A slope deformation prediction method based on multi-core TCN network under feature screening

Abstract

The embodiment of this specification provides a slope deformation prediction method of a multi-core TCN network under feature screening, the method comprising: determining the slope deformation variable and deformation influence factor corresponding to the target monitoring area of ​​the slope to be monitored, wherein the slope deformation variable is used to characterize the deformation variable of the target monitoring area at the current moment compared to the previous moment, and the deformation influence factor is used to characterize the factors affecting the slope deformation; based on the mutual information value of the slope deformation variable and the deformation influence factor, feature screening is performed on the deformation influence factor to obtain a strong correlation factor; based on the strong correlation factor, a corresponding time series feature is constructed, and based on the time series feature, the slope deformation variable prediction result of the target monitoring area at the next moment is determined. The degree of dependence between the deformation influence factor and the slope deformation variable can be fully considered in the prediction process, thereby improving the prediction accuracy.

Description

Technical Field

[0001] The embodiments of this specification relate to the technical field of geological disaster safety monitoring, and in particular to a slope deformation prediction method of a multi-core TCN network under feature screening. Background Art

[0002] Slope deformation monitoring is an important means to ensure the safety of engineering construction and geographical environment. Through slope deformation monitoring, the deformation of the slope can be discovered in time, providing a basis for decision-making in engineering construction, reducing engineering risks, and ensuring the safety and stability of the project.

[0003] In the existing technology, slope deformation prediction and analysis pays more attention to the slope displacement, and uses mathematical methods to explore the law of slope deformation change with time series, but ignores the influence of some natural factors on slope deformation, resulting in inaccurate slope deformation prediction and poor reference value of prediction results in practical applications.

[0004] Therefore, a solution is provided to solve the above problems. Summary of the invention

[0005] In view of this, an embodiment of this specification provides a slope deformation prediction method of a multi-core TCN network under feature screening. One or more embodiments of this specification also involve a computing device, a computer-readable storage medium and a computer program to solve the technical defects existing in the prior art.

[0006] According to a first aspect of an embodiment of this specification, a slope deformation prediction method of a multi-core TCN network under feature screening is provided, comprising:

[0007] Determine the slope deformation amount and deformation influencing factor corresponding to the target monitoring area of ​​the slope to be monitored, wherein the slope deformation amount is used to characterize the deformation amount of the target monitoring area at the current moment compared with the previous moment, and the deformation influencing factor is used to characterize the factors affecting the slope deformation;

[0008] Based on the mutual information value of the slope deformation amount and the deformation influencing factor, feature screening is performed on the deformation influencing factor to obtain a strong correlation factor;

[0009] A TCN network with multiple convolution kernels in parallel is constructed, the strong correlation factor is input into the TCN network with multiple convolution kernels in parallel to obtain corresponding time series features, and based on the time series features, the slope shape variable prediction result of the target monitoring area at the next moment is determined.

[0010] In one embodiment of the present specification, before determining the slope deformation amount and deformation influencing factor corresponding to the target monitoring area of ​​the slope to be monitored, the method further includes:

[0011] Obtaining slope point cloud data of the slope to be monitored, selecting at least one key point from the slope point cloud data, dividing the slope with the key point as the center and a preset length as the radius to obtain at least one monitoring area;

[0012] The target monitoring area is selected from the at least one monitoring area.

[0013] In one embodiment of the present specification, determining the slope deformation amount corresponding to the target monitoring area of ​​the slope to be monitored includes:

[0014] Determine the position information of the target key point corresponding to the target monitoring area, wherein the position information includes the first position information of the target key point at the current moment and the second position information at the previous moment;

[0015] The slope deformation amount of the target monitoring area is calculated according to the first position information and the second position information.

[0016] In one embodiment of the present specification, determining the deformation influence factor corresponding to the target monitoring area of ​​the slope to be monitored includes:

[0017] Acquire target terrain features and target environmental features corresponding to target key points in the target monitoring area;

[0018] Aggregating the neighborhood points and target key points in the target monitoring area, and performing feature extraction to obtain target feature information corresponding to the target monitoring area;

[0019] The deformation influence factor corresponding to the target key point is constructed according to the target terrain feature, the target environment feature and the target feature information.

[0020] In one embodiment of the present specification, based on the mutual information value of the slope deformation amount and the deformation influencing factor, feature screening is performed on the deformation influencing factor to obtain a strong correlation factor, including:

[0021] Calculating the first mutual information between the deformation influencing factor and the slope deformation amount;

[0022] A maximum mutual information model is established according to the first mutual information, and a strong correlation factor of the target monitoring area is determined based on an information growth rate criterion.

[0023] In one embodiment of the present specification, the step of establishing a maximum mutual information model according to the first mutual information and determining a strong correlation factor of the monitoring area based on an information growth rate criterion includes:

[0024] Selecting a deformation influencing factor corresponding to a maximum value in the first mutual information as a candidate correlation factor, and adding the candidate correlation factor to a variable candidate set of the target monitoring area;

[0025] Determine the second mutual information between the deformation influencing factor and the slope deformation amount under the condition of the variable candidate set;

[0026] Selecting a deformation influence factor corresponding to a maximum value in the second mutual information as a candidate correlation factor, and adding it to the variable candidate set;

[0027] The step of determining the second mutual information between the deformation influencing factor and the slope deformation amount under the variable candidate set condition is continued until the variable candidate set satisfies the number constraint condition, and the candidate correlation factors included in the variable candidate set are used as the strong correlation factors.

[0028] In one embodiment of the present specification, constructing a corresponding time series feature based on the strong correlation factor includes:

[0029] The strongly correlated factors corresponding to the target monitoring area are used as a strongly correlated multivariate set;

[0030] The strongly correlated multi-variable set is input into the multi-convolution kernel parallel TCN network, and multi-convolution kernel parallel processing and multi-dimensional feature extraction processing of at least two residual modules are performed to obtain the time series features corresponding to the strongly correlated factors.

[0031] In one embodiment of the present specification, based on the time series characteristics, determining the slope deformation prediction result of the target monitoring area at the next moment includes:

[0032] Inputting the time series features into a pre-established prediction model, wherein the prediction model includes a feature attention mechanism, a fully connected layer, and an activation function;

[0033] Processing the time series features through a feature attention mechanism to obtain weights corresponding to the strong correlation factors;

[0034] Normalizing the weights to obtain a feature attention distribution, and weighting the corresponding temporal features based on the feature attention distribution to obtain a correlation feature;

[0035] The associated features are input into the fully connected layer, and nonlinear transformation is performed through the activation function to obtain the slope shape variable prediction result of the target monitoring area at the next moment.

[0036] In one embodiment of the present specification, the associated features are input into the fully connected layer, and nonlinear transformation is performed through an activation function to obtain the slope shape variable prediction result of the target monitoring area at the next moment, including:

[0037] Input the associated features into the fully connected layer to obtain the hidden layer state value corresponding to the current moment;

[0038] The slope deformation prediction result of the target monitoring area at the next moment is calculated based on the hidden layer state value.

[0039] According to a second aspect of an embodiment of this specification, a computing device is provided, including:

[0040] Memory and processor;

[0041] The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening are implemented.

[0042] According to a third aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions. When the instructions are executed by a processor, the steps of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening are implemented.

[0043] An embodiment of the present specification can determine the deformation variable and the factors affecting the deformation variable in the monitoring area by determining the slope deformation sample set and the multivariate candidate set. By screening out a strongly correlated multivariate set from the multivariate candidate set based on the mutual information value of the slope deformation variable sample set and the multivariate candidate set, the degree of mutual influence between the slope deformation variable and the factors affecting the slope deformation can be taken into account, making the data source more reliable, predicting the slope deformation variable based on the strongly correlated multivariate set, improving the accuracy of the prediction result, and improving the reference value of the slope deformation prediction result in practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a flow chart of a slope deformation prediction method of a multi-core TCN network under feature screening provided by an embodiment of this specification;

[0045] Figure 2 It is a flowchart of feature optimization and prediction in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of this specification;

[0046] Figure 3 It is a schematic diagram of the variation of information gain with the number of factors in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of this specification;

[0047] Figure 4 It is a schematic diagram of a TCN network with multiple convolution kernels in parallel in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of the present specification;

[0048] Figure 5It is a schematic diagram of a feature attention mechanism in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of this specification;

[0049] Figure 6 It is a structural schematic diagram of a slope deformation prediction device provided by an embodiment of this specification;

[0050] Figure 7 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0051] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0052] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0053] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0054] First, the terms involved in one or more embodiments of this specification are explained.

[0055] Mutual Information: Mutual Information is an important concept in information theory, which measures the degree of mutual dependence between two random variables.

[0056] In this specification, a slope deformation prediction method of a multi-core TCN network under feature screening is provided. This specification also involves a slope deformation prediction device of a multi-core TCN network under feature screening, a computing device, and a computer-readable storage medium, which are described in detail one by one in the following embodiments.

[0057] See also Figure 1 , Figure 1 It is a flowchart of a slope deformation prediction method of a multi-core TCN network under feature screening provided by an embodiment of this specification, which specifically includes the following steps.

[0058] Step 102: Determine the slope deformation amount and deformation influencing factor corresponding to the target monitoring area of ​​the slope to be monitored, wherein the slope deformation amount is used to characterize the deformation amount of the target monitoring area at the current moment compared with the previous moment, and the deformation influencing factor is used to characterize the factors affecting the slope deformation.

[0059] Specifically, the slope to be monitored refers to the area where the slope deformation needs to be predicted in actual application. The slope to be monitored includes the target monitoring area. In order to facilitate calculation and improve the accuracy of prediction, the slope to be monitored needs to be divided into one or more monitoring areas. During the prediction process, each area can be processed in turn, or multiple monitoring areas can be processed in parallel. The embodiment of this specification takes processing each area once as an example, and the monitoring area currently being processed among multiple monitoring areas is used as the target monitoring area.

[0060] Based on this, before predicting slope deformation, it is necessary to first determine the slope deformation variable and deformation influencing factors of the slope to be monitored in the target monitoring area.

[0061] More specifically, the slope deformation amount and deformation influencing factor obtained in the above steps may be the target key point corresponding to the target monitoring area at the current moment. In the embodiment of the present specification, each monitoring area preferably corresponds to a target key point, and each target key point may correspond to one or more deformation influencing factors and a deformation amount at one moment.

[0062] Furthermore, the data information collected by the slope terrain monitoring equipment is a massive amount of point cloud data. If the slope deformation prediction is directly performed on the original point cloud data, it will be inefficient or infeasible on existing computing equipment.

[0063] In order to improve the efficiency and accuracy of slope deformation prediction under slope terrain monitoring equipment, it is necessary to divide the monitored slope into monitoring areas and aggregate its features, which can not only reduce the dimensionality of the original point cloud data, but also preserve the feature information to the greatest extent.

[0064] Therefore, before determining the slope deformation amount and deformation influencing factor corresponding to the target monitoring area of ​​the slope to be monitored, it also includes: obtaining slope point cloud data of the slope to be monitored, selecting at least one key point from the slope point cloud data, dividing the slope with the key point as the center and a preset length as the radius to obtain at least one monitoring area; and selecting the target monitoring area from the at least one monitoring area.

[0065] Furthermore, obtaining the slope point cloud data of the slope to be monitored includes: first obtaining the original point cloud data of the slope to be monitored, and preprocessing the original point cloud data to obtain a plurality of slope point cloud data corresponding to different times, and forming a point cloud data set from the plurality of slope point cloud data.

[0066] Specifically, the slope to be monitored is provided with a monitoring device for monitoring the slope topography, such as a slope radar, and the original point cloud data of the slope to be monitored is obtained from the monitoring device. The pre-processing may be, for example, a median filter process.

[0067] In one embodiment of the present specification, the original point cloud data corresponding to a preset number T moments within the preset time period can be determined according to the preset time period, and the point cloud data sets corresponding to the T moments can be obtained after preprocessing the original point cloud data. m (t)|t=1,2,…,T; m=1,2,…,M}, where, P m (t) represents the mth point cloud data at the tth time point, and That is, each slope point cloud data contains the corresponding location information and terrain features. The location information represented by Indicates P m The three-dimensional coordinate value of (t), Indicates P m (t); M represents the number of point clouds.

[0068] Furthermore, at least one key point may be selected from the point cloud data set corresponding to a certain moment by using a uniform network sampling method, and the specific number of key points selected may be preset.

[0069] For example, if the number of selected key points is set to K in advance, K key points can be selected from the point cloud data set corresponding to the t-th moment according to the uniform network sampling method, and form a key point set {P m,k (t)|k=1,2,…,K},P m,k (t) represents the kth key point at the tth moment.

[0070] Select a target key point from K key points, take the target key point as the center of the circle, and search the neighborhood radius corresponding to the target key point by fixing the radius r. Among them, d represents the Euclidean distance. Indicates P m,k The location information of the i-th neighboring point of (t), Indicates P m,k (t), M represents the total number of neighborhood points, and the neighborhood radius R m,k (t) preset length. In addition, the neighborhood radius can also be obtained for all key points at the same time.

[0071] The area with the key point as the center and the preset length as the radius is used as the monitoring area, that is, each key point corresponds to a monitoring area. Therefore, the number of monitoring areas is the same as that of key points, which can be one or more.

[0072] A target monitoring area is selected from at least one monitoring area so as to process the monitoring areas in an orderly manner.

[0073] In another embodiment of the present specification, after acquiring the slope point cloud data, the slope point cloud data can be directly divided into different point cloud data groups according to the position information of the slope point cloud data, and the range corresponding to a point cloud data group is taken as a corresponding monitoring area, and multiple monitoring areas are obtained by analogy. A point is selected from the multiple slope point cloud data corresponding to each monitoring area as the target key point corresponding to the monitoring area, and one of the multiple monitoring areas is selected as the target monitoring area.

[0074] In the embodiments of the present specification, by preprocessing massive amounts of original point cloud data to obtain slope point cloud data and selecting key points from the slope point cloud data, the amount of data processed can be effectively reduced and prediction efficiency can be improved. At the same time, dividing the monitoring area according to key points can not only achieve dimensionality reduction of point cloud data, but also preserve feature information to the greatest extent.

[0075] In addition, in another embodiment of the present specification, the target monitoring area may not be selected. In this case, multiple monitoring areas need to be processed simultaneously later.

[0076] Furthermore, determining the slope deformation amount corresponding to the target monitoring area of ​​the slope to be monitored includes: determining the position information of the target key point corresponding to the target monitoring area, wherein the position information includes the first position information of the target key point at the current moment and the second position information at the previous moment; calculating the slope deformation amount of the target monitoring area according to the first position information and the second position information.

[0077] Specifically, the position information of the target key point can be expressed in the form of two-dimensional coordinates or three-dimensional coordinates. In order to facilitate the calculation of the deformation amount of the target key point at the current moment relative to the previous moment, the position information can include the first position information corresponding to the target key point at the current moment and the second position information corresponding to the previous moment.

[0078] Based on this, the slope deformation amount of the target monitoring area is calculated according to the first position information at the current moment and the second position information corresponding to the previous moment.

[0079] For example, the position information of the target key point in the target monitoring area at the current moment is Indicates P m,k The slope shape variable D of the kth target monitoring area at the tth time point (or current moment) compared with the t-1th time point (or previous moment) is calculated using formula (1): m,k (t), thus obtaining the slope deformation sample set {D m,k (t)|t=1,2,…,T;k=1,2,…,K}.

[0080]

[0081] Furthermore, the deformation influence factor corresponding to the target monitoring area of ​​the slope to be monitored is determined, including: obtaining the target terrain characteristics and target environmental characteristics corresponding to the target key points in the target monitoring area; aggregating the neighborhood points and the target key points in the target monitoring area, and performing feature extraction to obtain the target feature information corresponding to the target monitoring area; and constructing the deformation influence factor corresponding to the target key points according to the target terrain characteristics, the target environmental characteristics and the target feature information.

[0082] Specifically, the key points corresponding to the target monitoring area are target key points. Since the target key points are selected from the slope point cloud data, each slope point cloud data contains the corresponding target position information and target terrain features. Therefore, the target key points also contain the corresponding position information and terrain features.

[0083] Based on this, the corresponding terrain features are extracted from the target key, and the terrain features may be, for example, the curvature, slope, and slope direction of the terrain. The environmental features may be obtained from the environmental monitoring equipment of the slope to be monitored, such as temperature monitoring equipment, humidity monitoring equipment, or precipitation monitoring equipment, and the corresponding environmental features may include temperature, humidity, precipitation, etc. It should be noted that there may be one or more environmental features or terrain features corresponding to any target monitoring point at any time.

[0084] The neighborhood points and target key points in the target monitoring area are aggregated, and feature extraction is performed to obtain the target feature information corresponding to the target monitoring area.

[0085] In one embodiment of the present specification, the target feature information of the target key point in the kth monitoring area can be calculated by formula (2):

[0086]

[0087] The deformation influencing factors corresponding to the target key points at the current moment are constructed according to the target terrain features, target environment features and target feature information.

[0088] In one embodiment of the present specification, the multivariate candidate set is composed of N deformation influencing factors of all monitoring areas. in, It represents the nth deformation impact factor of the kth monitoring area at the tth time point.

[0089] It should be noted that there may be one or more deformation influencing factors corresponding to any target key point at any time, but there is only one slope deformation amount corresponding to any target key point at any time.

[0090] In the embodiments of this specification, by obtaining the slope deformation amount and the corresponding deformation influence factor of the target area, it is convenient to link the dependency relationship between the slope deformation amount and the corresponding deformation influence factor when calculating the slope deformation prediction result in the subsequent calculation, providing a basis for subsequent calculation and improving calculation efficiency. By forming the deformation influence factor of the target key point according to the target terrain characteristics, target environmental characteristics and target feature information of the target key point, it is possible to combine multiple deformation influence factors, so that the obtained deformation influence factor is more accurate, and thus the subsequent prediction results can also be more accurate.

[0091] Step 104: Based on the mutual information value between the slope deformation amount and the deformation influencing factor, feature screening is performed on the deformation influencing factor to obtain a strong correlation factor.

[0092] Specifically, the mutual information value refers to the mutual dependence between the slope deformation variable and each deformation influencing factor.

[0093] Based on this, the mutual information value between the slope deformation variable and the corresponding deformation influencing factor in the target monitoring area is calculated. Since any target key point has one corresponding slope deformation variable at any time, and there may be multiple deformation influencing factors, there may be multiple mutual information values ​​obtained from the calculation point. The deformation influencing factors are feature screened according to the mutual information value, and the qualified deformation influencing factors are selected as strong correlation factors.

[0094] The embodiments of this specification calculate the mutual information value between the slope deformation variable and the deformation influencing factor, so as to accurately determine the degree of dependence between the slope deformation variable and the deformation influencing factor, and then screen out the strongly correlated factors that have a greater impact on the slope deformation variable. Deformation prediction is performed based on the strongly correlated factors, which can make the prediction result more accurate.

[0095] Furthermore, based on the mutual information value of the slope deformation amount and the deformation influencing factor, feature screening is performed on the deformation influencing factor to obtain a strong correlation factor, including: calculating the first mutual information between the deformation influencing factor and the slope deformation amount; establishing a maximum mutual information model according to the first mutual information, and determining the strong correlation factor of the target monitoring area based on the information growth rate criterion.

[0096] Specifically, the first mutual information refers to the mutual information value obtained by calculating the mutual information of the slope deformation variable and the deformation influencing factor without any preconditions. The maximum mutual information model is used to select a maximum mutual information from multiple calculated mutual information values.

[0097] Based on this, the first mutual information between the deformation influencing factor and the slope deformation variable is calculated, the maximum mutual information model is established according to the first mutual information, and the strong correlation factor of the target monitoring area is determined based on the growth rate criterion.

[0098] See also Figure 2 , Figure 2 It is a flowchart of feature optimization and prediction in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of this specification. In one embodiment of this specification, Figure 2 The feature optimization module in is equivalent to the above-mentioned maximum mutual information model. The prediction object is the slope deformation variable. The deformation influencing factor and the slope deformation variable are input into the maximum mutual information model, and the deformation influencing factor is calculated by formula (3): and the corresponding slope shape variable D m,k The first mutual information between (t)

[0099]

[0100] In formula (3), express and D m,k The joint probability distribution of (t); Respectively and D m,k The marginal probability density function of (t).

[0101] According to mutual information The maximum joint mutual information model is established, and based on the information growth rate criterion, a strongly correlated multivariate set of the kth monitoring area is determined. It represents the gth strongly correlated factor affecting the slope deformation variable in the kth monitoring area at the tth time point.

[0102] Furthermore, the maximum mutual information model is established according to the first mutual information, and the strong correlation factor of the monitoring area is determined based on the information growth rate criterion, including: selecting the deformation influence factor corresponding to the maximum value in the first mutual information as the candidate correlation factor, and adding the candidate correlation factor to the variable candidate set of the target monitoring area; determining the second mutual information between the deformation influence factor and the slope deformation variable under the condition of the variable candidate set; selecting the deformation influence factor corresponding to the maximum value in the second mutual information as the candidate correlation factor, and adding it to the variable candidate set; continuing to execute the step of determining the second mutual information between the deformation influence factor and the slope deformation variable under the condition of the variable candidate set until the variable candidate set meets the number constraint, and taking the candidate correlation factor included in the variable candidate set as the strong correlation factor.

[0103] Specifically, since the mutual information value can reflect the degree of dependence between the slope deformation variable and the deformation influencing factor, the larger the mutual information value, the greater the influence of the corresponding deformation influencing factor on the slope deformation variable. Therefore, it is necessary to select the deformation influencing factor corresponding to the maximum value in the first mutual information as the candidate correlation factor, and add the candidate correlation factor to the candidate variable set.

[0104] Under the condition of the candidate variable set, the second mutual information between the deformation influencing factor and the slope deformation variable is calculated. The second mutual information refers to the mutual information value between the deformation influencing factor and the slope deformation variable calculated under the condition of the candidate variable set. The deformation influencing factor corresponding to the maximum value in the second mutual information is selected as the candidate variable factor and added to the candidate variable set. That is, the candidate variables in the candidate variable set are the deformation influencing factors with the greatest influence on the slope deformation variable calculated under different conditions.

[0105] Repeat the previous step until the candidate variable factors in the candidate variable set reach a preset number, at which point the screening is stopped, and the candidate variable factors in the candidate variable set are taken as strongly correlated factors, and the candidate variable set is taken as a strongly correlated variable set.

[0106] In one embodiment of the present specification, the deformation influence factor is calculated. and slope shape D m,k The first mutual information of (t) Thus, we can obtain the N and first mutual information of N deformation influencing factors and slope deformation And select the deformation influence factor corresponding to the maximum value As a candidate variable factor, add it to the candidate variable set corresponding to the kth monitoring area In

[0107] Calculated in Under the conditional deformation influence factor and slope shape D m,k The second mutual information of (t) Thus we get Under the condition of N deformation influencing factors and D m,k (t) N second mutual information And select the deformation influence factor corresponding to the maximum value and add it to Follow the steps in this process until Until the number of variables in reaches a constant G, the candidate variable set As a set of strongly correlated variables, the candidate variable factors are taken as strongly correlated factors. It represents the g-th strongly correlated factor affecting the slope deformation variable in the k-th monitoring area at the t-th time point, thereby obtaining the strongly correlated multivariate set at T time points:

[0108] See also Figure 3 , Figure 3 It is a schematic diagram of the variation of information gain with the number of factors in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of the present specification. In one embodiment of the present specification, the maximum joint mutual information (JMI) is adopted, and JMI is compared with partial mutual information (PMI) and correlation coefficient method (CC).

[0109] The embodiment of this specification can reflect the influence of the deformation influencing factor on the slope deformation according to the values ​​of the first mutual information and the second mutual information, and select the deformation influencing factor with the greatest influence as the strong correlation factor, thereby making the subsequent prediction results calculated based on the strong correlation factor more accurate.

[0110] Step 106: construct a TCN network with multiple convolution kernels in parallel, input the strong correlation factor into the TCN network with multiple convolution kernels in parallel to obtain corresponding time series features, and determine the slope shape variable prediction result of the target monitoring area at the next moment based on the time series features.

[0111] Specifically, time series characteristics refer to the trend of strong correlation factors changing over time, showing the same nature of rising, falling, and staying. The next moment is relative to the current moment, that is, the time point after the current moment.

[0112] Based on this, a multi-convolution kernel processing model is constructed to process the strong correlation factors to obtain time series features.

[0113] See also Figure 2 In one embodiment of the present specification, the multi-core convolution processing model may be a TCN network with multiple convolution kernels in parallel. Construct a TCN network with multiple convolution kernels in parallel, and Processing to obtain the time series feature set in, It represents the time series characteristics of the g-th strongly correlated factor in the k-th monitoring area at the t-th time point; the prediction result is obtained after the time series characteristics are processed by the dilated convolution layer, the fully connected layer, the recurrent layer, etc.

[0114] See also Figure 4 , Figure 4 It is a schematic diagram of a TCN network with multiple convolution kernels in parallel in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of this specification. This embodiment of the specification constructs a TCN network with multiple convolution kernels in parallel, inputs multiple time series features into the TCN network with multiple convolution kernels in parallel, performs convolution operations in each convolution layer, and the convolution layer is composed of multiple convolution kernels in parallel. At the same time, multiple general residual blocks are designed and combined to perform multi-dimensional feature extraction. Each layer of residual blocks contains two layers of dilated causal convolution and activation function ReLU, and weight normalization is applied to the convolution kernel. In addition, dropout is added after each dilated causal convolution for regularization to avoid overfitting.

[0115] By processing the strongly correlated factors through a TCN network with multiple convolution kernels in parallel, it is possible to take into account the characteristics of each strongly correlated factor changing with the time series, improve the accuracy of subsequent prediction results, and multi-core parallel processing can make the processing efficiency higher when facing massive data.

[0116] Furthermore, in order to improve the prediction accuracy and prediction efficiency, based on the time series characteristics, the prediction result of the slope shape variable of the target monitoring area at the next moment is determined, including: inputting the time series characteristics into a pre-established prediction model, the prediction model includes a feature attention mechanism, a fully connected layer and an activation function; processing the time series characteristics through the feature attention mechanism to obtain the weights corresponding to the strong correlation factors; normalizing the weights to obtain the feature attention distribution, and weighting the corresponding time series characteristics based on the feature attention distribution to obtain the associated features; inputting the associated features into the fully connected layer, performing nonlinear transformation through the activation function, and obtaining the prediction result of the slope shape variable of the target monitoring area at the next moment.

[0117] Specifically, a prediction model is established in advance, which is used to process the time series features of strongly correlated factors, so as to predict the slope deformation variable. The time series features are weighted using the feature attention mechanism to quantify the correlation between the time series features and the deformation variable, so as to obtain the slope deformation variable prediction result, which can significantly improve the accuracy and efficiency of slope deformation prediction.

[0118] See also Figure 5 , Figure 5 is a schematic diagram of a feature attention mechanism in a slope deformation prediction method of a multi-core TCN network under feature screening provided by another embodiment of this specification. In one embodiment of this specification, the feature attention mechanism is used to select a time series feature set. Perform optimization processing, that is, calculate the feature weights of the time series features to obtain strong correlation factors G×T weights And use weights After weighting, the associated features are obtained, which are input into a fully connected layer and transformed nonlinearly through the activation function to obtain the slope deformation prediction result of the kth monitoring area at the t+1 time point.

[0119] The embodiments of this specification calculate the weights of relevant factors and weight the time series features to highlight or suppress the importance of different features, thereby affecting the prediction model's attention to different time series features, enhancing the weights of key time series features, and further improving the fusion efficiency of time series features and the accuracy of slope shape variable prediction results.

[0120] Furthermore, the associated feature set is input into the fully connected layer, and a nonlinear transformation is performed through an activation function to obtain a prediction result of the slope deformation variable of the target monitoring area at the next moment, including: inputting the associated features into the fully connected layer to obtain a hidden layer state value corresponding to the current moment; and calculating the slope deformation prediction result of the target monitoring area at the next moment based on the hidden layer state value.

[0121] Specifically, the fully connected layer also includes a hidden layer. Therefore, after the associated features are input into the fully connected layer, it is necessary to calculate the hidden layer state value corresponding to the current moment.

[0122] In one embodiment of the present specification, the strongly correlated multivariate set A variable matrix with a total time span of T under the kth monitoring area is formed Will The input is sent to the TCN network, and after parallel processing of multiple convolution kernels and multi-dimensional feature extraction processing of multiple universal residual blocks, the time series feature of the g-th strong correlation factor is obtained using formula (4): Then we get the time series feature set

[0123]

[0124] In formula (4), Activation is the activation function, + is the residual connection, and Q represents the output of the residual block.

[0125] The feature attention mechanism uses formula (5) to obtain the attention weight of the g-th temporal feature at the t-th time point: Thus, the weights of each feature at the tth time point are obtained

[0126]

[0127] In formula (5), V e , W e and U e are the three weights of the neuron, is the bias coefficient, tanh() is the excitation function, h t-1 Represents the state value of the hidden layer at the previous time point t-1.

[0128] Using formula (6), we can get the characteristics of the g-th feature at the t-th time point: Attention Distribution Thus, the feature attention distribution set at the tth time point is obtained

[0129]

[0130] In formula (6), softmax() is the normalization function;

[0131] Using formula (7), we can get the correlation features at the tth time point:

[0132]

[0133] Using formula (8), we can update the state value h of the hidden layer at the tth time point. t .

[0134]

[0135] In formula (8), f 1 It is a TCN network unit.

[0136] Using formula (9), we can get the slope deformation prediction result of the kth monitoring area at time point t+1: Therefore, predictions can be made based on the slope monitoring data at previous historical time points to obtain the slope shape variables predicted at subsequent time points.

[0137]

[0138] Furthermore, after step 106, the benchmark models ARIMAX, XGBoost, DeepAR, and LSTM can be selected to compare and analyze the slope deformation prediction results with the JMI-TCN method, and the mean square error (MSE) and mean absolute error (MAE) are used as evaluation indicators of the model performance results. The comparative analysis of the experimental results in Table 1 shows that the JMI-TCN model has a good prediction effect in slope deformation prediction.

[0139] Table 1

[0140]

[0141] The embodiments of this specification solve the limitation that previous detection equipment can only monitor in real time but cannot predict data, and realize real-time monitoring with high accuracy and strong timeliness. The embodiments of this specification take into account the influence between slope monitoring induced variable data and deformation prediction result data, use the maximum joint mutual information to screen out environmental factor variables that are more related to slope deformation, and input these environmental factor variables into the TCN model, thereby providing a high-quality data source for deformation prediction model training.

[0142] In order to overcome the problem of low computational efficiency of deep learning models when faced with a large amount of slope monitoring data, the embodiments of this specification propose a TCN network improved by a multi-convolution kernel parallel network, which can obtain a prediction model with higher fit in a shorter training time, thereby improving the computational efficiency of model prediction.

[0143] The embodiments of this specification have a wide range of applications: without adding additional auxiliary information, the method can be widely used in geological disaster deformation monitoring, provides a scientific basis for landslide disaster prevention and mitigation, and meets the needs of practical engineering applications.

[0144] It should be noted that the slope deformation prediction method provided in this specification is applicable to any environment where deformation prediction is required, such as slope deformation prediction in mines, slope deformation prediction in water conservancy projects, or slope deformation prediction in transportation projects.

[0145] Corresponding to the above method embodiment, this specification also provides a slope deformation prediction device embodiment, Figure 6 1 is a schematic diagram of a slope deformation prediction device of a multi-core TCN network under feature screening provided by an embodiment of this specification. Figure 6 As shown, the device comprises:

[0146] The variable determination module 602 is configured to determine the slope deformation amount and deformation influencing factor corresponding to the target monitoring area of ​​the slope to be monitored, wherein the slope deformation amount is used to characterize the deformation amount of the target monitoring area at the current moment compared with the previous moment, and the deformation influencing factor is used to characterize the factors affecting the slope deformation;

[0147] A feature screening module 604 is configured to perform feature screening on the deformation influencing factor to obtain a strong correlation factor based on the mutual information value between the slope deformation amount and the deformation influencing factor;

[0148] The prediction module 606 constructs a TCN network with multiple convolution kernels in parallel, inputs the strong correlation factor into the TCN network with multiple convolution kernels in parallel to obtain corresponding time series features, and determines the slope shape variable prediction result of the target monitoring area at the next moment based on the time series features.

[0149] Optionally, the slope deformation prediction device of the multi-core TCN network under feature screening further includes:

[0150] The area division module is configured to obtain slope point cloud data of the slope to be monitored, select at least one key point from the slope point cloud data, divide the slope with the key point as the center and a preset length as the radius, and obtain at least one monitoring area;

[0151] The area determination module is configured to select the target monitoring area from the at least one monitoring area.

[0152] Optionally, the variable determination module 602 includes:

[0153] A position information acquisition submodule is configured to determine the position information of the target key point corresponding to the target monitoring area, wherein the position information includes the first position information of the target key point at the current moment and the second position information at the previous moment;

[0154] The deformation calculation submodule is configured to calculate the slope deformation amount of the target monitoring area according to the first position information and the second position information.

[0155] Optionally, the variable determination module 602 includes:

[0156] A feature extraction submodule is configured to obtain target terrain features and target environment features corresponding to target key points in the target monitoring area;

[0157] The feature calculation submodule is configured to aggregate the neighborhood points and target key points in the target monitoring area and perform feature extraction to obtain target feature information corresponding to the target monitoring area;

[0158] The deformation factor generation submodule is configured to form a deformation influence factor corresponding to the target key point according to the target terrain feature, the target environment feature and the target feature information.

[0159] Optionally, the feature screening module 604 includes:

[0160] A mutual information calculation submodule, configured to calculate a first mutual information between the deformation influencing factor and the slope deformation amount;

[0161] The strong correlation factor determination submodule is configured to establish a maximum mutual information model according to the first mutual information, and determine the strong correlation factor of the target monitoring area based on an information growth rate criterion.

[0162] Optionally, the strong correlation factor determination submodule includes:

[0163] A first screening submodule is configured to select a deformation influencing factor corresponding to a maximum value in the first mutual information as a candidate correlation factor, and add the candidate correlation factor to a variable candidate set of the target monitoring area;

[0164] A second mutual information calculation submodule is configured to determine the second mutual information between the deformation influencing factor and the slope deformation amount under the condition of the variable candidate set;

[0165] A second screening submodule is configured to select a deformation influencing factor corresponding to a maximum value in the second mutual information as a candidate correlation factor, and add it to the variable candidate set;

[0166] The conditional judgment submodule is configured to continue to execute the step of determining the second mutual information between the deformation influencing factor and the slope deformation amount under the condition of the variable candidate set until the variable candidate set meets the number constraint condition, and use the candidate correlation factor included in the variable candidate set as the strong correlation factor.

[0167] Optionally, the prediction module 606 includes:

[0168] A model building submodule uses the strongly correlated factors corresponding to the target monitoring area as a strongly correlated multivariate set;

[0169] The feature extraction submodule is configured to input the strongly correlated multi-variable set into the multi-convolution kernel parallel TCN network, perform multi-convolution kernel parallel processing and multi-dimensional feature extraction processing of at least two residual modules, and obtain the time series features corresponding to the strongly correlated factors.

[0170] Optionally, the prediction module 606 includes:

[0171] An input submodule, configured to input the time series features into a pre-established prediction model, wherein the prediction model includes a feature attention mechanism, a fully connected layer, and an activation function;

[0172] A weight calculation submodule is configured to process the time series features through a feature attention mechanism to obtain weights corresponding to the strong correlation factors;

[0173] A weighted processing submodule is configured to perform normalization processing on the weights to obtain a feature attention distribution, and perform weighted processing on the corresponding temporal features based on the feature attention distribution to obtain a correlation feature;

[0174] The result prediction submodule is configured to input the associated features into the fully connected layer, perform nonlinear transformation through an activation function, and obtain the slope shape variable prediction result of the target monitoring area at the next moment.

[0175] Optionally, the result prediction submodule includes:

[0176] A first calculation submodule calculates the associated features corresponding to the target key point at the current moment, inputs the associated features into the fully connected layer, and obtains the hidden layer state value corresponding to the current moment;

[0177] The second calculation submodule is configured to calculate the slope deformation prediction result of the target monitoring area at the next moment based on the hidden layer state value.

[0178] The above is a schematic scheme of a slope deformation prediction device of a multi-core TCN network under feature screening of this embodiment. It should be noted that the technical scheme of the slope deformation prediction device of the multi-core TCN network under feature screening and the technical scheme of the slope deformation prediction method of the multi-core TCN network under feature screening belong to the same concept. The details not described in detail in the technical scheme of the slope deformation prediction device of the multi-core TCN network under feature screening can be found in the description of the technical scheme of the slope deformation prediction method of the multi-core TCN network under feature screening.

[0179] Figure 7 7 is a block diagram of a computing device provided by an embodiment of the present specification. The components of the computing device 700 include but are not limited to a memory 710 and a processor 720. The processor 720 is connected to the memory 710 via a bus 730, and the database 750 is used to store data.

[0180] The computing device 700 also includes an access device 740 that enables the computing device 700 to communicate via one or more networks 760. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 740 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0181] In one embodiment of the present specification, the above components of the computing device 700 and Figure 7 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 7 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0182] The computing device 700 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 700 may also be a mobile or stationary server.

[0183] The processor 720 is used to execute the following computer executable instructions, which, when executed by the processor, implement the steps of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening.

[0184] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening belong to the same concept. For details not described in detail in the technical scheme of the computing device, please refer to the description of the technical scheme of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening.

[0185] An embodiment of the present specification also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the slope deformation prediction method of a multi-core TCN network under the above-mentioned feature screening.

[0186] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening.

[0187] An embodiment of the present specification also provides a computer program, wherein when the computer program is executed in a computer, the computer is instructed to execute the steps of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening.

[0188] The above is a schematic scheme of a computer program of this embodiment. It should be noted that the technical scheme of the computer program and the technical scheme of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening belong to the same concept. For details not described in detail in the technical scheme of the computer program, please refer to the description of the technical scheme of the slope deformation prediction method of the multi-core TCN network under the above-mentioned feature screening.

[0189] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0190] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0191] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0192] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0193] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A slope deformation prediction method based on a multi-core TCN network under feature screening, characterized in that: include: Determine the slope deformation amount and deformation influence factor corresponding to the target monitoring area of ​​the slope to be monitored, wherein the slope deformation amount is used to characterize the deformation amount of the target monitoring area at the current moment compared with the previous moment, and the deformation influence factor is used to characterize the factors affecting the slope deformation; Determine the deformation influence factor corresponding to the target monitoring area of ​​the slope to be monitored, including: obtaining the target terrain characteristics and target environmental characteristics corresponding to the target key points in the target monitoring area; aggregate the neighborhood points and the target key points in the target monitoring area, and perform feature extraction to obtain the target feature information corresponding to the target monitoring area; construct the deformation influence factor corresponding to the target key point according to the target terrain characteristics, the target environmental characteristics and the target feature information; Based on the mutual information value of the slope deformation amount and the deformation influencing factor, the deformation influencing factor is subjected to feature screening to obtain a strong correlation factor, wherein the mutual information value refers to the mutual dependence degree between the slope deformation amount and each deformation influencing factor; Constructing a multi-convolution kernel parallel TCN network, inputting the strong correlation factor into the multi-convolution kernel parallel TCN network to obtain corresponding time series features, and determining the slope shape variable prediction result of the target monitoring area at the next moment based on the time series features; Based on the time series features, the prediction result of the slope shape variable of the target monitoring area at the next moment is determined, including: inputting the time series features into a pre-established prediction model, the prediction model including a feature attention mechanism, a fully connected layer and an activation function; processing the time series features through the feature attention mechanism to obtain the weights corresponding to the strong correlation factors; normalizing the weights to obtain a feature attention distribution, and weighting the corresponding time series features based on the feature attention distribution to obtain associated features; inputting the associated features into the fully connected layer, performing nonlinear transformation through the activation function, and obtaining the prediction result of the slope shape variable of the target monitoring area at the next moment.

2. The slope deformation prediction method of multi-core TCN network under feature screening according to claim 1 is characterized in that: Before determining the slope deformation amount and deformation influencing factor corresponding to the target monitoring area of ​​the slope to be monitored, it also includes: Obtaining slope point cloud data of the slope to be monitored, selecting at least one key point from the slope point cloud data, dividing the slope with the key point as the center and a preset length as the radius to obtain at least one monitoring area; The target monitoring area is selected from the at least one monitoring area.

3. The slope deformation prediction method of multi-core TCN network under feature screening according to claim 1 is characterized in that: Determine the slope deformation corresponding to the target monitoring area of ​​the slope to be monitored, including: Determine the position information of the target key point corresponding to the target monitoring area, wherein the position information includes the first position information of the target key point at the current moment and the second position information at the previous moment; The slope deformation amount of the target monitoring area is calculated according to the first position information and the second position information.

4. The slope deformation prediction method of multi-core TCN network under feature screening according to claim 1 is characterized in that: Based on the mutual information value of the slope deformation amount and the deformation influencing factor, feature screening is performed on the deformation influencing factor to obtain a strong correlation factor, including: Calculating the first mutual information between the deformation influencing factor and the slope deformation amount; A maximum mutual information model is established according to the first mutual information, and a strong correlation factor of the target monitoring area is determined based on an information growth rate criterion.

5. The slope deformation prediction method of multi-core TCN network under feature screening according to claim 4 is characterized in that: The step of establishing a maximum mutual information model according to the first mutual information and determining a strong correlation factor of the monitoring area based on an information growth rate criterion includes: Selecting a deformation influencing factor corresponding to a maximum value in the first mutual information as a candidate correlation factor, and adding the candidate correlation factor to a variable candidate set of the target monitoring area; Determine the second mutual information between the deformation influencing factor and the slope deformation amount under the condition of the variable candidate set; Selecting a deformation influence factor corresponding to a maximum value in the second mutual information as a candidate correlation factor, and adding it to the variable candidate set; The step of determining the second mutual information between the deformation influencing factor and the slope deformation amount under the variable candidate set condition is continued until the variable candidate set satisfies the number constraint condition, and the candidate correlation factors included in the variable candidate set are used as the strong correlation factors.

6. The slope deformation prediction method of multi-core TCN network under feature screening according to claim 1 is characterized in that: Inputting the strong correlation factor into the TCN network with multiple convolution kernels in parallel to obtain the corresponding time series features, including: The strongly correlated factors corresponding to the target monitoring area are used as a strongly correlated multivariate set; The strongly correlated multi-variable set is input into the multi-convolution kernel parallel TCN network, and multi-convolution kernel parallel processing and multi-dimensional feature extraction processing of at least two residual modules are performed to obtain the time series features corresponding to the strongly correlated factors.

7. The slope deformation prediction method of multi-core TCN network under feature screening according to claim 1 is characterized in that: The step of inputting the associated features into the fully connected layer and performing nonlinear transformation through an activation function to obtain the slope shape variable prediction result of the target monitoring area at the next moment includes: Input the associated features into the fully connected layer to obtain the hidden layer state value corresponding to the current moment; The slope deformation prediction result of the target monitoring area at the next moment is calculated based on the hidden layer state value.

8. A computing device, characterized in that include: Memory and processor; The memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the slope deformation prediction method of the multi-core TCN network under feature screening according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which, when executed by a processor, can implement the steps of the slope deformation prediction method of a multi-core TCN network under feature screening as described in any one of claims 1 to 7.

10. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of the slope deformation prediction method of a multi-core TCN network under feature screening as described in any one of claims 1 to 7.

Images