Coal and gas outburst disaster prediction method and device based on multi-physics field coupling

Abstract

The invention relates to the technical field of disaster prediction, and discloses a coal and gas outburst disaster prediction method and device based on multi-physics field coupling. The method comprises the steps of performing data enhancement and normalization processing on a training sample set to obtain a model training data set, and performing pattern recognition and parameter optimization on the model training data set to obtain a prominent disaster prediction model; carrying out probability distribution calculation on output data of the outburst disaster prediction model to obtain a confidence interval, and carrying out multi-dimensional cross validation on the confidence interval to obtain a prediction reliability evaluation result; and carrying out threshold grading and risk grade division on the prediction reliability evaluation result to obtain an early warning grade judgment result, and carrying out space-time correlation analysis on the early warning grade judgment result to obtain target early warning decision information. According to the method, the coupling relationship among the multi-physics field parameters is analyzed, the multi-physics field coupling prediction model is established, and accurate prediction of coal and gas outburst disasters is realized, so that the reliability and accuracy of prediction are improved.

Description

Technical Field

[0001] This application relates to the field of disaster prediction, and particularly to a method and device for predicting coal and gas outburst disasters based on multi-physical field coupling. Background Art

[0002] Coal and gas outburst is one of the most serious dynamic disasters in coal mine production, posing a major threat to coal mine safety production. Currently, the prediction of coal and gas outburst mainly analyzes based on single-physical field monitoring data, such as stress monitoring, gas pressure monitoring, temperature monitoring, etc. With the development of monitoring technology, a variety of sensors are applied to coal mine safety monitoring, including stress sensors, gas concentration sensors, temperature sensors, and acoustic emission sensors, etc. These sensors can collect multi-physical field parameter data in real time. However, the existing prediction methods mainly focus on the change characteristics of single-physical field parameters, and fail to fully consider the coupling relationship between multi-physical fields and its comprehensive impact on the evolution process of coal and gas outburst disasters.

[0003] The existing technologies have the following deficiencies: First, single-physical field monitoring is difficult to comprehensively reflect the gestation process of coal and gas outburst, resulting in low reliability of prediction results; Second, although the multi-physical field monitoring data contains rich disaster omen information, due to the lack of effective multi-physical field coupling analysis methods, it is difficult to extract key feature information from the massive monitoring data; Third, the existing prediction methods generally have problems such as insufficient data preprocessing, incomplete feature extraction, and insufficient generalization ability of prediction models, affecting the accuracy and practicality of prediction. Summary of the Invention

[0004] This application provides a method and device for predicting coal and gas outburst disasters based on multi-physical field coupling, which is used to establish a multi-physical field coupling prediction model by analyzing the coupling relationship between multi-physical field parameters, so as to achieve accurate prediction of coal and gas outburst disasters, thereby improving the reliability and accuracy of prediction.

[0005] In a first aspect, this application provides a method for predicting coal and gas outburst disasters based on multi-physical field coupling. The method for predicting coal and gas outburst disasters based on multi-physical field coupling includes: collecting and performing data standardization processing on the original data of the stress field, gas field, temperature field, and acoustic emission field at the coal mine site to obtain an initial multi-physical field dataset, and performing outlier detection and data filling processing on the initial multi-physical field dataset to obtain a preprocessed multi-physical field parameter dataset; Performing physical field component extraction and feature decomposition on the preprocessed multi-physical field parameter dataset to obtain the basic parameter matrix of each physical field, and performing cross-correlation analysis and coupling factor calculation on the basic parameter matrix to obtain a physical field coupling relationship matrix; Extract the time-domain features and frequency-domain features of the physical field coupling relationship matrix to obtain a multi-dimensional feature vector, and perform redundant feature elimination and feature reconstruction on the multi-dimensional feature vector to obtain an optimized feature parameter set; Perform time-series segmentation and sample annotation on the optimized feature parameter set to obtain a training sample set, perform data augmentation and normalization processing on the training sample set to obtain a model training data set, and perform pattern recognition and parameter optimization on the model training data set to obtain a prominent disaster prediction model; Calculate the probability distribution of the output data of the prominent disaster prediction model to obtain the confidence interval of the prediction result, and perform multi-dimensional cross-validation on the confidence interval to obtain the prediction reliability evaluation result; Perform threshold grading and risk level division on the prediction reliability evaluation result to obtain the early warning level determination result, and perform spatio-temporal correlation analysis on the early warning level determination result to obtain the target early warning decision-making information.

[0006] In a second aspect, the present application provides a device for predicting coal and gas outburst disasters based on multi-physical field coupling. The device for predicting coal and gas outburst disasters based on multi-physical field coupling includes: An acquisition module, configured to acquire and perform data standardization processing on the original data of the stress field, gas field, temperature field, and acoustic emission field at the coal mine site to obtain an initial multi-physical field data set, and perform outlier detection and data filling processing on the initial multi-physical field data set to obtain a preprocessed multi-physical field parameter data set; A decomposition module, configured to perform physical field component extraction and feature decomposition on the preprocessed multi-physical field parameter data set to obtain a basic parameter matrix of each physical field, and perform cross-correlation analysis and coupling factor calculation on the basic parameter matrix to obtain a physical field coupling relationship matrix; An extraction module, configured to extract time-domain features and frequency-domain features of the physical field coupling relationship matrix to obtain a multi-dimensional feature vector, and perform redundant feature elimination and feature reconstruction on the multi-dimensional feature vector to obtain an optimized feature parameter set; A processing module, configured to perform time-series segmentation and sample annotation on the optimized feature parameter set to obtain a training sample set, perform data augmentation and normalization processing on the training sample set to obtain a model training data set, and perform pattern recognition and parameter optimization on the model training data set to obtain a prominent disaster prediction model; A calculation module, configured to calculate the probability distribution of the output data of the prominent disaster prediction model to obtain the confidence interval of the prediction result, and perform multi-dimensional cross-validation on the confidence interval to obtain the prediction reliability evaluation result; A partitioning module is used to perform threshold grading and risk level partitioning on the predicted reliability evaluation results, obtain an early warning level determination result, and perform spatio-temporal correlation analysis on the early warning level determination result to obtain target early warning decision-making information.

[0007] In the technical solution provided by this application, through multi-physical field data collection and preprocessing of the stress field, gas field, temperature field, and acoustic emission field at the coal mine site, the noise and outliers in the original data are effectively eliminated, and the data quality is improved. Through physical field component extraction and eigen decomposition, the basic characteristics of each physical field are deeply analyzed, and through cross-correlation analysis and coupling factor calculation, the coupling relationship between multi-physical fields is comprehensively revealed, providing more comprehensive characteristic information for outburst disaster prediction. By extracting time-domain and frequency-domain characteristics of the physical field coupling relationship matrix and combining redundant feature elimination and feature reconstruction techniques, an optimized feature parameter set is obtained, enhancing the representativeness and effectiveness of the features. Through time series segmentation and sample annotation, a standardized training sample set is established, and data augmentation and normalization processing methods are used to enhance the representativeness of the samples and the generalization ability of the model. Based on the outburst disaster prediction model of the deep neural network, accurate prediction of outburst risks is achieved. Through probability distribution calculation and multi-dimensional cross-validation, the reliability of the prediction results is evaluated, ensuring the credibility of the prediction results. Finally, through threshold grading and risk level partitioning, combined with spatio-temporal correlation analysis, scientific early warning decision-making information is formed, providing important decision-making support for coal mine safety production. This method overcomes the defect of incomplete information in traditional single-physical field monitoring methods, improves the accuracy and reliability of prediction, and realizes accurate early warning of coal and gas outburst disasters. Brief Description of the Drawings

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0009] Figure 1 It is a schematic diagram of an embodiment of the coal and gas outburst disaster prediction method based on multi-physical field coupling in the embodiments of this application; Figure 2 It is a schematic diagram of an embodiment of the coal and gas outburst disaster prediction device based on multi-physical field coupling in the embodiments of this application. Detailed Embodiments

[0010] The embodiments of the present application provide a method and device for predicting coal and gas outburst disasters based on multi-physical field coupling. The terms "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described here can be implemented in an order other than those illustrated or described here. In addition, the term "including" or "having" and any deformation thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0011] For ease of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 , an embodiment of the method for predicting coal and gas outburst disasters based on multi-physical field coupling in the embodiments of the present application includes: Step S101: Collect and standardize the original data of the stress field, gas field, temperature field and acoustic emission field at the coal mine site to obtain an initial multi-physical field dataset, and perform outlier detection and data filling on the initial multi-physical field dataset to obtain a preprocessed multi-physical field parameter dataset; Step S102: Extract physical field components and perform feature decomposition on the preprocessed multi-physical field parameter dataset to obtain the basic parameter matrices of each physical field, and perform cross-correlation analysis and coupling factor calculation on the basic parameter matrices to obtain a physical field coupling relationship matrix; Step S103: Extract time-domain features and frequency-domain features from the physical field coupling relationship matrix to obtain a multi-dimensional feature vector, and perform redundant feature elimination and feature reconstruction on the multi-dimensional feature vector to obtain an optimized feature parameter set; Step S104: Perform time series segmentation and sample annotation on the optimized feature parameter set to obtain a training sample set, perform data augmentation and normalization on the training sample set to obtain a model training dataset, and perform pattern recognition and parameter optimization on the model training dataset to obtain an outburst disaster prediction model; Step S105: Calculate the probability distribution of the output data of the outburst disaster prediction model to obtain the confidence interval of the prediction result, and perform multi-dimensional cross-validation on the confidence interval to obtain a prediction reliability evaluation result; Step S106: Perform threshold classification and risk level division on the prediction reliability evaluation result to obtain an early warning level determination result, and perform spatio-temporal correlation analysis on the early warning level determination result to obtain target early warning decision-making information.

[0012] It can be understood that the execution entity of this application can be a coal and gas outburst disaster prediction device based on multi-physical field coupling, or it can also be a terminal or a server, and specific details are not limited here. In this embodiment of the application, the server is used as the execution entity for illustration.

[0013] Specifically, multi-physical field data is collected through a sensor network deployed on the coal mine working face. Specifically, it includes: stress sensors collect data on the stress changes of the coal body, gas sensors collect data on gas concentration and pressure, temperature sensors collect data on the temperature changes of the coal body, and acoustic emission sensors collect data on the acoustic emission signals of coal body fractures. The collected raw data is standardized to eliminate the dimensional differences of different physical quantities. Abnormal data is processed through peak normalization and median filling, and the sliding window technique is used for data segmentation, with the window size set to 1 hour. The interquartile range method is used to detect outliers, and the data points outside the normal range are marked. The marked abnormal data is supplemented using linear interpolation, and then the noise components in the data are removed through wavelet transform. All physical field data is aligned to a unified time series, and equidistant resampling is used to ensure the time consistency of the data.

[0014] Physical field component extraction is performed on the preprocessed data to obtain the feature matrices of the stress field, gas field, temperature field, and acoustic emission field respectively. The main feature components of each physical field are extracted through the principal component analysis method, the eigenvalues and eigenvectors are calculated, and the principal components with a cumulative contribution rate reaching 85% are selected as the key features. Cross-correlation analysis is performed on the extracted features, the coupling coefficients between physical fields are calculated, and a coupling relationship matrix is constructed. Time-domain and frequency-domain feature extraction is performed on the coupling relationship matrix. Time-domain features are obtained through wavelet transform, and frequency-domain features are obtained through Fourier transform. The extracted features are dimensionally concatenated to form a multi-dimensional feature vector, and redundancy features are eliminated using correlation analysis and hierarchical clustering methods to obtain an optimized feature parameter set.

[0015] Training samples are constructed based on the optimized feature parameter set, and time series segmentation is performed through a sliding window. The window size is set to 24 hours, and the step size is set to 1 hour. The segmented samples are labeled into two categories: "outburst danger" and "outburst safety". Data augmentation is performed using data rotation and noise injection methods to improve the representativeness of the samples. A prediction model is constructed using a deep neural network, and the model performance is improved through parameter optimization. The reliability of the prediction results is evaluated. The confidence interval is calculated through kernel density estimation, and the stability of the model is evaluated using K-fold cross-validation (K = 5). The early warning level threshold is determined according to the evaluation results, and the early warning level is divided into three levels: "high risk", "medium risk", and "low risk". The spatial distribution characteristics and evolution trends of the dangerous areas are determined through spatio-temporal correlation analysis.

[0016] For example, the gas pressure data collected at a coal mine working face shows that the normal value range is between 0.74 - 2.35 MPa, and an outlier of 3.86 MPa is detected through outlier detection. After data filling and processing, this outlier is corrected to 2.28 MPa. At the same time, the stress monitoring data shows that the vertical stress increases from 18.5 MPa to 26.7 MPa, the temperature increases from 24.3 °C to 28.9 °C, and the acoustic emission event frequency increases from 12 times per hour to 45 times per hour. Through multi - physical - field coupling analysis, it is found that the abnormal changes of these parameters are significantly correlated, the coupling coefficient reaches 0.87, and the prediction model gives a "high - risk" warning, indicating that there is a greater outburst risk in this area.

[0017] In the embodiment of the present application, through multi - physical - field data collection and pre - processing of the stress field, gas field, temperature field, and acoustic emission field at the coal mine site, the noise and outliers in the original data are effectively eliminated, and the data quality is improved. Through physical - field component extraction and eigen - decomposition, the basic characteristics of each physical field are deeply analyzed, and through cross - correlation analysis and coupling factor calculation, the coupling relationship between multi - physical fields is comprehensively revealed, providing more comprehensive characteristic information for outburst disaster prediction. By extracting time - domain and frequency - domain characteristics of the physical - field coupling relationship matrix, combined with redundant - feature elimination and feature reconstruction technologies, an optimized set of characteristic parameters is obtained, enhancing the representativeness and effectiveness of the characteristics. Through time - series segmentation and sample annotation, a standardized training sample set is established, and data augmentation and normalization processing methods are used to enhance the representativeness of the samples and the generalization ability of the model. The outburst disaster prediction model based on a deep neural network realizes the accurate prediction of outburst risk. Through probability distribution calculation and multi - dimensional cross - validation, the reliability of the prediction results is evaluated, ensuring the credibility of the prediction results. Finally, through threshold grading and risk - level division, combined with spatio - temporal correlation analysis, scientific warning decision - making information is formed, providing important decision - making support for coal mine safety production. This method overcomes the defect of incomplete information in traditional single - physical - field monitoring methods, improves the accuracy and reliability of prediction, and realizes the accurate warning of coal - and - gas outburst disasters.

[0018] In a specific embodiment, the process of executing step S101 may specifically include the following steps: (1) Collect and perform data standardization processing on the original data of the stress field, gas field, temperature field, and acoustic emission field at the coal mine site to obtain an initial multi - physical - field data set (2) Perform peak normalization processing on the initial multi - physical - field data set to obtain a normalized data set, and perform median filling on the normalized data set to obtain the supplemented initial data; (3) Perform sliding window segmentation on the supplemented initial data to obtain a data segment sequence, and perform interquartile range method anomaly detection on the data segment sequence to obtain a marked data sequence; (4) Perform linear interpolation on the outliers in the marked data sequence to obtain an imputed data sequence, and perform wavelet denoising on the imputed data sequence to obtain a denoised data sequence; (5) Perform time window synchronization on the denoised data sequence to obtain an aligned data sequence, and perform data resampling on the aligned data sequence to obtain an equally spaced data sequence; (6) Perform Z-score standardization on the equally spaced data sequence to obtain a standardized data sequence, and perform batch verification on the standardized data sequence to obtain a multi-physical field initial data set; (7) Perform 3σ criterion anomaly detection on the multi-physical field initial data set to obtain an anomaly point marking sequence, and perform spline interpolation on the anomaly point marking sequence to obtain a data filling sequence; (8) Perform moving average filtering on the data filling sequence to obtain a smoothed data sequence, and perform data quality assessment on the smoothed data sequence to obtain a data quality index; (9) Perform threshold screening on the data quality index to obtain a qualified data sequence, and perform outlier detection on the qualified data sequence to obtain a detection result sequence; (10) Perform data verification on the detection result sequence to obtain a verified data sequence, and perform data format unification on the verified data sequence to obtain a preprocessed multi-physical field parameter data set.

[0019] Specifically, at the coal mine site, multi-physical field data is collected through a variety of sensors deployed. Among them, stress sensors collect coal body stress data, including vertical stress, horizontal stress, etc.; gas sensors collect gas concentration, pressure and other data; temperature sensors collect coal body temperature change data; acoustic emission sensors collect acoustic emission signal data generated when the coal body fractures. The collection frequency is set to 1 time per second, and continuous collection for 24 hours forms the original data set.

[0020] First, perform peak normalization on the original data. For missing values in the data, use the median filling method to supplement. Use the sliding window technique to segment the supplemented data. The window size is set to 1 hour, and the sliding step is 10 minutes to obtain a data segment sequence.

[0021] Perform interquartile range method anomaly detection on each data segment. The calculation formula is: ; ; ; where is the lower quartile, is the upper quartile, is the interquartile range, and are the upper and lower threshold values respectively. Data points outside the threshold range are marked as outliers.

[0022] The marked outliers are corrected using linear interpolation, and then wavelet decomposition is used to denoise the data. All physical field data are aligned to a unified time series through time window synchronization processing, and equidistant resampling is used to ensure the time consistency of the data. The resampled data are normalized by Z-score to eliminate the dimensional differences between different physical quantities, and then outlier detection is performed using the 3σ criterion. Spline interpolation is performed on the data points outside the normal range. Finally, moving average filtering is performed on the processed data with a window size of 5 data points to obtain a smooth data sequence. Quality assessment is carried out by calculating indicators such as data integrity, consistency, and accuracy, and threshold values are set to screen out qualified data. Final data format unification processing is performed on the screened data to form a preprocessed multi-physical field parameter dataset.

[0023] For example, in the gas pressure data collected from a coal mine working face, the original data range is 0.5 - 3.8 MPa. After peak normalization processing, the data is mapped to between 0 and 1. An outlier 3.8 MPa (normalized value 1.0) is detected by the interquartile range method, where = 0.8 MPa (normalized value 0.21), = 2.1 MPa (normalized value 0.55), = 1.3 MPa, and is calculated to be 4.05 MPa, = -1.15 MPa. Since 3.8 MPa is less than the upper threshold value of 4.05 MPa, this data point is retained. After subsequent data processing steps, a standardized and smooth data sequence is obtained.

[0024] In a specific embodiment, the process of executing step S102 may specifically include the following steps: (1) Perform physical field type division processing on the preprocessed multi-physical field parameter dataset to obtain a stress field data subset, a gas field data subset, a temperature field data subset, and an acoustic emission field data subset, and perform time series sorting processing on each data subset to obtain an ordered data sequence; (2) Perform principal component decomposition processing on the ordered data sequence to obtain main characteristic components, and calculate the variance contribution rate of the main characteristic components to obtain a characteristic weight sequence; (3) Perform threshold screening on the feature weight sequence to obtain the key feature sequence, and perform unit unification processing on the key feature sequence to obtain the standardized feature sequence; (4) Perform matrix reconstruction on the standardized feature sequence to obtain the basic parameter matrix of each physical field, and perform data integrity verification on the basic parameter matrix to obtain the verified parameter matrix; (5) Calculate the Pearson correlation coefficient for the verified parameter matrix to obtain the correlation coefficient matrix, and perform significance test on the correlation coefficient matrix to obtain the effective correlation terms; (6) Calculate the coupling degree for the effective correlation terms to obtain the coupling factor sequence, and perform normalization processing on the coupling factor sequence to obtain the standardized coupling factor; (7) Perform matrix mapping on the standardized coupling factor to obtain the initial coupling relationship matrix, and perform symmetry test on the initial coupling relationship matrix to obtain the physical field coupling relationship matrix.

[0025] Specifically, first perform physical field type division on the preprocessed multi-physical field parameter dataset, and divide the data into four subsets: the stress field data subset (including data such as vertical stress and horizontal stress), the gas field data subset (including data such as gas concentration and pressure), the temperature field data subset (including temperature change data), and the acoustic emission field data subset (including acoustic emission signal data). Sort each data subset according to the time stamp to generate ordered time series data. After obtaining the ordered data sequence, use the principal component analysis method for feature decomposition. By calculating the eigenvalues and eigenvectors of the data covariance matrix, the main feature components are obtained. Calculate the variance contribution rate for each feature component, that is, the ratio of the eigenvalue corresponding to this feature component to the sum of all eigenvalues, so as to obtain the feature weight sequence. Set the variance contribution rate threshold to 85%, screen out the key features whose cumulative contribution rate reaches the threshold to form the key feature sequence. Perform unit unification processing on the physical quantities in the key feature sequence to eliminate the dimensional differences between different physical quantities.

[0026] Perform matrix reconstruction on the standardized feature sequence, and organize the key features of each physical field into a basic parameter matrix according to the time series. Ensure the quality of the matrix data by checking the integrity, continuity, and consistency of the data, and obtain the verified parameter matrix. Calculate the Pearson correlation coefficient for the verified parameter matrix to analyze the correlation between different physical field parameters. Through significance testing (confidence level set to 95%), screen out the parameter pairs with significant correlation to form valid correlation terms. Calculate the coupling degree for the valid correlation terms to characterize the interaction strength between different physical field parameters. Normalize the calculated coupling factors to unify their value ranges to between 0 and 1. Construct an initial coupling relationship matrix for the standardized coupling factors and ensure the symmetry of the matrix through symmetry testing to obtain the physical field coupling relationship matrix.

[0027] It should be noted that when constructing the physical field coupling relationship matrix, for each pair of physical field parameters (such as the stress field and the gas field, the temperature field and the acoustic emission field, etc.), not only the overall correlation coefficient and coupling factor are calculated, but also the dynamic change process in the time series needs to be recorded. First, perform the extended construction of the coupling relationship matrix. For each pair of physical field parameters, calculate their dynamic coupling coefficients within a continuous time window, and measure the interaction strength between the two physical fields in this time window through covariance and standard deviation. The values of each physical field parameter represent the measured values of the physical field at different time points. Then construct a three-dimensional coupling relationship matrix, which includes the physical field type dimension and the time series dimension. Each element in the matrix represents the coupling coefficient of the corresponding physical field pair at a specific time point. In this way, the coupling relationship matrix contains information in the time dimension. For time-domain feature extraction, perform wavelet transform on the three-dimensional coupling relationship matrix in the time dimension.

[0028] For example, in the processing of monitoring data from a coal mine working face, after dividing the collected data into physical fields, the stress field data includes vertical stress (15 - 25 MPa) and horizontal stress (12 - 20 MPa), the gas field data includes gas pressure (0.74 - 2.35 MPa) and concentration (0.5% - 1.8%), the temperature field data ranges from 24 to 29 °C, and the acoustic emission field data includes the acoustic emission event frequency (10 - 50 times per hour). Through principal component analysis, it is found that the variance contribution rate of the first principal component (vertical stress change) in the stress field is 76%, and the variance contribution rate of the second principal component (horizontal stress change) is 21%, and the cumulative contribution rate of the two reaches 97%. Therefore, these two principal components are retained as key features. Through correlation analysis, it is found that the correlation coefficient between the vertical stress and the gas pressure is 0.82, indicating a significant correlation. Through coupling degree calculation, the coupling factor between the stress field and the gas field is obtained as 0.78, indicating a strong coupling relationship between the two physical fields.

[0029] In a specific embodiment, the process of performing step S103 may specifically include the following steps: (1) Perform wavelet transform processing on the physical field coupling relationship matrix to obtain a time-domain coefficient sequence, and calculate statistical features of the time-domain coefficient sequence to obtain a time-domain feature set; (2) Perform Fourier transform processing on the time-domain feature set to obtain a frequency-domain feature sequence, and calculate the band energy of the frequency-domain feature sequence to obtain a frequency band energy distribution; (3) Perform feature point extraction processing on the frequency band energy distribution to obtain a feature point sequence, and perform dimension splicing processing on the feature point sequence to obtain a multi-dimensional feature vector; (4) Perform correlation analysis processing on the multi-dimensional feature vector to obtain a feature correlation matrix, and perform hierarchical clustering processing on the feature correlation matrix to obtain a feature cluster sequence; (5) Perform information gain calculation on the feature cluster sequence to obtain a feature importance sequence, and perform threshold screening processing on the feature importance sequence to obtain a candidate feature set; (6) Perform orthogonal transform processing on the candidate feature set to obtain a feature basis vector, and perform linear combination processing on the feature basis vector to obtain an optimized feature parameter set.

[0030] Specifically, wavelet transform can provide local features of a signal at different time scales. By setting appropriate scale parameters and translation parameters, the time series is decomposed using a wavelet basis function. Then, statistical features of the wavelet coefficients are calculated, including mean, standard deviation, energy, and entropy values. The mean reflects the overall level of coupling strength, the standard deviation represents the degree of fluctuation of the coupling strength, the energy feature reflects the signal strength, and the entropy value represents the complexity of the signal. For frequency-domain feature extraction, Fourier transform is performed on the coefficients after wavelet transform to convert the time-domain signal to the frequency domain. In the frequency domain, the frequency range is divided into several frequency bands, and the energy distribution within each frequency band is calculated. By analyzing the distribution characteristics of energy in different frequency bands, the periodic change law of the coupling relationship can be obtained. Finally, feature fusion is performed to combine the extracted time-domain and frequency-domain features to form a feature vector. When optimizing the features, first calculate the correlation between features to remove redundant features. Then, use the principal component analysis method for dimensionality reduction, and select the features with a cumulative contribution rate reaching a predetermined threshold to form the final optimized feature set. By analyzing the time-varying characteristics of the coupling relationship matrix, not only the coupling relationship information between physical fields is retained, but also the dynamic characteristics of this coupling relationship changing with time are captured. For each pair of physical fields, the finally obtained feature vector not only contains the coupling strength information between physical fields, but also contains the dynamic change characteristics of the coupling relationship.

[0031] Perform wavelet transform processing on the physical field coupling relationship matrix and extract time-domain features using the multi-resolution analysis method. Wavelet transform is a time-frequency localization analysis method that can provide local feature information of signals in both the time domain and the frequency domain simultaneously. The calculation formula of wavelet transform is: ; where, is the wavelet coefficient, is the scale parameter (controlling the stretching and shrinking of the wavelet), b is the translation parameter (controlling the translation of the wavelet), is the input signal, is the wavelet basis function. Calculate the statistical features of the obtained wavelet coefficients: ; where, is the k-th statistical moment, n is the number of samples, is the wavelet coefficient. When k = 1, it is the mean value; when k = 2, it is the variance; when k = 3, it is the skewness; when k = 4, it is the kurtosis.

[0032] Perform Fourier transform on the time-domain feature set to convert the time-domain signal into a frequency-domain representation: ; where, is the frequency-domain feature sequence, is the time-domain signal, is the angular frequency, is the imaginary unit. Calculate the energy distribution of different frequency bands: ; where, is the energy of the i-th frequency band, and are the upper and lower limits of the frequency band.

[0033] It should be noted that in the embodiments of the present application, the data after wavelet transform is the result of decomposing the original signal at different scales. For each scale level, a coefficient sequence that varies with time can be obtained. These coefficient sequences are still essentially time-domain signals, except that they respectively represent the performance of the original signal at different scales (which can be understood as different frequency bands). When performing Fourier transform on these coefficient sequences, it is actually analyzing the frequency characteristics of the signal at each scale. Because any time-domain signal can be Fourier-transformed to obtain its frequency characteristics. In this way, a more detailed analysis result can be obtained: First, wavelet transform helps to decompose the signal into different scale spaces, which is equivalent to performing frequency band division; then, performing Fourier transform on the signal within each frequency band can obtain more accurate frequency structure information within that frequency band. Such an analysis method provides a multi-level signal feature extraction approach, which not only retains the ability of wavelet transform to depict local features, but also obtains the periodic features in the frequency domain through Fourier transform.

[0034] Feature point extraction is performed on the energy distribution of frequency bands, including the energy peak, mean, and variance of each frequency band. All features are concatenated into a vector form by dimension: ; Among them, is a multi-dimensional feature vector, is the i-th feature component. Calculate the correlation coefficient matrix between features: ; Among them, is the correlation coefficient between features i and j , is the covariance, is the variance.

[0035] Among them, performing feature extraction and analysis again is based on the idea of multi-level feature mining: The feature extraction in the first stage is aimed at the original data of the physical field, and a coupling relationship matrix is obtained by analyzing the interaction between different physical fields. This matrix reflects the intensity of the mutual influence between physical fields; the second stage is to perform in-depth time-frequency feature analysis on this coupling relationship matrix, obtain time-domain features through wavelet transform, and then obtain frequency-domain features through Fourier transform, and extract key feature points from them. These feature points reflect the variation law of the coupling relationship with time and frequency. Such a progressive analysis method enables the coupling mechanism between physical fields to be understood from different dimensions: First, understand the physical fields with coupling, then further analyze how the coupling relationship evolves over time and is manifested at different frequencies, and finally, through the correlation analysis and clustering of these time-frequency features, screen out the most representative feature combinations to form an optimized feature parameter set.

[0036] Group the features by hierarchical clustering method and calculate the distance between features: ; where is the Euclidean distance between features p and q. Calculate the information gain for the feature cluster sequence and perform feature selection. Finally, obtain the eigenbasis vectors through orthogonal transformation and perform linear combination to obtain the optimized feature parameter set.

[0037] For example, in the monitoring of coal mine working faces, analyze the collected gas pressure data (sampling frequency 100Hz). First, obtain the wavelet coefficients at different scales through wavelet transform. The mean value is 0.85MPa, the variance is 0.12, the skewness is 0.34, and the kurtosis is 2.87 calculated at scale 1. After Fourier transform, divide the frequency range into low-frequency band (0 - 10Hz), medium-frequency band (10 - 50Hz), and high-frequency band (50 - 100Hz). The energy ratios of each frequency band are calculated as 65%, 25%, and 10% respectively. Extract the energy peaks (0.82, 0.45, 0.21), mean values (0.56, 0.32, 0.15), and variances (0.08, 0.05, 0.03) for each frequency band to form an 18-dimensional feature vector. Through correlation analysis and hierarchical clustering, merge the features with a correlation coefficient greater than 0.85 to obtain 8 key feature parameters with strong independence, which constitute the optimized feature parameter set.

[0038] In a specific embodiment, the process of executing step S104 may specifically include the following steps: (1) Perform a sliding window process on the optimized feature parameter set to obtain a sequence of time series segments, and calculate the overlap degree of the time series segment sequence to obtain a set of segmentation points; (2) Perform boundary optimization on the set of segmentation points to obtain a time series segmentation sequence, and perform feature statistics on the time series segmentation sequence to obtain a sample feature matrix; (3) Perform label assignment on the sample feature matrix to obtain initial training samples, and perform balanced sampling on the initial training samples to obtain a training sample set; (4) Perform data rotation on the training sample set to obtain an augmented sample sequence, and perform noise injection on the augmented sample sequence to obtain an enhanced data set; (5) Perform maximum-minimum normalization on the enhanced data set to obtain a standardized data sequence, and perform data verification on the standardized data sequence to obtain a model training data set; (6) Perform parameter space partitioning on the model training data set to obtain a parameter grid sequence, and perform deep neural network processing on the parameter grid sequence to obtain a prominent disaster prediction model.

[0039] Specifically, the processing of the optimized feature parameter set first uses the sliding window technique for time series segmentation. The window size is set to 24 hours, and the sliding step is 1 hour. The segmentation points are determined by calculating the data overlap degree between adjacent windows. The overlap degree calculation is based on the number of data points jointly covered by adjacent windows divided by the window size. The overlap degree threshold is set to 50%. When the overlap degree is lower than the threshold, it is determined as a segmentation point. The boundary of the obtained segmentation point set is optimized by detecting the mutation degree of the data before and after the segmentation point to adjust the boundary position. Statistical features are calculated for each time series segment, including statistics such as mean, standard deviation, skewness, kurtosis, etc., as well as the change rate and trend features of physical field parameters. These features are organized into a sample feature matrix. Each row of the matrix represents the feature vector of a time window, and each column represents a type of feature.

[0040] Based on historical outburst event records and expert annotations, label assignment is performed on the sample feature matrix. The labels are divided into two categories: "outburst danger" and "outburst safety". Among them, the samples within 24 hours before the occurrence of the outburst event are marked as "outburst danger", and other samples are marked as "outburst safety". Since the number of "outburst danger" samples in the actual data is small, the SMOTE (Synthetic Minority Over-sampling Technique) method is used for balanced sampling. New samples are generated by interpolating between minority class samples to balance the number of the two types of samples.

[0041] Data augmentation processing is performed on the balanced training sample set. First, through the data rotation method, the feature vectors are rotated by a small angle to generate new samples. The rotation angle range is set to -15° to 15°, and the step is 5°. Then, Gaussian noise with a mean of 0 and a standard deviation of 0.01 is added to the rotated samples to enhance the anti-noise ability of the model. The enhanced data set is normalized by the maximum and minimum values, and all feature values are mapped to the interval [0, 1] to eliminate the dimensional difference between different features. Data verification is carried out to ensure that the normalized data distribution is reasonable without outliers and missing values.

[0042] Finally, a deep neural network model is constructed, using a multi-layer perceptron structure, including an input layer (feature dimension), three hidden layers (the number of nodes is 128, 64, 32 respectively), and an output layer (2 nodes, corresponding to two types of labels). The ReLU activation function is used in the hidden layer, and the Softmax function is used in the output layer. The model parameters are optimized by the grid search method, including hyperparameters such as learning rate (range: 0.0001 - 0.1), batch size (range: 32 - 256), dropout rate (range: 0.1 - 0.5), etc.

[0043] For example, in the outburst prediction of a coal mine working face, multi-physical field monitoring data for 30 consecutive days was collected. Through sliding window processing, the data was segmented into 720 time windows (24 hours / window, step size 1 hour). 15 segmentation points were calculated, and the position with the lowest overlap occurred on the 8th day, with an overlap of 45%. 20 statistical features were extracted for each time window to form a 720×20 feature matrix. According to historical records, 2 outburst events occurred during this period, and 48 samples within 24 hours before the event were marked as "outburst danger", and the remaining 672 samples were marked as "outburst safety". 624 artificial "outburst danger" samples were generated by the SMOTE method, making the number of samples in both categories 672. After data augmentation, the total number of training samples reached 4032. The model was trained with a batch size of 128, a learning rate of 0.001, and a dropout rate of 0.3. The final model achieved a prediction accuracy of 90% on the test set.

[0044] In a specific embodiment, the process of performing step S105 may specifically include the following steps: (1) Perform kernel density estimation processing on the output data of the outburst disaster prediction model to obtain a probability density function, and perform integral calculation on the probability density function to obtain a cumulative distribution sequence; (2) Calculate the confidence level for the cumulative distribution sequence to obtain a confidence boundary sequence, and perform interval estimation processing on the confidence boundary sequence to obtain the confidence interval of the prediction result; (3) Perform K-fold cross-validation processing on the confidence interval of the prediction result to obtain a validation index sequence, and perform leave-one-out validation processing on the validation index sequence to obtain a validation result set; (4) Perform stability analysis processing on the validation result set to obtain a stability index sequence, and perform consistency test processing on the stability index sequence to obtain a test result matrix; (5) Perform comprehensive scoring processing on the test result matrix to obtain an evaluation score sequence, and perform weight assignment processing on the evaluation score sequence to obtain the prediction reliability evaluation result.

[0045] Specifically, the output data of the outburst disaster prediction model is first subjected to kernel density estimation processing, and the probability density function of the predicted value is obtained through a non-parametric estimation method. The Gaussian kernel function is used as the kernel function for kernel density estimation, and the bandwidth parameter is determined by the cross-validation method. The calculation formula for kernel density estimation is: ; where, is the probability density function, is the predicted output value, n is the number of samples, B is the bandwidth parameter, and K is the Gaussian kernel function. Numerical integration calculation is performed on the probability density function to obtain the cumulative distribution function:

[0046] Among them, is the cumulative distribution function, is the probability density function.

[0047] Calculate the confidence level for the cumulative distribution sequence, set the confidence level to 95%, and determine the upper and lower boundary values according to the confidence interval theory. The K-fold cross-validation adopts the setting of K = 5, randomly divides the data set into 5 subsets, uses 4 subsets to train the model each time, and the remaining 1 subset is used for verification. Repeat this process 5 times to obtain the complete verification index sequence. The verification indicators include accuracy, precision, recall, and F1 score. Through the leave-one-out cross-validation, each sample is used as the verification sample in turn, and the remaining samples are used for training to obtain a more detailed verification result set. Conduct a stability analysis on the verification results and calculate the fluctuation degree of the prediction results in different verification processes. The stability indicators include statistical quantities such as standard deviation and coefficient of variation. The consistency test uses Cohen's Kappa coefficient to evaluate the consistency level of the prediction results in different verification processes. In the comprehensive scoring stage, assign weights to indicators such as accuracy, stability, and consistency according to their importance to obtain the final prediction reliability evaluation result.

[0048] For example, in the outburst prediction of a certain coal mine working face, process the monitoring data for 30 consecutive days. The distribution of the predicted values obtained by kernel density estimation shows that 85% of the predicted values fall within the interval of 0.2 - 0.8, and the median is 0.45. At the 95% confidence level, the confidence interval is [0.15, 0.85]. Through 5-fold cross-validation, the average accuracy is 92.5%, and the standard deviation is 2.3%. The precisions of each fold of verification are 90.2%, 91.5%, 89.8%, 92.1%, and 91.8% respectively, and the recalls are 88.5%, 89.2%, 87.9%, 90.3%, and 89.6% respectively. The leave-one-out verification results show that 97.2% of the sample prediction results are consistent with the original labels. The stability analysis shows that the coefficient of variation of the prediction results in different verification processes is 0.068, and the Kappa coefficient is 0.856, indicating that the prediction results have high stability and consistency. Finally, the accuracy, stability, and consistency indicators are weighted according to the weight ratio of 0.4:0.3:0.3 to obtain a comprehensive evaluation score of 0.895, indicating that the prediction model has high reliability.

[0049] In a specific embodiment, the process of executing step S106 may specifically include the following steps: (1) Calculate the distribution characteristics of the prediction reliability evaluation results to obtain a characteristic statistical sequence, and perform clustering analysis on the characteristic statistical sequence to obtain a hierarchical classification sequence; (2) Calculate the threshold intervals for the grading sequence to obtain the risk level boundaries, and perform fuzzy classification on the risk level boundaries to obtain the early warning level determination results; (3) Conduct time series analysis on the early warning level determination results to obtain the time evolution sequence, and perform spatial interpolation on the time evolution sequence to obtain the spatial distribution matrix; (4) Conduct regional correlation analysis on the spatial distribution matrix to obtain the correlation feature sequence, and perform trend prediction on the correlation feature sequence to obtain the evolution trend sequence; (5) Conduct risk propagation analysis on the evolution trend sequence to obtain the propagation path sequence, and perform decision optimization on the propagation path sequence to obtain the target early warning decision information.

[0050] Specifically, first calculate the distribution characteristics of the prediction reliability evaluation results. By calculating statistical characteristics such as the distribution density, distribution skewness, and dispersion degree of the evaluation values, obtain the characteristic statistical sequence. Use the K-means clustering algorithm to analyze the characteristic statistical sequence, set the number of clusters K = 3, and divide the data points into three levels: high risk, medium risk, and low risk, forming the grading sequence. The selection of the cluster center is based on the natural grouping characteristics of the data distribution, and classification is performed by calculating the Euclidean distance from the sample points to the cluster center. Calculate the threshold intervals for the grading sequence, and use the method based on the probability density function to determine the risk level boundaries. By analyzing the probability distribution characteristics of the data at each level and combining expert experience, set the threshold intervals: the high-risk threshold is set at the 90th percentile, and the low-risk threshold is set at the 10th percentile. Use the fuzzy classification method to process the boundary values, introduce the membership function to describe the degree to which the data points belong to different risk levels, and obtain a more reasonable early warning level determination result.

[0051] Conduct time series analysis on the early warning level determination results, and use the time series decomposition method to extract the trend term, cycle term, and random term. Identify the change rules of the time series through autocorrelation analysis and cross-correlation analysis to obtain the time evolution sequence. Use the Kriging spatial interpolation method to extend the early warning level information of discrete monitoring points to the entire monitoring area and generate a continuous spatial distribution matrix. The interpolation process takes into account the spatial position correlation and the weights of each monitoring point. Conduct regional correlation analysis on the spatial distribution matrix to calculate the risk propagation relationship between different regions. Evaluate the risk spatial aggregation degree by calculating the spatial autocorrelation index and identify the spatial distribution characteristics of high-risk regions. According to the change rules of the risk levels in adjacent regions, extract the correlation feature sequence, and use the time series prediction method to predict the risk evolution trend. The prediction uses the sliding time window method, with the window size set to 24 hours and the prediction step size set to 1 hour.

[0052] Conduct risk propagation analysis on the evolution trend sequence, and calculate the diffusion path and speed of risks based on the risk propagation network model. By analyzing the propagation characteristics of historical prominent events, establish a risk propagation equation to calculate the probability and intensity of risk diffusion from high-risk areas to the surrounding areas. Finally, combine the risk level, propagation path, and evolution trend to generate partition warning decision information.

[0053] For example, in the outburst prediction of a certain coal mine working face, analyze the evaluation results of the prediction reliability for 7 consecutive days. Use K-means clustering to divide the evaluation values into three categories: high risk (evaluation value > 0.85), medium risk (0.45 - 0.85), and low risk (< 0.45). Among the 15 monitoring points arranged in the working face, it is detected that the evaluation values of 3 points exceed 0.85. Through spatial interpolation analysis, it is found that these high-risk points are mainly concentrated near the fault zone of the working face. Time series analysis shows that the risk level in the high-risk area shows a gradually increasing trend, and the risk growth rate within 24 hours reaches 15%. Risk propagation analysis shows that the risk propagation speed from the fault zone to both wings is 2 meters per hour, and it is expected that the affected range will expand to adjacent areas within 48 hours. Based on these analysis results, partition warning decision information is generated: issue a red warning for the area near the fault zone, requiring immediate shutdown and evacuation of personnel; issue a yellow warning for the adjacent areas, and increase the monitoring frequency; maintain normal monitoring for other areas.

[0054] The above describes the method for predicting coal and gas outburst disasters based on multi-physical field coupling in the embodiments of the present application. Next, the device for predicting coal and gas outburst disasters based on multi-physical field coupling in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the device for predicting coal and gas outburst disasters based on multi-physical field coupling in the embodiments of the present application includes: A collection module 201, configured to collect and perform data standardization processing on the original data of the stress field, gas field, temperature field, and acoustic emission field at the coal mine site to obtain an initial multi-physical field data set, and perform outlier detection and data filling processing on the initial multi-physical field data set to obtain a preprocessed multi-physical field parameter data set; A decomposition module 202, configured to perform physical field component extraction and feature decomposition on the preprocessed multi-physical field parameter data set to obtain a basic parameter matrix for each physical field, and perform cross-correlation analysis and coupling factor calculation on the basic parameter matrix to obtain a physical field coupling relationship matrix; An extraction module 203, configured to extract time-domain features and frequency-domain features from the physical field coupling relationship matrix to obtain a multi-dimensional feature vector, and perform redundant feature elimination and feature reconstruction on the multi-dimensional feature vector to obtain an optimized feature parameter set; A processing module 204, configured to perform temporal segmentation and sample annotation on the optimized feature parameter set to obtain a training sample set, perform data augmentation and normalization processing on the training sample set to obtain a model training data set, and perform pattern recognition and parameter optimization on the model training data set to obtain a prominent disaster prediction model; A calculation module 205, configured to perform probability distribution calculation on the output data of the prominent disaster prediction model to obtain a confidence interval of the prediction result, and perform multi-dimensional cross-validation on the confidence interval to obtain a prediction reliability evaluation result; A division module 206, configured to perform threshold grading and risk level division on the prediction reliability evaluation result to obtain an early warning level determination result, and perform spatio-temporal correlation analysis on the early warning level determination result to obtain target early warning decision-making information.

[0055] Through the collaborative cooperation of the above-mentioned various components, by collecting and preprocessing multi-physical field data of the stress field, gas field, temperature field, and acoustic emission field at the coal mine site, the noise and outliers in the original data are effectively eliminated, and the data quality is improved. Through physical field component extraction and feature decomposition, the basic characteristics of each physical field are deeply analyzed, and through cross-correlation analysis and coupling factor calculation, the coupling relationship between multi-physical fields is comprehensively revealed, providing more comprehensive feature information for prominent disaster prediction. By extracting time-domain features and frequency-domain features from the physical field coupling relationship matrix, combined with redundant feature elimination and feature reconstruction techniques, an optimized feature parameter set is obtained, improving the representativeness and effectiveness of the features. Through temporal segmentation and sample annotation, a standardized training sample set is established, and data augmentation and normalization processing methods are used to enhance the representativeness of the samples and the generalization ability of the model. Based on the deep neural network-based prominent disaster prediction model, accurate prediction of prominent risks is realized. Through probability distribution calculation and multi-dimensional cross-validation, the reliability of the prediction result is evaluated, ensuring the credibility of the prediction result. Finally, through threshold grading and risk level division, combined with spatio-temporal correlation analysis, scientific early warning decision-making information is formed, providing important decision-making support for coal mine safety production. This method overcomes the defect of incomplete information in traditional single-physical field monitoring methods, improves the accuracy and reliability of prediction, and realizes accurate early warning of coal and gas outburst disasters.

[0056] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system, system, and unit can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0057] As described above, the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for predicting coal and gas outburst disasters based on multi-physical field coupling, characterized in that: The coal and gas outburst disaster prediction method based on multi-physical field coupling includes: The original data of the stress field, gas field, temperature field and acoustic emission field at the coal mine site are collected and processed for data standardization to obtain a multi-physical field initial data set, and outlier detection and data completion processing are performed on the multi-physical field initial data set to obtain a pre-processed multi-physical field parameter data set; Performing physical field component extraction and feature decomposition on the preprocessed multi-physical field parameter data set to obtain a basic parameter matrix of each physical field, and performing cross-correlation analysis and coupling factor calculation on the basic parameter matrix to obtain a physical field coupling relationship matrix; Extracting time domain features and frequency domain features of the physical field coupling relationship matrix to obtain a multidimensional feature vector, and performing redundant feature elimination and feature reconstruction on the multidimensional feature vector to obtain an optimized feature parameter set; Performing time series segmentation and sample labeling on the optimized feature parameter set to obtain a training sample set, performing data enhancement and normalization processing on the training sample set to obtain a model training data set, and performing pattern recognition and parameter optimization on the model training data set to obtain a prominent disaster prediction model; Performing probability distribution calculation on the output data of the prominent disaster prediction model to obtain a confidence interval of the prediction result, and performing multi-dimensional cross-validation on the confidence interval to obtain a prediction reliability evaluation result; The prediction reliability assessment result is subjected to threshold classification and risk level classification to obtain a warning level determination result, and the warning level determination result is subjected to spatiotemporal correlation analysis to obtain target warning decision information.

2. The method for predicting coal and gas outburst disasters based on multi-physical field coupling according to claim 1 is characterized in that: The original data of the stress field, gas field, temperature field and acoustic emission field at the coal mine site are collected and processed for data standardization to obtain a multi-physical field initial data set, and outlier detection and data completion processing are performed on the multi-physical field initial data set to obtain a pre-processed multi-physical field parameter data set, including: The original data of stress field, gas field, temperature field and acoustic emission field at the coal mine site are collected and standardized to obtain the initial data set of multi-physics fields. Performing peak normalization processing on the multi-physics field initial data set to obtain a normalized data set, and performing median filling on the normalized data set to obtain supplemented initial data; Performing sliding window segmentation processing on the supplemented initial data to obtain a data segment sequence, and performing interquartile range method anomaly detection on the data segment sequence to obtain a marked data sequence; Performing linear interpolation processing on the abnormal values ​​in the marked data sequence to obtain an interpolated data sequence, and performing wavelet denoising processing on the interpolated data sequence to obtain a denoised data sequence; Performing time window synchronization processing on the noise reduction data sequence to obtain an aligned data sequence, and performing data resampling processing on the aligned data sequence to obtain an equally spaced data sequence; Performing Z-score standardization processing on the equally spaced data sequence to obtain a standardized data sequence, and performing batch verification processing on the standardized data sequence to obtain a multi-physics field initial data set; Performing 3σ criterion anomaly detection on the initial data set of the multi-physics field to obtain an abnormal point marking sequence, and performing spline interpolation processing on the abnormal point marking sequence to obtain a data completion sequence; Performing moving average filtering on the data completion sequence to obtain a smoothed data sequence, and performing data quality assessment on the smoothed data sequence to obtain a data quality index; Performing threshold screening processing on the data quality indicators to obtain a qualified data sequence, and performing outlier detection on the qualified data sequence to obtain a detection result sequence; The detection result sequence is subjected to data verification processing to obtain a verified data sequence, and the verified data sequence is subjected to data format unification processing to obtain a preprocessed multi-physical field parameter data set.

3. The method for predicting coal and gas outburst disasters based on multi-physical field coupling according to claim 1 is characterized in that: The preprocessed multi-physical field parameter data set is subjected to physical field component extraction and feature decomposition to obtain a basic parameter matrix of each physical field, and the basic parameter matrix is ​​subjected to cross-correlation analysis and coupling factor calculation to obtain a physical field coupling relationship matrix, including: Performing physical field type classification processing on the preprocessed multi-physical field parameter data set to obtain a stress field data subset, a gas field data subset, a temperature field data subset and an acoustic emission field data subset, and performing time series sorting processing on each data subset to obtain an ordered data sequence; Performing principal component decomposition processing on the ordered data sequence to obtain main characteristic components, and calculating variance contribution rates of the main characteristic components to obtain a characteristic weight sequence; Performing threshold screening processing on the feature weight sequence to obtain a key feature sequence, and performing physical quantity unit unification processing on the key feature sequence to obtain a standardized feature sequence; Performing matrix reconstruction processing on the standardized characteristic sequence to obtain a basic parameter matrix of each physical field, and performing data integrity check on the basic parameter matrix to obtain a verified parameter matrix; Calculating the Pearson correlation coefficient of the verified parameter matrix to obtain a correlation coefficient matrix, and performing a significance test on the correlation coefficient matrix to obtain effective correlation items; Calculating the coupling degree of the effective correlation items to obtain a coupling factor sequence, and normalizing the coupling factor sequence to obtain a standardized coupling factor; The standardized coupling factors are subjected to matrix mapping processing to obtain an initial coupling relationship matrix, and the initial coupling relationship matrix is ​​subjected to symmetry testing to obtain a physical field coupling relationship matrix.

4. The method for predicting coal and gas outburst disasters based on multi-physical field coupling according to claim 1 is characterized in that: The time domain feature and frequency domain feature extraction of the physical field coupling relationship matrix to obtain a multidimensional feature vector, and the redundant feature elimination and feature reconstruction of the multidimensional feature vector to obtain an optimized feature parameter set include: Performing wavelet transform processing on the physical field coupling relationship matrix to obtain a time domain coefficient sequence, and performing statistical feature calculation on the time domain coefficient sequence to obtain a time domain feature set; Performing Fourier transform processing on the time domain feature set to obtain a frequency domain feature sequence, and performing frequency band energy calculation on the frequency domain feature sequence to obtain frequency band energy distribution; Performing feature point extraction processing on the frequency band energy distribution to obtain a feature point sequence, and performing dimension splicing processing on the feature point sequence to obtain a multi-dimensional feature vector; Performing correlation analysis on the multidimensional feature vector to obtain a feature correlation matrix, and performing hierarchical clustering on the feature correlation matrix to obtain a feature cluster sequence; Performing information gain calculation on the feature cluster sequence to obtain a feature importance sequence, and performing threshold screening processing on the feature importance sequence to obtain a candidate feature set; The candidate feature set is orthogonally transformed to obtain feature basis vectors, and the feature basis vectors are linearly combined to obtain an optimized feature parameter set.

5. The method for predicting coal and gas outburst disasters based on multi-physical field coupling according to claim 1 is characterized in that: The optimized feature parameter set is subjected to time series segmentation and sample labeling to obtain a training sample set, and the training sample set is subjected to data enhancement and normalization processing to obtain a model training data set, and the model training data set is subjected to pattern recognition and parameter optimization to obtain a prominent disaster prediction model, including: Performing sliding window processing on the optimized feature parameter set to obtain a time sequence segment sequence, and performing overlap calculation on the time sequence segment sequence to obtain a segmentation point set; Performing boundary optimization processing on the segmentation point set to obtain a time series segment sequence, and performing feature statistics processing on the time series segment sequence to obtain a sample feature matrix; Performing label assignment processing on the sample feature matrix to obtain initial training samples, and performing balanced sampling processing on the initial training samples to obtain a training sample set; Performing data rotation processing on the training sample set to obtain an amplified sample sequence, and performing noise injection processing on the amplified sample sequence to obtain an enhanced data set; Performing maximum and minimum value normalization processing on the enhanced data set to obtain a standardized data sequence, and performing data verification processing on the standardized data sequence to obtain a model training data set; The model training data set is processed by parameter space partition to obtain a parameter grid sequence, and the parameter grid sequence is processed by deep neural network to obtain a prominent disaster prediction model.

6. The method for predicting coal and gas outburst disasters based on multi-physical field coupling according to claim 1 is characterized in that: The probability distribution calculation is performed on the output data of the prominent disaster prediction model to obtain the confidence interval of the prediction result, and the confidence interval is cross-validated in multiple dimensions to obtain the prediction reliability evaluation result, including: Performing kernel density estimation processing on the output data of the prominent disaster prediction model to obtain a probability density function, and performing integral calculation on the probability density function to obtain a cumulative distribution sequence; Calculating the confidence level of the cumulative distribution sequence to obtain a confidence boundary sequence, and performing interval estimation processing on the confidence boundary sequence to obtain a confidence interval of the prediction result; Performing K-fold cross validation on the confidence interval of the prediction result to obtain a validation index sequence, and performing leave-one-out validation on the validation index sequence to obtain a validation result set; Performing stability analysis processing on the verification result set to obtain a stability index sequence, and performing consistency test processing on the stability index sequence to obtain a test result matrix; The inspection result matrix is ​​subjected to comprehensive scoring processing to obtain an evaluation score sequence, and the evaluation score sequence is subjected to weight distribution processing to obtain a prediction reliability evaluation result.

7. The method for predicting coal and gas outburst disasters based on multi-physical field coupling according to claim 1 is characterized in that: The prediction reliability assessment result is subjected to threshold classification and risk level classification to obtain a warning level determination result, and the warning level determination result is subjected to spatiotemporal correlation analysis to obtain target warning decision information, including: Calculating the distribution characteristics of the prediction reliability evaluation results to obtain a characteristic statistical sequence, and performing cluster analysis on the characteristic statistical sequence to obtain a graded classification sequence; Performing threshold interval calculation on the level division sequence to obtain risk level boundaries, and performing fuzzy classification processing on the risk level boundaries to obtain warning level determination results; Performing time series analysis on the warning level determination result to obtain a time evolution sequence, and performing spatial interpolation processing on the time evolution sequence to obtain a spatial distribution matrix; Performing regional correlation analysis processing on the spatial distribution matrix to obtain a correlation feature sequence, and performing trend prediction processing on the correlation feature sequence to obtain an evolution trend sequence; The evolution trend sequence is subjected to risk propagation analysis to obtain a propagation path sequence, and the propagation path sequence is subjected to decision optimization to obtain target early warning decision information.

8. A coal and gas outburst disaster prediction device based on multi-physical field coupling, used to implement the coal and gas outburst disaster prediction method based on multi-physical field coupling as described in any one of claims 1 to 7, characterized in that: The coal and gas outburst disaster prediction device based on multi-physical field coupling includes: The acquisition module is used to collect and standardize the original data of the stress field, gas field, temperature field and acoustic emission field at the coal mine site to obtain a multi-physical field initial data set, and perform outlier detection and data completion processing on the multi-physical field initial data set to obtain a pre-processed multi-physical field parameter data set; A decomposition module is used to perform physical field component extraction and feature decomposition on the preprocessed multi-physical field parameter data set to obtain a basic parameter matrix of each physical field, and to perform cross-correlation analysis and coupling factor calculation on the basic parameter matrix to obtain a physical field coupling relationship matrix; An extraction module is used to extract time domain features and frequency domain features of the physical field coupling relationship matrix to obtain a multi-dimensional feature vector, and to eliminate redundant features and reconstruct features of the multi-dimensional feature vector to obtain an optimized feature parameter set; A processing module, used to perform time series segmentation and sample labeling on the optimized feature parameter set to obtain a training sample set, perform data enhancement and normalization processing on the training sample set to obtain a model training data set, perform pattern recognition and parameter optimization on the model training data set to obtain a prominent disaster prediction model; A calculation module, used to perform probability distribution calculation on the output data of the prominent disaster prediction model to obtain a confidence interval of the prediction result, and to perform multi-dimensional cross-validation on the confidence interval to obtain a prediction reliability evaluation result; The classification module is used to perform threshold classification and risk level classification on the prediction reliability evaluation result to obtain a warning level determination result, and perform spatiotemporal correlation analysis on the warning level determination result to obtain target warning decision information.

Images