Online intelligent detection method and system for ecological slope support

Abstract

The invention relates to an online intelligent detection method and system for ecological slope support, and the method comprises the steps: collecting ecological parameters and structural parameters, carrying out the causal inspection of the two types of parameters, and distinguishing a causal index set and an independent index; then, feature vectors are constructed by adopting different processing strategies and causal index sets, and a first result is obtained by using a first risk prediction model after deep learning model fusion; the independent indexes are subjected to correlation analysis and classification, and a second result is obtained through a second risk prediction model. And finally, combining the two results to obtain a total risk prediction result of the target slope. The system is used for executing the method. The stability risk of the slope is predicted by combining the causal relationship between the ecological parameters and the structural parameters of the ecological slope and the relevance between the indexes, so that the causal relationship and the relevance between the indexes can be fully considered during risk prediction, and the accuracy of stability risk prediction can be improved; and thus, the accuracy and effectiveness of online intelligent detection of the ecological slope are improved.

Description

Technical Field

[0001] This application relates to the technical field of slope support detection, and particularly relates to an online intelligent detection method and system for ecological slope support. Background Art

[0002] Traditional support technologies include gravity retaining walls, counterfort retaining walls, cantilever supports, etc. The slope is reinforced by setting support structures such as retaining plates, anchor cables, and anchor bolts on the slope. Ecological slope support refers to the combination of ecological planting technology on the basis of traditional support. Greening plants are planted on the slope surface. On the one hand, the greening plants can beautify the slope surface, and on the other hand, they can also play a role in strengthening the slope and reducing soil erosion on the slope. Compared with single traditional support, ecological slope support has better support effect.

[0003] The detection of slope support is an important task in the slope operation and maintenance process. By real-time monitoring parameters such as the stress of anchor bolts / anchor cables, the bending moment, shear force of the retaining plate, and the displacement of the support structure, the real-time state of the slope support structure can be effectively detected, and then the slope stability can be detected.

[0004] Most traditional slope support detections only focus on the monitoring of the physical and mechanical properties of the support structure. However, for ecological slopes, the traditional support structure is only a part of the ecological slope system. Ecological parameters of the ecological slope system, such as vegetation coverage, soil humidity, soil density, etc., are equally important for the ecological slope system. More importantly, the causality between ecological parameters and the physical and mechanical properties of the support structure is of great significance for the real-time detection and risk prediction of the ecological slope system. The existing slope support detection systems do not pay attention to this characteristic, resulting in low detection accuracy for ecological slope support, which needs to be improved. Summary of the Invention

[0005] In order to solve the above problems existing in the prior art, the purpose of this application is to provide an online intelligent detection method and system for ecological slope support. This application combines the causal relationship between the ecological parameters and structural parameters of the ecological slope and the correlation between indicators to predict the stability risk of the slope, so that the causal relationship and correlation between indicators can be fully considered during risk prediction, which is beneficial to improving the accuracy of stability risk prediction, and further improving the accuracy and effectiveness of online intelligent detection of ecological slopes.

[0006] An online intelligent detection method for ecological slope support described in this application includes the following steps:

[0007] S1. Collect the ecological parameters and structural parameters of the target slope. The ecological parameters include several ecological indicators related to the ecological state of the target slope, and the structural parameters include several structural indicators related to the stability of the support structure of the target slope;

[0008] S2. Conduct a causal test on the ecological parameters and the structural parameters, and record the ecological indicators and structural indicators with a causal relationship as a causal indicator set. The causal indicator set also includes the lag order; record the ecological indicators and structural indicators without a causal relationship as independent indicators;

[0009] S3. Process the causal indicator set and the independent indicators respectively using a first processing strategy and a second processing strategy;

[0010] The first processing strategy includes:

[0011] Based on the causal indicator set, construct a feature vector corresponding to the causal indicator set;

[0012] Input the feature vector into a pre-configured deep learning model for multi-source data fusion to obtain the fused feature data;

[0013] Input the fused feature data into a pre-configured first risk prediction model to obtain a first risk prediction result for the target slope;

[0014] The second processing strategy includes:

[0015] Conduct a correlation analysis on each of the independent indicators, classify the independent indicators with a correlation into one category to obtain several different types of independent indicator sets, and input the independent indicator sets into a pre-configured second risk prediction model to obtain a second risk prediction result for the target slope;

[0016] S4. Combine the first risk prediction result and the second risk prediction result to obtain an overall risk prediction result for the target slope.

[0017] Preferably, in step S1, the ecological indicators include vegetation coverage, average vegetation height, normalized difference vegetation index, soil moisture, soil density, surface runoff direction, and surface runoff velocity, and the structural indicators include anchor cable stress, shear force of retaining plate, displacement of support structure, damage rate of slope protection net, and crack size of lattice beam.

[0018] Preferably, step S2 specifically includes:

[0019] Conduct data preprocessing on the ecological parameters and the structural parameters;

[0020] Determine that the order range of the lag order is [ord min , ord max , set the lag order of the structural indicator Y itself as p, and the lag order of the ecological indicator X itself as q. Both p and q belong to the order range [ord min , ordmax ;

[0021] For each pair of ecological index X and structural index Y, an unrestricted model and a restricted model are constructed respectively:

[0022] The unrestricted model is expressed as:

[0023]

[0024] The restricted model is expressed as:

[0025]

[0026] where t represents time, Y t represents the value of the structural index at the current moment, Y t-i represents the lag value of the structural index Y in the past i time steps, a i represents the corresponding coefficient, X t-j represents the lag value of the ecological index X in the past j time steps, β i represents the corresponding coefficient, ∈ i represents the random error term;

[0027] Perform regression analysis on the unrestricted model and the restricted model respectively to obtain the residual sum of squares RSS U of the unrestricted model, and the residual sum of squares RSS R of the restricted model;

[0028] Calculate the F statistic according to the following formula:

[0029]

[0030] where n represents the sample size, that is, the length of the time series data;

[0031] At the given significance level, obtain the corresponding critical value by looking up the F statistic distribution table. If the calculated F statistic is greater than the critical value, it is determined that the ecological index X is the Granger cause of the ecological index Y, and record the corresponding lag order ord X-Y , construct the causal index set expressed as (X, Y, ord X-Y ), if the calculated F statistic is not greater than the critical value, it is determined that there is no causal relationship between the ecological index X and the ecological index Y. If an ecological index or a structural index has no causal relationship with any other index, mark the ecological index or the structural index as an independent index.

[0032] Preferably, in step S3, the training process of the deep learning model includes the following steps:

[0033] Construct an ordered cascaded CNN module and LSTM module;

[0034] Collect multiple sets of time series data on ecological indicators and structural indicators of the slope;

[0035] Use the Granger causality analysis method to conduct causality analysis on the ecological indicators and structural indicators in the same group, and screen out the ecological indicators and structural indicators with causal relationships, which are recorded as the sample causal data set;

[0036] Divide the sample causal data set into a training set, a validation set, and a test set, and based on the sample causal data set, construct a feature vector corresponding to the sample causal data set;

[0037] Input the feature vector corresponding to the training set into the CNN module, and perform convolution operations and pooling operations in sequence to extract feature data;

[0038] Input the obtained feature data into the LSTM module, and the LSTM module fuses the extracted feature data and outputs a fusion feature containing the causal relationship and lag order between indicators;

[0039] Update the parameters of the deep learning model through the loss function and the backpropagation algorithm;

[0040] Repeat the training steps until the model converges or the number of iterations is satisfied, and the deep learning model is obtained.

[0041] Preferably, in step S3, the training process of the first risk prediction model includes the following steps:

[0042] Construct a multi-layer fully connected neural network including an input layer, a hidden layer, and an output layer, and label the corresponding slope risk level for each pair of ecological indicators and structural indicators in the sample causal data set, which is recorded as the true risk level;

[0043] Input the fusion feature output by the LSTM module into the input layer of the multi-layer fully connected neural network;

[0044] Then sequentially transfer the fusion feature received by the input layer to the multi-layer hidden layers. In the hidden layer, perform non-linear transformation through the activation function and calculate the activation value. Integrate and transform the fusion feature through the multi-layer hidden layers, and sequentially calculate the activation values of the multi-layer hidden layers;

[0045] Transfer the obtained activation values of the multi-layer hidden layers to the output layer. The output layer combines the activation values of the multi-layer hidden layers and outputs the predicted slope risk level, which is recorded as the predicted risk level;

[0046] Based on the difference between the true risk level and the predicted risk level, update the parameters of the multi-layer fully connected neural network through a loss function and a backpropagation algorithm;

[0047] Repeat the training step until the network converges or the number of iterations is satisfied, and the multi-layer fully connected neural network is obtained.

[0048] Preferably, in step S3, the Pearson correlation coefficient method, the K-Means clustering algorithm or the principal component analysis method is used to perform correlation analysis and classification on each of the independent indicators, and the second risk prediction model is a decision tree model;

[0049] Step S4 includes the following steps:

[0050] Perform numerical normalization on the first risk prediction result and the second risk prediction result, and perform a summation calculation on the results of the numerical normalization processing by means of weighted summation. The calculated result is the total risk prediction result.

[0051] Preferably, the method further includes: presetting that the initial acquisition frequencies of the ecological parameters and the structural parameters are both the first frequency; judging whether the target slope has a stability risk according to the total risk prediction result;

[0052] In response to the target slope having a stability risk, obtain the causal relationship and the lag order ord of the ecological indicators and the structural indicators in the causal indicator set X-Y , increase the acquisition frequency of the ecological indicators belonging to the causal indicator set to the second frequency, and combine the lag order ord of the ecological indicators and the structural indicators X-Y , and after the first time interval Time, increase the acquisition frequency of the corresponding structural indicators to the third frequency, where the third frequency is greater than the second frequency, and the first time interval Time ∈ [75% * ord X-Y , ord X-Y .

[0053] An online intelligent detection system for ecological slope support of the present application includes:

[0054] An acquisition module, which is used to acquire the ecological parameters and the structural parameters of the target slope. The ecological parameters include a number of ecological indicators related to the ecological state of the target slope, and the structural parameters include a number of structural indicators related to the stability of the support structure of the target slope;

[0055] A causal test module, which is used to perform a causal test on the ecological parameters and the structural parameters, record the ecological indicators and the structural indicators with a causal relationship as a causal indicator set, and the causal indicator set also includes a lag order; record the ecological indicators and the structural indicators without a causal relationship as independent indicators;

[0056] A processing module, which is used to process the causal index set and the independent index respectively by using a first processing strategy and a second processing strategy;

[0057] The first processing strategy includes:

[0058] Based on the causal index set, construct a feature vector corresponding to the causal index set;

[0059] Input the feature vector into a pre-configured deep learning model for multi-source data fusion to obtain fused feature data;

[0060] Input the fused feature data into a pre-configured first risk prediction model to obtain a first risk prediction result for the target slope;

[0061] The second processing strategy includes:

[0062] Conduct a correlation analysis on each of the independent indicators, classify the independent indicators with correlations into one category to obtain several different types of independent indicator sets, input the independent indicator sets into a pre-configured second risk prediction model to obtain a second risk prediction result for the target slope;

[0063] A combination prediction module, which is used to combine the first risk prediction result and the second risk prediction result to obtain a total risk prediction result for the target slope.

[0064] A computer device of the present application includes a processor and a memory connected by signals. At least one instruction or at least one program segment is stored in the memory. When the at least one instruction or the at least one program segment is loaded by the processor, it executes the ecological slope support online intelligent detection method as described above.

[0065] A computer-readable storage medium of the present application stores at least one instruction or at least one program segment thereon. When the at least one instruction or the at least one program segment is loaded by a processor, it executes the ecological slope support online intelligent detection method as described above.

[0066] An online intelligent detection method and system for ecological slope support according to the present application has the following advantages. By performing causal analysis on the ecological parameters and structural parameters of the target slope, the causal relationship between the ecological parameters and structural parameters is used as the input information for the first risk prediction model. By performing correlation analysis between independent indicators, the correlation analysis result is used as the input information for the second risk prediction model. The first risk prediction result and the second risk prediction result are respectively obtained, and the total risk prediction result is obtained by combination. Thus, when predicting the slope stability risk, the causal relationship and hysteresis between the ecological parameters and structural parameters are comprehensively considered, and the correlation between each independent indicator is considered, making the risk prediction of the ecological slope more comprehensive and more suitable for the actual situation of the ecological slope, improving the accuracy of the risk prediction of the ecological slope, and further improving the accuracy and effectiveness of the online intelligent detection of the ecological slope. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 is a flowchart of the steps of an online intelligent detection method for ecological slope support according to the present application;

[0068] Figure 2 is a schematic structural diagram of the computer device in this embodiment.

[0069] Description of reference numerals: 101 - processor, 102 - memory. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0070] As Figure 1 shown, an online intelligent detection method for ecological slope support according to the present application includes the following steps:

[0071] S1. Collect the ecological parameters and structural parameters of the target slope. The ecological parameters include several ecological indicators related to the ecological state of the target slope, and the structural parameters include several structural indicators related to the stability of the support structure of the target slope;

[0072] Further, the ecological indicators include vegetation coverage, average vegetation height, normalized difference vegetation index, soil moisture, soil density, surface runoff direction, and surface runoff velocity, and the structural indicators include anchor cable / bolt stress, retaining plate shear force, support structure displacement, slope protection net damage rate, and lattice beam crack size.

[0073] Specifically, the vegetation coverage is calculated by vegetation coverage area / slope total area. The vegetation coverage area can be obtained by acquiring an image of the slope (such as an unmanned aerial vehicle aerial image). The image of the vegetation-covered area in the image is significantly different from the image of the exposed area of the slope. Thus, the area of the vegetation-covered area can be obtained through an image analysis algorithm.

[0074] The average vegetation height is mainly obtained through manual measurement. During the operation and maintenance of the slope, the operation and maintenance personnel will conduct regular inspections on the slope. During the inspection, the growth height of the plants can be manually measured, and the average value can be obtained by measuring the plant heights in different areas of the slope, which is the average vegetation height during this period.

[0075] The Normalized Difference Vegetation Index (NDVI) is obtained by means of unmanned aerial vehicle (UAV) remote sensing. After obtaining multiple UAV image data, they are stitched, orthorectified and other processed to ensure the spatial accuracy of the images. Then, the red light band and near-infrared band data in the images are identified and extracted, and then substituted into the NDVI for calculation to generate the NDVI results reflecting the vegetation conditions of the slope.

[0076] The soil moisture can be collected by humidity sensors placed in the slope soil.

[0077] The soil density mainly depends on manual collection during inspections. For example, the soil density of soil samples can be calculated by the core cutter method, wax sealing method, etc. And because the change frequency of the soil density is relatively low, the second collection and measurement can be carried out at a relatively long interval after the single collection of the soil density.

[0078] The flow direction of surface runoff can be obtained through manual observation input, and the flow velocity of surface runoff can be collected by flow velocity sensors placed in the fluid.

[0079] The stress of anchor bolts / cables and the shear force of retaining plates can be measured by arranging force sensors at the monitoring positions, such as the shear stress sensors commonly used in engineering monitoring.

[0080] The displacement of the support structure can be monitored by installing displacement sensors at the monitoring positions of the support structure.

[0081] The damage rate of the slope protection net can be obtained by using the aerial images of the aforementioned slope. The damaged areas in the slope protection net are identified through image recognition algorithms, and the damage rate of the slope protection net can be calculated by extracting the area of the damaged areas and dividing it by the total area of the slope.

[0082] The crack size of the lattice beam is calculated by manually taking surface images of the lattice beam and using existing image crack calculation methods.

[0083] In summary, various ecological indicators and structural indicators required for the monitoring of ecological slopes can be obtained.

[0084] S2. Conduct a causality test on the ecological parameters and the structural parameters, and record the ecological indicators and structural indicators with causal relationships as the causal indicator set, and the causal indicator set also includes the lag order; record the ecological indicators and structural indicators without causal relationships as independent indicators;

[0085] In step S2, it specifically includes:

[0086] Perform data preprocessing on the ecological parameters and the structural parameters; specifically, each of the aforementioned ecological parameters and structural parameters is time-series data with a certain time length, ensuring that each parameter is within the same acquisition period. For parameters with a relatively low change frequency, such as soil density, the parameter is considered unchanged within the acquisition period.

[0087] Since different ecological indicators and structural indicators may have different dimensions and value ranges, in order to eliminate the influence of dimensional differences on the causal test results, it is necessary to standardize the data of each indicator. In this embodiment, the maximum-minimum normalization method is used to normalize each parameter, so that the values of all indicators fall within the interval of [0, 1], which is convenient for subsequent calculations.

[0088] Determine that the order range of the lag order is [ord min , ord max , for example, [1 week, 52 weeks]. Set the lag order of the structural indicator Y itself as p, and the lag order of the ecological indicator X itself as q. Both p and q belong to the order range [ord min , ord max ;

[0089] For each pair of ecological indicator X and structural indicator Y, construct an unrestricted model and a restricted model respectively:

[0090] The unrestricted model is expressed as:

[0091]

[0092] The restricted model (assuming that X is not the Granger cause of Y) is expressed as:

[0093]

[0094] where t represents time, Y t represents the value of the structural indicator at the current moment, Y t-i represents the lag value of the structural indicator Y at the past i time steps, a i represents the corresponding coefficient, X t-j represents the lag value of the ecological indicator X at the past j time steps, β i represents the corresponding coefficient, ∈ i represents the random error term;

[0095] Perform regression analysis on the unrestricted model and the restricted model respectively to obtain the residual sum of squares RSS U of the unrestricted model, and the residual sum of squares RSS R of the restricted model;

[0096] Calculate the F statistic according to the following formula:

[0097]

[0098] Among them, n represents the number of samples, that is, the length of the time series data;

[0099] At a given significance level, such as the commonly used 5% significance level, the corresponding critical value is obtained by looking up the F-statistic distribution table. If the calculated F-statistic is greater than the critical value, the null hypothesis (assuming that X is not the Granger cause of Y) is rejected, and it is determined that the ecological index X is the Granger cause of the ecological index Y, and the corresponding lag order q at this time is recorded. This lag order q is the lag order ord between the ecological index X and the corresponding structural index Y X-Y , indicating that when there is an obvious change in the ecological index X, it will have an impact on the structural index Y after ord X-Y days. The constructed causal index set is expressed as (X, Y, ord X-Y ).

[0100] For example, perform a causal analysis on the soil moisture in the above ecological index and the bolt stress in the structural index.

[0101] Collect real-time soil moisture and bolt stress. For example, through a humidity sensor, the soil moisture value in a certain monitoring period is obtained as 25%. Assume that the minimum value of the soil moisture is 10% and the maximum value is 40%. Use the maximum-minimum normalization method for normalization:

[0102]

[0103] Similarly, normalize the bolt stress by the maximum-minimum normalization method.

[0104] Let the soil moisture be X and the bolt stress be Y, and construct the above unrestricted model and restricted model. Preset the lag order values p = 3, q = 2. Calculate the sum of squared residuals of the two models and calculate the F-statistic for critical value judgment according to the above steps to judge whether there is a Granger causal relationship between the soil moisture X and the bolt stress Y. Among them, through the information criterion method, such as the Akaike information criterion (AIC) or the Schwarz information criterion (SIC), calculate the information criterion values under different lag orders one by one. Based on the obtained information criterion values, determine the optimal lag order. Generally speaking, select the lag order with the smallest information criterion value as the optimal lag order. For example, the above lag order values p = 3, q = 2.

[0105] Through the above steps, it can be determined that the soil moisture is the Granger cause of the bolt stress, and the lag order is 2 weeks. In this embodiment, only the lag effect of the ecological index on the structural index is analyzed, so p = 3 is discarded.

[0106] If the calculated F statistic is not greater than the critical value, it is determined that there is no causal relationship between the ecological indicator X and the ecological indicator Y. If an ecological indicator or a structural indicator has no causal relationship with any other indicator, then mark this ecological indicator or structural indicator as an independent indicator.

[0107] S3. Process the causal indicator set and the independent indicator respectively using a first processing strategy and a second processing strategy;

[0108] The first processing strategy includes:

[0109] Based on the causal indicator set, construct a feature vector corresponding to the causal indicator set;

[0110] Input the feature vector into a pre-configured deep learning model for multi-source data fusion to obtain fused feature data;

[0111] Furthermore, in step S3, the training process of the deep learning model includes the following steps:

[0112] 1. Construct a cascaded CNN module and LSTM module in sequence.

[0113] CNN module: The Convolutional Neural Network (CNN) is mainly used to extract local features in data. When constructing the CNN module, it is necessary to determine the number and parameters of the convolutional layer and the pooling layer.

[0114] Convolutional layer: The convolutional layer contains multiple convolutional kernels. Each convolutional kernel performs a sliding convolutional operation on the input data to extract different features. The size, number, and stride of the convolutional kernel are important parameters that need to be set. For example, the convolutional kernel size can be set to 3x3 or 5x5, and the number of convolutional kernels can be adjusted according to the complexity of the data and the requirements of feature extraction.

[0115] Pooling layer: The pooling layer is used to reduce the dimension of the feature map, reduce the amount of calculation, and enhance the robustness of the model. Common pooling operations include max pooling and average pooling. The size and stride of the pooling window are also parameters that need to be set.

[0116] LSTM module: The Long Short-Term Memory (LSTM) is a special recurrent neural network that can handle long-term dependencies in sequence data. When constructing the LSTM module, it is necessary to determine the number of LSTM units, that is, the number of neurons in the hidden layer. The LSTM unit controls the flow of information through a gating mechanism (input gate, forget gate, and output gate), thereby effectively capturing the temporal dependencies in the sequence data.

[0117] Cascade mode: The output of the CNN module is used as the input of the LSTM module to achieve sequential cascading of the two modules. The CNN module first extracts local features from the input data, and then the LSTM module processes and fuses the extracted features in the time series.

[0118] 2. Collect multiple groups of time series data on ecological indicators and structural indicators of the slope.

[0119] Ecological indicators: Include vegetation coverage, average vegetation height, normalized difference vegetation index, soil moisture, soil density, surface runoff direction, surface runoff velocity, etc.

[0120] Structural indicators: Include anchor cable stress, retaining plate shear force, support structure displacement, slope protection net damage rate, lattice beam crack size, etc.

[0121] Time series data: Collect continuous observation values of these indicators over a period of time to form time series data. The time interval of the data can be set according to actual needs and the accuracy of the monitoring equipment. For example, data is collected once a day, once a week, or once a month.

[0122] 3. Use Granger causality analysis method to conduct causality analysis on the ecological indicators and structural indicators of the same group, and screen out the ecological indicators and structural indicators with causal relationships, which are recorded as the sample causal data set.

[0123] The method of determining Granger causality analysis and lag order can be understood with reference to the above description and will not be elaborated here.

[0124] Sample causal data set: Screen out the ecological indicators and structural indicators with causal relationships, and combine their corresponding time series data into the sample causal data set.

[0125] 4. Divide the sample causal data set into a training set, a validation set, and a test set, and based on the sample causal data set, construct the feature vectors corresponding to the sample causal data set.

[0126] Data set division: Divide the sample causal data set into a training set, a validation set, and a test set according to a certain ratio. Common division ratios are 70%-15%-15% or 80%-10%-10%. The training set is used for parameter learning of the model, the validation set is used to adjust the hyperparameters of the model (such as learning rate, batch size, etc.), and the test set is used to evaluate the final performance of the model.

[0127] Feature vector construction: Construct the corresponding feature vectors according to the ecological indicators and structural indicators in the sample causal data set.

[0128] 5. Input the feature vectors corresponding to the training set into the CNN module, and perform convolution operation and pooling operation in sequence to extract feature data.

[0129] Convolution operation: The feature vectors corresponding to the training set are input into the convolutional layer of the CNN module. Each convolutional kernel performs a sliding convolution operation on the feature vectors to generate corresponding feature maps. The convolution operation can extract local features in the feature vectors, and different convolutional kernels can extract different types of features.

[0130] Pooling operation: The feature maps output by the convolutional layer undergo a pooling operation through the pooling layer to reduce the dimension of the feature maps. For example, the max pooling operation will select the maximum value in each pooling window as the output, thereby reducing the size of the feature maps while retaining important feature information.

[0131] Feature data extraction: After multiple convolution and pooling operations, the CNN module outputs the extracted feature data. These feature data contain local feature information in the sample causal dataset and provide input for the subsequent LSTM module.

[0132] 6. Input the obtained feature data into the LSTM module, and the LSTM module fuses the extracted feature data to output fusion features containing the causal relationship between indicators and the lag order.

[0133] LSTM cell processing: Input the feature data output by the CNN module into the LSTM module. The LSTM module processes the input feature data through a gating mechanism. The forget gate determines how much information from the previous cell state needs to be forgotten, the input gate determines how much information from the current input needs to be added to the cell state, and the output gate determines how much information from the current cell state needs to be output.

[0134] Feature fusion: During the process of processing sequence data, the LSTM module fuses the feature data at different time steps to capture the temporal dependence and causal relationship between indicators. By continuously updating the cell state and hidden state, the LSTM module can learn the long-term dependence information in the sequence data.

[0135] Fusion feature output: The LSTM module finally outputs fusion features containing the causal relationship between indicators and the lag order. These fusion features integrate the local features extracted by the CNN module and the temporal dependence relationship captured by the LSTM module, and can more comprehensively reflect the feature information of the sample causal dataset.

[0136] 7. Update the parameters of the deep learning model through the loss function and the backpropagation algorithm.

[0137] Loss function: Select an appropriate loss function to measure the difference between the model's prediction results and the true labels. For regression problems, commonly used loss functions include mean squared error (MSE), mean absolute error (MAE), etc.; for classification problems, commonly used loss functions include cross-entropy loss function, etc. Select an appropriate loss function according to specific task requirements.

[0138] Backpropagation algorithm: The backpropagation algorithm is a method for calculating gradients and updating model parameters. Calculate the loss value of the model according to the loss function, and then calculate the gradient of the loss function with respect to the model parameters through the backpropagation algorithm. According to the calculated gradients, use an optimization algorithm (such as stochastic gradient descent, Adam, etc.) to update the model parameters, so that the value of the loss function gradually decreases.

[0139] 8. Repeat the training steps until the model converges or reaches the number of iterations, and the deep learning model is obtained.

[0140] Iterative training: Repeat steps 5-7, continuously input the feature vectors of the training set into the model for training, and update the model parameters. Each iteration will gradually improve the performance of the model and gradually reduce the value of the loss function.

[0141] Convergence judgment: During the training process, judge whether the model converges by observing the value of the loss function and the performance metrics of the validation set (such as accuracy, mean squared error, etc.). If the value of the loss function no longer decreases significantly, or the performance metrics of the validation set no longer improve, it means that the model has converged.

[0142] Iteration number limit: In addition to judging whether the training ends according to the convergence situation, the maximum number of iterations can also be set. When the training reaches the maximum number of iterations, even if the model has not fully converged, the training stops. Finally, the trained deep learning model is obtained, and this model can use the causal relationship and lag order information between ecological indicators and structural indicators to perform slope-related prediction and analysis.

[0143] Through the above steps, the deep learning model can be constructed. When actually performing slope detection, by inputting the feature vectors corresponding to the causal index set after Granger analysis, the deep learning model can capture the causal relationship in the data and generate fusion feature information containing the causal relationship and lag order between indicators.

[0144] Input the fused feature data into a pre-configured first risk prediction model to obtain a first risk prediction result for the target slope;

[0145] The first risk prediction model is constructed according to the following steps:

[0146] 1. Construct a multi-layer fully connected neural network and label the true risk levels.

[0147] Construct a multi-layer fully connected neural network.

[0148] The multi-layer fully connected neural network consists of an input layer, hidden layers, and an output layer.

[0149] Input layer: The number of its neurons depends on the dimension of the input data. In this embodiment, the input data is the fused features output by the LSTM module, so the number of neurons in the input layer is the same as the dimension of the fused features. For example, if the fused feature is a 3-dimensional vector, the number of neurons in the input layer is 3.

[0150] Hidden layers: Multiple hidden layers can be set. The number of hidden layers and the number of neurons in each hidden layer are hyperparameters that need to be adjusted. Generally, increasing the number of hidden layers and neurons can improve the expressive ability of the model, but it will also increase the risk of overfitting. Usually, it is possible to first try a smaller number of hidden layers (such as 2 - 3 layers) and a moderate number of neurons (such as 30 - 100 neurons per layer), and then adjust according to the training effect.

[0151] Output layer: The number of neurons in the output layer depends on the number of categories of the prediction task. Since here we are predicting the slope risk level, assuming the slope risk level is divided into 3 levels: high, medium, and low, the number of neurons in the output layer is 3.

[0152] For each pair of ecological indicators and structural indicators in the sample causal dataset, label the corresponding slope risk level according to professional knowledge, historical data, or expert experience, and record these labeled risk levels as the true risk levels. For example, when the soil humidity is too high and the bolt stress increases abnormally, according to experience, it is judged that the slope is at a high risk level at this time, and the sample corresponding to this pair of indicators is labeled as "high risk".

[0153] 2. Input the fused features into the input layer.

[0154] Input the fused features output by the LSTM module into the input layer of the multi-layer fully connected neural network. Each fused feature vector corresponds to a sample, and each neuron in the input layer receives the value of the corresponding dimension in the fused feature vector.

[0155] 3. Transmission and transformation of the fused features in the hidden layers.

[0156] Calculation process of the hidden layers.

[0157] The fused features received by the input layer are sequentially transmitted to the multi-layer hidden layers. In each hidden layer, the neurons perform the following calculations:

[0158] Weighted summation: Each neuron receives the outputs of all neurons in the previous layer, multiplies these outputs by the corresponding weights, then adds the products together, and then adds a bias term. The mathematical expression is:

[0159]

[0160] Among them, z j is the weighted sum of the current neuron, w ij is the weight from the i-th neuron in the previous layer to the current neuron, x i is the output of the i-th neuron in the previous layer, b j is the bias term of the current neuron, and n is the number of neurons in the previous layer.

[0161] Nonlinear transformation: The weighted sum z j will undergo a nonlinear transformation through an activation function to obtain the activation value a of the current neuron j . Common activation functions include ReLU (Rectified Linear Unit), Sigmoid function, Tanh function, etc.

[0162] The multi-layer hidden layers will gradually integrate and transform the fused features. Each hidden layer will perform calculations based on the output of the previous layer to extract higher-level feature representations. For example, the first hidden layer may extract some basic feature combinations, and subsequent hidden layers will further combine and abstract these basic features to form more representative features. In this way, the multi-layer hidden layers can deeply mine and transform the fused features.

[0163] 4. The output layer outputs the predicted risk level.

[0164] The activation values of the multi-layer hidden layers are passed to the output layer. Each neuron in the output layer will perform a weighted sum of the input activation values, and then convert the output into a probability distribution through a suitable activation function (such as the Softmax function). Suppose the output layer has 3 neurons corresponding to high, medium, and low risk levels respectively. The Softmax function will convert the output of each neuron into a probability value, and the sum of these 3 probability values is 1. The risk level corresponding to the neuron with the largest probability value is the predicted slope risk level, which is denoted as the predicted risk level.

[0165] 5. Update network parameters based on the difference.

[0166] Calculate the loss function:

[0167] To measure the difference between the true risk level and the predicted risk level, a suitable loss function needs to be selected. For multi-classification problems, the commonly used loss function is the cross-entropy loss function.

[0168] Update parameters using the backpropagation algorithm:

[0169] Based on the calculated loss function value, use the backpropagation algorithm to calculate the gradients of the loss function with respect to all the parameters (including weights and biases) in the neural network. The backpropagation algorithm starts from the output layer and calculates the gradients layer by layer through the chain rule.

[0170] 6. Repeat the training until convergence or the number of iterations is satisfied.

[0171] Repeat steps 2 - 5, continuously input the fused features into the network for training, and update the parameters of the network. During the training process, it is necessary to pay attention to the value of the loss function and the performance of the model on the validation set.

[0172] Convergence judgment: If the value of the loss function no longer decreases significantly after multiple iterations, or the performance of the model on the validation set (such as accuracy, F1 value, etc.) no longer improves, it indicates that the model has converged.

[0173] Iteration number limit: A maximum number of iterations can also be set. When the training reaches the maximum number of iterations, even if the model has not fully converged, the training is stopped. Finally, a trained multi - layer fully - connected neural network is obtained, and this model can be used to predict the risk level of the slope.

[0174] Through the above steps, the first risk prediction model can be constructed. When actually conducting slope detection, input the fused feature information extracted by the aforementioned deep - learning model into the first risk prediction model, and the risk level of the target slope can be predicted.

[0175] The second processing strategy includes:

[0176] Conduct a correlation analysis on each of the independent indicators, group the related independent indicators into one category to obtain several different types of independent indicator sets, and input the independent indicator sets into a pre - configured second risk prediction model to obtain a second risk prediction result for the target slope; among them, use the Pearson correlation coefficient method, K - Means clustering algorithm or principal component analysis method to conduct a correlation analysis and classification on each of the independent indicators, and the second risk prediction model is a decision - tree model.

[0177] Specifically, the decision - tree model selects information gain as the splitting criterion. For example, at the root node, calculate the information gain of each indicator. Assume that the current data set has 80 sample points, and the slope risks are divided into 3 levels: high, medium, and low, with the numbers being 20, 30, and 30 respectively. Taking the average vegetation height as an example, use it as the splitting feature, divide the data set into subsets according to different average vegetation height thresholds, and calculate the information entropy and information gain before and after the division. Assume that two subsets are obtained after the division. Subset 1 has 30 sample points, and the number of risk levels is 5, 15, and 10 respectively; Subset 2 has 50 sample points, and the number of risk levels is 15, 15, and 20 respectively.

[0178] Calculate the information entropy before partitioning and the conditional entropy after partitioning respectively, calculate the information gain, compare the information gains of each metric, select the metric with the largest information gain as the partitioning feature of the root node, and then recursively construct the sub-tree.

[0179] Continue to partition each subset until the stopping condition is met, such as the number of subset samples is less than 5 or all samples belong to the same category. To prevent overfitting, adopt the post-pruning method. After the decision tree is constructed, prune the tree according to the performance of the test set.

[0180] Input the data of the independent metric set of the training set into the decision tree model for training, and the model learns the relationship between different combinations of independent metrics and the slope risk. For example, when the average height of the vegetation is relatively high, the soil density is relatively large, the shear force of the retaining plate is moderate, and the damage rate of the slope protection net is relatively low, the model may judge that the slope is in a low-risk level (similarly, for each group of sample data, it is pre-annotated through expert experience).

[0181] Use the test set to evaluate the trained decision tree model, and calculate metrics such as accuracy, recall rate, and F1 value. Adjust the parameters of the decision tree model according to the evaluation results until the evaluation results meet the requirements, and then obtain the second risk prediction model.

[0182] S4. Combine the first risk prediction result and the second risk prediction result to obtain the total risk prediction result for the target slope.

[0183] Perform numerical normalization on the first risk prediction result and the second risk prediction result, and perform summation calculation on the results of numerical normalization by means of weighted summation. The calculated result is the total risk prediction result.

[0184] Comprehensively combine the prediction results of the two models by means of weighted summation. Among them, the first risk prediction result focuses on the causal relationship between ecological metrics and structural metrics, and the second risk prediction result focuses on the correlation between each independent metric. Therefore, the total risk prediction result can comprehensively consider the correlation between each metric, making the prediction result more accurate and comprehensive.

[0185] Furthermore, the method further includes: presetting that the initial acquisition frequencies of the ecological parameters and the structural parameters are both the first frequency, that is, the initial acquisition frequency; according to the total risk prediction result, judge whether the target slope has stability risks, for example, by numerically comparing the total risk prediction result with a preset risk threshold. If it is greater than the risk threshold, it is judged that the target slope has stability risks.

[0186] In response to the stability risk existing in the target slope, it is predicted that the target slope may be damaged in a certain period in the future. At this time, the causal relationship and the lag order ord between the ecological index and the structural index in the causal index set are obtained. X-Y The acquisition frequency of the ecological index belonging to the causal index set is increased to the second frequency, and the lag order ord between the ecological index and the structural index is combined. X-Y After the first time interval Time, the acquisition frequency of the corresponding structural index is increased to the third frequency, where the third frequency is greater than the second frequency, and the first time interval Time ∈ [75% * ord X-Y , ord X-Y . In this step, when it is predicted that there may be a risk of damage to the target slope, it is necessary to increase the acquisition frequency of the monitoring data to ensure that slope anomalies can be detected in time. Since there is a causal relationship between the ecological index and the structural index in the causal index set, for example, soil moisture and anchor stress, the soil moisture will induce changes in the anchor stress. Therefore, first increase the acquisition frequency of the soil moisture located upstream. For example, increase the acquisition frequency of the soil moisture sensor. In addition, using the lag order between the soil moisture and the anchor stress, the timing of increasing the acquisition frequency of the anchor stress can be set. For example, according to the lag order analysis, the change in soil moisture will induce a significant change in the anchor stress after 2 weeks. Therefore, control the acquisition frequency of the anchor stress to increase when it is close to 2 weeks after the current moment. And the anchor stress located downstream can more intuitively reflect the stability of the target slope. Therefore, make its acquisition frequency greater than the acquisition deviation of the soil moisture. Through the design of this step, the acquisition frequencies of each index can be reasonably adjusted according to the causal relationship and the lag order between the indexes, and the acquisition frequency can be reasonably controlled on the premise of timely adjusting the acquisition frequency according to the predicted risk situation to effectively monitor the target slope, and the unnecessary energy consumption of the detection system can be reduced.

[0187] In this application, through causal analysis of the ecological parameters and structural parameters of the target slope, the causal relationship between the ecological parameters and the structural parameters is used as the input information of the first risk prediction model. Through the correlation analysis between independent indexes, the correlation analysis result is used as the input information of the second risk prediction model. The first risk prediction result and the second risk prediction result are obtained respectively, and the total risk prediction result is obtained by combination. Thus, when predicting the stability risk of the slope, the causal relationship and hysteresis between the ecological parameters and the structural parameters are comprehensively considered, and the correlation between each independent index is considered, so that the risk prediction of the ecological slope is more comprehensive, more adapted to the actual situation of the ecological slope, can improve the accuracy of the risk prediction of the ecological slope, and then improve the accuracy and effectiveness of the online intelligent detection of the ecological slope.

[0188] Through causal analysis of the ecological parameters and structural parameters of the target slope, the causal relationship between the ecological parameters and the structural parameters is used as the input information of the first risk prediction model. Through correlation analysis between independent indicators, the correlation analysis results are used as the input information of the second risk prediction model. The first risk prediction result and the second risk prediction result are obtained respectively, and the total risk prediction result is obtained by combination. Thus, when predicting the slope stability risk, the causal relationship and hysteresis between the ecological parameters and the structural parameters are comprehensively considered, and the correlation between each independent indicator is considered, making the risk prediction of the ecological slope more comprehensive, more adaptable to the actual situation of the ecological slope, improving the accuracy of the risk prediction of the ecological slope, and further improving the accuracy and effectiveness of the online intelligent detection of the ecological slope.

[0189] The present application also provides an online intelligent detection system for ecological slope support, including:

[0190] An acquisition module, which is used to acquire the ecological parameters and structural parameters of the target slope. The ecological parameters include several ecological indicators related to the ecological state of the target slope, and the structural parameters include several structural indicators related to the stability of the support structure of the target slope;

[0191] A causal test module, which is used to conduct a causal test on the ecological parameters and the structural parameters, record the ecological indicators and structural indicators with a causal relationship as a causal indicator set, and the causal indicator set also includes the lag order; record the ecological indicators and structural indicators without a causal relationship as independent indicators;

[0192] A processing module, which is used to process the causal indicator set and the independent indicators respectively by using a first processing strategy and a second processing strategy;

[0193] The first processing strategy includes:

[0194] Based on the causal indicator set, construct a feature vector corresponding to the causal indicator set;

[0195] Input the feature vector into a pre-configured deep learning model for multi-source data fusion to obtain fused feature data;

[0196] Input the fused feature data into a pre-configured first risk prediction model to obtain a first risk prediction result regarding the target slope;

[0197] The second processing strategy includes:

[0198] Perform a correlation analysis on each of the independent indicators, classify the independent indicators with correlations into one category, obtain several different types of independent indicator sets, and input the independent indicator sets into a pre-configured second risk prediction model to obtain a second risk prediction result for the target slope;

[0199] A prediction combining module is configured to combine the first risk prediction result and the second risk prediction result to obtain a total risk prediction result for the target slope.

[0200] An ecological slope support online intelligent detection system of the present application belongs to the same inventive concept as the above method and can be understood by referring to the above description, and will not be elaborated herein.

[0201] As Figure 2 shown, this embodiment also provides a computer device, including a processor 101 and a memory 102 connected by a bus signal. At least one instruction or at least one program segment is stored in the memory 102. When the at least one instruction or the at least one program segment is loaded by the processor 101, it executes the ecological slope support online intelligent detection method as described above. The memory 102 can be used to store software programs and modules. The processor 101 executes various functional applications by running the software programs and modules stored in the memory 102. The memory 102 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device. In addition, the memory 102 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices. Correspondingly, the memory 102 can also include a memory controller to provide the processor 101 with access to the memory 102.

[0202] The method embodiments provided by the embodiments of the present application can be executed on a computer terminal, a server, or a similar computing device, that is, the above computer device can include a computer terminal, a server, or a similar computing device. The internal structure of the computer device can include, but is not limited to: a processor, a network interface, and a memory. Among them, the processor, network interface, and memory in the computer device can be connected by a bus or other means.

[0203] Among them, the processor 101 (or CPU (Central Processing Unit)) is the computing core and control core of the computer device. The network interface may optionally include a standard wired interface, a wireless interface (such as WI-FI, mobile communication interface, etc.). The memory 102 (Memory) is the memory device in the computer device, used to store programs and data. It can be understood that the memory 102 here can be a high-speed RAM storage device, or a non-volatile memory device, such as at least one disk storage device; optionally, it can also be at least one storage device located far from the aforementioned processor 101. The memory 102 provides a storage space, and the operating system of the electronic device is stored in this storage space, which may include but is not limited to: Windows system (an operating system), Linux (an operating system), Android (a mobile operating system) system, IOS (a mobile operating system) system, etc., and this application does not make any limitations in this regard; and, one or more instructions suitable for being loaded and executed by the processor 101 are also stored in this storage space, and these instructions can be one or more computer programs (including program codes). In the embodiments of this specification, the processor 101 loads and executes one or more instructions stored in the memory 102 to implement the online intelligent detection method for ecological slope support described in the above method embodiments.

[0204] The embodiments of this application also provide a computer-readable storage medium, on which at least one instruction or at least one segment of program is stored, and when the at least one instruction or the at least one segment of program is loaded by the processor 101, it executes the online intelligent detection method for ecological slope support as described above. The above computer-readable storage medium carries one or more programs, and when the one or more programs are executed, the method according to the embodiments of this application is implemented.

[0205] According to the embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium. For example, it may include but is not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this application, the computer-readable storage medium can be any tangible medium that contains or stores a program, and this program can be used by or in combination with an instruction execution system, device, or component.

[0206] Those skilled in the art can make various corresponding changes and deformations according to the technical solutions and concepts described above, and all such changes and deformations should fall within the protection scope of the claims of this application.

Claims

1. An online intelligent detection method for ecological slope support, characterized in that: The following steps are involved: S1. Collecting ecological parameters and structural parameters of the target slope, wherein the ecological parameters include several ecological indicators related to the ecological state of the target slope, and the structural parameters include several structural indicators related to the stability of the support structure of the target slope; S2. Performing a causal test on the ecological parameters and the structural parameters, marking the ecological indicators and structural indicators with causal relationship as a causal indicator set, wherein the causal indicator set also includes a lag order; marking the ecological indicators and structural indicators without causal relationship as independent indicators; S3, processing the causal indicator set and the independent indicator using a first processing strategy and a second processing strategy respectively; The first processing strategy includes: Based on the causal indicator set, construct a feature vector corresponding to the causal indicator set; Inputting the feature vector into a preconfigured deep learning model to perform multi-source data fusion to obtain fused feature data; Inputting the fused feature data into a preconfigured first risk prediction model to obtain a first risk prediction result about the target slope; The second processing strategy includes: Performing correlation analysis on each of the independent indicators, classifying the correlated independent indicators into one category, obtaining a number of different types of independent indicator sets, inputting the independent indicator sets into a preconfigured second risk prediction model, and obtaining a second risk prediction result for the target slope; S4. Combining the first risk prediction result and the second risk prediction result, obtain a total risk prediction result about the target slope.

2. The online intelligent detection method for ecological slope support according to claim 1 is characterized in that: In step S1, the ecological indicators include vegetation coverage, average vegetation height, normalized vegetation index, soil moisture, soil density, surface runoff direction and surface runoff velocity, and the structural indicators include anchor rod / anchor cable stress, retaining board shear force, support structure displacement, slope protection net damage rate and lattice beam crack size.

3. The online intelligent detection method for ecological slope support according to claim 1 or 2 is characterized in that: Step S2 specifically includes: Performing data preprocessing on the ecological parameters and the structural parameters; The order range of the lag order is determined as [ord min ,ord max ], the lag order of the structural indicator Y itself is set to p, and the lag order of the ecological indicator X itself is set to q, and both p and q belong to the order range [ord min ,ord max ]; For each pair of ecological indicator X and structural indicator Y, an unrestricted model and a restricted model are constructed respectively: The unrestricted model is expressed as: The restriction model is expressed as: Where t represents time, Y t Indicates the structural index value at the current moment, Y t-i represents the lag value of the structural index Y in the past i time steps, a i represents the corresponding coefficient, X t-j represents the lagged value of ecological indicator X in the past j time steps, β i represents the corresponding coefficient, ∈ i represents the random error term; Regression analysis was performed on the unrestricted model and the restricted model respectively to obtain the residual sum of squares RSS of the unrestricted model. U , and the residual sum of squares RSS of the restricted model R ; The F statistic is calculated as follows: Among them, n represents the number of samples, that is, the length of time series data; At a given significance level, the corresponding critical value is obtained by looking up the F statistic distribution table. If the calculated F statistic is greater than the critical value, the ecological indicator X is judged to be the Granger cause of the ecological indicator Y, and the corresponding lag order ord is recorded at this time. X-Y , construct the causal indicator set represented as (X, Y, ord X-Y ), if the calculated F statistic is not greater than the critical value, it is judged that there is no causal relationship between the ecological indicator X and the ecological indicator Y; if there is no causal relationship between an ecological indicator or a structural indicator and any other indicator, the ecological indicator or the structural indicator is marked as an independent indicator.

4. The online intelligent detection method for ecological slope support according to claim 3 is characterized in that: In step S3, the training process of the deep learning model includes the following steps: Construct sequentially cascaded CNN modules and LSTM modules; Collect multiple sets of time series data on ecological and structural indicators of slopes; Granger causality analysis was used to conduct causal analysis on the same group of ecological indicators and structural indicators, and the ecological indicators and structural indicators with causal relationships were screened out and recorded as sample causal data sets; Dividing the sample causal data set into a training set, a validation set, and a test set, and constructing a feature vector corresponding to the sample causal data set based on the sample causal data set; Input the feature vector corresponding to the training set into the CNN module, perform convolution operation and pooling operation in sequence, and extract feature data; The obtained feature data is input into the LSTM module, the extracted feature data is fused through the LSTM module, and the fused features including the causal relationship and lag order between indicators are output; Updating the parameters of the deep learning model through a loss function and a back-propagation algorithm; The training steps are repeated until the model converges or the number of iterations is met, thereby obtaining the deep learning model.

5. The online intelligent detection method for ecological slope support according to claim 4 is characterized in that: In step S3, the training process of the first risk prediction model includes the following steps: A multi-layer fully connected neural network including an input layer, a hidden layer and an output layer is constructed, and the corresponding slope risk level is marked for each pair of ecological indicators and structural indicators in the sample causal data set, which is recorded as the real risk level; Inputting the fused features output by the LSTM module into the input layer of the multi-layer fully connected neural network; Then, the fusion features received by the input layer are sequentially transmitted to the multi-layer hidden layer, nonlinear transformation is performed in the hidden layer through the activation function and activation value is calculated, the fusion features are integrated and transformed through the multi-layer hidden layer, and activation values ​​of the multi-layer hidden layer are sequentially calculated; The activation values ​​of the obtained multi-layer hidden layers are transferred to the output layer, and the output layer combines the activation values ​​of the multi-layer hidden layers to output the predicted slope risk level, which is recorded as the predicted risk level; Based on the difference between the actual risk level and the predicted risk level, updating the parameters of the multi-layer fully connected neural network through a loss function and a back propagation algorithm; The training steps are repeated until the network converges or the number of iterations is met, thereby obtaining the multi-layer fully connected neural network.

6. The online intelligent detection method for ecological slope support according to claim 5 is characterized in that: In step S3, the Pearson correlation coefficient method, K-Means clustering algorithm or principal component analysis method are used to perform correlation analysis and classify the independent indicators, and the second risk prediction model is a decision tree model; Step S4 includes the following steps: The first risk prediction result and the second risk prediction result are numerically normalized, and the numerical normalization results are summed up using a weighted summation method, and the calculated result is the total risk prediction result.

7. The online intelligent detection method for ecological slope support according to claim 6 is characterized in that: Also includes: Presetting the initial acquisition frequencies of the ecological parameters and the structural parameters to be the first frequency; According to the total risk prediction result, determining whether the target slope has a stability risk; In response to the existence of stability risk of the target slope, the causal relationship and hysteresis order of the ecological index and the structural index in the causal index set are obtained. X-Y , the collection frequency of ecological indicators belonging to the causal indicator set is increased to the second frequency, and the lag order ord of the ecological indicators and structural indicators is combined X-Y , after the first time interval Time, the collection frequency of the corresponding structural indicators is increased to a third frequency, wherein the third frequency is greater than the second frequency, and the first time interval Time∈[75%*ord X-Y , ord X-Y ].

8. An online intelligent detection system for ecological slope support, characterized in that: include: A collection module, which is used to collect ecological parameters and structural parameters of the target slope, wherein the ecological parameters include several ecological indicators related to the ecological state of the target slope, and the structural parameters include several structural indicators related to the stability of the support structure of the target slope; A causal test module, which is used to perform a causal test on the ecological parameters and the structural parameters, marking the ecological indicators and structural indicators with causal relationship as a causal indicator set, and the causal indicator set also includes a lag order; marking the ecological indicators and structural indicators without causal relationship as independent indicators; A processing module, which is used to process the causal indicator set and the independent indicator using a first processing strategy and a second processing strategy respectively; The first processing strategy includes: Based on the causal indicator set, construct a feature vector corresponding to the causal indicator set; Inputting the feature vector into a preconfigured deep learning model to perform multi-source data fusion to obtain fused feature data; Inputting the fused feature data into a preconfigured first risk prediction model to obtain a first risk prediction result about the target slope; The second processing strategy includes: Performing correlation analysis on each of the independent indicators, classifying the correlated independent indicators into one category, obtaining a number of different types of independent indicator sets, inputting the independent indicator sets into a preconfigured second risk prediction model, and obtaining a second risk prediction result for the target slope; A combined prediction module is used to combine the first risk prediction result and the second risk prediction result to obtain a total risk prediction result about the target slope.

9. A computer device comprising a processor and a memory connected in a signal connection, characterized in that: The memory stores at least one instruction or at least one program, and when loaded by the processor, the at least one instruction or the at least one program executes the online intelligent detection method for ecological slope support as described in any one of claims 1-7.

10. A computer-readable storage medium having at least one instruction or at least one program stored thereon, characterized in that: When the at least one instruction or the at least one program is loaded by the processor, the online intelligent detection method for ecological slope support as described in any one of claims 1-7 is executed.

Images