A tunnel blasting charge structure identification recording method, device, equipment and medium

Abstract

The application provides a tunnel blasting charge structure identification recording method, device, equipment and medium, comprising: acquiring a blasting signal set generated by a wireless vibration measuring unit during tunnel peripheral charge structure blasting; processing the blasting signal set to generate multiple sets of dimension reduction data; calling a trained KNN model to process the dimension reduction data to generate a charge structure type corresponding to the blasting signal set, wherein the KNN model automatically records vibration data corresponding to the charge structure type generated during tunnel blasting. In addition, most of the methods on the market for identifying and recording charge structure types and vibration data according to vibration characteristics are manually recorded and supervised in real time, which is labor-intensive and prone to omissions, errors, type identification errors and other problems, which may affect the subsequent tunnel blasting excavation vibration reduction and loss reduction effect, thereby reducing the self-stability of the surrounding rock.

Description

Technical Field

[0001] This invention relates to the field of tunnel blasting construction, specifically to a method, device, equipment, and medium for identifying and recording the structure of tunnel blasting charges. Background Technology

[0002] The main method for tunnel excavation in the market is drill-and-blast construction. In tunnel blasting construction, how to use a scientific and reasonable charge structure to control the over- and under-excavation of the surrounding area and reduce the damage to the surrounding rock is a key focus for engineers. Among them, adjusting and optimizing the charge structure of the holes around the tunnel can significantly reduce the disturbance of the blasting to the remaining surrounding rock of the tunnel, which is crucial for controlling the quality of tunnel blasting.

[0003] Currently, there is a lack of quantitative indicators for the vibration characteristics of different charge structures in tunnel perimeter boreholes. Traditional tunnel vibration monitoring adopts a "test as you go" approach, which has drawbacks such as frequent setting up of measuring points and poor data acquisition timeliness. Most existing methods for identifying and recording charge structure types and vibration data based on vibration characteristics rely on manual real-time recording and supervision. This is labor-intensive, prone to omissions, errors, and incorrect type identification, and is susceptible to the subjective influence of the recorders, failing to effectively guide production practice. This falls far short of meeting the urgent requirements of engineering and management personnel for identifying and classifying charge structures in perimeter boreholes, and can easily lead to problems such as affecting the vibration reduction and damage reduction effects of subsequent tunnel blasting and excavation, thereby reducing the self-stabilizing capacity of the surrounding rock.

[0004] In view of the above, this application is hereby submitted. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method, device, equipment and medium for identifying and recording the structure of blasting charges in tunnels. This method can effectively solve the problem that most existing methods for identifying and recording the type of blasting charge structure and vibration data based on vibration characteristics rely on manual real-time recording and supervision. This is labor-intensive and prone to omissions, errors, and incorrect type identification, which can affect the vibration reduction and loss reduction effect of subsequent tunnel blasting excavation and reduce the self-stabilizing capacity of the surrounding rock.

[0006] This invention provides a method for identifying and recording the structure of explosive charges in tunnel blasting, comprising:

[0007] Acquire the blasting signal set generated by the blasting of the explosive charge structure around the tunnel, collected by the wireless vibration measurement unit;

[0008] The blasting signal set is processed to generate multiple sets of dimensionality-reduced data;

[0009] The trained KNN model is called to process the dimension-reduced data, to generate a charge structure type corresponding to the blasting signal set, wherein an automatic acquisition unit of the KNN model automatically records vibration data corresponding to the charge structure type generated in the tunnel blasting process.

[0010] Preferably, the blasting signal set is processed to generate multiple groups of dimension-reduced data, specifically:

[0011] AOK time-frequency spectrum calculation is performed on the data of the blasting signal set to generate a time-frequency image.

[0012] The time-frequency image is calculated using a gray level co-occurrence algorithm to generate a gray level co-occurrence matrix corresponding to the time-frequency image.

[0013] The gray level co-occurrence matrix is subjected to PCA dimension reduction processing to establish three principal component feature vectors reflecting the characteristics of the time-frequency image after dimension reduction, and multiple groups of dimension-reduced data are generated.

[0014] Preferably, the time-frequency image is calculated using a gray level co-occurrence algorithm to generate a gray level co-occurrence matrix corresponding to the time-frequency image, specifically:

[0015] The time-frequency image is subjected to adaptive optimal kernel analysis to generate an amplitude gray scale of the time-frequency spectrum image.

[0016] According to the formula m2 = m1 + Δm, n2 = n1 + Δn, the amplitude gray scale of the time-frequency spectrum image is calculated to generate a matrix co-occurrence, wherein i, j = 1, 2,..., L, < > is a co-occurrence value generated at element (i, j) in the image matrix, Δm and Δn are the displacements of the element to the left and right and up and down of element (i, j) respectively, f(x, y) is the gray scale of the blasting signal in the time-frequency domain two-dimensional time-frequency spectrum digital image, and its size is L m × L n , L m and L n are the spatial positions of the time-frequency spectrum pixels, L m = 1, 2,..., N m , L n = 1, 2,..., N n , (m1, n1) is the coordinate of a point in the image; (m2, n2) is the coordinate of another point.

[0017] Each value of the co-occurrence matrix obtained by calculating I(i, j) is divided by the maximum element in the matrix to perform co-occurrence matrix normalization processing, to generate a gray level co-occurrence matrix of the time-frequency spectrum image matrix of the blasting vibration signal of different charge structures.

[0018] Preferably, the gray-level co-occurrence matrix is ​​subjected to PCA dimensionality reduction processing, specifically as follows:

[0019] According to the formula Calculate the mean vector of the blasting signal set, where i = 1, 2, ..., N is the number of feature directions extracted from the gray-level co-occurrence matrix of the time-frequency image; k = 1, 2, ..., P is the number of features in each feature direction of the gray-level co-occurrence matrix of the time-frequency image;

[0020] According to the formula Calculate the standard matrix of the original high-dimensional data after zero-mean normalization, where,

[0021] According to the formula Calculate the covariance matrix of the blasting signal set;

[0022] Calculate the eigenvalues ​​i = λ1, λ2, ..., λ of the covariance matrix. P ,λ1,>>...>>λ P and their corresponding attribute vectors ∝1, ∝2, ..., ∝ P And sort the eigenvalues ​​from largest to smallest, according to the formula. Calculate the contribution rate σ of each principal component. i and cumulative contribution rate δ;

[0023] Take the d largest attribute feature values ​​(d < P), and assign the corresponding attribute feature vectors ∝1, ∝2, ..., ∝ d Form the transformation matrix A = [∝1, ∝2, ..., ∝ d ], and according to the formula Y = A T X calculates the image feature matrix data X and reduces it to d-dimensional data Y, generating multiple sets of dimensionality-reduced data.

[0024] Preferably, before calling the trained KNN model to process the dimensionality-reduced data, the method further includes:

[0025] Acquire multiple sets of blasting signals of different charge structure types generated by the blasting of different charge structures around the tunnel, collected by the wireless vibration measurement unit;

[0026] The blasting signal sets of each of the different charge structure types are processed to generate multiple sets of original datasets for recognition, classification, image analysis, and dimensionality reduction.

[0027] Randomly select multiple sets of original datasets for identification and classification image analysis and dimensionality reduction with different charge structure types and the same preset number, and call a PSO-ELM-based KNN model to train each of the original datasets for identification and classification image analysis and dimensionality reduction, generate training error values, until the training error values ​​are less than a preset value, and generate a KNN training model.

[0028] Randomly screen different charge structure types and preset the same number of groups of recognition classification image analysis dimension reduction original data sets, and call the KNN training model to perform accuracy verification calculation to generate an average accuracy until the average accuracy reaches a preset value, and generate a KNN model.

[0029] Preferably, the blasting signal set of different charge structure types includes a continuous charge structure blasting signal set, a water interval charge structure blasting signal set, and an air interval charge structure blasting signal set.

[0030] The application further provides a tunnel blasting charge structure identification recording device, which comprises:

[0031] A data acquisition unit is configured to acquire a blasting signal set generated by a wireless vibration measurement unit during tunnel peripheral charge structure blasting;

[0032] A data processing unit is configured to process the blasting signal set to generate a plurality of groups of dimension reduction data.

[0033] An identification recording unit is configured to call a trained KNN model to process the dimension reduction data to generate a charge structure type corresponding to the blasting signal set, wherein the automatic acquisition unit of the KNN model automatically records vibration data corresponding to the charge structure type generated during tunnel blasting.

[0034] Preferably, the data processing unit is specifically configured to:

[0035] perform AOK time-frequency spectrum calculation on the data of the blasting signal set to generate a time-frequency image;

[0036] perform calculation on the time-frequency image by using a gray level co-occurrence algorithm to generate a gray level co-occurrence matrix corresponding to the time-frequency image;

[0037] perform PCA dimension reduction processing on the gray level co-occurrence matrix to establish three principal component feature vectors reflecting characteristics of the time-frequency image after dimension reduction, and generate a plurality of groups of dimension reduction data.

[0038] The application further provides a tunnel blasting charge structure identification recording device, which comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the tunnel blasting charge structure identification recording method according to any one of the above when executing the computer program.

[0039] The application further provides a readable storage medium storing a computer program, and the computer program can be executed by a processor of a device where the storage medium is located to implement the tunnel blasting charge structure identification recording method according to any one of the above.

[0040] To sum up, the tunnel blasting charge structure identification recording method, device, equipment and medium provided by the embodiment, after collecting the blasting signal, the tunnel blasting charge structure identification recording method adopts AOK time-frequency spectrum calculation, gray level co-occurrence algorithm and PCA dimension reduction processing in sequence to process the data, and calls the trained KNN model to calculate the processed data, the KNN classifier in the KNN model can identify the type of charge structure used according to the collected blasting signal, and the automatic acquisition unit in the KNN model can automatically record the vibration data generated in the blasting process, without the need for personnel to independently identify and record data, thereby reducing the tedious work steps of personnel to a certain extent. Thus, the problem that the method for identifying and recording the charge structure type and vibration data according to the vibration characteristics in the prior art mostly adopts manual real-time recording and supervision, has high labor intensity, is prone to missing recording, wrong recording, type identification error and the like, and is prone to affecting subsequent tunnel blasting excavation vibration reduction and loss reduction, thereby reducing the self-stability of surrounding rock. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is a flowchart of a tunnel blasting charge structure identification recording method provided by the embodiment of the present application.

[0042] Figure 2 is a schematic diagram of three charge structures of a tunnel peripheral hole provided by the embodiment of the present application.

[0043] Figure 3 is a schematic diagram of a tunnel peripheral reserved light blasting layer blast hole arrangement provided by the embodiment of the present application.

[0044] Figure 4 is a schematic diagram of a drilling jumbo blasting operation process provided by the embodiment of the present application.

[0045] Figure 5 is a schematic diagram of a blasting vibration signal waveform curve of different peripheral hole charge structures provided by the present application.

[0046] Figure 6 is a schematic diagram of a blasting vibration signal AOK time-frequency spectrum of different peripheral hole charge structures provided by the embodiment of the present application.

[0047] Figure 7 is a schematic diagram of four generation directions of a gray level co-occurrence matrix provided by the embodiment of the present application.

[0048] Figure 8 is a schematic diagram of a PCA dimension reduction process of a blasting signal time-frequency image feature provided by the present application.

[0049] Figure 9It is the KNN algorithm verification sample classification schematic diagram based on the PSO-ELM feature mapping provided in the first aspect of the present application.

[0050] Figure 10 It is the KNN algorithm verification sample classification schematic diagram based on the PSO-ELM feature mapping provided in the second aspect of the present application.

[0051] Figure 11 It is the module schematic diagram of the tunnel blasting charge structure identification recording device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0053] The specific embodiments of the present application will be described in detail below with reference to the drawings.

[0054] The present application discloses a tunnel blasting charge structure identification recording method, device, equipment and medium, which at least solves the deficiency of the prior art to some extent.

[0055] Please refer to Figure 1 The first embodiment of the present application provides a tunnel blasting charge structure identification recording method, which can be executed by a tunnel blasting charge structure identification recording device (hereinafter referred to as an identification recording device), in particular, by one or more processors in the identification recording device, to implement the following steps:

[0056] S101, acquiring a set of blasting signals generated by the tunnel peripheral charge structure blasting collected by the wireless vibration measurement unit;

[0057] Specifically, in the present embodiment, the specific forms of the three common tunnel peripheral hole charge structures on the market are continuous charge, water interval charge and air interval charge. Please refer to Figure 2(a), the continuous charging is designed to have a single cycle of 1.2 m, the blast hole is continuously charged, and the blast hole is divided into a stemming section and a charging section from the blast hole mouth to the bottom. The length of the stemming section is 400 mm, and the length of the charging section is 800 mm. The blast hole is filled with 2# low-density rock emulsion explosive, the blast hole diameter is 40 mm, the cartridge diameter is 32 mm, the length is 200 mm, the single cartridge mass is 80 g, and the radial decoupling coefficient is 1.25. In order to improve the blasting efficiency, the bottom of the blast hole is reversely initiated by using a digital electronic detonator. Please refer to Figure 2 (b), the water interval charging is designed to have a single cycle of 1.2 m, the blast hole is three-section charged, and the blast hole is divided into a stemming section, a first charging section, a first water interval section, a second charging section, a second water interval section, and a bottom charging section from the blast hole mouth to the bottom. The length of the stemming section is 400 mm, the length of the first charging section is 80 mm, the length of the first water interval section is 240 mm, the length of the second charging section is 80 mm, the length of the second water interval section is 240 mm, and the length of the bottom charging section is 160 mm. The blast hole is filled with 2# rock emulsion explosive, the blast hole diameter is 40 mm, the cartridge diameter is 32 mm, the length is 200 mm, the single cartridge mass is 200 g, the radial decoupling coefficient is 1.25, and the axial decoupling coefficient is 2.5. In order to improve the blasting efficiency, the bottom of the blast hole is reversely initiated by using a digital electronic detonator, and the amount of explosive in the bottom charging section is twice that in the first and second charging sections. Please refer to Figure 2(c), the air gap charge is designed to have a single cycle of blasting with a penetration of 1.2 m, the blast hole is a three-section charge structure, and the blast hole is divided into a stemming section with a length of 400 mm, a first charge section with a length of 80 mm, a first air gap section with a length of 240 mm, a second charge section with a length of 80 mm, a second air gap section with a length of 240 mm, and a bottom charge section with a length of 160 mm from the blast hole mouth to the bottom of the blast hole. The blast hole is filled with 2# rock emulsion explosive, the blast hole diameter is 40 mm, the cartridge diameter is 32 mm, the cartridge length is 200 mm, the single cartridge mass is 200 g, the radial decoupling coefficient is 1.25, and the axial decoupling coefficient is 2.5. In order to improve the blasting efficiency, the bottom of the blast hole is reversely initiated, a digital electronic detonator is used for initiation, and the explosive amount of the bottom charge section is 2 times that of the first and second charge sections. The stemming used for plugging the blast hole is prepared by clay, medium-coarse sand and water in a weight ratio of 0.75:0.10:0.15, and an appropriate amount of polyacrylamide is added to the water to increase the adhesion of the stemming. The stemming is prepared by a PNJ series stemming machine, and the main technical parameters of the stemming machine are as follows: production efficiency 600-1000 per hour, stemming diameter 30-50 mm, stemming density 1.9 g / cm3, and a layer of silicate cement powder is uniformly wrapped on the surface of the stemming during use. When charging, the bundled explosive is sent to the bottom of the blast hole, the left hand pulls the fuse to send the stemming into the blast hole, and a wooden tamping rod is used to push in, and when the explosive is fully sent to the bottom of the blast hole, the stemming is tamped with the tamping rod. After using the stemming of this type, the explosive consumption can be reduced by about 20%, the number of drill holes can be reduced, the drill steel can be saved by about 15%, the penetration can be increased by about 5%, in addition, the reduction of air shock wave shortens the throwing distance of the blasted rock, increases the safety of blasting, reduces the content of smoke and toxic gas, reduces the difficulty of long-distance ventilation, improves the tunnel operation environment, reduces the size of the blasted rock, reduces the secondary crushing rate, reduces the cost of cleaning the rock by about 15%, and has good plugging effect.

[0058] In the embodiment, before collecting the blasting signal, the staff needs to reserve a light blasting layer first, that is, the blasting adopts a form of reserving a light blasting layer, as shown in Figure 3As shown, the reserved light blasting layer thickness is 500 mm, the excavated part is mechanically excavated by EBH35 series hard rock tunneling machine, the peripheral hole is drilled by drilling jumbo, the drilling angle of the peripheral hole is 3°, all the hole bottoms fall in the same plane, the single hole charge is 0.32 kg, the total number of holes is 37, 30 holes are evenly arranged on both sides of the waist and arch, the hole spacing is 500 mm, the number of hole spacing bottom holes is 7, the hole spacing is 960 mm, the total charge is 11.84 kg, the blasting network adopts cluster connection mode, all the holes are divided into 6 clusters, each cluster uses 2 MS1 section millisecond delay detonator, and finally the detonation tube is connected by detonation tube four-way connector, and the detonation tube is initiated by the detonator once. The tunnel surrounding rock is weakly weathered granite, the density is 2.657 kg / cm3, the elastic modulus E is 80 GPa, and the Poisson's ratio is 0.29. In order to ensure the drilling accuracy, the above three kinds of charging structure holes are drilled by ZYS113 three-arm drilling jumbo. Compared with the traditional air leg drilling manual drilling, the drilling jumbo has stronger adaptability to high hardness surrounding rock, has the advantages of fast drilling speed and low energy consumption, can strictly follow the design drawing to carry out automatic hole drilling, and can effectively set the drilling parameters, automatically drill and clean the hole. In order to fully guarantee the drilling according to the relevant design, strengthen the control of the drilling angle and depth of the hole, draw 5 control lines along the tunnel face, 1 arch shoulder and 4 slope lines, and accurately position the hole center position. The blasting operation process of the drilling jumbo is shown in Figure 4 .

[0059] The monitoring and analysis of the blasting vibration signals generated by the blasting tunnel have always been an important basis for evaluating the blasting effect and optimizing the parameters. The traditional tunnel vibration monitoring adopts the "measuring while walking" monitoring mode, which has defects such as frequent arrangement of measuring points and poor timeliness of data collection. The tunnel blasting charge structure identification recording method monitors the tunnel blasting vibration by using remote wireless network automatic monitoring. The monitoring system mainly consists of cloud vibration data acquisition box, three-direction vibration velocity sensor, public network server data center and client. Before the blasting site test, the sensor can be fixed by customizing the clamp and filling the gap between the sensor and the tunnel lining with adhesive. After turning on the instrument power and ensuring network connection, the site can be left. After completing parameter setting and starting collection through a remote computer, the instrument enters the working state. Before the test, according to the characteristics of tunnel blasting, the sampling frequency of the vibration meter is set to 8000 Hz, the sampling time is 0.1 s, and the trigger mode is internal trigger. When the blasting vibration triggers the collection sensor, the system will automatically record the entire dynamic waveform, convert it into a digital signal and transmit it to the cloud server data center, realizing real-time collection and storage of data. Test personnel can operate and analyze data and monitor the working state of the control system in real time through the terminal software away from the blasting site. The blasting vibration signal waveform curves of different peripheral hole charge structures are shown in Figure 5As shown in the figure, Figure 5 (a) is a continuous charge structure blasting vibration signal waveform curve, Figure 5 (b) is a water interval charge structure blasting vibration signal waveform curve, Figure 5 (c) is an air interval charge structure blasting vibration signal waveform curve.

[0060] S102, processing the blasting signal set to generate a plurality of groups of reduced dimension data;

[0061] Specifically, step S102 includes: performing AOK time-frequency spectrum calculation on the data of the blasting signal set to generate a time-frequency image;

[0062] The gray level co-occurrence algorithm is used to calculate the time-frequency image to generate a gray level co-occurrence matrix corresponding to the time-frequency image;

[0063] The gray level co-occurrence matrix is subjected to PCA dimension reduction processing to establish three principal component feature vectors reflecting the characteristics of the time-frequency image after dimension reduction, and a plurality of groups of reduced dimension data are generated.

[0064] Specifically, in this embodiment, the optimal kernel time-frequency distribution method is used to analyze the blasting signal set. The optimal kernel time-frequency distribution is a main time-frequency analysis method, which has the characteristics of high time-frequency resolution. However, due to its bilinear table transformation, there is serious cross-term interference when analyzing multi-component signals, which affects the final analysis result. In order to suppress the interference of cross terms and separate the signal's own amount, people have improved the optimal kernel time-frequency distribution. The improved time-frequency analysis is collectively referred to as Cohen class bilinear time-frequency distribution, and its expression is: In the formula, Φ(θ,τ) is a low-pass kernel function, A(θ,τ) is a blur function of the blasting signal s(t), t is the duration of the blasting signal, ω is the frequency, θ is the frequency shift, and τ is the time shift, which is defined as: By designing different kernel functions, different distribution characteristics can be obtained. In order to eliminate cross components and only retain self components in the blur function domain, the kernel function is defined as a two-dimensional function of Gauss type along any radial section. The expression is: In the formula, is a control radial Gaussian kernel function in the radial angle direction, is the angle between the horizontal direction and the radial direction.

[0065] The kernel function is defined as a two-dimensional function of Gauss type along any radial profile, which can effectively remove the cross-component and retain the auto-component, but it is not suitable for the analysis of long-time non-stationary signals generated by multi-segment differential blasting. The short-time ambiguity function A(t, θ, τ) is defined in the adaptive optimum kernel (AOK) time-frequency distribution, which is the ambiguity function of a small segment of signal truncated by a window function: In the formula, ω(u) is a symmetric window function, and t is the center position of ω(u). Let |u|>T (T is the window length), ω(u)=0, then at any time, only the signal in the range [t-T, t+T] can calculate its ambiguity function. For any detailed part of the blasting signal, the short-time ambiguity function can be accurately described, and the corresponding adaptive optimum kernel function Φ opt (t, θ, τ) can be obtained. The adaptive optimum kernel time-frequency distribution of the blasting signal in the time period [t-T, t+T] is: The AOK time-frequency spectrum of the typical blasting vibration signals under the above three different charge structures is shown in Figure 6 , wherein, Figure 6 (a) is the AOK time-frequency spectrum of the blasting vibration signal of the continuous charge structure, Figure 6 (b) is the AOK time-frequency spectrum of the blasting vibration signal of the water interval charge structure, Figure 6 (c) is the AOK time-frequency spectrum of the blasting vibration signal of the air interval charge structure.

[0066] In this embodiment, the gray level co-occurrence matrix (GLCM) can reflect the texture features of the time-frequency spectrum of the blasting signal, and in combination with the correlation eigenvalues, it can be used for the discrimination of the charge structure. The specific steps of calculating the gray level co-occurrence matrix and the eigenvalues are as follows:

[0067] Firstly, the amplitude of the AOK time-frequency spectrum image of the blasting signal is grayed (i.e. the amplitude is converted to the range of 0-256).

[0068] Secondly, the gray level co-occurrence matrix of the time-frequency spectrum image is calculated, which mainly represents the probability distribution of the gray values i and j of the points Δm and Δn in the image: m2 = m1 + Am, n2 = n1 + An, where i, j = 1, 2,..., L, < > represents the co-occurrence value generated at the element (i, j) of the image matrix; Am and An are the displacements of the element to the left and right and above and below the element (i, j), respectively; f(x, y) is the gray scale of the time-frequency spectrum digital image of the blasting signal in the two dimensions of time-frequency domain, and the size is L m × L n , L m and L n are the spatial positions of the time-frequency spectrum pixels, L m = 1, 2,..., N m , L n = 1, 2,..., N n , (m1, n1) is the coordinate of a point in the image; (m2, n2) is the coordinate of another point. The co-occurrence matrix is essentially a joint histogram of two pixel points, and for fine and regular textures in the time-frequency spectrum image, the two-dimensional histogram of paired pixel points tends to be uniformly distributed; for coarse and regular textures, it tends to be diagonally distributed.

[0069] Again, divide each value of the co-occurrence matrix obtained by calculating I(i, j) by the maximum element in the matrix to perform co-occurrence matrix normalization processing.

[0070] Finally, the gray level co-occurrence matrix (GLCM) of the time-frequency spectrum image matrix of the blasting vibration signal of different charge structures of the blast hole is calculated, and six-dimensional feature vectors of energy value, contrast, correlation, entropy value, inverse variance, and inertia moment of the time-frequency spectrum are extracted.

[0071] Among them, the energy value is defined as: The energy value reflects the uniformity of the image gray scale distribution and the fineness of the texture. If the element values of the gray level co-occurrence matrix are similar, the energy is small, indicating that the texture is fine; if some element values are large and other values are small, the energy value is large. A large energy value indicates a more uniform texture pattern. The contrast is defined as: CON = åi å j (i-j) 2 I(i, j, d, q), the contrast is the inertia moment near the main diagonal of the gray level co-occurrence matrix, which reflects how the values of the matrix are distributed, and reflects the sharpness of the image and the depth of the texture groove. The correlation is defined as: CORRLN = [å i å j (ij)I(i, j, d, q) - m x m y ] / s x s y , the correlation reflects the similarity of the spatial gray level co-occurrence matrix elements in the row or column direction, and reflects the local gray correlation of the image. The entropy value is defined as: ENT = - å iI(i, j, d, θ) log I(i, j, d, θ), the entropy value reflects the randomness of the image texture. If all the values in the co-occurrence matrix are equal, the maximum value is obtained; if the values in the co-occurrence matrix are uneven, the value will become very small. The inverse variance is defined as: The inverse variance reflects the clarity and regularity of the texture. The clearer and more regular the texture, the easier to describe, the larger the value. The inertia moment is defined as: MOM =∑ i ∑ j I(i,j|d,θ)(i-j) 2 The inertia moment is a second statistical quantity of the gray level co-occurrence matrix. The inertia moment can pull apart the spatial distribution of the image gray level, and can better distinguish the complexity of the spatial distribution of the image gray level. The larger the inertia moment value, the more obvious the image groove, and the clearer the image; on the contrary, the smaller the inertia moment, the weaker the image texture groove, and the more blurred the image. According to the energy value, the contrast, the correlation, the entropy value, the inverse variance and the inertia moment generated by the above time-frequency image gray level co-occurrence matrix, a six-dimensional feature vector f i (i = 1, 2, 3, 4, 5, 6) is established.

[0072] The gray level co-occurrence matrix reflects the spatial relationship of the pixels under a certain texture mode. This spatial relationship contains three aspects of direction θ, image gray level g and distance s. When generating the gray level co-occurrence matrix, the statistics is performed in 0°, 45°, 90° and 135° directions, the distance is uniformly taken as s = 1, and the gray level g = 16, so that a 4 × 6 feature matrix X is obtained. In order to facilitate the statistics, the above matrix is normalized to obtain a normalized feature matrix X. The feature values under this construction method are stable, the change is small, and the probability of special value occurrence is minimum. The four generation directions of the blasting time-frequency image gray level co-occurrence matrix are shown in Figure 7 The normalized feature matrix of the typical blasting signal time-frequency image gray level co-occurrence is shown in Table 1.

[0073] Table 1

[0074]

[0075]

[0076] In this embodiment, the feature space dimension of the blasting vibration signal time-frequency spectrum image gray level co-occurrence matrix of different charge structures is too high, which is easy to cause data redundancy, and causes difficulty in subsequent recognition and classification. The high-dimensional matrix data obtained is reduced in dimension, which can not only make full use of the rich data amount, but also reduce the data processing cost, solve the problem of too high feature space dimension, and also retain the necessary image information, improve the accuracy and efficiency of the time-frequency spectrum image recognition and classification.

[0077] The PCA algorithm is a feature extraction technique based on linear mapping. Through a certain transformation, high-dimensional image data is transformed into a new low-dimensional space, so that the maximum variance of the high-dimensional data is projected on the first low-dimensional space coordinate (i.e. the first principal component), the second largest variance is projected on the second low-dimensional space coordinate (the second principal component), and so on. The original high-dimensional image data is retained to the maximum extent by using a few principal components, and the first principal component contains most of the information in the original high-dimensional image data. The PCA algorithm mainly uses the property that the covariance matrix is a real diagonal matrix, i.e. variance maximization and covariance minimization, to reduce the dimension. The PCA algorithm is used to reduce the dimension of the different charge structure feature vectors of the blast hole extracted from the gray level co-occurrence matrix, and to remove the redundant vectors in the vector set. Assuming that there are P characteristic values of the gray level co-occurrence matrix in each direction data, the data is represented as: X = (x1, x2,...,x N ) = (X1, X2,...,X P ) T , where X k is an N x 1 column vector (N is the number of gray level co-occurrence matrix directions of the time-frequency image), and X is a P x N matrix. The low-dimensional output dimension after dimension reduction is d (d << P). The main steps are as follows:

[0078] First, the mean vector of all data is calculated: In the formula, i = 1, 2,..., N represents the number of feature directions extracted from the gray level co-occurrence matrix of the time-frequency image; k = 1, 2,..., P represents the number of features in each feature direction of the gray level co-occurrence matrix of the time-frequency image.

[0079] Secondly, the standard matrix of the original high-dimensional data zero mean is calculated, In the formula,

[0080] Thirdly, the covariance matrix of the data sample to be analyzed is calculated:

[0081] Fourthly, the eigenvalues i = λ1, λ2,..., λ P , λ1, > > > λ P and the corresponding attribute vectors ∝1, ∝2,..., ∝ P of the covariance matrix are calculated, and the eigenvalues are sorted from large to small, and the attribute vectors are arranged in turn with the sorting of the eigenvalues. Through the obtained attribute eigenvalues, the contribution rate σ i of each principal component and the cumulative contribution rate δ are calculated:

[0082] Finally, the largest d attribute eigenvalues (d << P) are taken, and the corresponding attribute eigenvectors ∝1, ∝2,..., ∝d The transformation matrix A = [∝1,∝2,...,∝] is formed. d The image feature matrix data X is reduced to d-dimensional data Y using the following formula: Y = A T X. After transformation, most of the information in the original image data is found in the first few principal component components, while the numerous later components contain primarily noise. The dimensionality reduction flowchart of the PCA algorithm is as follows: Figure 8 As shown in Table 2, the contribution rates and cumulative contribution rates of each principal component in the PCA algorithm are presented below.

[0083] Table 2

[0084]

[0085] S103, the trained KNN model is called to process the dimensionality reduction data to generate the charge structure type corresponding to the blasting signal set. The automatic acquisition unit of the KNN model will automatically record the vibration data generated during the tunnel blasting process that corresponds to the charge structure type.

[0086] Specifically, in this embodiment, after PCA dimensionality reduction, the first three principal component components of the original blasting signal time-frequency image grayscale co-occurrence data can express more than 95% of the information in the entire data. The PCA algorithm can optimize the data and achieve the purpose of data compression and dimensionality reduction.

[0087] Ninety-one sets of blasting vibration signals under three different peripheral hole charging configurations were collected at the same detonation center distance. The AOK time-frequency spectrum of these signals was calculated to generate classification labels E = {e1, e2, e3}, typically assigned as three simple classification categories (1, 2, and 3) representing continuous charge structure, water-gap charge structure, and air-gap charge structure, respectively. A gray-level co-occurrence algorithm was used to obtain a gray-level co-occurrence matrix reflecting the features of the time-frequency images. PCA was then used to reduce the dimensionality of this feature matrix, forming target matrices for different charge configurations. Gray-level co-occurrence calculations were performed on each time-frequency image, and the PCA-reduced features were defined as a three-dimensional feature vector Y = {y1, y2, y3}, corresponding to the first principal component PC1, the second principal component PC2, and the third principal component PC3, respectively, thus obtaining the finite-dimensional digital information features of the time-frequency images.

[0088] Particle Swarm Optimization (PSO) is a collaborative stochastic search algorithm developed by simulating the foraging behavior of bird flocks. It is widely used in fields such as parameter optimization and system control. In PSO, each particle represents a point in space, and its corresponding spatial coordinates represent a set of solutions to the problem. The fitness of each particle can be calculated using a fitness function, and each particle also has a velocity attribute that determines its direction and distance. Particles search for the optimal solution in the solution space by sharing information about the best particle. The PSO algorithm first calculates the fitness of each particle, selects the point with the best fitness, and moves all other points towards this point, iteratively searching for the optimal solution. In each iteration, the particle first updates its velocity towards the current optimal solution, and then updates its current position. The formula for particle velocity update is: v id =v id +c1·rand(p id -x id )+c2·rand(p gd -x gd Establish the particle position update formula: x id =x id +v id In the formula: p i =(p i1 ,p i2 ,...p iD ), representing the local optimal solution for the i-th particle, i.e., the best point that the i-th particle has experienced in each iteration to its current position; p g =(p g1 ,p g2 ,...p gD ), represents the optimal solution of the particle swarm, that is, the solution with the best fitness found by the particle swarm so far; c1 and c2 are learning factors, which are constants greater than zero; rand means generating random numbers in [0,1].

[0089] Extreme Learning Machine (ELM) is a single-hidden-layer feedforward neural network that allows for arbitrary selection of the number of hidden neurons and activation function types to construct different network structures. Hidden layer biases and input layer weights can be randomly generated, and the output layer weights are calculated using the least squares method based on the output layer expectation. This algorithm offers advantages such as high computation speed and good generalization ability, avoiding the problems of long computation time and susceptibility to local optima inherent in traditional neural networks using gradient algorithms. It currently has wide applications in system identification, pattern recognition, and function approximation.

[0090] KNN(K-Nearest Neighbor) algorithm as a lazy learning model, also known as instance-based learning model. The idea of KNN algorithm is as follows: if a data sample to be analyzed in the feature space of K most similar (K nearest neighbor in the feature space) of the majority of the analyzed data sample belongs to a certain class, the data sample to be analyzed is classified into the class. When the training accuracy reaches the set threshold, when the test data to be classified is input, the classification result can be predicted according to the training data set. Minkowski distance is one of the widely used distance measures in traditional KNN. If the feature of the data to be analyzed has the same size and data distribution, the Minkowski distance can represent the actual distance between two points. When the number of feature points is m, the Euclidean distance between the training data sample P=(x1, x2,...x m ) and the test data sample Q=(y1, y2,...y m ) is calculated as: The shorter the distance between two points, the greater the similarity between them. After calculating the distance of all point pairs, the class of the test data sample can be determined. KNN classification algorithm is easy to implement and has good classification performance, mainly manifested as: the performance of the algorithm is only related to the training set, and when the test data sample changes, it will not have an impact, and the larger the data sample set, the better the performance.

[0091] The tunnel blasting charge structure identification recording method uses KNN algorithm based on PSO-ELM feature mapping to identify and classify image data sets. The algorithm uses PSO optimization strategy to replace the random generation process of ELM algorithm input layer weight and hidden layer bias. The optimized ELM input layer and hidden layer activation function are used for feature mapping of the data sample to be analyzed, and then the mapped feature data sample is input into the KNN algorithm for classification. For a given n-dimensional input classification data sample x=[x1, x2,...x n ] T , first, a set of input layer weights ω i and hidden biases b i are randomly generated by using particle swarm optimization algorithm. Through the ELM with hidden layer neurons and the activation function G(x), the feature vector of the data sample to be analyzed after feature mapping is: In the formula: G(ω i *x)+b i is the output of the hidden layer node, ω i is the dimensional weight vector connecting n input nodes and the i-th hidden layer node; b iBias of the i-th hidden layer node. Commonly used activation functions include Hardlim function, Sigmoid function, Sine function, etc. The h(x) after the feature change is input into the KNN algorithm, and the accuracy of the classification of the data sample to be analyzed is calculated. The classification accuracy is returned to the particle swarm optimization algorithm as the fitness of each particle. Through multiple iterations of PSO, a set of input layer weights ω i and hidden layer bias b i .

[0092] The specific steps of the algorithm are as follows:

[0093] Step 1, randomly generate a particle swarm, and the position coordinates of each particle represent the input layer weights ω i and hidden layer bias b i .

[0094] Step 2, select each particle in turn, and perform feature mapping on the input data sample x i to obtain h(x), and input h(x) into KNN as a new data sample to be analyzed, and calculate the classification accuracy as the fitness of the corresponding particle;

[0095] Step 3, update the local optimal solution P i and the global optimal solution P g of each particle according to the fitness;

[0096] Step 4, update the speed and position of each particle according to the particle update speed and update position;

[0097] Step 5, judge whether the maximum number of iterations is reached, if the maximum number of iterations is reached, output the optimal particle position, i.e. the optimal input layer weights ω i and hidden layer b i and the classification accuracy, otherwise return to step 2.

[0098] In this embodiment, the tunnel blasting charge structure identification recording method calculates the dimensionality reduction data according to the established particle swarm, and the extreme learning machine algorithm and the KNN algorithm, and finally the KNN classifier in the KNN model generates a charge structure type corresponding to the blasting signal set, and the automatic acquisition unit in the KNN model automatically records the vibration data corresponding to the charge structure type generated in the tunnel blasting process, without the need for personnel to independently identify and record data, thereby reducing the tedious work steps of personnel to a certain extent.

[0099] In a possible embodiment of the present application, before calling the trained KNN model to process the dimensionality reduction data, the following steps are further included:

[0100] acquire a plurality of different charge structure type blasting signal sets generated by different tunnel peripheral charge structure blasting collected by the wireless vibration measurement unit;

[0101] process each of the different charge structure type blasting signal sets to generate a plurality of recognition classification image analysis dimension reduction original data sets;

[0102] randomly select a plurality of recognition classification image analysis dimension reduction original data sets of different charge structure types and with a preset number of the same, and call a PSO-ELM-based KNN model to train and process each of the recognition classification image analysis dimension reduction original data sets to generate a training error value, until the training error value is less than a preset value, and generate the KNN model.

[0103] randomly select a plurality of recognition classification image analysis dimension reduction original data sets of different charge structure types and with a preset number of the same, and call the KNN model to perform accuracy verification calculation to generate an average accuracy, until the average accuracy reaches a preset value, and generate the KNN model.

[0104] In a possible embodiment of the present application, the different charge structure type blasting signal set includes a continuous charge structure blasting signal set, a water interval charge structure blasting signal set, and an air interval charge structure blasting signal set.

[0105] Please refer to Figure 9 to Figure 10 , specifically, in the present embodiment, before the KNN model is used for identification, it needs to be trained, for example, 300 groups (a total of 900 groups) of blasting vibration signals of continuous charge, water interval charge and air interval charge of different peripheral hole charge types at different blast center distances are randomly selected and three types of time-frequency image gray level co-occurrence matrices are established and dimension reduction processing is performed, and according to the ratio of 7:2:1, they are divided into a training set (210 groups in total, 630 groups in total), a verification set (60 groups in total, 180 groups in total) and a test set (30 groups in total, 90 groups in total), and 210 images of the image set under three different blasting schemes are randomly selected as training data input into the aforementioned PSO-ELM-based KNN classifier, and the classifier is repeatedly trained until the training classification error is less than 10-6. Randomly select 60 groups (a total of 180 groups) of data to verify the classification accuracy, and when the average accuracy of the verification set reaches more than 95%, it is considered that the accuracy meets the engineering requirements, and the remaining 30 groups (a total of 90 groups) of data under different peripheral charge forms are input into the trained KNN classifier to determine the given different categories (e1, e2, e3) as the final result of the identification and classification of the different charge structure of the peripheral hole of the tunnel blasting, and the discrimination and classification of the continuous charge, water interval charge and air interval charge of the three different peripheral charge modes are realized.

[0106] The model recognition accuracy refers to the percentage of the number of model prediction correct data samples in the total number of observation values in the batch of data. In the formula, P i is the i-th image recognition accuracy, T i is the number of i-th image recognition correct data samples, F i is the number of i-th image recognition error data samples; the average recognition accuracy P is obtained by averaging the recognition accuracy of each type of image sample. The tunnel blasting charge structure recognition recording method has an accuracy of more than 97% in discriminating and classifying three different types of charge structures, namely continuous charge, water interval charge and air interval charge, and provides a convenient, efficient, intelligent prediction, discrimination, classification and recording method for tunnel structure safety and fine design construction and optimization selection of blast hole charge form. The KNN model training set accuracy and test set accuracy comparison table is shown in Table 3:

[0107] Table 3

[0108]

[0109] Please refer to Figure 11 , the second embodiment of the present application provides a tunnel blasting charge structure recognition recording device, comprising:

[0110] The data acquisition unit 101 is used for acquiring the blasting signal set generated by the tunnel peripheral charge structure blasting collected by the wireless vibration measurement unit;

[0111] The data processing unit 102 is used for processing the blasting signal set to generate multiple groups of dimensionality reduction data;

[0112] The recognition recording unit 103 is used for calling the trained KNN model to process the dimensionality reduction data to generate the charge structure type corresponding to the blasting signal set, wherein the automatic acquisition unit of the KNN model automatically records the vibration data corresponding to the charge structure type generated in the tunnel blasting process.

[0113] In a possible embodiment of the present application, the data processing unit is specifically used for:

[0114] Perform AOK time-frequency spectrum calculation on the data of the blasting signal set to generate a time-frequency image;

[0115] The gray level co-occurrence algorithm is used to calculate the time-frequency image to generate a gray level co-occurrence matrix corresponding to the time-frequency image;

[0116] The gray level co-occurrence matrix is subjected to PCA dimensionality reduction processing to establish three principal component feature vectors reflecting the characteristics of the time-frequency image after dimensionality reduction, and multiple groups of dimensionality reduction data are generated.

[0117] The third embodiment of the present application provides a tunnel blasting charge structure identification recording device, comprising a processor, a memory and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the tunnel blasting charge structure identification recording method according to any one of the above when executing the computer program.

[0118] The fourth embodiment of the present application provides a readable storage medium, which stores a computer program capable of being executed by a processor of a device where the readable storage medium is located to implement the tunnel blasting charge structure identification recording method according to any one of the above.

[0119] Exemplarily, the computer program in the third embodiment and the fourth embodiment of the present application can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the tunnel blasting charge structure identification recording device. For example, the device in the second embodiment of the present application.

[0120] The processor can be a central processing unit (CPU), and can also be 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. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the tunnel blasting charge structure identification recording method, and is connected with all parts of the tunnel blasting charge structure identification recording method through various interfaces and lines.

[0121] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the tunnel blasting charge structure identification recording method by running or executing the computer program and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application program required by a function (such as a sound playing function, a character conversion function, etc.), and the like; and the data storage area can store data created according to the use of the mobile phone (such as audio data, text message data, etc.), and the like. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0122] Among them, the implemented module, if in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The computer program can implement the steps of each method embodiment when executed by a processor. Among them, the computer program includes computer program code, which can be in the form of source code, object code, executable file or some intermediate form, etc. The computer readable medium can include any entity or device, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. that can carry the computer program code. It should be noted that the contents included 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, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0123] It should be noted that the apparatus embodiments described above are merely illustrative, and the units described as separate units can or can not be physically separate, and the units shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. In addition, the connection relationship between the modules in the apparatus embodiment provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0124] The above is only a preferred embodiment of the present application, and the protection scope of the present application is not limited to the above-mentioned embodiments. Any technical solution falling within the concept of the present application shall fall within the protection scope of the present application.

Claims

1. A method of identifying a record of a tunnel shot charge structure, characterized by, The method comprises the following steps: acquiring a blasting signal set generated by a wireless vibration measuring unit during blasting of a tunnel peripheral charge structure; processing the blasting signal set to generate multiple sets of dimension-reduced data, specifically: performing AOK time-frequency spectrum calculation on the data of the blasting signal set to generate a time-frequency image; calculating the time-frequency image using a gray level co-occurrence algorithm to generate a gray level co-occurrence matrix corresponding to the time-frequency image; performing PCA dimension reduction processing on the gray level co-occurrence matrix to establish three principal component feature vectors reflecting the characteristics of the time-frequency image after dimension reduction, and generate multiple sets of dimension-reduced data; calling a trained KNN model to process the dimension-reduced data to generate a charge structure type corresponding to the blasting signal set, wherein the automatic acquisition unit of the KNN model automatically records vibration data corresponding to the charge structure type generated during tunnel blasting.

2. A method of identifying records of a tunnel shot charge structure according to claim 1, characterized in that, The gray level co-occurrence algorithm is used to calculate the time-frequency image to generate a gray level co-occurrence matrix corresponding to the time-frequency image, specifically: performing adaptive optimal kernel analysis on the time-frequency image to generate an amplitude gray level of the time-frequency spectrum image; According to the formula , m, The amplitude of the time-frequency spectrum image is calculated, and a matrix co-occurrence is generated, wherein , is the co-occurrence value generated at the element (i, j) of a certain element pair in the image matrix, and are the displacements of the element to the left and right and up and down of the element (i, j), is the gray scale of the time-frequency spectrum digital image of the explosion signal in the two dimensions of time-frequency, and the size is , and are the spatial positions of the time-frequency spectrum pixels, , , is the coordinates of a point in the image; is the coordinates of another point; To calculate The obtained individual values of the co-occurrence matrix are divided by the maximum element in the matrix, and the co-occurrence matrix is normalized to generate a gray-level co-occurrence matrix of the time-frequency spectrum image matrix of the blasting vibration signals of different charge structures.

3. The method of claim 1, wherein, performing PCA dimension reduction processing on the gray level co-occurrence matrix, specifically: According to the formula a mean vector of the burst signal set is calculated, wherein the number of feature directions extracted for the gray level co-occurrence matrix of the time-frequency image; the number of features for each feature direction of the gray level co-occurrence matrix of the time-frequency image; According to the formula The standard matrix of the original high-dimensional data zero mean is calculated, wherein, ; According to the formula computing a covariance matrix of the set of burst signals; Eigenvalues of the covariance matrix are calculated , and their corresponding eigenvectors and the eigenvalues are sorted from large to small, and the contribution rate and cumulative contribution rate of each principal component are calculated according to the formula ;​​ Take the maximum d attribute feature values (d P), the corresponding attribute feature vector The conversion matrix , and according to the formula Calculate the image feature matrix data X to d-dimensional data Y, generate multiple sets of reduced dimension data.

4. The method of claim 1, wherein, Before calling the trained KNN model to process the dimension-reduced data, the method further comprises the following steps: acquiring multiple sets of different charge structure types of blasting signal sets generated by a wireless vibration measuring unit during blasting of different tunnel peripheral charge structures; processing each of the different charge structure types of blasting signal sets to generate multiple sets of recognition classification image analysis dimension-reduced original data sets; randomly selecting multiple sets of recognition classification image analysis dimension-reduced original data sets of different charge structure types and with a same preset number, and calling a KNN model based on PSO-ELM to train and process each of the recognition classification image analysis dimension-reduced original data sets to generate a training error value, until the training error value is less than a preset value, thereby generating a KNN training model; randomly selecting multiple sets of recognition classification image analysis dimension-reduced original data sets of different charge structure types and with a same preset number, and calling the KNN training model to perform accuracy verification calculation to generate an average accuracy, until the average accuracy reaches a preset value, thereby generating a KNN model.

5. A method of identifying records of a tunnel shot charge configuration according to claim 4, wherein, The different charge structure types of blasting signal sets include continuous charge structure blasting signal sets, water interval charge structure blasting signal sets, and air interval charge structure blasting signal sets.

6. A tunnel shot charge configuration identification recording device characterized by comprising: The method comprises the following steps: a data acquisition unit is configured to acquire a blasting signal set generated by a wireless vibration measuring unit during blasting of a tunnel peripheral charge structure; a data processing unit is configured to process the blasting signal set to generate multiple sets of dimension-reduced data; the data processing unit is specifically configured to: perform AOK time-frequency spectrum calculation on the data of the blasting signal set to generate a time-frequency image; calculate the time-frequency image using a gray level co-occurrence algorithm to generate a gray level co-occurrence matrix corresponding to the time-frequency image; perform PCA dimension reduction processing on the gray level co-occurrence matrix to establish three principal component feature vectors reflecting the characteristics of the time-frequency image after dimension reduction, and generate multiple sets of dimension-reduced data; An identification recording unit is configured to call the trained KNN model to process the dimension-reduced data, and generate a charge structure type corresponding to the blasting signal set, wherein the automatic acquisition unit of the KNN model automatically records vibration data corresponding to the charge structure type generated in the tunnel blasting process.

7. A tunnel blasting charge structure identification and recording device, characterized in that, The computer program product comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor executes the computer program to implement the tunnel blasting charge structure identification recording method according to any one of claims 1 to 5.

8. A readable storage medium, characterized by, The computer program product comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor executes the computer program to implement the tunnel blasting charge structure identification recording method according to any one of claims 1 to 5.

Images