted based on the structural density field and scale response parameters.

2. The method for predicting the strength of slip zone soil based on CT scan microstructure parameters according to claim 1, characterized in that, S1 specifically includes: The CT image of the slip zone soil is a three-dimensional gray volume dataset composed of continuous two-dimensional slices. The three-dimensional gray volume dataset uses voxels as the smallest data unit. Each voxel corresponds to an original gray value. The original gray values ​​are normalized to obtain standardized gray values.

3. The method for predicting the strength of slip zone soil based on CT scan microstructure parameters according to claim 2, characterized in that, S1 specifically includes: After completing the grayscale normalization process, a local spatial analysis window is constructed with the voxel as the center. For any voxel position, a cubic neighborhood is constructed around it. Within the cubic neighborhood, the grayscale changes are weighted and accumulated through a spatial weight function with orientation modulation to generate an orientation-scale joint structural response operator.

4. The method for predicting the strength of slip zone soil based on CT scan microstructure parameters according to claim 3, characterized in that, S1 specifically includes: The structure density values ​​of each voxel are obtained by power-weighted integration of the direction-scale joint structure response operator; a structure density field is constructed based on the structure density values ​​of all voxels.

5. The method for predicting the strength of slip zone soil based on CT scan microstructure parameters according to claim 1, characterized in that, S2 specifically includes: Based on the structural density field, the average structural density value of each voxel block at each spatial scale is calculated; based on the average structural density value of each voxel block at each spatial scale, combined with the deviation enhancement term and the normalized harmonic term, the normalized structural perturbation intensity index at each spatial scale is calculated.

6. The method for predicting the strength of slip zone soil based on CT scan microstructure parameters according to claim 5, characterized in that, S2 specifically includes: The normalized structural perturbation intensity indices at various spatial scales are integrated into a single scale response parameter through nonlinear mapping.

7. The method for predicting the strength of slip zone soil based on CT scan microstructure parameters according to claim 6, characterized in that, S2 specifically includes: Based on the structural density field, a local structural perturbation term is constructed; based on the structural density field and a single scale response parameter, a power-mean term of structural density is constructed.

8. The method for predicting the strength of slip zone soil based on CT scan microstructure parameters according to claim 7, characterized in that, S2 specifically includes: A structural response prediction model is constructed based on the local structural disturbance term and the power mean term of the structural density.

Images