A method and device for evaluating the reliability of slope support

Abstract

The present application relates to the technical field of slope stability analysis, and discloses a slope support reliability evaluation method and equipment. The present application fully considers the shear strength variability of rock-soil mass and support structure parameters, trains an efficient and robust safety factor proxy model, and is used for calculating the safety factor of a slope under different support structures and rock-soil mass parameters, thereby providing technical support for the rapid evaluation of the stability of slope engineering, the design of slope support structure and the like.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of slope stability analysis, in particular to a slope support reliability evaluation method and device. BACKGROUND

[0002] Slope disaster is one of the most common geological disasters, which seriously affects the safety of life and property and infrastructure construction. In order to evaluate the safety of the slope, the limit equilibrium method, finite element numerical simulation, strength reduction and other aspects are generally used to give the safety factor of the slope, but these deterministic analysis methods cannot consider the variability of rock and soil parameters. Due to the load history condition and physical and chemical process, the shear strength parameters of the slope soil show spatial variability. In traditional slope engineering analysis, the shear strength parameters cannot be accurately determined by investigation, but only the local mean value can represent the shear strength of the whole slope, and the safety factor of the slope is given by the limit equilibrium method, finite element numerical simulation, strength reduction and other deterministic analysis methods, but the variability of rock and soil parameters will affect the reliability of deterministic analysis, and the existing analysis methods lack considering the uncertainty of rock and soil to evaluate whether the supporting structure is reliable. SUMMARY

[0003] In order to overcome the influence of the variability of rock and soil parameters on the analysis results in the existing slope stability analysis, and the problem that the existing analysis method lacks considering the uncertainty of rock and soil to evaluate whether the supporting structure is reliable, the present application provides a slope support reliability evaluation method and device.

[0004] In order to achieve the above application purpose, the present application provides the following technical scheme:

[0005] In the first aspect, the present application provides a slope support reliability evaluation method, which comprises:

[0006] Based on the geological body parameters of the target slope, a random field simulation is performed to generate a plurality of first shear strength distribution samples, and a random extraction is performed based on the range of the support parameters of the target slope to generate a plurality of support structure combination samples;

[0007] The plurality of first shear strength distribution samples and the plurality of support structure combination samples are randomly combined to generate a plurality of slope samples, and a first safety factor of each slope sample is calculated through numerical simulation;

[0008] The geological body parameter distribution image, the geological body parameter average value and the support structure design parameter in the slope sample are extracted, the corresponding geological body parameter distribution image, the geological body parameter average value and the support structure design parameter are labeled according to the first safety factor, and a data set is generated;

[0009] training a proxy model according to the data set, to obtain a trained safety factor proxy model;

[0010] Based on the target slope to be evaluated support parameter combination, each support parameter combination to be evaluated is simulated based on the geological body parameter random field, and a plurality of second shear strength distribution samples are generated;

[0011] Each support parameter combination to be evaluated and the corresponding plurality of second shear strength distribution samples are input into the trained safety factor proxy model, to obtain a plurality of second safety factors of each support parameter combination to be evaluated.

[0012] According to the plurality of second safety factors, the reliability of each support parameter combination to be evaluated is calculated, and an evaluation result is obtained.

[0013] According to a specific embodiment, in the above evaluation method, the geological body parameters include the average value of cohesion, the coefficient of variation of cohesion, the fluctuation distance of cohesion, the average value of friction angle, the coefficient of variation of friction angle and the fluctuation distance of friction angle, the geological body parameter distribution image includes the cohesion distribution image and the friction angle distribution image, and the average value of the geological body parameter is the average value of cohesion and the average value of friction angle in the first shear strength distribution sample.

[0014] According to a specific embodiment, in the above evaluation method, the extraction of the geological body parameter distribution image in the slope sample includes:

[0015] The cohesion value and the friction angle value of each unit in the first shear strength distribution sample are derived respectively, to obtain the spatial distribution of cohesion and the spatial distribution of friction angle, and the values of each unit in the spatial distribution of cohesion and the spatial distribution of friction angle are converted into gray intensity respectively, to generate the cohesion distribution image and the friction angle distribution image.

[0016] According to a specific embodiment, in the above evaluation method, the safety factor proxy model includes a convolutional neural network and a fully connected neural network; wherein the convolutional neural network is used to extract features in the geological body parameter distribution image and output one-dimensional features; after the one-dimensional features, the average value of the geological body parameter and the support structure design parameter are spliced to form a fusion feature, the fusion feature is input into the fully connected neural network, and the safety factor is output.

[0017] According to a specific embodiment, in the above evaluation method, the convolutional neural network includes a channel attention module, a spatial attention module, two standard convolutional layers, two maximum pooling layers and an unfolding layer, the channel attention module includes a global pooling layer and a fully connected layer, and the spatial attention module includes a pooling layer and a one-dimensional convolutional layer.

[0018] According to a specific embodiment, in the above evaluation method, the convolutional neural network is configured to extract features in the geological body parameter distribution image and output two-dimensional features, and specifically includes:

[0019] The geological body parameter distribution image is input into a first standard convolutional layer, and then input into the channel attention module and the spatial attention module for selective processing of features, so as to improve the weights of key channels and key regions of the image input features.

[0020] The processed image input features are input into a first maximum pooling layer, a second standard convolutional layer, and a second maximum pooling layer, and then output one-dimensional features through an unfolding layer.

[0021] According to a specific embodiment, in the above evaluation method, the support parameters include support parameters of anchor rods or anchor cables and anti-slide pile parameters, and the combinations of support parameters to be evaluated include lengths of anchor rods or anchor cables, incident angles of anchor rods or anchor cables, prestresses of anchor rods or anchor cables, spacings of anchor rods or anchor cables, and lengths of anti-slide piles.

[0022] According to a specific embodiment, in the above evaluation method, the reliability of each combination of support parameters to be evaluated is calculated according to the second safety factors, and specifically includes:

[0023] The proportion of the second safety factors less than the preset threshold in the second safety factors is calculated to obtain the failure probability, and the formula is:

[0024] ;

[0025] The reliability of each combination of support parameters to be evaluated is calculated according to the failure probability, and the formula is:

[0026] ;

[0027] wherein, is the reliability, is the failure probability of each combination of support parameters to be evaluated, is the probability that the second safety factor is less than the preset threshold, is the number of second safety factors less than the preset threshold, is the number of second safety factors.

[0028] ​​In a second aspect, the present application provides an electronic device, characterized by comprising at least one processor, and a memory connected with the at least one processor in communication; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the slope support reliability evaluation method according to any one of the above.

[0029] Compared with the prior art, the present application has the following beneficial effects:

[0030] The present application fully considers the shear strength variability of the rock-soil body and the supporting structure parameters, trains an efficient and robust slope stability analysis proxy model, and is used for calculating the slope safety factor under different supporting structures and rock-soil body parameters, thereby providing technical support for the rapid evaluation of the stability of the slope engineering and the design of the slope supporting structure. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 A flowchart of a slope support reliability evaluation method provided by the embodiment of the present application is shown in the figure;

[0032] Figure 2 A slope sample generation schematic diagram provided by the embodiment of the present application is shown in the figure;

[0033] Figure 3 A flowchart of a slope support reliability evaluation method provided by the embodiment of the present application is shown in the figure;

[0034] Figure 4 A structure schematic diagram of a safety factor proxy model provided by the embodiment of the present application is shown in the figure;

[0035] Figure 5 An image processing flowchart provided by the embodiment of the present application is shown in the figure;

[0036] Figure 6 A safety factor proxy model and a general model prediction effect comparison schematic diagram provided by the embodiment of the present application is shown in the figure;

[0037] Figure 7 A to-be-optimized supporting parameter schematic diagram provided by the embodiment of the present application is shown in the figure;

[0038] Figure 8 A sample safety factor calculation result schematic diagram provided by the embodiment of the present application is shown in the figure;

[0039] Figure 9 A performance evaluation schematic diagram of a safety factor proxy model provided by the embodiment of the present application is shown in the figure;

[0040] Figure 10 A slope reliability evaluation result schematic diagram of different supporting parameter combinations provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0041] The application will be described in further detail below with reference to the test examples and specific embodiments. However, this should not be understood as limiting the scope of the above-mentioned subject matter of the application to the following examples only, but all technical solutions achieved on the basis of the content of the application belong to the scope of the application.

[0042] The terms "first", "second", and the like in the description and in the claims of the application and in the drawings of the application merely serve the purpose of distinguishing between described objects, and do not carry any relative implications either to the significance or to the sequence thereof. Furthermore, the terms "comprising" and "having" and any of their derivations, are intended to cover the inclusions not exclusively, for example, the inclusion of a series of steps or units. The method, system, product or device is not necessarily limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0043] In addition, any example or design scheme described as "exemplary" or "for example" in the embodiments of the application should not be interpreted as more preferred or more advantageous than other examples or design schemes. Rather, the use of "exemplary" or "for example" and the like is intended to present the relevant concept in a specific manner for ease of understanding.

[0044] The application is mainly used to solve the calculation efficiency problem of slope reliability analysis considering the shear strength variability of rock-soil mass, the problem that the machine learning agent model of slope pure data driving cannot consider the physical properties of geological body, and the problem that the general attention mechanism cannot capture the complex structure and multi-level features in the slope.

[0045] Based on the above problems, random reliability analysis can be used to estimate the reliability or failure probability of the slope from uncertain strength parameters. The basic idea is to introduce the variability of rock and soil parameter distribution in the slope model through random field simulation and Monte Carlo sampling, and finally to calculate the safety factor and reliability of the slope by using numerical simulation methods such as finite element and finite difference, in order to evaluate whether the support structure design meets the requirements under uncertain slope geology. The calculation accuracy of reliability requires a large number of Monte Carlo sampling as support, so we need to perform numerical simulation calculation for millions of times, but these numerical simulations usually require a lot of time and computing power. In order to improve the calculation efficiency, scholars use proxy models to predict the safety factor of the slope, such as using neural networks and support vector regression models to build the relationship between the slope parameters and the safety factor of the slope, which can greatly reduce the calculation amount of slope reliability analysis. Using neural networks and support vector regression methods can build a correlation between simple parameters of the slope (such as slope height, slope angle, average shear strength, etc.) and the safety factor, i.e. a pure data-driven proxy model. Pure data-driven machine learning methods cannot reflect the variability of shear strength parameters, and require a large number of training samples to improve model performance. Especially when there is a complex support structure, pure data-driven models ignore the physical mechanism between the support structure and the slope rock and soil, resulting in poor generalization ability and poor interpretability. Therefore, it is necessary to develop more robust and parameter space variability considering slope stability analysis proxy models.

[0046] The slope support reliability evaluation method provided by the embodiments of the present application will be described in detail below.

[0047] Please refer to Figure 1 which shows a flowchart of a slope support reliability evaluation method provided by an embodiment of the present application, the method comprising:

[0048] Step 1: based on the geology parameters of the target slope, perform random field simulation to generate a plurality of first shear strength distribution samples, and based on the range of support parameters of the target slope, perform random sampling to generate a plurality of support structure combination samples;

[0049] Step 2: randomly combine the plurality of first shear strength distribution samples and the plurality of support structure combination samples to generate a plurality of slope samples, and calculate the first safety factor of each slope sample through numerical simulation;

[0050] Step 3: extract the geology parameter distribution image, geology parameter average value and support structure design parameter in the slope sample, label the corresponding geology parameter distribution image, geology parameter average value and support structure design parameter according to the first safety factor, and generate a data set;

[0051] Step 4: training an agent model according to the data set to obtain a trained safety coefficient agent model;

[0052] Step 5: based on the target slope, a geological body parameter-based random field simulation is performed on each to-be-evaluated supporting parameter combination to generate a plurality of second shear strength distribution samples;

[0053] Step 6: each to-be-evaluated supporting parameter combination and the corresponding plurality of second shear strength distribution samples are input into the trained safety coefficient agent model to obtain a plurality of second safety coefficients of each to-be-evaluated supporting parameter combination.

[0054] Step 7: the reliability of each to-be-evaluated supporting parameter combination is calculated according to the plurality of second safety coefficients to obtain an evaluation result.

[0055] The embodiment of the present application fully considers the shear strength variability of the rock-soil body and the supporting structure parameters, trains an efficient and robust slope stability analysis agent model, and is used for calculating the safety coefficient of the slope under different supporting structures and rock-soil body parameters, thereby providing technical support for the safety rapid evaluation of the slope engineering and the design of the slope supporting structure.

[0056] Specifically, in step 1, please refer to Figure 2 which shows a sample generation schematic diagram provided by the embodiment of the present application. First, according to the fine finite difference numerical model of the typical section of the slope, according to the geological exploration report and the to-be-evaluated supporting design parameters, the numerical simulation of the slope under different working conditions is carried out.

[0057] For the geological body parameters, based on the average value, the variation coefficient, the fluctuation distance and other parameters of the cohesion and the friction angle of different soil layers, the shear strength distribution of the rock-soil body is simulated by using the random field simulation method to generate a plurality of first shear strength distribution samples.

[0058] For the supporting parameters, the supporting parameters include the supporting parameters of the anchor rod or anchor cable and the anti-slide pile parameters, the to-be-evaluated supporting parameter combination includes the length of the anchor rod or anchor cable, the incident angle of the anchor rod or anchor cable, the prestress of the anchor rod or anchor cable, the spacing of the anchor rod or anchor cable and the length of the anti-slide pile. After the range of the supporting parameters of the anchor rod or anchor cable and the anti-slide pile parameters is determined, the design parameters are randomly extracted by using the uniform sampling method to generate a plurality of supporting structure combination samples.

[0059] According to the random combination of the samples generated in the above steps, a plurality of slope samples are generated, each slope sample has different shear strength distribution and supporting structure combination, the safety coefficient is calculated by using the strength reduction method of numerical simulation, and the first safety coefficient of each slope sample is calculated.

[0060] Further, the cohesion distribution image and the friction angle distribution image in each slope sample are extracted and converted into a gray image as an image feature. The average value of the cohesion, the average value of the friction angle, and the support structure design parameters in each first shear strength distribution sample are used as numerical features. A convolutional neural network with a multi-level attention mechanism is used to extract image features. The processed image features, geotechnical parameters, support structure parameters, and other numerical features are spliced to form a fusion feature. A fully connected neural network is used to process the fusion feature, and finally the safety factor corresponding to each slope sample is output. The safety factor label of the data set is the first safety factor calculated by numerical simulation in step 1. According to the first safety factor label, the geological body parameter distribution image, the average value of the geological body parameter, and the support structure design parameter are labeled to generate a data set. The proxy model is trained according to the data set, and a physical information machine learning safety factor proxy model that can accurately predict the safety factor of the slope is established.

[0061] Further, different support structure combination parameters and random field simulation results are input into the proxy model to realize rapid evaluation of the reliability of the slope under different parameter combinations and assist in the design of the slope support structure. Please refer to Figure 3 Fig. 1 shows a flowchart of a slope support reliability evaluation method according to an embodiment of the present application.

[0062] Specifically, the physical information machine learning safety factor proxy model proposed in the embodiments of the present application mainly consists of a convolutional neural network with a multi-level attention mechanism and a fully connected neural network. In the safety factor proxy model, the distribution images of cohesion and friction angle are obtained from numerical simulation and belong to the category of physical information. At the same time, the average value of the cohesion, the average value of the friction angle, and the slope support structure parameters (such as pile length, anchor length, anchor incidence angle, etc.) are used to enhance the performance of the model from the perspective of data-driven. Such a combination of features not only utilizes the physical information brought by numerical simulation, but also improves the prediction ability of the model by introducing key statistical parameters and design parameters.

[0063] Specifically, a dual-channel gray image is formed by pre-processing the cohesion distribution image and the friction angle distribution image as physical information. The dual-channel gray image is first input into a standard convolution layer, and then the features are input into a multi-level attention module for selective feature processing to improve the weight of key channels and key regions of the image features. The image features processed by the multi-level attention module are output to a first max-pooling layer. Then, a second convolution layer and a second max-pooling layer are used. The output of the second max-pooling layer is unfolded into a one-dimensional array by an unfolding layer, and is connected with numerical input features (rock mass shear strength parameters, support structure design parameters, etc.) to form complete fusion features. The spliced feature vector is output to a neural network composed of multiple fully connected layers, and the layers are connected by a nonlinear activation function. Finally, the safety factor of the slope is output. Please refer to Figure 4 Fig. 3 shows a structural diagram of a safety factor proxy model according to an embodiment of the present application.

[0064] The color image is composed of three channels of RGB, and the gray image is composed of one channel of gray. The input image used by the proxy model proposed in the embodiment of the present application is a gray image. The gray intensity value of each pixel point in the gray image represents the color depth of the pixel point. According to this characteristic, the cohesion value and the friction angle value of each unit in the first shear strength distribution sample are derived respectively to obtain the cohesion spatial distribution and the friction angle spatial distribution. The values of each unit in the cohesion spatial distribution and the friction angle spatial distribution are converted into gray intensity respectively to generate a cohesion distribution image and a friction angle distribution image, which are then input into a convolutional neural network as features. Specifically, open the slope attribute view in the numerical analysis software. At this time, the color of each region in the numerical model displayed in the window represents the attribute value of the position. For example, open the cohesion attribute view. Different grid points have different colors, from blue to red, representing the cohesion from small to large, which is similar to the gray intensity value of the pixel point in the gray image.

[0065] Therefore, the cohesion distribution image and the friction angle distribution image are processed according to the following method in the embodiment of the present application:

[0066] (1) Random field discretization: first, based on the geological body parameters of the target slope, the shear strength distribution is obtained after random field simulation. The cohesion value and the friction angle value of each unit are derived respectively, that is, the cohesion and friction angle spatial distribution of the sample is obtained;

[0067] (2) Mapping to digital image: the above-mentioned cohesion and friction angle spatial distribution is converted into a gray digital image. Specifically, the coordinates of the numerical model unit correspond to the pixel coordinates of the image. The shear strength value (cohesion or friction angle) of the unit is converted into a gray intensity (0-255) respectively;

[0068] (3) Image generation: Create two gray-scale images representing the distribution of cohesion or friction angle, respectively;

[0069] (4) Additional numerical features: Calculate the average value of the cohesion or friction angle for each realization of the random field and use it as an additional numerical input feature.

[0070] This method enables the present application to leverage the powerful capabilities of convolutional neural networks in processing images, effectively capturing the spatial variability of the soil mass of the slope.

[0071] Further, the input includes two gray-scale digital images (representing the distribution of cohesion or friction angle, respectively) and two numerical features (average cohesion and friction angle) as well as the support structure design parameters, and the processing flow is as shown in Figure 5 .

[0072] It can be understood that the attention module is an important part of the convolutional neural network architecture design, similar to the selective processing of the human visual system for complex scenes, the convolutional neural network uses the attention mechanism to focus on specific parts of the input features. It enables the convolutional neural network to focus on key areas or channels of key images, reducing unnecessary calculations and improving efficiency.

[0073] The present application uses a multi-level attention mechanism to process the cohesion distribution image and the friction angle distribution image to enhance the model's ability to identify key areas such as potential sliding surfaces, in order to enhance the performance of the agent model in predicting the safety factor of the slope.

[0074] Specifically, in one possible implementation, let the input feature map be where H, W, and C represent the height, width, and number of channels of the feature map, respectively. The input feature map consists of two channels, which are the gray-scale images representing the cohesion distribution and friction angle distribution of the slope physical model.

[0075] (1) Channel attention module

[0076] Step 1: Global pooling

[0077] The features are first passed through a global pooling layer to identify which shear strength parameter (cohesion or friction angle) has a greater impact on the stability of the slope. First, perform global average pooling and global maximum pooling on the input feature map F:

[0078]

[0079]

[0080] where the global average pooling and global maximum pooling for the kth channel can be represented as:

[0081]

[0082]

[0083] After global average pooling, the results of all channels are summarized into a tensor with dimensions 1x1xC After global max pooling, the results of all channels are summarized into a tensor with dimensions 1x1xC .

[0084] Step 2: Weight Generation

[0085] The feature vectors after global max pooling and average pooling are input into a shared fully connected layer. This fully connected layer is used to learn the attention weights for each channel. Through learning, the network can adaptively decide which channels are more important for the current task.

[0086] For global average pooling:

[0087]

[0088] For global max pooling:

[0089]

[0090] where and are the weight matrices of the fully connected layer, r is the compression ratio, is the activation function, and the ReLU activation function is used for nonlinear transformation. To ensure that the attention weights are between 0 and 1, a Sigmoid activation function is applied to produce the channel attention weights. The weight matrices and are initialized when the fully connected layer is defined. During the entire training process, as each batch of data is input and backpropagated, the weight matrices and are gradually adjusted to minimize the loss function. After several training epochs, these weight matrices will gradually converge to values that best capture the important channels in the input features.

[0091] Step 3: Feature Weighting

[0092] Element-wise multiplication of the channel attention weights and the original feature map enhances the feature expression of key channels:

[0093]

[0094] where, denotes element-wise multiplication. The attention weight is multiplied with each channel of the original feature map, resulting in an attention-weighted channel feature map. This will emphasize the channels that are helpful for the current task and suppress the irrelevant ones.

[0095] (2) Spatial attention module

[0096] First step: Pooling operation

[0097] The channel-weighted feature map is averaged and max-pooled in the channel dimension. This operation takes the average of all channel values at each pixel position, resulting in an output feature map containing a single channel.

[0098]

[0099]

[0100] The average of all channel values at each spatial position (i,j) is taken:

[0101] ,

[0102] .

[0103] Second step: Weight generation

[0104] The pooled feature maps are concatenated in the channel dimension, and a two-dimensional convolutional layer (Conv2D) is used to process the concatenated feature maps. A two-dimensional convolutional layer with 1 filter is applied. This operation outputs a single-channel attention weight map by integrating information from the concatenated channels, generating spatial attention weights that are used to adjust the weights of each spatial position of the input feature map.

[0105] ,

[0106] .

[0107] Third step: Feature weighting

[0108] The original feature map is element-wise multiplied with the spatial attention weights:

[0109] ,

[0110] The final output is the feature map that has been weighted by both channel and spatial attention , which is used for further slope safety factor prediction analysis.

[0111] The multiple attention mechanism combines channel and spatial attention mechanisms, extracts and amplifies key information of feature maps in channel and spatial dimensions, and improves the identification ability of important shear strength parameters and potential sliding areas of the slope.

[0112] Specifically, the slope support reliability evaluation method provided by the embodiment of the present application uses a physical information-based machine learning neural network as a surrogate model for slope safety factor calculation, reducing the computational load of numerical simulation; uses the shear strength distribution image of the rock-soil mass as one of the input features of the model, which can represent the physical characteristics of the rock-soil mass, and integrates the physical feature constraint on the basis of pure data driving. A multi-level attention mechanism is used, and the model has a significant improvement in identifying key slope geological body areas and key shear strength parameters, providing higher accuracy and stronger generalization ability.

[0113] The safety factor surrogate model based on physical information machine learning provided by the embodiment of the present application exhibits excellent performance in slope safety factor prediction and reliability evaluation, and has obvious advantages compared with traditional methods:

[0114] Greatly improved computational efficiency: By introducing a physical information-based surrogate model, the present application greatly reduces the numerical calculation required in the traditional Monte Carlo simulation method. In a scenario that usually requires a million numerical simulations, the surrogate model of the present application only needs to perform a small amount of numerical simulation calculation to achieve the same calculation accuracy. This greatly shortens the evaluation time in slope engineering design and provides strong support for the rapid optimization of complex slope design.

[0115] Fully consider the spatial variability of rock-soil mass: The present application inputs the spatial variability of rock-soil mass shear strength into the model through physical information images, overcoming the limitation of traditional pure data-driven models that cannot reflect the physical characteristics of geological bodies. By inputting the distribution of shear strength in the form of a grayscale image into the model and combining a multi-level attention mechanism for feature extraction, the model's ability to identify complex structures and potential sliding surfaces in slopes is significantly enhanced, making safety factor prediction more accurate.

[0116] Improve the generalization ability and accuracy of the model: Compared with traditional neural network or support vector machine surrogate models, the present application combines physical information and geological design parameters and uses a multi-level attention mechanism to effectively capture key areas and parameters in complex slopes.

[0117] Better interpretability: By introducing the physical information of shear strength, the present application improves the interpretability of the model. The model output can not only directly predict the safety factor of the slope, but also be used to analyze the impact of support design parameters on the reliability of the slope. This combination of physical and data greatly enhances the reliability and acceptability of the model in practical engineering applications.

[0118] It should be noted that the explainability of the surrogate model is crucial in engineering applications, as it helps engineers to understand the prediction mechanism of the model and make necessary adjustments. Grad-CAM images can be used to qualitatively evaluate the explainability of the surrogate model in predicting the safety factor. Grad-CAM is a visualization method that clearly shows the attention area of the model by calculating the importance of spatial positions in the convolution layer using gradients.

[0119] In order to verify the improvement of the safety factor surrogate model provided by the embodiment of the present application in terms of explainability and accuracy, the Grad-CAM images of the safety factor predicted by the safety factor surrogate model are compared with the general model (LeNet5 convolutional neural network model without using multiple attention mechanisms), as shown in Figure 6 The two models are trained and tested on the same data set. Six samples (including three failure samples) are randomly selected from the validation data set, and Grad-CAM is used to generate heat maps of the feature weights of the output of the last convolutional layer to show the feature areas that the model focuses on when predicting the safety factor of the slope. At the same time, we also derive the maximum shear strain increment map when calculating the safety factor using numerical simulation. The shear strain increment represents the deformation degree of each region. The region with larger shear strain increment may indicate that significant shear deformation occurs in this region, so the local area of the shear strain increment can represent the potential sliding surface. Qualitatively compare the heat map generated by Grad-CAM with the maximum shear strain increment map. The safety factor surrogate model provided by the embodiment of the present application always focuses on the region with larger shear strain increment, which indicates potential slope instability, indicating that the safety factor surrogate model provided by the embodiment of the present application has better ability to identify key regions than the general convolutional neural network model. Moreover, the predicted safety factor value of the model is more aligned with the numerical simulation value, especially in samples with safety factor less than 1. Therefore, the safety factor surrogate model provided by the embodiment of the present application significantly enhances the ability of the model to identify key features of the slope, and improves the performance and explainability of the surrogate model.

[0120] In summary, the present application not only improves the efficiency and accuracy of slope stability analysis, but also provides a more intelligent decision support tool for slope engineering design, which has engineering application value and market potential.

[0121] The technical solutions provided by the present application will be further described below in combination with specific implementation cases.

[0122] Specifically, the present application intends to calculate the safety factor of the slope under different design parameters considering the variability of parameters, and provide reference for the slope support design. Therefore, in the embodiment, first, a refined finite difference numerical model of the typical section of the slope is established, and numerical simulation of the slope under different working conditions is carried out according to the geological exploration report and the support design parameters to be evaluated. For the parameters of the slope geological body, the shear strength distribution of the rock-soil body is simulated by using the random field simulation method based on the average value, coefficient of variation, fluctuation distance and other parameters of the cohesion and friction angle of different soil layers.

[0123] The design parameters to be optimized include the anchor rod length, anchor rod incidence angle, anchor rod prestress, anchor rod spacing, anti-slide pile length, etc. as shown in Figure 7 Specifically, in a possible implementation manner, the support parameters include: anchor rod length L = 15-60m, anchor rod incidence angle α = 20-40°, anchor rod spacing (out of plane) s = 2.5~6 m, anchor rod prestress F = 600~800 kN, anti-slide pile length D = 15~30 m.

[0124] After determining the range of the design parameters, the design parameters are randomly extracted by using uniform sampling. In summary, the automatic process is realized by using the python-flac3d interface, 500 groups of random working condition slope samples are randomly generated, and the safety factor is calculated by using the strength reduction method of numerical simulation. The safety factor calculation results of the 500 groups of slope samples are subject to normal distribution as shown in Figure 8 The rock-soil body parameters and support structure parameters are shown in the following table.

[0125] Table 1 is an example table of rock-soil body parameters

[0126]

[0127] Table 2 is an example table of support parameters

[0128]

[0129] The 500 groups of slope samples obtained according to the above steps are used as the training samples of the surrogate model, and the safety factor surrogate model is obtained after training. The 10000 groups of labeled samples not participating in the training are input into the surrogate model, and the error between the safety factor output by the surrogate model and the safety factor calculated by numerical simulation (as the true value) is compared to evaluate the performance of the model. The results show that (as shown in Figure 9As shown), the model performs well on the validation set, with a variance of 0.00961 and a mean absolute deviation of 0.07739 for the predicted safety factor. 100 different combinations of the aforementioned support parameters to be evaluated are generated, and each combination is subjected to a random field simulation based on geological parameters, resulting in 500 sets of second shear strength distribution samples. Based on the trained safety factor proxy model, each combination of support parameters to be evaluated and its corresponding second shear strength distribution sample are simultaneously input, yielding 500 second safety factors for each of the 500 sets of second shear strength distribution samples. The reliability of each combination of support parameters to be evaluated is then calculated based on these second safety factors. Specifically, if m out of M second safety factors are less than a preset threshold, then when M is sufficiently large, according to the law of large numbers, the failure probability P and reliability index of the slope are:

[0130] , ;

[0131] in, For reliability, The failure probability for each combination of support parameters to be evaluated. Second safety factor Less than the preset threshold The probability, The number of second safety factors that are less than a preset threshold. This represents the number of the second safety factor. In this embodiment, for better illustration, a preset threshold of 1 is set, and the slope reliability corresponding to the above 100 combinations of support parameters to be evaluated is plotted as follows: Figure 10 As shown, the slope reliability is highest under parameter combination 53.

[0132] It is understood that the number or combination of the above-mentioned samples should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the quantities mentioned above are intended to present the relevant concepts in a specific manner for ease of understanding.

[0133] On the other hand, embodiments of the present invention also provide an electronic device, which includes a processor, a network interface, and a memory, wherein the processor, the network interface, and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the above-described slope support reliability assessment method.

[0134] In an embodiment of the present application, the processor can be an integrated circuit chip with the processing capability. The processor can be a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components.

[0135] The disclosed methods, steps, and logic block diagrams in the embodiments of the present application can be implemented or executed by using a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof. The steps of the methods disclosed in the embodiments of the present application can be directly embodied as one or more hardware coding processes executed by the processor, or embodied as software modules executed by the coding processor. The software modules can be located in the memory, the flash memory, the Read Only Memory (ROM), the Programmable Read Only Memory (PROM), the Electrically Programmable Read Only Memory (EPROM), the Electrically Erasable Programmable Read Only Memory (EEPROM), the registers, or other forms of the storage medium in the art. The processor reads information in the storage medium and combines the hardware to implement the steps of the above methods.

[0136] The storage medium can be a memory, for example, can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories.

[0137] The non-volatile memory can be a Read Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically EPROM (EEPROM), or a flash memory.

[0138] The volatile memory can be a Random Access Memory (RAM), which is used as an external cache. By way of example, and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Sync Link DRAM (SLDRAM), and Direct Rambus RAM (DRRAM).

[0139] The storage media described in the embodiments of the present application is intended to include, but not limited to, these and any other suitable types of memory.

[0140] It should be understood that the system disclosed by the present application can be implemented in other ways. For example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the communication connection between the modules can be indirect coupling or communication connection through some interfaces, servers or units, which can be electrical or other forms.

[0141] In addition, each functional module in each embodiment of the present application can be integrated in one processing unit, or each module can exist physically separately, or two or more modules can be integrated in one processing unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0142] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application, essentially or in other words, the part that contributes to the prior art or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in various embodiments of the present application. The aforementioned storage medium includes: a U disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0143] The above merely describes preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for assessing the reliability of slope protection, characterized in that, The method includes: Random field simulation is performed based on the geological parameters of the target slope to generate several first shear strength distribution samples. Based on the range of the support parameters of the target slope, several support structure combination samples are generated by random sampling. The plurality of first shear strength distribution samples and the plurality of support structure combination samples are randomly combined to generate a plurality of slope samples, and the first safety factor of each slope sample is calculated by numerical simulation. Extract the geological body parameter distribution image, the average geological body parameter, and the support structure design parameters from the slope sample. Generate a dataset based on the geological body parameter distribution image, the average geological body parameter, and the support structure design parameters labeled according to the first safety factor. Train the proxy model based on the dataset to obtain a well-trained proxy model with a high security coefficient; Based on the combination of support parameters to be evaluated for the target slope, a random field simulation based on geological parameters is performed for each combination of support parameters to be evaluated to generate several second shear strength distribution samples. Each combination of support parameters to be evaluated and its corresponding number of second shear strength distribution samples are input into the trained safety factor proxy model to obtain a number of second safety factors for each combination of support parameters to be evaluated. The reliability of each support parameter combination to be evaluated is calculated based on the aforementioned second safety factors to obtain the evaluation result; The geological parameters include the average cohesion, the coefficient of variation of cohesion, the cohesion fluctuation distance, the average friction angle, the coefficient of variation of friction angle, and the friction angle fluctuation distance. The geological parameter distribution images include a cohesion distribution image and a friction angle distribution image. The average geological parameters are the average cohesion and the average friction angle in the first shear strength distribution sample. Extracting the geological parameter distribution image from the slope sample, including: The cohesion and friction angle values ​​of each unit in the first shear strength distribution sample are exported to obtain the spatial distribution of cohesion and the spatial distribution of friction angle. The values ​​of each unit in the spatial distribution of cohesion and the spatial distribution of friction angle are converted into grayscale intensity to generate cohesion distribution image and friction angle distribution image.

2. The method for assessing the reliability of slope protection according to claim 1, characterized in that, The safety factor proxy model includes a convolutional neural network and a fully connected neural network; wherein, the convolutional neural network is used to extract features from the geological body parameter distribution image and output one-dimensional features; the one-dimensional features, the average value of the geological body parameters and the support structure design parameters are spliced ​​together to form a fused feature, which is then input into the fully connected neural network to output the safety factor.

3. The method for assessing the reliability of slope protection according to claim 2, characterized in that, The convolutional neural network includes a channel attention module, a spatial attention module, two standard convolutional layers, two max pooling layers, and an unfolded layer. The channel attention module includes a global pooling layer and a fully connected layer, and the spatial attention module includes a pooling layer and a two-dimensional convolutional layer.

4. The slope support reliability assessment method according to claim 3, characterized in that, The convolutional neural network is used to extract features from the geological body parameter distribution image and output one-dimensional features, specifically including: After the geological body parameter distribution image is input into the first standard convolutional layer, it is then input into the channel attention module and the spatial attention module for feature selective processing, which is used to increase the weight of key channels and key regions of the distribution image input features; The processed distribution image input features are fed into the first max pooling layer, the second standard convolutional layer, and the second max pooling layer, and then output as one-dimensional features after being unfolded.

5. The method for assessing the reliability of slope protection according to claim 1, characterized in that, The support parameters include the support parameters of anchor bolts or anchor cables and the parameters of anti-slide piles. The combination of support parameters to be evaluated includes the length of anchor bolts or anchor cables, the incident angle of anchor bolts or anchor cables, the prestress of anchor bolts or anchor cables, the spacing of anchor bolts or anchor cables, and the length of anti-slide piles.

6. The method for assessing the reliability of slope protection according to claim 1, characterized in that, The reliability of each combination of support parameters to be evaluated is calculated based on the aforementioned second safety factors, specifically including: The failure probability is obtained by calculating the proportion of the second safety factors that are less than a preset threshold among the plurality of second safety factors, using the following formula: ； The reliability of each support parameter combination to be evaluated is calculated based on the failure probability, using the following formula: ； in, For reliability, The failure probability for each combination of support parameters to be evaluated. Second safety factor Less than the preset threshold The probability, The number of second safety factors that are less than a preset threshold. This represents the number of the second safety factor.

7. An electronic device, characterized in that, It includes at least one processor and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to perform a slope support reliability assessment method as described in any one of claims 1 to 6.

Images