An intelligent evaluation method for working face rock burst danger based on principal component analysis-probabilistic neural network

Abstract

This invention provides an intelligent evaluation method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks, belonging to the field of rockburst hazard evaluation technology. An evaluation index system is established and data is collected by combining rockburst hazard influencing factors, drill cuttings monitoring, critical stress index monitoring, and actual field conditions. Principal component analysis (PCA) is used to simplify the original evaluation index data, resulting in comprehensive evaluation index data containing information from the original evaluation indicators. The comprehensive evaluation index data is divided into a training set and a test set. A probabilistic neural network (PNN) is used to train the evaluation model on the data in the training set, and then the performance of the evaluation model is tested using data from the test set, and its accuracy is calculated. If the accuracy is greater than or equal to 90%, the evaluation model is considered acceptable; if the accuracy is less than 90%, it is considered unacceptable, and the PNN's smoothing factor needs to be modified and retrained until the accuracy of the evaluation result meets the set value.

Description

Technical Field

[0001] This invention relates to the field of rockburst hazard assessment technology, and in particular to an intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks. Background Technology

[0002] Rockburst refers to the sudden, rapid, and violent destructive phenomenon that occurs during mining operations when the elastic deformation energy of the coal and rock mass surrounding the mine roadway or working face is released instantaneously. With the increasing depletion of shallow resources, resource extraction is gradually shifting to deeper strata. The "high altitude, high temperature, high humidity, and high risk of disturbance" characteristics of deep engineering projects make rockburst one of the major geological hazards affecting mine safety. Rockburst hazard assessment is a powerful means of preventing and controlling rockburst disasters. Current assessment methods are mainly divided into two categories: single-factor methods and multi-factor comprehensive methods. Compared with single-factor methods, multi-factor comprehensive methods consider more comprehensive factors and are more accurate in their judgments, thus possessing certain advantages. Due to the suddenness, destructiveness, and complexity of rockbursts, assessment presents significant challenges. Currently, most rockburst hazard assessment methods are single-factor methods, based on a single indicator, and their consideration of factors is insufficient. While commonly used comprehensive index assessment methods are multi-factor comprehensive methods, some indicators are not quantified, and the assignment of indicator values ​​has a certain degree of subjectivity. Therefore, it is necessary to introduce new prediction models to conduct intelligent assessments of rockburst hazards. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks. This method alleviates the deficiencies of existing technologies by providing more comprehensive evaluation indicators, more objective indicator assignments, shorter evaluation time, more intelligent evaluation process, and more accurate evaluation results.

[0004] On the one hand, this invention provides an intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks, comprising the following steps:

[0005] Step 1: Establish a risk assessment index system for rockburst at the working face and collect rockburst risk assessment indexes to construct an original database;

[0006] The specific data for evaluating the risk of rockburst include mining depth, uniaxial compressive strength, elastic energy index, impact energy index, dynamic failure time, distance between fault and working face, distance between thick rock layer and coal seam, working face length, width of section coal pillar, thickness of bottom coal seam, pressure relief rate of protected layer, stress index ratio, and drilling powder rate index.

[0007] Step 2: Apply Principal Component Analysis (PCA) to reduce the dimensionality of the data to obtain comprehensive evaluation index data that includes the original evaluation index information;

[0008] Let x ij Let x1, x2, ..., xn be the value of the j-th indicator corresponding to the i-th sample in a real mine, where i = 1, 2, ..., m, m is the number of samples, and j = 1, 2, ..., n, n is the number of evaluation indicators; then the original evaluation indicators are x1, x2, ..., xn. n Let the comprehensive index obtained through principal component analysis, i.e., the principal components, be y1, y2, ..., y3. n Then the principal components are linear combinations of the original evaluation indicators after standardization, with a sum of squares of coefficients of 1 and each principal component being uncorrelated.

[0009] Step 2.1: Standardize the collected data on actual mine rockburst hazard indicators to obtain standardized evaluation indicators. As shown in the following formula:

[0010] ;

[0011] In the formula, Let s represent the mean of the j-th indicator. j The variance of the j-th index is expressed as follows:

[0012] ;

[0013] ;

[0014] Step 2.2: Calculate the correlation coefficients among the standardized indicators to obtain the correlation coefficient matrix R, and calculate the eigenvalues ​​of the correlation coefficient matrix R. , , ..., and normalized eigenvectors , , ..., The correlation coefficient matrix R is as follows:

[0015] ;

[0016] In the formula, This represents the correlation coefficient between the k-th indicator and the l-th indicator, where k = 1, 2, ..., n, l = 1, 2, ..., n. Correlation coefficient The calculation formula is as follows:

[0017] ;

[0018] Step 2.3: Calculate the variance contribution rate of each principal component. Calculate the cumulative contribution rate of the first w principal components. Where s = 1, 2, ..., n, w = 1, 2, ..., n, as shown in the following formula:

[0019] ;

[0020] ;

[0021] Step 2.4: Calculate the scores of the determined principal components; select the top w principal components with eigenvalues ​​greater than 1 and cumulative contribution rates reaching 85%~95%, and calculate their corresponding scores, as shown in the following formula:

[0022] ;

[0023] In the formula, This represents the normalized eigenvector corresponding to the t-th principal component. , , ..., Let represent n standardized evaluation indicators, where t = 1, 2, ..., w.

[0024] The scores of each principal component obtained after the original indicator data is processed by Principal Component Analysis (PCA) for dimensionality reduction are the comprehensive evaluation indicator data containing information from the original indicator data.

[0025] Step 3: Divide the comprehensive evaluation index data obtained by principal component dimensionality reduction into training set and test set according to the ratio. Use the comprehensive evaluation index data as the input layer of the probabilistic neural network (PNN) and learn and train it through the training set data to obtain the trained probabilistic neural network.

[0026] The probabilistic neural network (PNN) includes an input layer, a hidden layer, a summation layer, and an output layer, as detailed below:

[0027] Step A1: Determine the number of neurons in the input layer; the number of neurons in the input layer is the number of principal components selected after PCA processing. The comprehensive evaluation index data obtained after dimensionality reduction by principal component analysis is the input layer vector.

[0028] Step A2: Determine the number of neurons in the hidden layer and calculate the input-output relationship of neurons in each mode; the number of neurons in the hidden layer is the number of samples contained in the training set. After receiving the input from the input layer, the hidden layer calculates the distance between the input vector and the feature vector of the training sample and returns a standard value, as shown in the following formula:

[0029] ;

[0030] In the formula, This represents the relationship between the input and output determined by the j-th neuron of the i-th pattern, σ represents the smoothing factor of the probabilistic neural network (PNN), d represents the dimension of the sample space data, and x represents the feature vector of the test set data or the specific data of the object to be evaluated. ab Let represent the feature vector of the b-th training sample of the a-th class stored in the hidden layer, where a = 1, 2, ..., M, and M represents the total number of classes in the training set.

[0031] Step A3: Determine the number of neurons in the summation layer and perform a weighted average of the outputs of neurons of the same category; the number of neurons in the summation layer is the total number of sample patterns. The summation layer performs a weighted average of the outputs of neurons belonging to the same category in the hidden layer, as shown in the following formula:

[0032] ;

[0033] In the formula, v i Let L represent the output value of the i-th class, and L represent the number of neurons in the i-th class.

[0034] Step A4: Determine the number of neurons in the output layer and output the corresponding results; the number of neurons in the output layer is the total number of patterns in the training samples in the training set; according to the intensity of the rockburst hazard, the rockburst hazard is recorded as level 0, 1, 2, and 3, where 0 represents no impact and 3 represents a strong impact. The output result is in the form of a probability vector, and the sum of each element is 1; after the PNN outputs the probability, the maximum a posteriori probability decision (MAP) is added based on the Bayesian rule, that is, the neurons with the maximum a posteriori probability density output 1, and the rest output 0;

[0035] Step 4: Test the performance of the trained probabilistic neural network using test set data and calculate the accuracy of the prediction results. When the accuracy is greater than or equal to the set value, the model is considered to meet the standard and the prediction model is obtained. When the accuracy is less than the set value, the smoothing factor of the PNN needs to be adjusted and the training set needs to be relearned until the accuracy reaches the set value.

[0036] Step 5: Use the obtained prediction model to conduct an intelligent assessment of the rockburst hazard at the working face.

[0037] On the other hand, this application proposes a computer-readable storage medium storing executable instructions that, when executed, cause a processor to perform the intelligent assessment method for working face rockburst hazard based on principal component analysis-probabilistic neural network.

[0038] Thirdly, this application proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned intelligent assessment method for working face rockburst hazard based on principal component analysis-probabilistic neural network.

[0039] The beneficial effects of adopting the above technical solution are as follows:

[0040] This invention provides an intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks. Compared with existing assessment methods, the advantages of this invention are as follows:

[0041] 1. Based on the influencing factors of rockburst risk at the working face and the commonly used drill cuttings monitoring and critical stress index monitoring, a comprehensive evaluation index is developed, taking into account multiple factors. Furthermore, for some qualitative evaluation indicators, the most critical and core influencing factors are quantified, thereby eliminating the subjective bias of traditional qualitative evaluations and making the evaluation results more objective and reliable.

[0042] 2. Based on the PCA-PNN principle, this approach combines indicator data preprocessing with intelligent classification prediction, thus avoiding the limitations of a single prediction model. PCA reduces the dimensionality of the evaluation indicator data, simplifying the computational data required by the subsequent PNN model, thereby shortening the input data and computation time, while ensuring the indicators input into the PNN are highly representative. Simultaneously, PNN models are easy to train, have a simple structure, fast convergence speed, and high fault tolerance; given a sufficient number of samples, they can obtain the most effective classification results under the Bayesian criterion. Therefore, combining these two approaches avoids interference from high-dimensional redundant indicator data on the classification results and effectively improves the efficiency of the prediction process, as well as the accuracy and stability of the prediction results. Attached Figure Description

[0043] Figure 1 The overall flowchart of the intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural network in this invention is shown in the following embodiment.

[0044] Figure 2 Probabilistic Neural Network (PNN) topology diagram according to an embodiment of the present invention;

[0045] Figure 3 Schematic diagram of the stress-strain measurement system of the testing machine according to an embodiment of the present invention;

[0046] Figure 4 Schematic diagram of a standard specimen for uniaxial compression test according to an embodiment of the present invention;

[0047] Wherein (a) - front view of the standard specimen for uniaxial compression test, (b) - front view of the standard specimen for uniaxial compression test, and (c) - top view of the standard specimen for uniaxial compression test;

[0048] Figure 5 A schematic diagram illustrating the determination of impact tendency indicators in an embodiment of the present invention;

[0049] (a) is a schematic diagram for determining uniaxial compressive strength; (b) is a schematic diagram for determining impact energy index; (c) is a schematic diagram for determining elastic energy index; and (d) is a schematic diagram for determining dynamic failure time. Detailed Implementation

[0050] The specific implementation methods of this application will be further described in detail below with reference to the accompanying drawings and embodiments.

[0051] Example 1:

[0052] On the one hand, this invention provides an intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks, comprising the following steps:

[0053] Step 1: Establish a risk assessment index system for rockburst at the working face and collect rockburst risk assessment indexes to construct an original database;

[0054] The specific data for evaluating the risk of rockburst include mining depth, uniaxial compressive strength, elastic energy index, impact energy index, dynamic failure time, distance between fault and working face, distance between thick rock layer and coal seam, working face length, width of section coal pillar, thickness of bottom coal seam, pressure relief rate of protected layer, stress index ratio, and drilling powder rate index.

[0055] The main influencing factors of rockburst risk at the working face include one or more of the following: mining depth, coal seam rockburst tendency, geological structure, and mining technology conditions. Coal seam rockburst tendency is an inherent mechanical property of the coal body, directly determining the potential for rockburst under external stress disturbances. Coal seams with higher rockburst tendency are more likely to experience rockburst risk under stress disturbances. The stress on the working face increases with mining depth, and stress concentration becomes increasingly significant. Rockburst occurs when the stress on the working face exceeds its critical stress. Poor geological conditions can easily lead to stress concentration in localized areas, forming stress anomaly zones, thus significantly increasing the probability of rockburst occurrence. Mining technology conditions can directly alter the stress distribution at the working face; unreasonable arrangement of a particular technology or the cumulative effect of multiple factors can increase the probability of rockburst. Furthermore, current methods for assessing rockburst risk mainly include drill cuttings monitoring and critical stress index monitoring. The specific quantitative values ​​of the above evaluation indicators in this embodiment are shown in Table 1.

[0056] Table 1: Quantitative values ​​of rockburst hazard assessment indicators:

[0057] The mining depth h refers to the vertical distance from the ground surface to the top of the mining face, that is, the mining depth is the sum of the ground elevation and the depth of the mining layer;

[0058] The evaluation indicators for coal seam impact tendency include uniaxial compressive strength, elastic energy index, impact energy index, and dynamic failure time. The first three indicators are determined based on the stress-strain curves of the entire uniaxial compression test of the specimen, while the dynamic failure time is determined based on the stress-time curve of the entire uniaxial compression test of the specimen. Standard cylindrical specimens of Φ50mm×100mm are prepared from rock, uniaxial compressive loads are applied, and the stress-strain curves and stress-time curves of the entire process are recorded.

[0059] In this embodiment, the rock is first processed as follows: Figure 4 The standard cylindrical specimen with a diameter of 50mm × 100mm shown is as follows: Figure 4 (a) is a full view of the standard specimen for uniaxial compression testing; Figure 4 (b) Main (side) view; Figure 4 (c) is a top view, with the sample and load cell then placed... Figure 3 On the spherical base of the measurement system shown, adjust the spherical base to ensure uniform force on the sample. Then, connect the load sensor, displacement sensor, dynamic resistance strain gauge, and data acquisition system sequentially. The strain rate is controlled within the quasi-static strain rate range, such as 0.5 × 10⁻⁶. -5 The test machine is started again when the specimen comes into contact with the upper bearing plate at a speed of mm / s. The displacement sensor is installed between the upper and lower bearing plates and the test machine is started again to apply the load. The measured signal is transmitted to the data acquisition system through the dynamic resistance strain gauge, thereby recording the stress-strain curve and stress-time curve of the whole process.

[0060] Uniaxial compressive strength σ C Uniaxial compressive strength refers to the maximum load per unit area that a coal sample can withstand under unconfined axial pressure until failure. The stress value corresponding to the peak failure point of the stress-strain curve is the uniaxial compressive strength of the sample. Figure 5 As shown in (a), the calculation formula is shown in equation (1):

[0061] (1);

[0062] In the formula, P C The load (N) corresponding to the uniaxial compression failure of the specimen is represented by A, and the bearing area of ​​the specimen is represented by ( ). ).

[0063] Elasticity index (W) et The elastic strain energy (TSE) refers to the ratio of the elastic deformation energy to the plastic deformation energy of a coal sample under uniaxial compression when the stress reaches a certain value before failure and is unloaded. The elastic strain energy is the area under the unloading curve. Figure 5 (c) The area shown represents the plastic strain energy, which is the area enclosed by the loading and unloading curves.Figure 5 (c) The area shown is given. The calculation formulas are shown in equations (2) and (3).

[0064] (2);

[0065] (3);

[0066] In the formula, Represents the total strain energy. Represents plastic deformation energy. It represents the elastic deformation energy.

[0067] Impact Energy Index K e This refers to the ratio of the deformation energy accumulated before the peak to the deformation energy lost after the peak in the stress-strain curve of a coal sample under uniaxial compression. The area enclosed by the curve before the peak point of the stress-strain curve is the deformation energy accumulated before the peak point, and the area enclosed by the curve after the peak point is the deformation energy lost after the peak point. The calculation formula is shown in equation (4).

[0068] (4);

[0069] In the formula, A S A represents the deformation energy accumulated before the peak. X This represents the deformation energy lost after the peak value.

[0070] Dynamic destruction time (D) T ( ) refers to the time taken for a test specimen to develop from its ultimate strength to complete failure under uniaxial compressive load, such as Figure 5 As shown in (d).

[0071] Distance between fault and working face (L) d The fault line (FLP) is the vertical distance from the leading edge of the working face or a specific measuring point to the intersection of the fault plane and the coal seam. Faults are one of the main factors inducing rockburst hazards. When a fault is close to the working face, stress concentration occurs in the coal and rock mass between them. When the accumulated stress exceeds its critical stress, rockburst disasters may occur. As the distance between the fault and the working face decreases, the impact risk from fault activation on the working face increases.

[0072] The distance between the hard, thick rock strata in the roof and the coal seam (L) h The uniaxial compressive strength is a core influencing factor affecting the risk of rock bursts at the working face. It refers to the vertical distance between the lower boundary of the hard rock strata and the upper boundary of the coal seam. Rock strata with a strength of ≥60MPa are called hard, thick rock strata. The smaller the distance between the hard, thick rock strata in the roof and the coal seam, the greater the load exerted on the coal seam by the weight of the thick rock strata, the more energy is accumulated in the coal seam, and the greater the impact of the rock strata fracturing on the coal seam, making it more prone to rockburst disasters.

[0073] The working face length (L) refers to the effective horizontal mining width of the longwall face along the dip direction of the coal seam, from the coal wall of the return airway to the coal wall of the transport roadway. An increase in working face length is one of the reasons for the increased intensity of rockburst risk. As the working face length increases, the maximum concentrated stress in front of the coal wall gradually increases, leading to an increased risk of rockburst.

[0074] The width (d) of a coal pillar section refers to the horizontal width of the coal body that is reserved between two adjacent longwall faces and is not mined. Its main function is to isolate adjacent mining spaces and bear mining stress. Therefore, as the width of the coal pillar section increases, the risk of rock bursts gradually decreases.

[0075] Thickness of the coal seam at the bottom (t) d The bottom coal layer (BCL) refers to the vertical thickness of the coal seam that has not been mined and is left on the bottom of the mining face. When a thicker BCL is left during the mining process, the stress of the BCL will change when affected by external factors. When the accumulated elastic energy is suddenly released, it will cause rock bursts. The greater the thickness of the BCL, the greater the possibility of rock bursts.

[0076] Protective layer mining is an economical and effective stress relief and rockburst prevention technology in coal mine rockburst control. The most critical and core indicator for evaluating the degree of stress relief of the protective layer is the stress reduction rate of the protected layer. ), which is the ratio of the original rock stress of the protected layer to the residual stress after decompression. Its calculation formula is shown in equation (5).

[0077] (5);

[0078] In the formula, This indicates that the original rock stress of the protected layer was measured using the stress relief method. This indicates that the residual stress after depressurization corresponds to the stress value after depressurization stabilization, which is monitored by a borehole stress gauge.

[0079] The stress index ratio (η) is the ratio of the critical stress index to the rockburst occurrence index. The rockburst occurrence index is determined in the same way as the critical stress index. The critical stress index of a region that has experienced rockburst hazard and has the same rockburst tendency as the evaluation region is selected as the rockburst occurrence index of the evaluation region. The calculation formula is shown in equation (6).

[0080] (6);

[0081] In the formula, The critical stress index is expressed in equation (7). This indicates the rockburst occurrence index.

[0082] (7);

[0083] In the formula, P represents the actual stress of the coal seam, R represents the measured radius of the roadway, and P cr The critical stress for rockburst is shown in equation (8), R. cr The formula for calculating the depth of the critical resistance zone of the roadway is shown in equation (9).

[0084] (8);

[0085] (9);

[0086] In the formula, P represents uniaxial compressive strength, K represents impact energy index, a represents tunnel radius, and n represents tunnel shape variation coefficient: n=1 for circular tunnels, n=1.1~1.3 for semi-circular tunnels, n=1.3~1.5 for rectangular tunnels, and n=1.5~1.8 for trapezoidal tunnels. S The formula for calculating the support stress is shown in equation (10).

[0087] (10);

[0088] In the formula, F represents the sum of the support resistance of each support component, and S represents the support area.

[0089] The drill cuttings method involves drilling 42mm diameter, 12m deep, and 20m spacing coal boreholes in a single row within the coal seam. The relationship between the amount of drill cuttings removed during drilling and the normal amount of drill cuttings is used to monitor the risk of rockburst at the working face. Normal drill cuttings refers to the amount of drill cuttings measured under normal stress conditions. The drill cuttings ratio (ζ) is a key indicator for monitoring and evaluating the risk of rockburst at the working face; it is the ratio of actual drill cuttings per meter to normal drill cuttings per meter. The calculation formula is shown in equation (11).

[0090] (11);

[0091] In the formula, G p G0 represents the actual amount of drill cuttings per meter, while G0 represents the normal amount of drill cuttings per meter.

[0092] Step 2: Apply Principal Component Analysis (PCA) to reduce the dimensionality of the data to obtain comprehensive evaluation index data that includes the original evaluation index information;

[0093] Let x ijLet x1, x2, ..., xn be the value of the j-th indicator corresponding to the i-th sample in a real mine, where i = 1, 2, ..., m, m is the number of samples, and j = 1, 2, ..., n, n is the number of evaluation indicators; then the original evaluation indicators are x1, x2, ..., xn. n Let the comprehensive index obtained through principal component analysis, i.e., the principal components, be y1, y2, ..., y3. n Then the principal components are linear combinations of the original evaluation indicators after standardization, with a sum of squares of coefficients of 1 and each principal component being uncorrelated.

[0094] Step 2.1: Standardize the collected data on actual mine rockburst hazard indicators to obtain standardized evaluation indicators. The specific calculation formula is shown in equation (12):

[0095] (12);

[0096] In the formula, s represents the mean of the j-th indicator, calculated using formula (13). j Let represent the variance of the j-th index, calculated using formula (14):

[0097] (13);

[0098] (14);

[0099] Step 2.2: Calculate the correlation coefficients among the standardized indicators to obtain the correlation coefficient matrix R, and calculate the eigenvalues ​​of the correlation coefficient matrix R. , , ..., and normalized eigenvectors , , ..., The formula for calculating the correlation coefficient matrix R is shown in equation (15):

[0100] (15);

[0101] In the formula, This represents the correlation coefficient between the k-th indicator and the l-th indicator, where k = 1, 2, ..., n, l = 1, 2, ..., n. Correlation coefficient The calculation formula is shown in equation (16).

[0102] (16);

[0103] Step 2.3: Calculate the variance contribution rate of each principal component. The calculation formula is shown in equation (17), which calculates the cumulative contribution rate of the first w principal components. The calculation formula is shown in equation (18), where s = 1, 2, ..., n, w = 1, 2, ..., n;

[0104] (17);

[0105] (18);

[0106] Step 2.4: Calculate the scores of the determined principal components; select the top w principal components with eigenvalues ​​greater than 1 and cumulative contribution rates of 85%~95% and calculate their corresponding scores. The calculation formula is shown in equation (19):

[0107] (19);

[0108] In the formula, This represents the normalized eigenvector corresponding to the t-th principal component. , , ..., Let represent n standardized evaluation indicators, where t = 1, 2, ..., w.

[0109] The scores of each principal component obtained after the original indicator data is processed by Principal Component Analysis (PCA) for dimensionality reduction are the comprehensive evaluation indicator data containing information from the original indicator data.

[0110] Step 3: Divide the comprehensive evaluation index data obtained by principal component dimensionality reduction into training set and test set according to the ratio. Use the comprehensive evaluation index data as the input layer of the probabilistic neural network (PNN) and learn and train it through the training set data to obtain the trained probabilistic neural network. In this embodiment, the division ratio is 7:3.

[0111] The probabilistic neural network (PNN) includes an input layer, a hidden layer, a summation layer, and an output layer, as detailed below:

[0112] Step A1: Determine the number of neurons in the input layer; the number of neurons in the input layer is the number of principal components selected after PCA processing. The comprehensive evaluation index data obtained after dimensionality reduction by principal component analysis is the input layer vector.

[0113] Step A2: Determine the number of neurons in the hidden layer and calculate the input-output relationship of neurons in each mode; the number of neurons in the hidden layer is the number of samples contained in the training set. After receiving the input from the input layer, the hidden layer calculates the distance between the input vector and the feature vector of the training sample and returns a standard value. The calculation formula is shown in equation (20):

[0114] (20);

[0115] In the formula, This represents the relationship between the input and output determined by the j-th neuron of the i-th pattern, σ represents the smoothing factor of the probabilistic neural network (PNN), d represents the dimension of the sample space data, and x represents the feature vector of the test set data or the specific data of the object to be evaluated. ab Let represent the feature vector of the b-th training sample of the a-th class stored in the hidden layer, where a = 1, 2, ..., M, and M represents the total number of classes in the training set.

[0116] Step A3: Determine the number of neurons in the summation layer and perform a weighted average of the outputs of neurons of the same category; the number of neurons in the summation layer is the total number of sample patterns. The summation layer performs a weighted average of the outputs of neurons belonging to the same category in the hidden layer. The calculation formula is shown in equation (21):

[0117] (twenty one);

[0118] In the formula, v i Let L represent the output value of the i-th class, and L represent the number of neurons in the i-th class.

[0119] Step A4: Determine the number of neurons in the output layer and output the corresponding results; the number of neurons in the output layer is the total number of patterns in the training samples in the training set; since the rockburst hazard is divided into four categories: no impact, weak impact, moderate impact, and strong impact, the rockburst hazard is recorded as level 0, 1, 2, and 3 according to the intensity of the rockburst hazard, where 0 represents no impact and 3 represents strong impact. The output result is in the form of a probability vector, and the sum of each element is 1; after the PNN outputs the probability, the maximum a posteriori probability decision (MAP) is added based on the Bayesian rule, that is, the neuron with the maximum a posteriori probability density outputs 1, and the rest output 0. In this embodiment, the output vector of the probabilistic neural network PNN is a row vector similar to (1, 0, 0, 0), indicating that the rockburst hazard of the sample is level 0, that is, no impact hazard.

[0120] Step 4: Test the performance of the trained probabilistic neural network using test set data and calculate the accuracy of the prediction results. When the accuracy is greater than or equal to the set value, the model is considered to meet the standard and the prediction model is obtained. When the accuracy is less than the set value, the smoothing factor of the PNN needs to be adjusted and the training set needs to be relearned until the accuracy reaches the set value. In this embodiment, the set value is 90%.

[0121] Step 5: Use the obtained prediction model to conduct an intelligent assessment of the rockburst hazard at the working face.

[0122] Example 2:

[0123] This embodiment proposes a computer-readable storage medium that stores executable instructions. When these instructions are executed, if they are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0124] The computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the intelligent assessment method for rockburst hazard of working face based on principal component analysis-probabilistic neural network described in the various embodiments of this application.

[0125] The aforementioned storage media include: flash memory, hard disk, multimedia card, card-type memory (e.g., SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR) memory, random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, disk, optical disk, server, APP (Application) application store, and other media capable of storing program verification codes. These media store computer programs, and when executed by a processor, they can implement the various steps of the aforementioned intelligent assessment method for working face rockburst hazard based on principal component analysis-probabilistic neural network.

[0126] Example 3:

[0127] This embodiment proposes a computer program product, including a computer program or instructions, which, when executed by a processor, implements the aforementioned intelligent assessment method for working face rockburst hazard based on principal component analysis-probabilistic neural network.

[0128] Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or part of the technical solution, can be embodied in the form of a computer program product.

[0129] The various embodiments in this application are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0130] The scope of protection of this application is not limited to the embodiments described above. Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from the scope and spirit of this disclosure. If such modifications and variations fall within the scope of the methods disclosed herein and their equivalents, then the intent of this disclosure also includes such modifications and variations.

Claims

1. A method for intelligent assessment of rockburst hazard at working faces based on principal component analysis-probabilistic neural networks, characterized in that, Includes the following steps: Step 1: Establish a risk assessment index system for rockburst at the working face and collect rockburst risk assessment indexes to construct an original database; Step 2: Apply Principal Component Analysis (PCA) to reduce the dimensionality of the data to obtain comprehensive evaluation index data that includes the original evaluation index information; Step 3: Divide the comprehensive evaluation index data obtained by principal component dimensionality reduction into training set and test set according to the ratio. Use the comprehensive evaluation index data as the input layer of the probabilistic neural network (PNN) and learn and train it through the training set data to obtain the trained probabilistic neural network. Step 4: Test the performance of the trained probabilistic neural network using test set data and calculate the accuracy of the prediction results. When the accuracy is greater than or equal to the set value, the model is considered to meet the standard and the prediction model is obtained. When the accuracy is less than the set value, the smoothing factor of the PNN needs to be adjusted and the training set needs to be relearned until the accuracy reaches the set value. Step 5: Use the obtained prediction model to conduct an intelligent assessment of the rockburst hazard at the working face.

2. The intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural network according to claim 1, characterized in that, The rockburst hazard assessment index data mentioned in step 1 specifically include mining depth, uniaxial compressive strength, elastic energy index, impact energy index, dynamic failure time, distance between fault and working face, distance between thick rock layer and coal seam, working face length, section coal pillar width, bottom coal seam thickness, pressure relief rate of protected layer, stress index ratio, and drill dust rate index.

3. The intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural network according to claim 1, characterized in that, Step 2 specifically involves: Let x ij Let x1, x2, ..., xn be the value of the j-th indicator corresponding to the i-th sample in a real mine, where i = 1, 2, ..., m, m is the number of samples, and j = 1, 2, ..., n, n is the number of evaluation indicators; then the original evaluation indicators are x1, x2, ..., xn. n Let the comprehensive index obtained through principal component analysis, i.e., the principal components, be y1, y2, ..., y3. n Then the principal components are linear combinations of the original evaluation indicators after standardization, with a sum of squares of coefficients of 1 and each principal component being uncorrelated.

4. The intelligent assessment method for rockburst hazard at a working face based on principal component analysis-probabilistic neural network according to claim 3, characterized in that, Step 2 specifically includes the following steps: Step 2.1: Standardize the collected data on actual mine rockburst hazard indicators to obtain standardized evaluation indicators. As shown in the following formula: ； In the formula, Let s represent the mean of the j-th indicator. j The variance of the j-th index is expressed as follows: ； ； Step 2.2: Calculate the correlation coefficients among the standardized indicators to obtain the correlation coefficient matrix R, and calculate the eigenvalues ​​of the correlation coefficient matrix R. , , ..., and normalized eigenvectors , , ..., The correlation coefficient matrix R is as follows: ； In the formula, This represents the correlation coefficient between the k-th indicator and the l-th indicator, where k = 1, 2, ..., n, l = 1, 2, ..., n. Correlation coefficient The calculation formula is as follows: ； Step 2.3: Calculate the variance contribution rate of each principal component. Calculate the cumulative contribution rate of the first w principal components. Where s = 1, 2, ..., n, w = 1, 2, ..., n, as shown in the following formula: ； ； Step 2.4: Calculate the scores of the determined principal components; select the top w principal components with eigenvalues ​​greater than 1 and cumulative contribution rates reaching 85%~95%, and calculate their corresponding scores, as shown in the following formula: ； In the formula, This represents the normalized eigenvector corresponding to the t-th principal component. , , ..., Let represent n standardized evaluation indicators, where t = 1, 2, ..., w; The scores of each principal component obtained after the original indicator data is processed by Principal Component Analysis (PCA) for dimensionality reduction are the comprehensive evaluation indicator data containing information from the original indicator data.

5. The intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural network according to claim 1, characterized in that, The probabilistic neural network (PNN) described in step 3 includes an input layer, hidden layers, a summation layer, and an output layer. The specific process is as follows: Step A1: Determine the number of neurons in the input layer; the number of neurons in the input layer is the number of principal components selected after PCA processing, and the comprehensive evaluation index data obtained after dimensionality reduction by principal component analysis is the input layer vector; Step A2: Determine the number of neurons in the hidden layer and calculate the input-output relationship of neurons in each mode; the number of neurons in the hidden layer is the number of samples contained in the training set. After receiving the input from the input layer, the hidden layer calculates the distance between the input vector and the feature vector of the training sample and returns a standard value, as shown in the following formula: ； In the formula, This represents the relationship between the input and output determined by the j-th neuron of the i-th pattern, σ represents the smoothing factor of the probabilistic neural network (PNN), d represents the dimension of the sample space data, and x represents the feature vector of the test set data or the specific data of the object to be evaluated. ab Let represent the feature vector of the b-th training sample of class a stored in the hidden layer, where a = 1, 2, ..., M, and M represents the total number of classes in the training set; Step A3: Determine the number of neurons in the summation layer and perform a weighted average of the outputs of neurons of the same category; the number of neurons in the summation layer is the total number of sample patterns. The summation layer performs a weighted average of the outputs of neurons belonging to the same category in the hidden layer, as shown in the following formula: ； In the formula, v i L represents the output value of the i-th class, and L represents the number of neurons in the i-th class. Step A4: Determine the number of neurons in the output layer and output the corresponding results; the number of neurons in the output layer is the total number of patterns in the training samples in the training set; according to the intensity of the rockburst hazard, the rockburst hazard is recorded as level 0, 1, 2, and 3, where 0 represents no impact and 3 represents a strong impact. The output result is in the form of a probability vector, and the sum of each element is 1; after the PNN outputs the probability, the maximum a posteriori probability decision (MAP) is added based on the Bayesian rule, that is, the neurons with the maximum a posteriori probability density output 1, and the rest output 0.

6. A computer-readable storage medium, characterized in that, The system stores executable instructions that, when executed, cause the processor to perform the intelligent assessment method for rockburst hazard at the working face based on principal component analysis-probabilistic neural network as described in any one of claims 1-5.

7. A computer program product, characterized in that, Includes a computer program or instructions that, when executed by a processor, implement the intelligent assessment method for rockburst hazard at working faces based on principal component analysis-probabilistic neural networks as described in any one of claims 1-5.

Images