An intelligent analysis algorithm for gas outburst pre-warning of a coal mine tunneling working face

Abstract

The present application relates to a kind of intelligent analysis algorithm for coal mine tunneling working face gas outburst early warning, it is related to gas early warning data processing system field, including the following steps: first, the gas parameter data obtained is normalized, the correlation degree of each data index is sorted using grey correlation analysis method, and the data index with correlation degree greater than 0.65 is retained;Subsequently, the SVM model of gas outburst early warning is established, and RBF kernel function is selected;Second, the parameters of SVM model are calculated using adaptive particle swarm algorithm, based on the obtained calculation parameters, the sample data obtained in the corresponding step is imported, and SVM early warning model training is carried out.Last, the trained SVM early warning model is used to realize the gas outburst early warning of the current tunneling working face position.The present application uses improved particle swarm algorithm, realizes the automatic optimization selection of the calculation parameters of SVM model, carries out mining analysis to multidimensional, multi-source gas data, and realizes the stable and reliable coal mine tunneling working face gas outburst early warning.

Description

Technical Field

[0001] This invention relates to an intelligent analysis algorithm for early warning of gas outbursts in coal mine tunneling faces, and relates to the field of gas early warning data processing systems. Background Technology

[0002] In recent years, with the extension of coal mining levels and the increase in mining depth, gas outbursts have had an increasingly significant impact on coal mine safety. Gas outbursts threaten the lives of coal miners, and once an accident occurs, it can cause significant economic losses to the country. The prediction of gas outbursts has always been a key research focus. Gas outbursts are a nonlinear dynamic process coupled by multiple factors; relying solely on one or a few indicators for prediction has low accuracy and limitations. Traditional data mining algorithms heavily depend on parameter selection during the calculation process, often relying on experience to set parameters. This leads to the calculation process getting trapped in local minima, preventing the attainment of globally optimal results and thus hindering the improvement of the stability and reliability of early warning results.

[0003] During the mining process, coal mine tunneling faces reveal a large amount of real data, such as geological structure, coal seam occurrence conditions, and gas parameters. How to effectively utilize the large amount of real historical data from multiple dimensions and sources to analyze and achieve early warning of gas outbursts is a key research point. Summary of the Invention

[0004] Purpose of the invention: To propose an intelligent analysis algorithm for early warning of gas outbursts in coal mine tunneling faces, so as to solve the above-mentioned problems in the existing technology and realize stable and reliable early warning of gas outbursts in coal mine tunneling faces.

[0005] Technical Solution: A smart analysis algorithm for early warning of gas outbursts in coal mine tunneling faces is provided. The method includes the following steps:

[0006] Step 1: Obtain gas parameter data from manual detection and monitoring systems at the coal mine tunneling face. This data includes training data across multiple dimensions such as distance from geological structural zones, coal seam gas content, initial gas emission velocity, K1 gas dissipation rate, drill cuttings volume, coal seam depth, coal seam thickness, coal failure type, coal body firmness coefficient, and gas concentration. Each set of data corresponds to a gas outburst hazard level, categorized into three levels: safe, moderate, and severe, represented by 0, 0.5, and 1 respectively. The acquired data is then normalized using the extreme value method.

[0007] Step 2: Perform correlation analysis on the data of each indicator; based on the normalized data, use grey relational analysis to analyze the correlation between the data of each indicator and the experimental results, and retain the data of indicators with a correlation degree greater than a predetermined value; remove the data of other indicators to improve the training effect of the model.

[0008] Step 3: Establish an SVM training model for early warning of gas outbursts in coal mine tunneling faces;

[0009] Step 4: Obtain the corresponding kernel function parameters σ and b using the SVM model for early warning of gas outbursts in coal mine tunneling faces. * ;

[0010] Step 5: Based on the optimal model parameters σ and b obtained in Step 4 * To train the SVM model for gas outbursts in coal mine tunneling faces, the gas index data samples obtained in step 2 are imported into the model. After training, a gas outburst early warning model for coal mine tunneling faces that can be used for prediction is generated.

[0011] Step 6: Use the early warning model obtained in Step 5 to conduct early warning of gas outbursts in coal mine working faces; input the newly collected gas index data into the early warning model, and the model will automatically give the corresponding early warning risk level for this set of data.

[0012] The specific calculation method of the extremum method described in step 1 is shown in one of the equations:

[0013]

[0014] In the formula: p is the evaluation object, q is the evaluation parameter, i = 1, 2, ..., m, k = 1, 2, ..., n, x max and x min These are the maximum and minimum values ​​of the sequence, respectively.

[0015] Step 2: Based on the normalized data, the grey relational analysis method of Equations 2 and 3 is used to analyze the correlation between the data of each indicator and the experimental results. The data of indicators with a correlation degree greater than 0.65 are retained.

[0016]

[0017] In the formula ξ p (q) is the grey relational coefficient of the p-th evaluation object and the q-th parameter; ρ (0<ρ<1) is the discrimination coefficient. The smaller ρ is, the greater the difference between the correlation coefficients and the stronger the discrimination ability.

[0018]

[0019] r p is the average value of a certain indicator data series, representing the correlation between the indicator and the experimental results; n is the number of data points for a certain indicator.

[0020] Equation 4 is the objective function of the SVM model in step 3, and the kernel function is the RBF kernel function of Equation 5.

[0021]

[0022]

[0023] In the formula, σ and b * The parameters that need to be set for calculating the SVM model.

[0024] The kernel function parameters σ and b of the SVM model for early warning of gas outbursts in coal mine tunneling faces. * The method to obtain it is as follows:

[0025] σ and b are calculated using the improved particle swarm optimization algorithm based on equations 6, 7, and 8. * Set penalty factor b * The range of variation is (0.1, 100), the range of variation of kernel function parameter σ is (0.01, 1000), c1 is 1.5, c2 is 1.7, the maximum number of generations is 200, and the population size is 20.

[0026] The position x updated each time id The two parameters of SVM are used for sample training, and the average error of the training samples is used as the optimization objective of the particle swarm optimization algorithm.

[0027] v id (t+1)=wv id (t)+c1r1[p id (t)-x id (t)]+c2r2[p gd (t)-x id (t)] Equation 6

[0028] x id (t+1)=x id (t)+v id (t+1) Equation 7

[0029] In the formula, t is the number of iterations; d is the search dimension. In this paper, the parameters σ and C of the SVM are optimized, which are two in total, so d = 2; I = 1, 2, ..., m is the number of particles;

[0030] c1 is the individual learning factor, c2 is the social learning factor; r1 and r2 are random numbers distributed in the interval [0, 1]; p id (t) represents the historical best position of the particle during the t-th iteration; p gd (t) represents the historical best position of the population in the t-th iteration; v id (t) represents the velocity of the i-th particle in d-dimensional space, vid ∈[v min v max ];x id Let be the position of the i-th particle in d-dimensional space; w is the inertial weight, which is used to adjust the particle's flight speed and position by introducing an adaptive inertial weight, as expressed in Equation 8:

[0031]

[0032] Beneficial Effects: An intelligent analysis algorithm for early warning of gas outbursts in coal mine tunneling faces includes the following steps: First, the acquired gas parameter data is normalized, and the correlation degree of each data indicator is ranked using grey relational analysis, retaining data indicators with a correlation degree greater than 0.65; then, an SVM model for gas outburst early warning is established, and the RBF kernel function is selected; next, the parameters of the SVM model are calculated using an adaptive particle swarm optimization algorithm, and based on the obtained calculation parameters, the sample data obtained in the corresponding steps is imported to train the SVM early warning model. Finally, the trained SVM early warning model is used to achieve early warning of gas outbursts at the current tunneling face location. Addressing the problem that traditional gas data mining algorithms rely on empirically selected calculation parameters, resulting in insufficient stability and reliability of early warning results, this patent employs an improved particle swarm optimization algorithm to automatically optimize the selection of calculation parameters for the SVM model, performing mining and analysis on multi-dimensional, multi-source gas data to achieve stable and reliable early warning of gas outbursts in coal mine tunneling faces. Attached Figure Description

[0033] Figure 1 This is a flowchart illustrating an intelligent analysis algorithm for early warning of gas outbursts in coal mine tunneling faces. Detailed Implementation

[0034] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0035] The applicant argues that traditional data mining algorithms heavily rely on parameter selection during computation, often setting parameters based on experience. This leads to the computation process getting trapped in local minima, preventing the attainment of globally optimal results and thus hindering the improvement of the stability and reliability of early warning outcomes.

[0036] Therefore, this case aims to design an intelligent analysis algorithm for early warning of gas outbursts in coal mine tunneling faces. It addresses the problem of how to effectively utilize the large amount of current, multi-dimensional, and multi-source actual historical data to analyze and implement early warning of gas outbursts, so as to achieve stable and reliable early warning of gas outbursts in coal mine tunneling faces.

[0037] An intelligent analysis algorithm for early warning of gas outbursts in coal mine tunneling faces includes the following steps:

[0038] Step 1: Obtain gas parameter data from manual detection and monitoring systems at the coal mine tunneling face. This data includes training data across multiple dimensions such as the distance to geological structural zones, coal seam gas content, initial gas emission velocity, K1 gas dissipation rate, drill cuttings volume, coal seam depth, coal seam thickness, coal failure type, coal body firmness coefficient, and gas concentration. Each set of data corresponds to a gas outburst hazard level, categorized into three levels: safe, moderate, and severe, represented by 0, 0.5, and 1 respectively. The obtained data is then normalized using the extremum method described in Equation 1.

[0039]

[0040] In the formula: p is the evaluation object, q is the evaluation parameter, i = 1, 2, ..., m, k = 1, 2, ..., n, x max and x min These are the maximum and minimum values ​​of the sequence, respectively.

[0041] Step 2: Perform correlation analysis on the data of each indicator. Based on the normalized data, the grey relational analysis method according to Equations 2 and 3 is used to analyze the correlation between the data of each indicator and the experimental results. The data of indicators with a correlation degree greater than 0.65 are retained. Other data of indicators are removed to improve the training effect of the model.

[0042]

[0043] In the formula ξ p (q) is the grey relational coefficient of the p-th evaluation object and the q-th parameter; ρ (0 < ρ < 1) is the discrimination coefficient. The smaller ρ is, the greater the difference between the correlation coefficients and the stronger the discrimination ability. Generally, ρ = 0.5 is taken.

[0044]

[0045] r p is the average value of a certain indicator data series, representing the correlation between the indicator and the experimental results. n is the number of data points for a certain indicator.

[0046] Step 3: Establish an SVM training model for early warning of gas outbursts in coal mine tunneling faces. Equation 4 is the objective function of the SVM model, and the kernel function is the RBF kernel function selected from Equation 5.

[0047]

[0048]

[0049] In the formula, σ and b * The parameters that need to be set for calculating the SVM model.

[0050] Step 4: Kernel function parameters σ, b of the SVM model for early warning of gas outbursts in coal mine tunneling faces * The method for obtaining σ and b is as follows: σ and b are calculated using the improved particle swarm optimization algorithm based on equations 6, 7, and 8. * Set penalty factor b * The range of values ​​is (0.1, 100), the range of kernel function parameter σ is (0.01, 1000), c1 is 1.5, c2 is 1.7, the maximum number of generations is 200, and the population size is 20. The position x updated each time... id The samples are used as two parameters for SVM training, and the average error of the training samples is used as the optimization objective of the particle swarm optimization algorithm.

[0051] v id (t+1)=wv id (t)+c1r1[p id (t)-x id (t)]+c2r2[p gd (t)-x id (t)] Equation 6

[0052] x id (t+1)=x id (t)+v id (t+1) Equation 7

[0053] In the formula, t is the number of iterations. d is the search dimension. This paper optimizes the SVM parameters σ and C, which are two in total, therefore d = 2. I = 1, 2, ..., m is the number of particles.

[0054] c1 is the individual learning factor, and c2 is the social learning factor. r1 and r2 are random numbers distributed in the interval [0, 1]. p id (t) represents the historical best position of the particle during the t-th iteration. gd (t) represents the best historical position of the population in the t-th iteration. id (t) represents the velocity of the i-th particle in d-dimensional space, v id ∈[v min v max ]. x idLet be the position of the i-th particle in d-dimensional space. w is the inertial weight, which adjusts the particle's velocity and position by introducing an adaptive inertial weight, expressed in Equation 8:

[0055]

[0056] Step 5: Based on the optimal model parameters σ and b obtained in Step 4 * To train a gas outburst SVM model for coal mine tunneling faces, the gas index data samples obtained in step 2 are imported into the model. After training, a gas outburst early warning model for coal mine tunneling faces that can be used for prediction is generated.

[0057] Step 6: Use the early warning model obtained in Step 5 to conduct early warning of gas outbursts at coal mine working faces. Input the newly collected gas index data into the early warning model, and the model will automatically give the corresponding early warning risk level for this set of data.

[0058] The research on intelligent analysis algorithms for early warning of gas outbursts in coal mine tunneling faces involves the following workflow: First, the acquired gas parameter data is normalized, and the correlation degree of each data indicator is ranked using grey relational analysis, retaining data indicators with a correlation degree greater than 0.65. Next, an SVM model for gas outburst early warning is established, selecting the RBF kernel function. Then, the parameters of the SVM model are calculated using an adaptive particle swarm optimization algorithm. Based on the obtained parameters, the sample data obtained in step 2 is imported to train the SVM early warning model. Finally, the trained SVM early warning model is used to achieve early warning of gas outbursts at the current tunneling face location.

[0059] As described above, although this embodiment has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the embodiment itself. Various changes in form and detail may be made to this embodiment without departing from the spirit and scope of the embodiment as defined in the appended claims.

Claims

1. An intelligent analysis algorithm for gas outburst early warning of a coal mine tunneling working face, characterized in that, Comprise the following steps: Step 1: obtain the manual detection gas parameter data of coal mine tunneling face and the gas parameter data of monitoring system, the indexes include geological structure belt distance, coal seam gas content, coal gas diffusion initial velocity, K1 gas analysis amount, drill cuttings, coal seam buried depth, coal seam thickness, coal damage type, coal body firmness coefficient, gas concentration multiple dimension training data and each group of data corresponding gas outburst danger grade, the danger grade is divided into safety, general and major three grades, respectively represented by 0, 0.5, 1; the obtained data is processed by dimensionless normalization method; Step 2: correlation analysis is carried out on each index data; based on the normalized data, the grey correlation analysis method is used to analyze the correlation between each index data and the experimental results, and the index data with correlation degree greater than the predetermined value is reserved; other index data is rejected to improve the training effect of the model; Step 3: establish the SVM training model of coal mine tunneling face gas outburst early warning; Step 4: Obtain the corresponding kernel function parameters through the SVM model for early warning of gas outburst at the coal mining face , ; Step 5: Obtain the optimal model parameters based on step 4 , Perform SVM model training for gas outburst of coal mine tunneling face, import the gas index data samples obtained in step 2 into the model, and generate a coal mine tunneling face gas outburst early warning model that can be used for prediction after training. Step 6: use the early warning model obtained in step 5 to carry out coal mine working face gas outburst early warning; input the newly collected gas index data into the early warning model, and the model automatically gives the corresponding early warning risk grade of the data; Formula four is the objective function of the SVM model in step 3, and the kernel function is the RBF kernel function of formula five; Formula 4 Formula Five In the formula , Parameters to be set for the calculation of the SVM model; Coal mine tunneling working face gas outburst early warning SVM model kernel function parameter ,The method for obtaining the is as follows: The improved particle swarm algorithm of formula six, formula seven and formula eight is used for calculation , , the change range of the penalty factor is (0.1, 100), the change range of the kernel function parameter is (0.01, 1000), is 1.5, is 1.7, the maximum evolution generation is 200, and the population quantity is 20; The position of each update The sample is trained as two parameters of SVM, and the average error of the training sample is taken as the optimization target of the particle swarm algorithm. Formula Six Formula Seven In the formula is the iteration number; is the search dimension, the parameters of the optimized SVM in this paper and C, a total of two, so = 2; is the number of particles; is an individual learning factor, is a social learning factor; and is a random number distributed in the interval ; is the historical best position of the particle at the th iteration; is the historical best position of the population at the th iteration; is the position of the th particle in the -dimensional space, ; is the position of the th particle in the -dimensional space; is the inertia weight, which adjusts the flying velocity and position of the particle by introducing an adaptive inertia weight, and its expression is formula eight: Formula Eight.

2. The intelligent analysis algorithm for gas outburst pre-warning of a coal mine tunneling working face according to claim 1, characterized in that, The specific operation mode of the extreme value method in step 1 is shown in formula one: Formula One Formula One wherein: is the evaluation object, is the evaluation parameter, , , and are the maximum and minimum values of the sequence, respectively.

3. The intelligent analysis algorithm for coal mine tunneling face gas outburst early warning according to claim 1, characterized in that: Step 2: based on the normalized data, the grey correlation analysis method of formula two and formula three is used to analyze the correlation between each index data and the experimental results, and the index data with correlation degree greater than 0.65 is reserved; Formula 2 In the formula For the first The first evaluation object Grey relational coefficients of each parameter; ( ) is the resolution coefficient, if The smaller the value, the greater the difference between the correlation coefficients, and the stronger the discrimination ability. Formula Three Formula Three is the average value of the data sequence of the index, representing the correlation degree of the index and the experimental result; n is the data number of the index.

Images