Coal and gas outburst danger prediction method based on HDAIA
Patent Information
- Application Number
- CN202610798056.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-04
- Publication Date
- 2026-08-28
AI Technical Summary
[0004]本发明的目的在于为了解决AIA的预测亲和度低,预测结果精度不足等问题,而提出一种基于HDAIA的煤与瓦斯突出危险性预测方法
[0028] After the antibody is generated, the affinity between the antigen and the antibody is randomly calculated to obtain the Hamming distance required for prediction.
Smart Images

Figure CN122654798A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent coal mine gas sequence prediction technology, specifically a method for predicting coal and gas outburst risk based on HDAIA. Background Technology
[0002] The mining of deep coal mines is proceeding in an orderly manner to meet coal supply needs. As the mining depth increases, the ground stress and gas pressure become increasingly greater. The impact of ground stress and other coal and rock mechanics factors on coal and gas outbursts is not yet fully understood, and the research on the mechanism of coal and gas outbursts is still in the hypothesis stage. Due to its suddenness and extremely high harmfulness, coal and gas outbursts remain the primary threat to coal mine production and a disaster problem that coal mine safety researchers urgently need to solve.
[0003] In the prevention and control of coal and gas outbursts, monitoring and prediction are the most crucial links. Timely detection and prediction, along with preventative measures to eliminate coal and gas outbursts in their nascent stage, are essential for ensuring the normal operation of underground mining workers. Researching and constructing effective methods for predicting the risk of coal and gas outbursts, and making advance predictions of potential outbursts, aligns with the safety requirements of coal mining under the new circumstances. Biological immune mechanisms share characteristics with the uncertainty and unpredictability of coal and gas outbursts; therefore, their immune mechanisms can be used for prediction. Immune algorithms, mimicking biological immunity, can be used for research on coal and gas outburst prediction. The principle of immune algorithms is to mimic the biological immune process through data analysis and processing to represent and infer the immune results—an artificial algorithm system. Immune algorithms have a wide range of applications, possessing characteristics such as fast global search speed and good convergence. They have achieved successful applications in data mining and classification in the field of coal and gas outburst prediction. However, they also suffer from insufficient accuracy in affinity calculations for predicting certain engineering problems. Summary of the Invention
[0004] The purpose of this invention is to address the problems of low predictive affinity and insufficient accuracy in AIA (Anti-Assay Algorithm) predictions by proposing a coal and gas outburst hazard prediction method based on HDAIA (High-Density AIA). This invention introduces a Hamming distance immune algorithm to predict the hazard of coal and gas outbursts, building upon existing basic immune algorithms. The Hamming distance immune algorithm is a type of immune algorithm that uses specific antibody affinity and multiple prediction indicators tailored to a particular problem. The algorithm first encodes the antigen data using binary 0s and 1s. After encoding, antibodies are generated based on antigen recognition. The success of antibody immunization is determined by comparing the number of bits in the antigen and antibody codes. Successfully immunized antibodies continue the calculation. Subsequently, the number of bits in the antigen and antibody codes is compared, using the number of different bits in the generated antibody codes as a reference value for affinity comparison. The minimum Hamming distance is obtained, and antibodies with high affinity are selected, leading to the optimal solution. This invention combines the advantages of antibody immunization and Hamming distance affinity calculation for antibody screening, and constructs a coal and gas outburst hazard prediction model based on the Hamming distance immunization algorithm. This has positive significance for enriching the coal and gas outburst prediction method system and has guiding value for engineering practice.
[0005] The objective of this invention can be achieved through the following technical solution: a coal and gas outburst hazard prediction method based on HDAIA, comprising:
[0006] Step 1: Collect indicator data of local coal and gas outbursts in underground coal mines. After processing the indicator data, use it as an antigen to perform membership processing and divide it into training set and prediction set.
[0007] Step 2: Using the artificial immune algorithm and Hamming distance, establish a coal and gas outburst hazard prediction model based on HDAIA;
[0008] Step 3: Input the training set into the coal and gas outburst hazard prediction model to train the algorithm, stimulate the algorithm to generate antibodies, calculate the antibody affinity through Hamming distance, and screen high affinity antibodies;
[0009] Step 4: Input the prediction set into the algorithm, calculate the antibody with the highest affinity to the antigen, input the converted antigen of the sample to be predicted into the algorithm, and calculate the minimum weighted Hamming distance between it and the set of antibodies with no prominent risk and the set of antibodies with prominent risk, respectively, to obtain the distance D. A And distance 2D B If D A If the value is greater than DB, it is determined to have a prominent hazard; otherwise, it is determined to have no hazard.
[0010] Step 5: If there is unpredicted data, repeat step 4 until there is no more predictable data; finally, output the coal and gas outburst risk prediction results.
[0011] In a preferred embodiment of the present invention, the index data includes the amount of drill cuttings and the desorption index value of drill cuttings gas.
[0012] In a preferred embodiment of the present invention, the specific steps for performing membership processing on the indicator data are as follows:
[0013] The membership values of the drill cuttings quantity and drill cuttings gas desorption index are processed to obtain the membership degree values;
[0014] In the prediction of coal and gas outburst risk, v prediction objects are set, and w prediction indicators are set for each prediction object. Then the original indicator data constitute a prediction indicator data matrix of order v×w.
[0015] If w predictive indicators have z-level predictive standards, then construct a z×w predictive standard matrix.
[0016] When performing normalization, the membership values of the predicted indicators corresponding to the predicted objects are obtained through interpolation function formulas, and the predicted indicator data matrix is converted into a consistent predicted indicator membership matrix.
[0017] For prediction objects with extremely high outburst risk, the prediction index is set to a membership degree of 1, while for prediction objects with weak or virtually no outburst risk, the prediction index is set to 0. This yields the membership function of the coal seam being predicted, which is then used to convert the prediction standard matrix into a membership matrix.
[0018] In a preferred embodiment of the present invention, the artificial immune algorithm includes: coal and gas outburst identification antigen: the index data of coal and gas outburst risk are encoded as the objective function, and a training set and a prediction set are randomly generated as antigens, waiting for input;
[0019] Generate initial antibodies: Initialize the antibodies for the randomly generated objective function of coal and gas outburst risk indicators to make them consistent. Train the algorithm using the training set to generate initial antibodies and check whether the generated antibodies are abnormal. Debug and improve the algorithm model. After the algorithm model matures and the error rate is reduced to an acceptable range, use the prediction set for prediction.
[0020] Calculate the affinity value;
[0021] Memory cell differentiation: During training using the training set, among the large number of antibodies generated, the antibody solution with the highest affinity to the indicator corresponding to the coal and gas outburst hazard will be added to the memory cells. At the same time, the antibody will recognize the memory cell with the highest affinity and replace it. When using the prediction set for prediction, the memory cells generated after training are activated and produce antibodies with high affinity.
[0022] Antibody promotion and inhibition: By calculating the expected value of the antibody, high-affinity antibodies are screened to obtain the desired antibody solution, while solutions that are lower than the set expected value are discarded.
[0023] Generate new antibodies: Among the generated antibodies, randomly select two antibodies with different affinities, and mutate and cross over these two antibodies according to a certain proportion and probability to obtain new antibodies. Repeat the operation until all F antibodies are generated to obtain the desired solution.
[0024] Hamming distance is used to compare the distances between antibodies: the Hamming distance algorithm is used to calculate the distance between the F generated antibodies and the antigen, and the optimal solution with the smallest distance is obtained;
[0025] Termination: When an optimal solution that satisfies the termination condition is found, the result is output, and the algorithm terminates.
[0026] In a preferred embodiment of the present invention, before calculating the affinity value, the method further includes: dividing the interval corresponding to the membership value into L-bit binary equal parts for encoding, converting the continuous membership values into binary strings, and then calculating the standard Hamming distance.
[0027] As a preferred embodiment of the present invention, the distance between the generated F antibodies and the antigen is calculated using the Hamming distance algorithm. Specifically, it includes: assuming the prediction space of coal and gas outburst is P, setting the coordinates of the antigen and antibody within the prediction space P, calculating the distance between the antigen and antibody, and then obtaining the Hamming distance.
[0028] After the antibody is generated, the affinity between the antigen and the antibody is randomly calculated to obtain the Hamming distance required for prediction.
[0029] Compared with existing technologies, the beneficial effects of this invention are as follows: By improving the Hamming distance to meet the needs of AIA for coal and gas outburst prediction, the affinity between antibodies is increased, thereby effectively predicting outburst anomalies. AIA has the advantages of fast convergence speed and no need for repeated encoding and decoding, which optimizes the immune mutation process and improves the shortcomings of the AIA clonal mutation process. Based on the combination of the two, a method for predicting coal and gas outburst risk based on HDAIA is established. HD is introduced into the mutation process of AIA, so that the mutation no longer depends on a large amount of computational binary encoding and decoding, and the antibodies generated in the mutation process can still show high affinity. Attached Figure Description
[0030] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.
[0031] Figure 1 This is a schematic diagram of the AIA process of the present invention;
[0032] Figure 2 This is a schematic diagram of the HDAIA model training process of the present invention;
[0033] Figure 3 This is a graph showing the predicted drill cuttings data from the 11192 longwall face from July to October 2018, according to the present invention.
[0034] Figure 4 This is a graph showing the predicted drill cuttings data from the 11192 longwall face from January to March 2019, according to the present invention.
[0035] Figure 5 This is a scatter plot of Hamming distance values for drill cuttings from the 11192 longwall face from July to October 2018, according to the present invention.
[0036] Figure 6 This is a scatter plot of Hamming distance values for drill cuttings from the 11192 longwall face from January to March 2019, according to the present invention. Detailed Implementation
[0037] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0038] Example 1
[0039] Please see Figure 1-6As shown, a coal and gas outburst hazard prediction method based on HDAIA is implemented using the Artificial Immune Algorithm (AIA) and Hamming distance (HD); it includes:
[0040] Step 1: Collect indicator data of local coal and gas outbursts in underground coal mines, including data on indicator drill cuttings S and drill cuttings gas desorption K1, etc. After processing, use the data as antigens, perform membership processing on the indicator data, and divide them into training set and prediction set; the specific steps are as follows:
[0041] Step 1.1: Collect data on two characteristic indicators of coal and gas outburst risk: the amount of drill cuttings (S) and the desorption of gas from drill cuttings (K1).
[0042] Step 1.2: The two characteristic indicators of coal and gas outburst risk, S (drill cuttings volume) and K1 (drill cuttings gas desorption), are processed to obtain membership values, all within the range [0, 1]. In the prediction of coal and gas outburst risk, v prediction objects are designed, and each prediction object has w prediction indicators. The original data can then form a v×w order prediction indicator data matrix:
[0043] (1)
[0044] In the formula, This is the w-th (w=1,2,3…) predictor index for the v-th (v=1,2,3…) predictor object.
[0045] Meanwhile, assuming there are z-level prediction standards for each of the w prediction indicators, meaning different critical values correspond to different risk levels, then for the v objects being predicted, we have a z×w prediction standard matrix:
[0046] (2)
[0047] When performing normalization, an interpolation function formula is introduced. Thus, in equation (1) Membership value:
[0048] (3)
[0049] In the formula, MAXm ij MINm represents the maximum value of a single predictive indicator in the data used. ij This represents the minimum value of a single predictive indicator in the data used. i represents the index of the predictive object, and j represents the index of the predictive indicator, i=1,2,...,v, j=1,2,...,w.
[0050] Use interpolation formula (3) to convert (1) into a consistent predictor membership matrix:
[0051] (4)
[0052] The prediction of coal and gas outburst risk in coal seams is uncertain. Before prediction, it is necessary to first consider whether the coal seam has the risk of coal and gas outburst. The prediction index of the predicted object with extremely high outburst risk is set to a membership degree of 1, while the prediction index of the predicted object with weak or basically no outburst risk is set to 0. Thus, the membership function of the predicted coal seam is obtained:
[0053] (5)
[0054] In the formula, n dj Let be the standard prediction value for the d-th level of the j-th predictive indicator, where d = 1, 2, ..., z. 1j Let n be the standard prediction value for the first level of the j-th predictive indicator; zj Let be the standard prediction value for the z-th level of the j-th prediction indicator. Let be the membership degree of the j-th predictive indicator at level d.
[0055] Using equation (5), the prediction standard matrix equation (2) is converted into a membership matrix. ,get:
[0056] (6)
[0057] Through the above function calculations, the original data is associated with the predicted object, transforming the acquired original data into a consistent function matrix with membership degrees to coal and gas outbursts, thus preparing for further prediction.
[0058] Step 1.3: Divide the membership-processed data into a training set and a prediction set.
[0059] Step 2: Use the immune algorithm and Hamming distance to establish a coal and gas outburst hazard prediction model based on HDAIA;
[0060] The specific steps are as follows:
[0061] Step 2.1: Specific steps of the immune algorithm:
[0062] Step 2.1.1: Coal and gas outburst identification antigen.
[0063] The relevant index data of coal and gas outburst risk are set as the objective function through coding, and training and prediction sets are randomly generated as antigens, waiting for input.
[0064] Step 2.1.2: Generate the initial antibody.
[0065] An antibody is randomly generated to initialize the objective function of coal and gas outburst hazard related indicators to ensure consistency. The algorithm is then trained using a training set to generate the initial antibody. The generated antibody is checked for abnormalities. The algorithm model is then debugged and improved. Once the algorithm model is mature and the error rate is reduced to an acceptable range, a prediction set is used for prediction.
[0066] Step 2.1.3 (see step 2.2.2 for details): Calculate the affinity value.
[0067] Before calculating affinity, the membership interval [0, 1] obtained in step 1.2 is divided into L bits for equal encoding (e.g., L=8). After converting the continuous membership values into binary strings, the standard Hamming distance is calculated using formula (13). This can significantly improve the speed of the algorithm in processing nonlinear data and maintain the logical consistency of Hamming distance in the immune algorithm.
[0068] (1) Assume the affinity between antigen a and antibody for:
[0069] (7)
[0070] (2) The weighted affinity between antigen a and antibody is: :
[0071] (8)
[0072] Step 2.1.4: Memory Cell Differentiation
[0073] During training using the training set, among the numerous antibodies generated, the antibody solutions with the highest affinity to the indicators corresponding to coal and gas outburst hazards are added to the memory cells. Simultaneously, the antibodies recognize the memory cells with the highest affinity and replace them. When using the prediction set for prediction, the memory cells generated after training are activated, producing antibodies with high affinity, thereby increasing the affinity of antibody solutions in the entire antibody set, thus saving solution time and reducing the overall number of low-affinity antibodies.
[0074] Step 2.1.5: Antibody Promotion and Inhibition
[0075] By calculating the expected value of antibody a, high-affinity antibodies are screened to obtain the desired antibody solution, while solutions below the set expected value are discarded. The expected value of antibody a is as follows:
[0076] (9)
[0077] density of antibody a for:
[0078] (10)
[0079] In the formula This refers to antibodies that have a high affinity for antibody k. This calculation can effectively improve the maturity and convergence of the algorithm. F represents the total number of antibodies.
[0080] Step 2.1.6: Generate new antibodies.
[0081] Among the generated antibodies, two antibodies with different affinities are randomly selected. These two antibodies are then mutated and crossovered according to a certain proportion and probability to obtain new antibodies. This process is repeated until all F antibodies are generated, yielding the desired solution. Mutation is achieved through point mutation. Within the algorithm's internal space, the basic mutation rate is set to R1, and the mutation probability is pm. During the mutation process, the affinity of antibody v1 is calculated, and random numbers are generated. ∈[0,1], if If ≤pm, then generate a random number a1 and a new antibody x;
[0082] Where, a1∈[-R1,R1], x=v1+a1;
[0083] Crossover is achieved through multi-point crossover, where multiple crossover points are randomly set, and the positions of the two parent antibodies before and after the crossover points are swapped.
[0084] Otherwise, keep the original antibody unchanged.
[0085] Step 2.1.7: Use Hamming distance to compare the distances between the antibodies.
[0086] The Hamming distance algorithm is used to calculate the distance between the generated F-number of antibodies and the antigen, and the optimal solution with the minimum distance is obtained.
[0087] Step 2.1.8: End.
[0088] Once an optimal solution that satisfies the termination condition is found, the result is output, and the algorithm terminates.
[0089] The specific steps of step 2.1.7 are as follows:
[0090] Step 2.2.1: Assume the prediction space for coal and gas outbursts is P. Within space P, assume the coordinates of the antigen t are:
[0091] (11)
[0092] The antibody b coordinate is:
[0093] (12)
[0094] Therefore, the distance between antigen t and antibody b is:
[0095] (13)
[0096] r = 1, 2, ..., L; L represents L-bit binary equally divided encoding.
[0097] Then, for antibodies b = 1, 2, ..., n, their Hamming distance D is:
[0098] (14)
[0099] The larger the Hamming distance D, the greater the difference in antigen-antibody coding, the lower the similarity and affinity between the two; the smaller the Hamming distance D, the more similar the two are, and the higher the affinity.
[0100] Step 2.2.2: After the antibody is generated, the affinity between the antigen and the antibody is randomly calculated to obtain the Hamming distance.
[0101] Based on the above principles, the similarity between the predicted indicator data and the set standards is calculated. For example, the predicted indicators are y1, y2, ..., y w There are 1, and the corresponding prediction criteria u also have the same number. Then, for the object being predicted, assuming the membership degree of the prediction criterion to the object is 1... The membership degree of the prediction standard is Therefore, the Hamming distance between the predicted object v and the predicted standard u is:
[0102] (15)
[0103] The Hamming distance is calculated by weighting different predictive indicators as follows:
[0104] (16)
[0105] In the formula The weight of the j-th prediction indicator is calculated using the following formula:
[0106] (17)
[0107] In the formula M represents the d-th (d=1,2,…,z)-th order prediction critical value for the j-th prediction index. 1×j is the membership weight correction coefficient for the j-th prediction index, used to adjust the weight mapping of different indices in the Hamming distance.
[0108] Therefore, the weighted Hamming distance between the membership matrix of the object to be predicted and the standard membership matrix is obtained as follows:
[0109] (18)
[0110] The Hamming distance required for prediction can be obtained from the above formula. The smaller the value, the smaller the difference between the coal and gas outburst risk and the prediction standard, and the more accurate and reliable the prediction value.
[0111] Step 3: Based on Step 1, input the training set into HDAIA to train the algorithm, stimulate the algorithm to generate antibodies, introduce Hamming distance to calculate antibodies with high affinity, and screen high affinity antibodies.
[0112] Step 4: Input the prediction set data into the algorithm, calculate the antibody with the highest affinity to the antigen, and input the converted antigen of the sample to be predicted into the algorithm. Calculate the minimum weighted Hamming distances DA and DB between the antigen and the set of antibodies with no prominent risk (Class A) and the set of antibodies with prominent risk (Class B), respectively. If DA > DB, the antibody is determined to have prominent risk; otherwise, it is determined to have no risk.
[0113] Step 5: If there is unpredicted data, repeat step 4 and output the coal and gas outburst risk prediction results.
[0114] Example 2
[0115] The specific steps of a coal and gas outburst hazard prediction method based on HDAIA are as follows:
[0116] Step 1: Collect index data of local coal and gas outbursts in coal mines, including the amount of drill cuttings S and the desorption of gas from drill cuttings K1, as antigens. Perform membership processing on the index data and divide them into training set and prediction set.
[0117] Based on the requirements of the "Detailed Rules for the Prevention and Control of Coal and Gas Outbursts" regarding the prediction method for local coal and gas outbursts, and using coal and gas outburst index data from a coal mine in Guizhou Province as the target of the prediction, it was found that the drilling gas volume S and drill cuttings gas desorption K1 indicators of this coal mine have long-term tracking and monitoring data and are more effective in predicting the risk of outbursts. Therefore, this paper uses the maximum values of the collected drill cuttings volume S and drill cuttings gas desorption index K1 to predict the risk of coal and gas outbursts in the tunneling face. A total of 910 sets of historical data from the 11192 longwall face of the No. 19 coal seam in Xiangshui Coal Mine, Guizhou Province, from January to October 2022 and from January to October 2025, were selected as the prediction sample. The selected data were randomly selected from the data measured by drilling at the 11192 longwall face, with each sequence number as a prediction object. The first 900 sets were used to train the model, and the last 10 sets were used for prediction. The collected raw data are shown in Tables 1-2, and the data of the prediction groups are shown in Tables 1-2. Figures 3-4 As shown.
[0118] Table 1. Historical Measured Data of Drill Cuttings from the 11192 Longwall Face, January to October 2022
[0119] Table 2. Historical Measured Data of Drill Cuttings from the 11192 Longwall Face, January to October 2025
[0120] Step 2: Using the immune algorithm and Hamming distance, establish a coal and gas outburst hazard prediction model based on HDAIA; The actual data of the two predicted indicators, namely, the amount of drill cuttings S and the drill cuttings gas desorption index K1, obtained from the table above during the tunneling process of the No. 11192 longwall face from January to October 2022, are used to construct a matrix:
[0121] Using the critical values of the indicators in the detailed rules, and taking the drill cuttings quantity S as 0 and the drill cuttings gas desorption index K1 value as 0 as the safest indicators for gas outburst risk, the prediction index matrix for coal and gas outburst risk can be obtained:
[0122] Suppose there are two indicators for the risk of coal and gas outbursts. The membership degree is 1 if the outburst threshold is exceeded, and 0 if the threshold is not exceeded. Therefore, the predictive indicator N can be... 2×2 The prediction standard matrix is converted into a membership index matrix:
[0123] The above formula predicts that M will be used as an indicator. 10×2 Convert to a membership matrix of the predicted objects:
[0124] Equation (23) yields the weight matrix of two indicators for the 10 prediction objects:
[0125] For K 10×2 The matrix is obtained by unification:
[0126] Step 3: Based on Step 1, input the training set into HDAIA to train the algorithm, stimulate the algorithm to generate antibodies, use Hamming distance to calculate antibodies with high affinity, and screen high affinity antibodies. Step 4: Input the prediction set data into the algorithm, calculate the antibody with the highest affinity to the antigen, and calculate the minimum weighted Hamming distance DA and DB between it and the set of antibodies with no prominent risk (Class A) and the set of antibodies with prominent risk (Class B), respectively. If DA > DB, it is determined to have prominent risk; otherwise, it is determined to have no risk. The Hamming distances for the above predicted objects were calculated, and the results are shown in Table 3. These results are based on ten sets of data measured from January to October 2022 at the No. 11192 longwall face of Xiangshui Coal Mine, and are used to predict the coal and gas outburst risk for the next one to two months. The Hamming distance calculation results are shown in the following table: Table 3 Calculated values of dynamic Hamming distance
[0127] Subsequently, the actual data of the two predicted indicators, namely, the amount of drill cuttings S and the drill cuttings gas desorption index K1, obtained from Table 2 during the tunneling process of the No. 11192 longwall face from January to October 2025, were used to construct a matrix:
[0128] M 10×2 Convert to a membership matrix of the predicted objects: Equation (14) yields the weight matrix of two indicators for the 10 prediction objects: The matrix obtained by unifying K10×2 is: The Hamming distances for the above predicted objects were calculated, and the results are shown in Table 4. These results are based on 10 sets of data collected from January to October 2025 at the No. 11192 longwall face of Xiangshui Coal Mine, used to predict the coal and gas outburst risk for the next one to two months. The calculated Hamming distances are as follows: Table 4 Calculated values of dynamic Hamming distance
[0129] Since a smaller Hamming distance indicates a higher precision and more accurate prediction, the Hamming distance is calculated from Table 3. Figure 5 The scatter plot shows that the Hamming distance values during the tunneling process of the No. 11192 longwall face from January to October 2022 were all less than 1. Furthermore, except for value 7, which was greater than Hamming distance B, all other Hamming distance A values were less than B values. Therefore, it is predicted that the No. 11192 longwall face at Xiangshui Coal Mine has no risk of coal or gas outbursts. Table 5 shows the drill cuttings volume S and drill cuttings gas analysis index K1 values obtained from the field measurements of the No. 11192 longwall face in November 2022 (observed every ten days starting from November 1st). This indicates that the prediction results are consistent with the field measurements, and there is no risk of coal or gas outbursts.
[0130] Table 5 Actual measured data of tunneling at longwall face No. 1192 on November 11, 2022
[0131] Meanwhile, from the Hamming distance calculation results in Table 4 and Figure 6 The scatter plot shows that from January to October 2025, the Hamming distance values for the No. 11192 longwall face were all less than 1, and the Hamming distance A value was greater than the B value. Therefore, it is predicted that there is a risk of coal and gas outburst in the period immediately following November 2025 for the No. 11192 longwall face of Xiangshui Coal Mine. Table 6 shows the drill cuttings volume S and drill cuttings gas analysis volume K1 values for November 2025 at the No. 11192 coal mine (observed every ten days starting from April 1st). As shown in Table 6, the K1 values for November 2025 obtained from the field are all greater than the outburst critical value of 0.5, indicating that there is a possibility of coal and gas outburst during the tunneling operation during this period. The prediction results are consistent with the actual measured data on the field regarding the risk of coal and gas outburst.
[0132] Table 6 Actual measured data of tunneling at longwall face No. 11192 on November 11, 2025
[0133] The consistency between the calculation results of the two sets of Hamming distance algorithms and the measured values of the coal seam outburst risk indicators indicates that the Hamming distance method can be used as a new, accurate, and effective method for predicting coal and gas outburst risks.
[0134] Step 5: If there is unpredicted data, repeat step 4 and output the coal and gas outburst risk prediction results.
[0135] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims
1. A method for predicting coal and gas outburst hazard based on HDAIA, characterized in that, include: Step 1: Collect indicator data of local coal and gas outbursts in underground coal mines. After processing the indicator data, use it as an antigen to perform membership processing and divide it into training set and prediction set. Step 2: Using the artificial immune algorithm and Hamming distance, establish a coal and gas outburst hazard prediction model based on HDAIA; Step 3: Input the training set into the coal and gas outburst hazard prediction model to train the algorithm, stimulate the algorithm to generate antibodies, calculate the antibody affinity through Hamming distance, and screen high affinity antibodies; Step 4: Input the prediction set into the algorithm, calculate the antibody with the highest affinity to the antigen, input the converted antigen of the sample to be predicted into the algorithm, and calculate the minimum weighted Hamming distance between it and the set of antibodies with no prominent risk and the set of antibodies with prominent risk, respectively, to obtain the distance D. A And distance 2D B If D A If the value is greater than DB, it is determined to have a prominent hazard; otherwise, it is determined to have no hazard. Step 5: If there is unpredicted data, repeat step 4 until there is no more predictable data; finally, output the coal and gas outburst risk prediction results.
2. The method for predicting coal and gas outburst risk based on HDAIA according to claim 1, characterized in that, The data includes the amount of drill cuttings and the value of drill cuttings gas desorption index.
3. The method for predicting coal and gas outburst risk based on HDAIA according to claim 2, characterized in that, The specific steps for performing membership processing on the aforementioned indicator data are as follows: The membership values of the drill cuttings quantity and drill cuttings gas desorption index are processed to obtain the membership degree values; In the prediction of coal and gas outburst risk, v prediction objects are set, and w prediction indicators are set for each prediction object. Then the original indicator data constitute a prediction indicator data matrix of order v×w. If w predictive indicators have z-level predictive standards, then construct a z×w predictive standard matrix. When performing normalization, the membership values of the predicted indicators corresponding to the predicted objects are obtained through interpolation function formulas, and the predicted indicator data matrix is converted into a consistent predicted indicator membership matrix. For prediction objects with extremely high outburst risk, the prediction index is set to a membership degree of 1, while for prediction objects with weak or virtually no outburst risk, the prediction index is set to 0. This yields the membership function of the coal seam being predicted, which is then used to convert the prediction standard matrix into a membership matrix.
4. The method for predicting coal and gas outburst risk based on HDAIA according to claim 1, characterized in that, The artificial immune algorithm includes: Coal and gas outburst identification antigen: The index data of coal and gas outburst hazard are set as the objective function through encoding, and training set and prediction set are randomly generated as antigens, waiting for input; Generate initial antibodies: Initialize the antibodies for the randomly generated objective function of coal and gas outburst risk indicators to make them consistent. Train the algorithm using the training set to generate initial antibodies and check whether the generated antibodies are abnormal. Debug and improve the algorithm model. After the algorithm model matures and the error rate is reduced to an acceptable range, use the prediction set for prediction. Calculate the affinity value; Memory cell differentiation: During training using the training set, among the large number of antibodies generated, the antibody solution with the highest affinity to the indicator corresponding to the coal and gas outburst hazard will be added to the memory cells. At the same time, the antibody will recognize the memory cell with the highest affinity and replace it. When using the prediction set for prediction, the memory cells generated after training are activated and produce antibodies with high affinity. Antibody promotion and inhibition: By calculating the expected value of the antibody, high-affinity antibodies are screened to obtain the desired antibody solution, while solutions that are lower than the set expected value are discarded. Generate new antibodies: Among the generated antibodies, randomly select two antibodies with different affinities, and mutate and cross over these two antibodies according to a certain proportion and probability to obtain new antibodies. Repeat the operation until all F antibodies are generated to obtain the desired solution. Hamming distance is used to compare the distances between antibodies: the Hamming distance algorithm is used to calculate the distance between the F generated antibodies and the antigen, and the optimal solution with the smallest distance is obtained; Termination: When an optimal solution that satisfies the termination condition is found, the result is output, and the algorithm terminates.
5. The method for predicting coal and gas outburst risk based on HDAIA according to claim 4, characterized in that, Before calculating the affinity value, the following steps are also included: The intervals corresponding to the membership values are divided into L-bit binary equal parts for encoding. After converting the continuous membership values into binary strings, the standard Hamming distance is calculated.
6. The method for predicting coal and gas outburst risk based on HDAIA according to claim 4, characterized in that, The Hamming distance algorithm is used to calculate the distance between the generated F antibodies and the antigen, specifically including: Assuming the prediction space for coal and gas outbursts is P, within the prediction space P, the coordinates of the antigen and antibody are set, the distance between the antigen and antibody is calculated, and then the Hamming distance is obtained. After the antibody is generated, the affinity between the antigen and the antibody is randomly calculated to obtain the Hamming distance required for prediction.
