Pre-warning method, system and equipment for deformation and damage of bedding rock slope and storage medium

Abstract

The invention relates to the technical field of geological disaster monitoring and early warning, in particular to a bedding rock slope deformation and damage early warning method, system and device and a storage medium, and the method comprises the steps: obtaining, preprocessing, feature extraction and optimal clustering of micro-seismic monitoring signals in sequence to obtain optimal clustering features; then, acquiring real-time monitoring data, converting the optimal clustering characteristics into signals according to the real-time monitoring data, outputting the signals, then calibrating, constructing a rock fracture signal identification model, acquiring rock fracture signals, and performing association rule mining on the rock fracture signals and the displacement monitoring data to obtain association rules; and finally, constructing an integrated prediction model by using the association rule and the displacement monitoring data. And real-time accurate early warning of deformation and damage of the bedding rock slope is realized. The problems of poor real-time performance and accuracy of deformation and damage early warning of the bedding rock slope based on micro-seismic monitoring in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster monitoring and early warning technology, specifically to a deformation and damage early warning method for a bedding rock slope, in particular to a deformation and damage early warning method, system, equipment and storage medium for a bedding rock slope based on microseismic monitoring signals. Background Art

[0002] Bedding rock slope refers to a layered rock slope with the rock strata approaching or being consistent with the slope surface and having the same inclination. This type of slope is more common in mountainous areas with complex geological structures and undulating terrain. Especially in the construction of linear projects such as highways and railways, the road cut slope formed by excavating the mountain is often a bedding rock slope. The formation of bedding rock slopes is mainly related to geological structure and crustal movement. During the long-term crustal movement and tectonic movement, the rock strata gradually lose balance after being subjected to geological mechanical effects such as tension, compression, and shear, causing the rock strata to tilt along the slope surface. At the same time, natural forces such as surface water and rainwater will also accelerate the formation and destruction of slopes. Bedding rock slopes have the characteristics of poor stability and difficulty in governance. With the expansion of cities and the exploitation of natural resources, the stability problem of rock slopes has become increasingly prominent, especially in areas with complex terrain and variable geological conditions. The deformation and destruction of rock slopes not only threatens the safety of infrastructure, but may also cause significant casualties and economic losses. Establishing a slope monitoring and early warning system to monitor the deformation and displacement of the slope in real time and promptly discover and deal with potential safety hazards are important means of managing bedding rock slopes.

[0003] At present, traditional slope monitoring methods mainly include surface displacement measurement, inclination measurement and groundwater level monitoring. These methods can provide key slope stability data, but they usually have response delays and are not comprehensive in spatial coverage. For example, although surface displacement measurement can accurately record the movement of specific points, it is difficult to capture the dynamic changes of the entire slope in real time. In this context, microseismic monitoring technology has attracted attention because of its ability to provide continuous and real-time geological dynamic data. By analyzing these real-time transmitted data, the deformation and failure process of rocks can be predicted. A major advantage of microseismic monitoring is its high sensitivity and ability to cover a wide area. Compared with traditional monitoring technologies, microseismic monitoring can detect smaller changes, thus providing the possibility of early warning. In addition, the high-frequency acquisition of microseismic data also enables researchers to use advanced data analysis methods, such as time-frequency analysis and machine learning algorithms, to more accurately interpret data and predict future slope behavior. Although microseismic monitoring has the above significant advantages, due to the high sensitivity of microseismic detection data, it retains other signals besides rock fracture signals, which makes the microseismic monitoring data complicated, thus affecting the accuracy and real-time nature of rock slope deformation and damage monitoring and early warning. Summary of the invention

[0004] In view of the problem of poor real-time and accuracy of deformation and damage warning of bedding rock slopes based on microseismic monitoring in the prior art, the present invention provides a deformation and damage warning method, system equipment and storage medium for bedding rock slopes.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The present invention provides a method for early warning of deformation and damage of a bedding rock slope, comprising:

[0007] Acquire microseismic monitoring signals and perform signal preprocessing on the microseismic monitoring signals;

[0008] Extract features from the microseismic monitoring signals after signal preprocessing to obtain multi-dimensional feature data;

[0009] Perform optimal clustering on multi-dimensional feature data to obtain optimal clustering features;

[0010] Acquire real-time monitoring data, convert the optimal clustering features into signal outputs according to the real-time monitoring data, and then calibrate them to build a rock fracture signal recognition model;

[0011] The rock fracture components in the microseismic monitoring signal are identified using the identification model to obtain the rock fracture signal;

[0012] Acquire displacement monitoring data, and perform association rule mining between rock fracture signals and displacement monitoring data to obtain association rules;

[0013] Using association rules and displacement monitoring data, an integrated prediction model is constructed;

[0014] An integrated prediction model is used to analyze real-time monitoring data and provide early warning of the dynamics of slope deformation and damage.

[0015] Furthermore, the method for obtaining the microseismic monitoring signal is: obtaining microseismic monitoring data monitored in the field, and converting the microseismic monitoring data monitored in the field into a signal form to obtain the microseismic monitoring signal.

[0016] Furthermore, the method for preprocessing the microseismic monitoring signal is as follows:

[0017] Normalizing the microseismic monitoring signal to obtain normalized data;

[0018] The normalized data is subjected to discrete Fourier transform to obtain the frequency spectrum characteristics of the microseismic monitoring signal and complete the signal preprocessing.

[0019] Preferably, the multidimensional feature data includes original signal kurtosis, original signal skewness, original signal energy, ratio of signal rise time to fall time, signal envelope mean, signal envelope maximum, signal envelope median, signal envelope kurtosis, signal envelope skewness, ratio of signal envelope median to maximum value, number of autocorrelation function peaks, energy of the first 1 / 3 of the autocorrelation function and energy of the remaining part of the autocorrelation function.

[0020] Furthermore, the method for optimally clustering the multidimensional feature data to obtain the optimal clustering feature is:

[0021] Use SOM to cluster multidimensional feature data;

[0022] Bayesian is used to optimize the number of clusters after clustering and label them to obtain the optimal clustering features.

[0023] Furthermore, the method of acquiring real-time monitoring data, converting the optimal clustering features into signal outputs according to the real-time monitoring data and then calibrating to construct a rock fracture signal recognition model is as follows:

[0024] Convert the best clustering features into signals and plot each type of signal;

[0025] Obtain real-time monitoring data, and correspond the drawn signals to the real-time monitoring data one by one to build a recognition model.

[0026] Furthermore, the method of acquiring the displacement monitoring data and mining the association rules between the rock fracture signal and the displacement monitoring data to obtain the association rules is as follows:

[0027] Obtain displacement monitoring data and cluster the displacement monitoring data using SOM;

[0028] The number of clusters after clustering is optimized using Bayesian, and labels are marked to obtain a set of deformation events;

[0029] The rock fracture signal is taken as the front data item set I of interest to the user in the rules to be mined. F ={i1,i2,...,i n},in, And I F ≠Φ;

[0030] The deformation event set is regarded as the set of subsequent data items that users are interested in in the rules to be mined. R ={i1,i2,...,i m},in, And I R ≠Φ; the front data item set and the rear data item set have no intersection, that is, I F ∩I R=Φ;

[0031] Let X F To store the rule front item set I F The set of all proper subsets of , x i For X F The elements in X F ≠φ;

[0032] Let Y R To store the rule posterior item set I R The set of all proper subsets of , y j Y R The elements in Y R ≠φ;

[0033] Rule front item set X F and the set of rule posterior itemsets Y R The potential transaction set composed of them is V, V = {X F ×Y R}, that is, V means storage X F and Y R The set of itemsets resulting from the Cartesian product;

[0034] Determine the attribute items of the front data item set and the back data item set of the rule, and store them in the front data item set I respectively F and the set of posterior data items I R ;

[0035] Calculate the front data item set I F All elements in , and scan the transaction database to count the front data item set I F The frequency of all elements in the transaction database D;

[0036] For the front data item set I F Filter all elements in , and store the filtered elements into the front item set X according to different factor types and factor levels. F in different spaces;

[0037] Store them separately in the front item set X F Cartesian products of different types of elements in different spaces are iterated until the front item set X is obtained. F The potential front maximum frequent item set is obtained, and the transaction database is scanned to count the frequency of all potential front maximum frequent item sets appearing in the transaction database;

[0038] Set the minimum support threshold of the front item set according to the requirements, filter all potential front maximum frequent item sets, and store the obtained front maximum frequent item sets;

[0039] Find the set of posterior data items I R All elements existing in the , the elements are stored in the posterior item set set Y according to different factor types and factor levels R in different spaces;

[0040] Store them separately in the posterior item set Y R Cartesian products of different types of elements in different spaces are iterated until the posterior term set Y is obtained. R The largest frequent itemset at the rear of ;

[0041] Calculate the Cartesian product between the front maximum frequent item set and the rear maximum frequent item set, and put all the resulting combined item sets into the transaction item set V. Then, by scanning the transaction database D again, count the frequency of all non-empty true subsets in the transaction item set V.

[0042] According to the requirements, the minimum support threshold of the frequent itemset is set, and the non-empty true subsets in the transaction item set V are screened to obtain the final frequent itemset;

[0043] Calculate the confidence level of the final frequent itemsets, and use the minimum confidence threshold set by the user as the standard to filter out the required strong association rules.

[0044] A deformation and damage early warning method system for a bedding rock slope, comprising:

[0045] Data acquisition and processing module: used to acquire microseismic monitoring signals and perform signal preprocessing on the microseismic monitoring signals;

[0046] Feature extraction module: used to extract features from microseismic monitoring signals after signal preprocessing to obtain multi-dimensional feature data;

[0047] Optimal clustering module: used to perform optimal clustering on multi-dimensional feature data to obtain optimal clustering features;

[0048] Rock fracture signal recognition model building module: used to obtain real-time monitoring data, convert the optimal clustering features into signal outputs according to the real-time monitoring data, and then calibrate them to build a rock fracture signal recognition model;

[0049] Signal recognition module: used to identify rock fracture components in microseismic monitoring signals using recognition models;

[0050] Association rule acquisition module: used to acquire displacement monitoring data, and perform association rule mining on the identified rock fracture signal and displacement monitoring data to obtain association rules;

[0051] Prediction model building module: used to build an integrated prediction model using association rules and displacement monitoring data;

[0052] Data analysis and early warning module: used to analyze real-time monitoring data using an integrated prediction model and issue early warnings on the dynamics of slope deformation and damage.

[0053] A terminal device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0054] The present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0055] Compared with the prior art, the present invention has the following beneficial effects:

[0056] The present invention discloses a method for early warning of deformation and destruction of a bedding rock slope. The method obtains microseismic monitoring signals and performs signal preprocessing; performs signal preprocessing, feature extraction, and optimal clustering on the microseismic monitoring signals after signal preprocessing in sequence to obtain optimal clustering features; then, obtains real-time monitoring data, and converts the optimal clustering features into signal outputs according to the real-time monitoring data and performs calibration to construct a rock fracture signal recognition model; realizes the identification of rock fracture components to obtain rock fracture signals; and performs association rule mining on rock fracture signals and displacement monitoring data to obtain association rules; uses association rules and displacement monitoring data to construct an integrated prediction model; and realizes early warning of the dynamic deformation and destruction of the slope. The method can efficiently and accurately identify rock fracture signals, and by performing association rule mining on rock fracture signals and displacement monitoring data to construct an integrated prediction model, it realizes real-time and accurate early warning of deformation and destruction of bedding rock slopes, provides a strong scientific basis for engineering managers and decision makers, helps to formulate more reasonable and effective slope management strategies, and effectively guarantees personnel safety. The method is simple and easy to implement.

[0057] The present invention also provides a deformation and damage warning method system for bedding rock slopes. Through the setting of data acquisition and processing module, feature extraction module, optimal clustering module, rock fracture signal recognition model construction module, signal recognition module, association rule acquisition module, prediction model construction module and data analysis and warning module, microseismic monitoring signal reading, preprocessing, feature extraction, feature optimal clustering, rock fracture signal recognition model construction, rock fracture signal feature and slope displacement correlation mining, monitoring and early warning model construction and real-time data monitoring and early warning are realized; the integrated model combines association rule mining and temporal convolutional network (Temporal Convolutional Networks, TCN), including early warning branch and displacement prediction branch. The early warning branch uses rock fracture signal features and historical association rules to predict displacement changes in real time and realize preliminary early warning. When the signal feature is detected to reach the threshold of sudden displacement, the model sends out an early warning signal and recommends emergency measures such as evacuation of personnel. At the same time, the displacement prediction branch uses TCN to model each type of deformation event separately, predicts future displacement changes by inputting displacement data, and provides quantitative displacement prediction values. After issuing the first warning, the model compares the displacement prediction results with the expected changes to assess whether further prevention and control measures are needed. If the displacement prediction results meet the signs of sudden displacement changes in the initial warning, a more comprehensive prevention and control plan should be implemented, achieving real-time and accurate warning of deformation and failure of the bedding rock slope.

[0058] The present invention also provides a terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program. The device has a simple structure, low modification cost, and small resource occupation.

[0059] A computer-readable storage medium stores a computer program, wherein the computer program implements the steps of the above method when executed by a processor. The storage medium has good portability and strong versatility. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The present invention is a flow chart of a method for early warning of deformation and damage of a bedding rock slope.

[0061] Figure 2 The present invention is a flowchart of a method for early warning of deformation and damage of a bedding rock slope according to an embodiment of the present invention.

[0062] Figure 3 The present invention is a structural diagram of a deformation and damage early warning system for a bedding rock slope according to an embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0064] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0065] The present invention is further described in detail below in conjunction with specific embodiments, which are intended to explain the present invention rather than to limit it.

[0066] Example 1 See also Figure 1 The present invention discloses a method for early warning of deformation and damage of a bedding rock slope, comprising:

[0067] S1: Acquire microseismic monitoring signals and perform signal preprocessing on the microseismic monitoring signals;

[0068] S2: extracting features from the microseismic monitoring signal after signal preprocessing to obtain multi-dimensional feature data;

[0069] S3: Optimally cluster the multi-dimensional feature data to obtain the optimal clustering features;

[0070] S4: acquiring real-time monitoring data, converting the optimal clustering features into signal outputs according to the real-time monitoring data, and then calibrating them to build a rock fracture signal recognition model;

[0071] S5: using the recognition model to identify the rock fracture components in the microseismic monitoring signal and obtain the rock fracture signal;

[0072] S6: Acquire displacement monitoring data, and perform association rule mining between rock fracture signals and displacement monitoring data to obtain association rules;

[0073] S7: Using association rules and displacement monitoring data, an integrated prediction model is constructed;

[0074] S8: Use the integrated prediction model to analyze the real-time monitoring data and issue early warnings on the dynamics of slope deformation and failure.

[0075] This method can efficiently and accurately identify rock fracture signals, and by mining the association rules between rock fracture signals and displacement monitoring data, an integrated prediction model is constructed to achieve real-time and accurate early warning of deformation and damage of bedding rock slopes, providing a strong scientific basis for project managers and decision makers, helping to formulate more reasonable and effective slope management strategies, and effectively ensuring personnel safety. The method is simple and easy to implement.

[0076] Example 2

[0077] See also Figure 2 The present invention provides a method for early warning of deformation and damage of a bedding rock slope, comprising:

[0078] S1: Obtain microseismic monitoring signals and perform signal preprocessing on the microseismic monitoring signals, specifically:

[0079] S1.1: Acquire microseismic monitoring data acquired in the field, and convert the microseismic monitoring data acquired in the field into a signal form to obtain a microseismic monitoring signal;

[0080] S1.2: normalize the microseismic monitoring signal to obtain normalized data, that is, normalize the microseismic monitoring signal to a value within a range of -1 to 1;

[0081] S1.3: Perform discrete Fourier transform on the normalized data to obtain the spectrum characteristics of the microseismic monitoring signal and complete signal preprocessing, that is, select a suitable small window to perform discrete Fourier transform (DFT) on the normalized data to obtain the spectrum characteristics of the signal. By analyzing the spectrum, the main frequency components and energy distribution in the signal can be identified, which helps to discover the characteristic frequency and amplitude of the microseismic event. The discrete Fourier transform method is:

[0082]

[0083] Where X(k) is the discrete Fourier transform (DFT) of the signal, x(n) is the input sequence, representing the signal in the time domain, n is the index in the time domain, ranging from 0 to N-1, N is the signal length, and i is the imaginary unit. is the normalization factor, ensuring the periodicity of the frequency, is a complex exponential function used to convert signals from the time domain to the frequency domain.

[0084] In actual operation, in order to perform discrete wavelet transform on the normalized data to achieve decomposition and reconstruction of the microseismic signal, the characteristics of the signal in different frequency ranges can be analyzed through the reconstruction process, specifically:

[0085]

[0086] Among them, A1(n) is the first-level approximation coefficient, D1(n) is the first-level detail coefficient, f(k) is the input microseismic signal, h(kn) is the low-pass filter (scaling function), g(kn) is the high-pass filter (wavelet function), and k is the number of discrete points.

[0087] S2: Extract features from the microseismic monitoring signal after signal preprocessing to obtain multi-dimensional feature data. The multi-dimensional feature data, that is, the acquisition method, is shown in the following table:

[0088]

[0089]

[0090]

[0091]

[0092] In the table, signal refers to the normalized microseismic monitoring signal. Matlab (MATrix LABoratory) is a mathematical computing software developed by MathWorks. It is mainly used for algorithm development, data visualization, data analysis and high-level technical computing language and interactive environment for numerical calculation. STFT is short-time Fourier transform; FFT is fast Fourier transform.

[0093] S3: Optimally cluster the multi-dimensional feature data to obtain the optimal clustering features, specifically:

[0094] S3.1: Clustering the multi-dimensional feature data using a self-organizing map (SOM), that is, the SOM includes an input layer and a competition layer, and the input layer uses the multi-dimensional feature data as an input vector x;

[0095] S3.1.1: Initialize the weight vector so that it has the same dimension as the input vector x. The method is: randomly initialize the weight vector to a two-dimensional array W = {w ij}, where each weight vector w ij is a vector with the same dimension as the input vector x. Assuming the dimension of the input vector is n, then w ij∈R n . Set the learning rate and neighborhood function: Set the initial learning rate a(0) and the neighborhood function θ(t);

[0096] S3.1.2: Input vector x;

[0097] S3.1.3: Compute the Euclidean distance between each weight vector and the input vector x as follows:

[0098]

[0099] Where n is the number of features, x k is the kth component of the input vector x, w ij,k is the weight vector w ij The kth component of ;

[0100] S3.1.4: Select the weight vector with the shortest Euclidean distance to the input vector x as the best matching unit (BMU);

[0101] S3.1.5: Update the weight vectors of the best matching unit and its neighboring domains by:

[0102] w ij (t+1)=w ij (t)+α(t)θ ij (t)(xw ij (t)) (4)

[0103] Among them, w ij (t) is the neuron weight vector at time t, w ij (t+1) is the updated neuron weight vector, x is the input vector, α(t) is the learning rate, and θ ij (t) is a neighborhood function; the neighborhood function is obtained by the following method:

[0104]

[0105] Among them, r ij and r i*j* are the coordinate positions of the weight vector, σ(t) is the neighborhood radius, which decreases with time;

[0106] S3.1.6: Calculate the Euclidean distance between each updated weight vector and the input vector x, and iterate until the number of iterations reaches a threshold or the weight vector after the update changes less than a certain threshold, and complete the SOM training, that is, iterate S3.1.3 and S3.1.3 until the number of iterations reaches a threshold or the weight vector after the update changes less than a certain threshold, and complete the SOM training;

[0107] S3.1.7: According to the trained SOM, each input vector x is assigned to the node corresponding to the weight vector with the smallest distance to it, thus completing the clustering of the multi-dimensional feature data;

[0108] S3.2: Use Bayesian to optimize the number of clusters after clustering and label them to obtain the optimal clustering features. The method is:

[0109] S3.2.1: Define the objective function as the silhouette coefficient that depends on the number of categories k, which is used to evaluate the clustering effect. The method is:

[0110]

[0111] Among them, a(i) is the average distance between data point i and other data points in the same cluster. b(i) is the average distance between data point i and the data point in the nearest cluster. The silhouette coefficient ranges from [-1,1]. Values ​​close to 1 indicate good clustering results, values ​​close to 0 indicate poor clustering results and possible overlaps, and values ​​close to -1 indicate that the data is incorrectly assigned to an incorrect cluster.

[0112] S3.2.2: Use Gaussian Process Regression (GPR) as a surrogate model to approximate the objective function by:

[0113] f(x)~GP(m(x),k(x,x′))(7)

[0114] Where GP stands for Gaussian process, m(x) is the mean function, f(x) is the random function, and k(x,x′) is the kernel function;

[0115] S3.2.3: Select the expected improvement as the acquisition function by:

[0116] EI(x)=E[max(f * -f(x),0)] (8)

[0117] Among them, f * is the currently known optimal value, f(x) is the predicted distribution of the objective function value at position x, and EI(x) is the acquisition function;

[0118] S3.2.4: Optimize the acquisition function on the surrogate model to find the next most promising evaluation point x next , update the model to obtain the new objective function value f(x next ) is used to update the Gaussian process model;

[0119] S3.2.5: Iteration, i.e. repeating S3.2.3 and S3.2.4 until the calculated silhouette coefficient is the largest, thus completing the optimization of the number of clusters after clustering;

[0120] S4: Obtain real-time monitoring data, convert the optimal clustering features into signal outputs according to the real-time monitoring data, and then calibrate them to build a rock fracture signal recognition model, specifically:

[0121] S4.1: Convert the optimal clustering features into signals and plot each type of signal;

[0122] S4.2: Acquire real-time monitoring data, and correspond the drawn signals to the real-time monitoring data one by one, and build a recognition model. For example, by comparing the drawn signals with the video monitoring data, the clustered categories 1 to 5 are labeled as workers working, excavators working, drills working, rock breaking, and vehicles, respectively, so that they correspond one to one;

[0123] Use random forest to build a recognition model: Use previously classified and labeled data and rock fracture signal data obtained through physical model experiments as the input layer (also called training set), create a 5000-tree random forest model, and then extract features from new microseismic monitoring signals and put them into the constructed random forest model as a new input layer to identify new signals. The specific algorithm is as follows:

[0124] Bagging: Random forest is a bagging algorithm that uses decision trees as estimators. The training set is D = {(x1, y1), (x2, y2), ..., (x n ,y n )}, extract all N samples (with replacement) to construct each decision tree T i , repeat 5000 times and get 5000 lesson trees.

[0125] Splitting of decision tree: Use the above tree to split. When each node splits, select m features from d (number of features) features for evaluation, and the feature that maximizes the Gini coefficient is used as the split feature.

[0126] Voting for classification problems: For new input x (new signal), each tree T i Will output a classification result h i (x), the final classification result H(x) of random forest is determined by the following formula (9):

[0127]

[0128] Where argmax is the value of C where the expression reaches its maximum value, I is the exponential function. When the condition in the brackets is true, the indicator function value is 1, otherwise it is 0, C is the set of possible classification result categories, k is the number of decision trees, and h is the number of possible classification result categories. i (x) is the i-th decision tree T i The classification result of input x, Ti The i-th decision tree.

[0129] S5: Use the identification model to identify the rock fracture components in the microseismic monitoring signal and obtain the rock fracture signal, specifically:

[0130] The elbow method is used to rank the importance of the features of microseismic monitoring signals, and the features before the inflection point are taken as important features:

[0131] The specific algorithm principle is as follows:

[0132] S5.1: Feature importance score sorting, for example: for 44 features, the feature importance scores are sorted in descending order as {I1,I2,…,I 44}; Among them, I i represents the importance score of the i-th microseismic signal feature;

[0133] S5.2: Calculate the first-order difference. The method for calculating the difference between adjacent feature scores is:

[0134] ΔI i =I i -I i+1 i=1,2,…,43 (10)

[0135] Among them, ΔI i represents the difference between the i-th feature score and the i+1-th feature score;

[0136] S5.3: Calculate the second-order difference and the rate of change of the score difference as follows:

[0137] Δ 2 I i =ΔI i -ΔI i+1 i=1,2,…,42 (11)

[0138] Among them, Δ 2 I i Represents the change in the rate of change of the i-th feature score.

[0139] S5.4: Find the maximum index point. Find the index point with the largest second-order difference. The method is:

[0140] k=argmax(Δ 2 I i )+1 (12)

[0141] Among them, k is the index point with the largest second-order difference;

[0142] S5.5: Select important features, use the maximum index point of the second-order difference found as the inflection point, and take the features before the inflection point as important features to complete the acquisition of rock fracture signals.

[0143] S6: Obtain displacement monitoring data, and perform association rule mining on rock fracture signals and displacement monitoring data to obtain association rules, specifically:

[0144] S6.1: Obtain displacement monitoring data and cluster the displacement monitoring data using SOM. The clustering method refers to the clustering method of multi-dimensional feature data using Self-Organizing Map (SOM) in S3.1;

[0145] S6.2: Optimize the number of clusters after clustering using Bayesian, and mark them with labels to obtain a set of deformation events; the optimization method refers to the method of optimizing the number of clusters after clustering using Bayesian, and marking them with labels to obtain the optimal clustering characteristics in S3.2; cluster the acceleration, cumulative displacement, vector angle and velocity information of the displacement monitoring data through the above two steps to obtain a set of deformation events, and distinguish the deformation events of the slope;

[0146] The improved Apriori algorithm is used to mine the correlation between the important features of rock fracture signals extracted from microseismic monitoring signals and slope deformation events.

[0147] S6.3: Rock fracture signals are used as the front data item set I of interest to the user in the rules to be mined F ={i1,i2,...,i n},in, And I F ≠Φ;

[0148] The deformation event set is regarded as the set of subsequent data items that users are interested in in the rules to be mined. R ={i1,i2,...,i m},in, And I R ≠Φ; the front data item set and the rear data item set have no intersection, that is, I F ∩I R =Φ;

[0149] Let X F To store the rule front item set I F The set of all proper subsets of , x i For X F The elements in X F ≠φ;

[0150] Let Y RTo store the rule posterior item set I R The set of all proper subsets of , y j Y R The elements in Y R ≠φ;

[0151] Rule front item set X F and the set of rule posterior itemsets Y R The potential transaction set composed of them is V, V = {X F ×Y R}, that is, V means storage X F and Y R The set of itemsets resulting from the Cartesian product is:

[0152]

[0153] S6.4: Determine the attribute items of the front data item set and the back data item set of the rule, and store them in the front data item set I respectively F and the set of posterior data items I R ;

[0154] S6.5: Calculate the front data item set I F All elements in , and scan the transaction database to count the front data item set I F The frequency of all elements in the transaction database D;

[0155] S6.6: For the front data item set I F Filter all elements in , and store the filtered elements into the front item set X according to different factor types and factor levels. F in different spaces;

[0156] S6.7: Store them into the front item set X F Cartesian products of different types of elements in different spaces are iterated until the front item set X is obtained. F The potential front maximum frequent item set is obtained, and the transaction database is scanned to count the frequency of all potential front maximum frequent item sets appearing in the transaction database;

[0157] S6.8: Set the minimum support threshold of the front item set according to the requirements, screen all potential front maximum frequent item sets, and store the obtained front maximum frequent item sets;

[0158] S6.9: Find the set of posterior data items I R All elements existing in the , the elements are stored in the posterior item set set Y according to different factor types and factor levels R in different spaces;

[0159] S6.10: Store them into the posterior item set Y R Cartesian products of different types of elements in different spaces are iterated until the posterior term set Y is obtained. R The largest frequent itemset at the rear of ;

[0160] S6.11: Calculate the Cartesian product between the front maximum frequent item set and the rear maximum frequent item set, and put all the resulting combined item sets into the transaction item set V, and count the frequency of all non-empty true subsets in the transaction item set V by scanning the transaction database D again;

[0161] S6.12: Set the minimum support threshold of the frequent itemset according to the requirements, filter the non-empty true subsets in the transaction item set V, and obtain the final frequent itemset;

[0162] S6.13: Calculate the confidence level of the final frequent itemset, and use the minimum confidence threshold set by the user as the standard to filter out the required strong association rules.

[0163] S7: Using association rules and displacement monitoring data, an integrated prediction model is constructed, namely:

[0164] Association rules are used to determine the rock fracture signal threshold corresponding to the deformation event; the integrated prediction model combines association rule mining and temporal convolutional network (TCN), and contains two key branches: the early warning branch and the displacement prediction branch: the early warning branch uses the characteristics of rock fracture signals and historical association rules to predict displacement changes in real time and achieve preliminary early warning. When the signal characteristics are detected to reach the threshold of sudden displacement, the model issues a warning signal and recommends emergency measures such as evacuation. At the same time, the displacement prediction branch uses TCN to model each type of deformation event separately, predicts future displacement changes by inputting displacement data, and provides quantitative displacement prediction values. After issuing the first step of early warning, the model compares the displacement prediction results with the expected changes to evaluate whether further prevention and control measures are needed. If the displacement prediction results meet the signs of sudden displacement in the preliminary warning, a more comprehensive prevention and control plan should be implemented. The specific methods are:

[0165] Input layer: input the processed rock fracture signal characteristic data and displacement data correlation analysis results.

[0166] Convolutional layer: Causal convolutional layer, which applies causal convolution kernel to extract time series features. The method is:

[0167]

[0168] Among them, O(t) is the convolution output, I(tk) is the input feature map, K(k) is the convolution kernel, and bk is bias;

[0169] Extended convolution layer: Introduce extended causal convolution, the method is:

[0170]

[0171] Among them, r is the expansion rate, and other symbols are the same as causal convolution;

[0172] Activation function: Use the ReLU activation function to introduce nonlinearity, the method is:

[0173] f(x)=max(0,x) (16)

[0174] Among them, f(x) is the output of the ReLU activation function, and x is the input of the activation function;

[0175] Pooling layer: The sampling method of introducing the pooling layer is:

[0176]

[0177] Among them, O(i,j,c) is the pooled output of the cth channel and position (i,j) in the pooling window, p is the pooling window size, m and n are the indices of traversing the pooling window, and I(p·i+m,p·j+n,c) corresponds to the element value of each position in the pooling window in the input feature map;

[0178] Residual connection: In order to speed up training and prevent gradient disappearance, residual connection is introduced as follows:

[0179] O(t)=f(x)+I(t) (18)

[0180] Among them, f(x) is the output after the convolution layer and activation function, and I(t) is the input feature map;

[0181] Fully connected layer: flattens the features of the previous layer and connects them to all output neurons by:

[0182] y=W·x+b (19)

[0183] Among them, W is the weight matrix, x is the flattened input feature, b is the bias, and y is the output prediction value, i.e., the deformation event.

[0184] Regression output layer: Outputs predicted deformation events and their probabilities.

[0185] S8: Use the integrated prediction model to analyze the real-time monitoring data and issue early warnings on the dynamics of slope deformation and failure.

[0186] Through real-time monitoring of microseismic signals and displacement data, slope deformation events are predicted and warned. First, the real-time microseismic signals are used to identify rock fracture signals through the constructed recognition model; then, key features are extracted from the identified rock fracture signals, and then combined with the strong association rules previously mined, slope deformation events and their probability of occurrence are warned; finally, the displacement data is input into the TCN model to predict the specific displacement value; check whether it is consistent with the first step of the warning situation, so as to take the next step.

[0187] Example 3

[0188] See also Figure 3 The present invention provides a method system for early warning of deformation and damage of bedding rock slope, comprising:

[0189] Data acquisition and processing module: used to acquire microseismic monitoring signals and perform signal preprocessing on the microseismic monitoring signals;

[0190] Feature extraction module: used to extract features from microseismic monitoring signals after signal preprocessing to obtain multi-dimensional feature data;

[0191] Optimal clustering module: used to perform optimal clustering on multi-dimensional feature data to obtain optimal clustering features;

[0192] Rock fracture signal recognition model building module: used to obtain real-time monitoring data, convert the optimal clustering features into signal outputs according to the real-time monitoring data, and then calibrate them to build a rock fracture signal recognition model;

[0193] Signal recognition module: used to identify rock fracture components in microseismic monitoring signals using recognition models;

[0194] Association rule acquisition module: used to acquire displacement monitoring data, and perform association rule mining on the identified rock fracture signal and displacement monitoring data to obtain association rules;

[0195] Prediction model building module: used to build an integrated prediction model using association rules and displacement monitoring data;

[0196] Data analysis and early warning module: used to analyze real-time monitoring data using an integrated prediction model and issue early warnings on the dynamics of slope deformation and damage.

[0197] The integrated model combines association rule mining and temporal convolutional network (TCN), and includes an early warning branch and a displacement prediction branch. The early warning branch uses rock fracture signal characteristics and historical association rules to predict displacement changes in real time and achieve preliminary early warning. When the signal characteristics are detected to reach the threshold of sudden displacement, the model issues an early warning signal and recommends emergency measures such as evacuation. At the same time, the displacement prediction branch uses TCN to model each type of deformation event separately, predicts future displacement changes by inputting displacement data, and provides quantitative displacement prediction values. After issuing the first step of early warning, the model compares the displacement prediction results with the expected changes to evaluate whether further prevention and control measures are needed. If the displacement prediction results meet the signs of sudden displacement in the preliminary warning, a more comprehensive prevention and control plan should be implemented, achieving real-time and accurate early warning of deformation and damage of bedding rock slopes.

[0198] The present invention provides a terminal device including: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned method embodiments are implemented. Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0199] The computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to accomplish the present invention.

[0200] The terminal device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0201] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0202] The memory may be used to store the computer programs and / or modules, and the processor implements various functions of the terminal device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory.

[0203] If the module / unit integrated in the terminal device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0204] In summary, the present invention provides a deformation and damage warning method, system, equipment and storage medium for bedding rock slopes. By establishing a deformation and damage warning model for bedding rock slopes, microseismic monitoring signal reading, preprocessing, feature extraction, feature optimal clustering, signal calibration, rock fracture signal recognition model construction, rock fracture signal feature and slope displacement correlation mining, etc. are realized, and finally real-time data monitoring and accurate warning are realized; the model has a high degree of automation, which improves the efficiency and accuracy of rock slope deformation and damage monitoring and warning, provides a strong scientific basis for project managers and decision makers, helps to formulate more reasonable and effective slope management strategies, and effectively ensures personnel safety.

[0205] The above description is only a preferred embodiment of the present invention and is not intended to impose any limitation on the technical solution of the present invention. Those skilled in the art should understand that, without departing from the spirit and principles of the present invention, the technical solution can also be subjected to several simple modifications and substitutions, and these modifications and substitutions are also within the scope of protection covered by the claims.

Claims

1. A method for early warning of deformation and damage of bedding rock slope, characterized in that: include: Acquire microseismic monitoring signals and perform signal preprocessing on the microseismic monitoring signals; Extract features from the microseismic monitoring signals after signal preprocessing to obtain multi-dimensional feature data; Perform optimal clustering on multi-dimensional feature data to obtain optimal clustering features; Acquire real-time monitoring data, convert the optimal clustering features into signal outputs according to the real-time monitoring data, and then calibrate them to build a rock fracture signal recognition model; The rock fracture components in the microseismic monitoring signal are identified using the identification model to obtain the rock fracture signal; Acquire displacement monitoring data, and perform association rule mining between rock fracture signals and displacement monitoring data to obtain association rules; Using association rules and displacement monitoring data, an integrated prediction model is constructed; An integrated prediction model is used to analyze real-time monitoring data and provide early warning of the dynamics of slope deformation and damage.

2. The method for early warning of deformation and damage of bedding rock slope according to claim 1, characterized in that: The method for obtaining the microseismic monitoring signal is: obtaining microseismic monitoring data monitored in the field, and converting the microseismic monitoring data monitored in the field into a signal form to obtain the microseismic monitoring signal.

3. The method for early warning of deformation and damage of bedding rock slope according to claim 1, characterized in that: The method for preprocessing the microseismic monitoring signal is as follows: Normalizing the microseismic monitoring signal to obtain normalized data; The normalized data is subjected to discrete Fourier transform to obtain the frequency spectrum characteristics of the microseismic monitoring signal and complete the signal preprocessing.

4. The method for early warning of deformation and damage of bedding rock slope according to claim 1, characterized in that: The multidimensional feature data includes original signal kurtosis, original signal skewness, original signal energy, ratio of signal rise time to fall time, signal envelope mean, signal envelope maximum, signal envelope median, signal envelope kurtosis, signal envelope skewness, ratio of signal envelope median to maximum value, number of autocorrelation function peaks, energy of the first 1 / 3 part of the autocorrelation function and energy of the remaining part of the autocorrelation function.

5. The method for early warning of deformation and damage of bedding rock slope according to claim 1, characterized in that: The method for optimally clustering multidimensional feature data to obtain optimal clustering features is: Use SOM to cluster multidimensional feature data; Bayesian is used to optimize the number of clusters after clustering and label them to obtain the optimal clustering features.

6. The method for early warning of deformation and damage of bedding rock slope according to claim 1, characterized in that: The method of acquiring real-time monitoring data, converting the optimal clustering features into signal outputs according to the real-time monitoring data and then calibrating the signal outputs to construct a rock fracture signal recognition model is as follows: Convert the best clustering features into signals and plot each type of signal; Obtain real-time monitoring data, and correspond the drawn signals to the real-time monitoring data one by one to build a recognition model.

7. The method for early warning of deformation and damage of bedding rock slope according to claim 1, characterized in that: The method of acquiring displacement monitoring data and mining association rules between rock fracture signals and displacement monitoring data to obtain association rules is as follows: Obtain displacement monitoring data and cluster the displacement monitoring data using SOM; The number of clusters after clustering is optimized using Bayesian, and labels are marked to obtain a set of deformation events; The rock fracture signal is taken as the front data item set I of interest to the user in the rules to be mined. F ={i1,i2,...,i n },in, And I F ≠Φ; The deformation event set is regarded as the set of subsequent data items that users are interested in in the rules to be mined. R ={i1,i2,...,i m },in, And I R ≠Φ; the front data item set and the rear data item set have no intersection, that is, I F ∩I R =Φ; Let X F To store the rule front item set I F The set of all proper subsets of , x i For X F The elements in X F ≠φ; Let Y R To store the rule posterior item set I R The set of all proper subsets of , y j Y R The elements in Y R ≠φ; Rule front item set set X F and the set of rule posterior itemsets Y R The potential transaction set composed of them is V, V = {X F ×Y R }, that is, V means storage X F and Y R The set of itemsets resulting from the Cartesian product; Determine the attribute items of the front data item set and the back data item set of the rule, and store them in the front data item set I respectively F and the set of posterior data items I R ; Calculate the front data item set I F All elements in , and scan the transaction database to count the front data item set I F The frequency of all elements in the transaction database D; For the front data item set I F Filter all elements in , and store the filtered elements into the front item set X according to different factor types and factor levels. F in different spaces; Store them separately in the front item set X F Cartesian products of different types of elements in different spaces are iterated until the front item set X is obtained. F The potential front maximum frequent item set is obtained, and the transaction database is scanned to count the frequency of all potential front maximum frequent item sets appearing in the transaction database; Set the minimum support threshold of the front item set according to the requirements, filter all potential front maximum frequent item sets, and store the obtained front maximum frequent item sets; Find the set of posterior data items I R All elements existing in the , the elements are stored in the posterior item set set Y according to different factor types and factor levels R in different spaces; Store them separately in the posterior item set Y R Cartesian products of different types of elements in different spaces are iterated until the posterior term set Y is obtained. R The largest frequent itemset at the rear of ; Calculate the Cartesian product between the front maximum frequent item set and the rear maximum frequent item set, and put all the resulting combined item sets into the transaction item set V. Then, by scanning the transaction database D again, count the frequency of all non-empty true subsets in the transaction item set V. According to the requirements, the minimum support threshold of the frequent itemset is set, and the non-empty true subsets in the transaction item set V are screened to obtain the final frequent itemset; Calculate the confidence level of the final frequent itemsets, and use the minimum confidence threshold set by the user as the standard to filter out the required strong association rules.

8. A method and system for early warning of deformation and damage of bedding rock slope, characterized in that: include: Data acquisition and processing module: used to acquire microseismic monitoring signals and perform signal preprocessing on the microseismic monitoring signals; Feature extraction module: used to extract features from microseismic monitoring signals after signal preprocessing to obtain multi-dimensional feature data; Optimal clustering module: used to perform optimal clustering on multi-dimensional feature data to obtain optimal clustering features; Rock fracture signal recognition model building module: used to obtain real-time monitoring data, convert the optimal clustering features into signal outputs according to the real-time monitoring data, and then calibrate them to build a rock fracture signal recognition model; Signal recognition module: used to identify rock fracture components in microseismic monitoring signals using recognition models; Association rule acquisition module: used to acquire displacement monitoring data, and perform association rule mining on the identified rock fracture signal and displacement monitoring data to obtain association rules; Prediction model building module: used to build an integrated prediction model using association rules and displacement monitoring data; Data analysis and early warning module: used to analyze real-time monitoring data using an integrated prediction model and issue early warnings on the dynamics of slope deformation and damage.

9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images