Multi-channel coal quality online detection system and method based on signal enhancement

Abstract

The invention discloses a multi-channel coal quality online detection system and method based on signal enhancement, belongs to the technical field of coal quality detection, and solves the problem that the accuracy and stability of a detection result are affected as a fixed threshold rule adopted by wavelet threshold denoising in the existing method cannot adapt to noise characteristic changes in different temperature intervals. The method comprises the steps of performing signal enhancement preprocessing on a real-time information set, performing initial feature selection on a signal enhancement set based on an index identification model, and performing index collaborative correction based on a drift risk amount of an initial index set and the initial index set; according to the method, signal enhancement preprocessing is carried out on coal quality multi-channel spectral information and associated environment information which are acquired in real time, interference of environmental factors on spectral signals is considered during signal enhancement preprocessing, and interference of the environmental factors on the signals can be effectively compensated through coupling of a field three-dimensional temperature field and a self-adaptive time period; therefore, the stability and the accuracy of the signal are improved.

Description

Technical Field

[0001] This invention belongs to the field of coal quality testing technology, specifically relating to a multi-channel online coal quality testing system and method based on signal enhancement. Background Technology

[0002] Coal quality testing technology, as a key link in quality control in the coal industry, has evolved from traditional laboratory analysis to rapid on-site testing. Traditional coal quality analysis methods mainly rely on laboratory chemical analysis, such as industrial analysis and elemental analysis. While these methods are highly accurate, they have significant drawbacks, including complex sample preparation, long analysis cycles, and inability to meet the real-time requirements of industrial production. With the development of spectroscopic analysis technology, near-infrared spectroscopy, laser-induced breakdown spectroscopy, and other techniques have been introduced into the field of coal quality testing, enabling non-contact and rapid detection.

[0003] Existing online coal quality monitoring systems face numerous technical bottlenecks in signal acquisition, especially in remote monitoring scenarios. Optical signals are susceptible to environmental interference during transmission, resulting in low signal-to-noise ratios and poor stability of the acquired spectral data. Furthermore, traditional optical acquisition systems often employ a separate design, with spatial misalignment between the excitation source and the signal collection optical path, leading to unstable light radiation energy reception and severely impacting the repeatability and accuracy of the detection results.

[0004] Chinese patent CN119290810B discloses an adaptive calibration method and apparatus for correcting environmental influences in coal quality testing. The adaptive calibration method includes spectral preprocessing and LIBS spectral characteristic peak intensity correction; the apparatus includes a control host, a spectrometer, a laser generator, and a rotary sample stage. However, existing methods use wavelet transform to denoise the spectral signal of the sample and reduce the interference of adverse factors on the spectral signal. Wavelet threshold denoising uses a fixed threshold rule that cannot adapt to the changes in noise characteristics in different temperature ranges, and does not consider the influence of associated environmental information (such as temperature) on the spectral signal, which leads to the measured spectral signal being prone to drift and distortion, thereby affecting the accuracy and stability of the detection results. To address the above problems, we propose a multi-channel online coal quality detection system and method based on signal enhancement. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing a multi-channel online coal quality detection system and method based on signal enhancement. This invention solves the problems of existing methods that, when using wavelet transform to denoise the spectral signal of a sample and reduce interference from adverse factors, employ fixed threshold rules for wavelet threshold denoising, which cannot adapt to changes in noise characteristics across different temperature ranges. Furthermore, these methods do not consider the influence of associated environmental information on the spectral signal, leading to drift and distortion in the measured spectral signal, thus affecting the accuracy and stability of the detection results.

[0006] This invention is implemented as follows: a multi-channel online coal quality detection method based on signal enhancement, the method comprising: Based on the real-time acquisition of multi-channel spectral information and associated environmental information of coal by coal quality testing equipment, the multi-channel spectral information and associated environmental information of coal quality are merged into a real-time information set, and the real-time information set is preprocessed for signal enhancement to output a signal enhancement set; A support vector machine-based index recognition model is pre-built, and the signal enhancement set is iteratively trained until convergence. The signal enhancement set is loaded, and initial feature selection is performed on the signal enhancement set based on the index recognition model. The feature importance is determined by improving the random forest algorithm, and a multi-channel weighted feature map combining the feature importance is constructed. The index recognition model outputs the initial index set by combining the multi-channel weighted feature map. An initial indicator set is obtained, and the detection confidence of the initial indicator set is determined based on a multi-head attention mechanism. The drift risk of the initial indicator set is dynamically evaluated, and the indicators are collaboratively corrected based on the drift risk of the initial indicator set and the initial indicator set. The coal quality indicator detection results after collaborative correction are output.

[0007] Preferably, the method for signal enhancement preprocessing of the real-time information set includes: Load the real-time information set, perform outlier cleaning and removal on the real-time information set based on the boundary hybrid sampling strategy, obtain the real-time information set after outlier cleaning and removal, and perform normalization processing on the real-time information set. The normalized real-time information set is obtained, the signal bands of the real-time information set are identified based on the spectral analysis method, the peak characteristics of the signal bands are determined, and the coal quality multi-channel spectral information in the real-time information set is adaptively extracted based on the peak characteristics of the signal bands. The time stamp alignment method is used to align the associated environmental information with multiple sets of the adaptive time periods to obtain the adaptive time period set. The adaptive time period set is dynamically filtered based on the BEADS algorithm combined with the SG filtering method, and the dynamically filtered adaptive time period set is output. Load the adaptive time period set, construct the on-site three-dimensional temperature field based on the adaptive time period set, couple the on-site three-dimensional temperature field with the adaptive time period to obtain the derived enhancement features, and merge the derived enhancement features, the on-site three-dimensional temperature field and the adaptive time period to obtain the signal enhancement set.

[0008] Preferably, the method for outlier cleaning and removal from real-time information sets based on a boundary hybrid sampling strategy includes: Acquire a real-time information set and use a boundary point detection algorithm to determine the coefficient of variation of data points in the real-time information set; The coefficient of variation for each data point is calculated using the following formula: in, Represents the coefficient of variation for data points. These represent the data point density and the distance between the data points and the real-time information concentration. The number of adjacent data points; Based on the coefficient of variation of data points, a variation threshold is set for the real-time information set. Data points with a coefficient of variation greater than the variation threshold in the real-time information set are classified as boundary data points, and the remaining data points are classified as center data points. Load boundary data points, preset the oversampling factor based on the weight of the boundary data points, and use synthetic minority oversampling technology to copy the boundary data points to obtain new boundary data points after copying. Load the center data points, preset the undersampling rate based on the density of the center data points, traverse the center data points, and use the Tomek-Links method combined with the undersampling rate to remove and clean the center data points to obtain the reconstructed center data. The copied boundary data points and the reconstructed center data are obtained and integrated into a real-time information set after outlier cleaning and removal. Load the real-time information set after outlier cleaning and removal, and normalize the real-time information set.

[0009] Preferably, the adaptive time-time set dynamic filtering method based on the BEADS algorithm combined with the SG filtering method includes: Load the adaptive time period set, initialize the parameters of the BEADS algorithm, and set the number of iterations; The differential operator of coal quality multi-channel spectral information in the adaptive time period is calculated based on the BEADS algorithm and the spectral information estimate is updated. The baseline component and noise component of coal quality multi-channel spectral information in the adaptive time period are determined after a preset number of iterations. The estimated baseline component and noise component are subtracted from the original signal to obtain the denoised adaptive time period; The window size and polynomial order of the SG filter corresponding to different adaptive time periods are determined based on the smoothness of the adaptive time periods. The SG filtering method achieves smoothing of coal quality multi-channel spectral information in the adaptive time periods through local polynomial fitting, and outputs the dynamically filtered adaptive time period set.

[0010] Preferably, the method for coupling the on-site three-dimensional temperature field with the adaptive time period includes: Load the adaptive time period set, extract the associated environmental information from the adaptive time period set, establish a coal field grid model based on the associated environmental information, divide the coal field grid model into subgrids, and obtain the internal temperature field space and boundary temperature field space of the subgrids of the coal field grid model. The internal temperature field space and boundary temperature field space of the sub-grid of the coal field grid model are obtained. Based on the equivalent principle of velocity gradient and temperature gradient, the dynamically filtered associated temperature parameters are allocated to the internal temperature field space and boundary temperature field space. The internal temperature field space and boundary temperature field space are integrated to obtain the three-dimensional temperature field of the field. A nonlinear relationship is established with the associated temperature parameter, the adaptive time period coal quality multi-channel spectral information as the output variable, and the derived enhancement feature as the output variable. The nonlinear relationship includes the temperature correlation coefficient and the spectral correlation coefficient. The weight coefficients of temperature correlation coefficient and spectral correlation coefficient are adjusted based on particle swarm optimization algorithm, and Gaussian perturbation and amplitude change rate are introduced on the basis of particle swarm optimization algorithm to optimize the weight coefficients of temperature correlation coefficient and spectral correlation coefficient. Iterative computation is used to minimize the Bayesian information criterion of the nonlinear relationship. Based on the minimization of the Bayesian information criterion, the combination of temperature correlation coefficient and spectral correlation coefficient that meet the convergence condition is determined to obtain the derived enhancement features. The derived enhancement features, the on-site three-dimensional temperature field, and the adaptive time period are merged to obtain the signal enhancement set.

[0011] Preferably, when pre-constructing an indicator recognition model based on a support vector machine, the indicator recognition model is based on the support vector machine model. The indicator recognition model also includes an input layer and an output layer. A spatiotemporal feature extraction operator layer is set between the input layer and the support vector machine. The spatiotemporal feature extraction operator layer introduces a singular value decomposition algorithm. An improved random forest algorithm is introduced into the support vector machine model. The industrial analysis indicators of the multi-channel weighted feature map are predicted by a parallel SVM prediction head, and the industrial analysis indicators are used as the initial indicator set output.

[0012] Preferably, the method for outputting an initial indicator set by combining the indicator recognition model with multi-channel weighted feature maps includes: Feature extraction is performed on the signal enhancement set to obtain a feature combination set including at least one set of index features. For the index features in the feature combination set, the mutual information between them and other index features is calculated. Using the mutual information between indicator features as the distance between indicator features, the M indicator features closest to the indicator features are determined. The feature weight update probability of the M indicator features is determined based on the Relief algorithm. The weights of the M indicator features are updated based on the feature weight update probability to obtain the initial feature set after weight update. Based on the improved random forest algorithm, a Chebyshev chaotic mapping network is used for the initial feature set. The Chebyshev chaotic mapping network maps the initial feature set to weights of 0 to 1 through an approximate impulse function, and eliminates redundant index features based on the preset weights and screening thresholds. The initial feature set, after eliminating redundant indicator features, is used to extract indicator features and construct at least one indicator decision tree. Based on the error value and maximum information coefficient of the indicator decision tree, a weighted feature map combining the indicator feature importance is constructed. For multi-channel weighted feature maps, spectral channel RBF kernel function, environmental channel linear kernel function, and derived channel polynomial kernel function are set. Based on the attention mechanism, the spectral channel RBF kernel function, environmental channel linear kernel function, and derived channel polynomial kernel function are weighted and fused. Then, the industrial analysis index of the multi-channel weighted feature map is predicted by the parallel SVM prediction head, and the industrial analysis index is used as the initial index set output.

[0013] On the other hand, the present invention also provides a multi-channel online coal quality detection system based on signal enhancement, the multi-channel online coal quality detection system based on signal enhancement comprising: The signal enhancement module is based on the real-time acquisition of multi-channel spectral information of coal quality and related environmental information by coal quality detection equipment. It merges the multi-channel spectral information of coal quality and related environmental information into a real-time information set, performs signal enhancement preprocessing on the real-time information set, and outputs the signal enhancement set. The multi-channel analysis module is used to pre-build an index recognition model based on support vector machine, iteratively train the signal enhancement set until convergence, load the signal enhancement set, perform initial feature selection on the signal enhancement set based on the index recognition model, determine the feature importance through an improved random forest algorithm, construct a multi-channel weighted feature map combining feature importance, and the index recognition model outputs the initial index set by combining the multi-channel weighted feature map. The index correction module is used to obtain an initial index set, determine the detection confidence of the initial index set based on a multi-head attention mechanism, dynamically evaluate the drift risk of the initial index set, perform collaborative index correction based on the drift risk of the initial index set and the initial index set, and output the coal quality index detection results after collaborative correction.

[0014] Preferably, the signal enhancement module includes: The signal cleaning unit is used to load the real-time information set, clean and remove outliers from the real-time information set based on the boundary mixing sampling strategy, obtain the real-time information set after outlier cleaning and removal, and perform normalization processing on the real-time information set. An adaptive truncation unit is used to acquire a normalized real-time information set, identify the signal bands of the real-time information set based on spectral analysis, determine the peak characteristics of the signal bands, and adaptively truncate the coal quality multi-channel spectral information in the real-time information set based on the peak characteristics of the signal bands. The unit also uses a timestamp alignment method to align the associated environmental information with multiple sets of the adaptive time periods to obtain an adaptive time period set. The dynamic filtering unit uses the BEADS algorithm combined with the SG filtering method to dynamically filter the adaptive time period set and outputs the dynamically filtered adaptive time period set. The signal coupling unit is used to load the adaptive time period set, construct the on-site three-dimensional temperature field based on the adaptive time period set, couple the on-site three-dimensional temperature field with the adaptive time period to obtain the derived enhancement features, and merge the derived enhancement features, the on-site three-dimensional temperature field and the adaptive time period to obtain the signal enhancement set.

[0015] Preferably, the adaptive truncation unit includes: The spectral analysis module is used to acquire the normalized real-time information set, identify the signal bands of the real-time information set based on spectral analysis, and determine the peak characteristics of the signal bands. The time segmentation module is used to adaptively segment the multi-channel spectral information of coal quality in real-time information set by combining the peak characteristics of the signal band. The time period alignment module uses a timestamp alignment method to align the associated environmental information with multiple sets of adaptive time periods to obtain an adaptive time period set.

[0016] Compared with the prior art, the embodiments of this application have the following main advantages: In this embodiment of the invention, signal enhancement preprocessing is performed on real-time acquired multi-channel spectral information of coal quality and associated environmental information. The interference of environmental factors on the spectral signal is considered during the signal enhancement preprocessing. By coupling the on-site three-dimensional temperature field with the adaptive time period, derived enhancement features are obtained, which can effectively compensate for the interference of environmental factors on the signal. The signal after environmental factor compensation is more stable, reducing signal drift and distortion caused by environmental changes, thereby improving the stability and accuracy of the signal.

[0017] In this embodiment of the invention, when performing signal enhancement preprocessing on the real-time information set, the outlier cleaning and removal based on the boundary mixing sampling strategy can effectively extract non-Gaussian distributed data, thereby ensuring that the samples in the real-time information set are representative of coal quality while avoiding the problem of misjudgment in data removal. Furthermore, by using the BEADS algorithm combined with the SG filtering method to dynamically filter the adaptive time period set, the signal becomes more stable, reducing signal fluctuations caused by baseline drift. By coupling the on-site three-dimensional temperature field with the adaptive time period, derived enhancement features can be obtained, further enhancing the feature expression capability of the signal.

[0018] In this embodiment of the invention, when performing outlier cleaning and removal on the real-time information set based on the boundary hybrid sampling strategy, the variation threshold of the real-time information set is set based on the variation coefficient of the data points, and the oversampling ratio is preset based on the weight of the boundary data points. On the one hand, the data points can be accurately and clearly divided into boundary data points and center data points. On the other hand, by synthesizing minority oversampling techniques, the quality of boundary data points can be increased and their representativeness can be improved. By using the Tomek-Links method combined with the undersampling rate, redundant data in the center data points is deleted, significantly optimizing the distribution of the center data points. By integrating the copied boundary new data points and the reconstructed center new data points, a high-quality real-time information set is formed, providing a reliable data foundation for subsequent signal enhancement preprocessing.

[0019] In this embodiment of the invention, when dynamically filtering an adaptive time period set based on the BEADS algorithm combined with the SG filtering method, the BEADS algorithm can accurately separate the baseline component and noise component in the spectral signal through differential operators and iterative updates, thereby reducing signal distortion. Combined with the SG filtering method, the parameters can be dynamically adjusted according to the smoothness of the adaptive time period, and smoothing is achieved through local polynomial fitting, reducing the interference of high-frequency noise, while preserving the characteristic peaks of the signal, enhancing signal clarity and effectively reducing noise interference.

[0020] In this embodiment of the invention, by coupling the three-dimensional temperature field of the site with an adaptive time period, the influence of environmental factors on the spectral signal can be fully considered, thereby improving the accuracy and reliability of signal processing. Furthermore, by accurately establishing a grid model of the coal site, the modeling accuracy of the three-dimensional temperature field is improved. Combining the equivalence principle of velocity gradient and temperature gradient, the accurate allocation of associated temperature parameters is achieved, ultimately comprehensively simulating the temperature distribution of the site, reducing the interference of temperature changes on the spectral signal, and enhancing the stability of the signal. Moreover, by minimizing the determined temperature correlation coefficient and spectral correlation coefficient combination through the Bayesian information criterion, derived enhancement features are generated, further enhancing the feature expression capability of the signal, providing richer feature information for subsequent signal enhancement and coal quality index detection, and improving the accuracy and reliability of detection.

[0021] In this embodiment of the invention, when the index recognition model outputs the initial index set by combining the multi-channel weighted feature map, the index recognition model sets a spatiotemporal feature extraction operator layer between the input layer and the support vector machine. This can extract the temporal and spatial features in the signal, enhancing the model's ability to process complex data. Calculating the mutual information between features and using it as the distance between features can more accurately assess the correlation between features. Based on the Relief algorithm, the feature weight update probability is determined, which can dynamically adjust the feature weights and update the features, enhancing the contribution of important features and reducing the impact of redundant features. Finally, the support vector machine model sets a spectral channel RBF kernel function, an environmental channel linear kernel function, and a derived channel polynomial kernel function for the multi-channel weighted feature map, which can better handle different types of features. Furthermore, based on the attention mechanism, the kernel functions of different channels are weighted and fused, which can dynamically adjust the contribution of each channel, improving the model's adaptability and predictive ability. Attached Figure Description

[0022] Figure 1 This is a schematic diagram illustrating the implementation process of the multi-channel online coal quality detection method based on signal enhancement provided by the present invention.

[0023] Figure 2 A schematic diagram of the implementation process of a signal enhancement preprocessing method for real-time information sets is shown.

[0024] Figure 3 A schematic diagram of the implementation process of an outlier cleaning and removal method for real-time information sets based on a boundary hybrid sampling strategy is shown.

[0025] Figure 4 A schematic diagram of the implementation process of the adaptive time-time set dynamic filtering method based on the BEADS algorithm combined with the SG filtering method is shown.

[0026] Figure 5 A schematic diagram of the implementation process of the coupling method between the on-site three-dimensional temperature field and the adaptive time period is shown.

[0027] Figure 6 The diagram illustrates the implementation process of the method for outputting an initial index set by combining the index recognition model with a multi-channel weighted feature map.

[0028] Figure 7 A schematic diagram of the architecture of a multi-channel online coal quality detection system based on signal enhancement is shown. Detailed Implementation

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0030] Existing methods use wavelet transform to denoise the spectral signals of samples and reduce interference from adverse factors. However, wavelet threshold denoising, with its fixed threshold rules, cannot adapt to changes in noise characteristics across different temperature ranges and fails to consider the influence of associated environmental information on the spectral signals. This leads to drift and distortion in the measured spectral signals, affecting the accuracy and stability of the detection results. To address these issues, we propose a multi-channel online coal quality detection system and method based on signal enhancement. In short, the method first performs signal enhancement preprocessing on the real-time information set. Then, based on an index recognition model, initial features are selected from the enhanced signal set. The importance of features is determined by an improved random forest algorithm, and a multi-channel weighted feature map combining feature importance is constructed. The index recognition model outputs an initial index set based on the multi-channel weighted feature map. Finally, index collaborative correction is performed based on the drift risk of the initial index set and the initial index set itself. In this embodiment of the invention, signal enhancement preprocessing is performed on real-time acquired multi-channel spectral information of coal quality and associated environmental information. The interference of environmental factors on the spectral signal is considered during the signal enhancement preprocessing. By coupling the on-site three-dimensional temperature field with the adaptive time period, derived enhancement features are obtained, which can effectively compensate for the interference of environmental factors on the signal. The signal after environmental factor compensation is more stable, reducing signal drift and distortion caused by environmental changes, thereby improving the stability and accuracy of the signal.

[0031] This invention provides a multi-channel online coal quality detection method based on signal enhancement. Figure 1 This diagram illustrates the implementation flow of a multi-channel online coal quality detection method based on signal enhancement. The method specifically includes: S10: Based on the real-time acquisition of multi-channel spectral information and associated environmental information of coal quality by coal quality detection equipment, the multi-channel spectral information and associated environmental information of coal quality are merged into a real-time information set, the real-time information set is preprocessed for signal enhancement, and the signal enhancement set is output. The coal quality testing equipment includes, but is not limited to, laser-induced breakdown spectroscopy (LIBS) equipment, near-infrared spectroscopy (NIR) equipment, X-ray fluorescence (XRF) equipment, and gas chromatography-mass spectrometry (GC-MS) equipment. The multi-channel spectral information of coal quality includes, but is not limited to, ultraviolet-visible spectral information, near-infrared spectral information, Raman spectral information, and fluorescence spectral information. The associated environmental information includes, but is not limited to, associated temperature parameters, associated humidity parameters, associated air pressure parameters, and coal sample location parameters.

[0032] S20: A support vector machine-based index recognition model is pre-built, and the signal enhancement set is iteratively trained until convergence. The signal enhancement set is loaded, and initial feature selection is performed on the signal enhancement set based on the index recognition model. The importance of features is determined by improving the random forest algorithm, and a multi-channel weighted feature map combining the feature importance is constructed. The index recognition model outputs the initial index set by combining the multi-channel weighted feature map. S30: Obtain the initial indicator set, determine the detection confidence of the initial indicator set based on the multi-head attention mechanism, dynamically evaluate the drift risk of the initial indicator set, perform collaborative correction of the indicators based on the drift risk of the initial indicator set and the initial indicator set, and output the coal quality indicator detection results after collaborative correction.

[0033] In this embodiment of the invention, signal enhancement preprocessing is performed on real-time acquired multi-channel spectral information of coal quality and associated environmental information. The interference of environmental factors on the spectral signal is considered during the signal enhancement preprocessing. By coupling the on-site three-dimensional temperature field with the adaptive time period, derived enhancement features are obtained, which can effectively compensate for the interference of environmental factors on the signal. The signal after environmental factor compensation is more stable, reducing signal drift and distortion caused by environmental changes, thereby improving the stability and accuracy of the signal.

[0034] This invention provides a method for signal enhancement preprocessing of real-time information sets. Figure 2 This diagram illustrates the implementation flow of a signal enhancement preprocessing method for real-time information sets. The method specifically includes: S101, Load the real-time information set, perform outlier cleaning and removal on the real-time information set based on the boundary hybrid sampling strategy, obtain the real-time information set after outlier cleaning and removal, and perform normalization processing on the real-time information set. S102, Obtain the normalized real-time information set, identify the signal bands of the real-time information set based on spectral analysis, determine the peak characteristics of the signal bands, and adaptively extract time periods for the coal quality multi-channel spectral information in the real-time information set based on the peak characteristics of the signal bands. Then, use the timestamp alignment method to align the associated environmental information with multiple sets of the adaptive time periods to obtain an adaptive time period set. In this case, adaptive time period extraction based on the peak characteristics of the signal bands can optimize the range of signal processing and reduce unnecessary calculations. Furthermore, adaptive time period extraction can ensure the stability and consistency of signal processing and reduce signal fluctuations caused by time differences.

[0035] S103, based on the BEADS algorithm combined with the SG filtering method, dynamically filters the adaptive time period set and outputs the dynamically filtered adaptive time period set. S104, Load the adaptive time period set, construct the on-site three-dimensional temperature field based on the adaptive time period set, couple the on-site three-dimensional temperature field with the adaptive time period to obtain the derived enhancement features, merge the derived enhancement features, the on-site three-dimensional temperature field and the adaptive time period to obtain the signal enhancement set.

[0036] In this embodiment of the invention, when performing signal enhancement preprocessing on the real-time information set, the outlier cleaning and removal based on the boundary mixing sampling strategy can effectively extract non-Gaussian distributed data, thereby ensuring that the samples in the real-time information set are representative of coal quality while avoiding the problem of misjudgment in data removal. Furthermore, by using the BEADS algorithm combined with the SG filtering method to dynamically filter the adaptive time period set, the signal becomes more stable, reducing signal fluctuations caused by baseline drift. By coupling the on-site three-dimensional temperature field with the adaptive time period, derived enhancement features can be obtained, further enhancing the feature expression capability of the signal.

[0037] This invention provides a method for outlier cleaning and removal from real-time information sets based on a boundary hybrid sampling strategy. Figure 3 This diagram illustrates the implementation flow of a method for outlier removal from real-time information sets based on a boundary hybrid sampling strategy. The method specifically includes: S1011, Obtain the real-time information set and use the boundary point detection algorithm to determine the coefficient of variation of the data points in the real-time information set; The coefficient of variation for each data point is calculated using the following formula: in, Represents the coefficient of variation for data points. These represent the data point density and the distance between the data points and the real-time information concentration. The number of adjacent data points; S1012, Based on the coefficient of variation of the data points, set the variation threshold of the real-time information set. Using the variation threshold of the real-time information set, data points with a coefficient of variation greater than the variation threshold are divided into boundary data points, and the remaining data points are divided into center data points. S1013, Load boundary data points, preset the oversampling ratio based on the weight of the boundary data points, and use synthetic minority oversampling technology to copy the boundary data points to obtain the copied new boundary data points; The oversampling factor is determined by the following formula: in, This is the oversampling factor. This represents the floor function. The data types described for each data point The weighting coefficients and the maximum weighting coefficients for data types. Total number of data types Indicates the total number of boundary data points. Represents data type The average coefficient of variation of the associated data points; S1014, Load the center data points, preset the undersampling rate based on the density of the center data points, traverse the center data points, and use the Tomek-Links method combined with the undersampling rate to delete and clean the center data points to obtain the reconstructed center data. The undersampling rate is preset using the following formula: in, Indicates the undersampling rate. These are the density of the center data point and the average density of the center data point, respectively. Density retention factor; S1015, Obtain the new boundary data points after replication and the new center data after reconstruction, and integrate the new boundary data points after replication and the new center data after reconstruction into a real-time information set after outlier cleaning and removal; S1016, Load the real-time information set after outlier cleaning and removal, and normalize the real-time information set.

[0038] In this embodiment of the invention, when performing outlier cleaning and removal on the real-time information set based on the boundary hybrid sampling strategy, the variation threshold of the real-time information set is set based on the variation coefficient of the data points, and the oversampling ratio is preset based on the weight of the boundary data points. On the one hand, the data points can be accurately and clearly divided into boundary data points and center data points. On the other hand, by synthesizing minority oversampling techniques, the quality of boundary data points can be increased and their representativeness can be improved. By using the Tomek-Links method combined with the undersampling rate, redundant data in the center data points is deleted, significantly optimizing the distribution of the center data points. By integrating the copied boundary new data points and the reconstructed center new data points, a high-quality real-time information set is formed, providing a reliable data foundation for subsequent signal enhancement preprocessing.

[0039] This invention provides a dynamic filtering method for adaptive time-time sets based on the BEADS algorithm combined with the SG filtering method. Figure 4 This diagram illustrates the implementation flow of an adaptive time-time set dynamic filtering method based on the BEADS algorithm combined with the SG filtering method. The method specifically includes: S1031, Load the adaptive time period set, initialize the parameters of the BEADS algorithm, and set the number of iterations. The parameters of the BEADS algorithm are as follows: , These are the initial regularization weights, first-order difference weights, and second-order difference weights, respectively. The number of iterations is 15-20. The BEADS algorithm allows parameter initialization and iteration count setting, enabling the algorithm to dynamically adjust parameters according to data characteristics to achieve optimal noise reduction. Compared to wavelet transform, the BEADS algorithm offers greater flexibility and can better adapt to different types of spectral signals. The initial regularization term weights are determined using the following formula: in, Indicates the initial regularization term weight. For adaptive signal-to-noise ratio over time period, This is the weight scaling factor; S1032, based on the BEADS algorithm, calculates the differential operator of the multi-channel spectral information of coal quality in the adaptive time period and updates the spectral information estimate. After a preset number of iterations, the baseline component and noise component of the multi-channel spectral information of coal quality in the adaptive time period are determined. The BEADS algorithm can accurately separate the baseline component and noise component in the spectral signal by calculating the differential operator and iterative updating, thereby improving the signal-to-noise ratio of the signal. S1033, subtract the estimated baseline component and noise component from the original signal to obtain the denoised adaptive time period; S1034, based on the smoothness of the adaptive time period, the window size and polynomial order of the SG filter corresponding to different adaptive time periods are determined. The SG filtering method achieves smoothing of the multi-channel spectral information of coal quality in the adaptive time period through local polynomial fitting, and outputs a dynamically filtered adaptive time period set. It should be noted that the choice of polynomial order should be dynamically adjusted according to the complexity of the signal. The higher the polynomial order, the stronger the fitting ability, but the higher the computational complexity, and it may lead to overfitting. In this embodiment, the window size and polynomial order of the SG filter corresponding to different adaptive time periods are determined based on the smoothness of the adaptive time period, which can realize adaptive smoothing processing of signals with different smoothness levels. This method can not only improve the smoothing effect of the signal, but also reduce signal distortion and improve the signal quality and stability.

[0040] In this embodiment of the invention, when dynamically filtering an adaptive time period set based on the BEADS algorithm combined with the SG filtering method, the BEADS algorithm can accurately separate the baseline component and noise component in the spectral signal through differential operators and iterative updates, thereby reducing signal distortion. Combined with the SG filtering method, the parameters can be dynamically adjusted according to the smoothness of the adaptive time period, and smoothing is achieved through local polynomial fitting, reducing the interference of high-frequency noise, while preserving the characteristic peaks of the signal, enhancing signal clarity and effectively reducing noise interference.

[0041] This invention provides a method for coupling a three-dimensional temperature field with an adaptive time period in the field. Figure 5 This diagram illustrates the implementation flow of the method for coupling the on-site three-dimensional temperature field with an adaptive time period. The method specifically includes: S1041, Load the adaptive time period set, extract the associated environmental information in the adaptive time period set. By extracting the associated environmental information in the adaptive time period set, the influence of environmental factors on the spectral signal can be fully considered, thereby improving the accuracy and reliability of signal processing. Based on the associated environmental information, establish a coal field grid model, divide the coal field grid model, and obtain the internal temperature field space and boundary temperature field space of the sub-grid of the coal field grid model. S1042: Obtain the internal and boundary temperature field spaces of the sub-grids of the coal field mesh model. Combining the equivalence principle of velocity gradient and temperature gradient, the dynamically filtered associated temperature parameters are allocated to the internal and boundary temperature field spaces. By integrating the internal and boundary temperature field spaces, the three-dimensional temperature field of the field is obtained. By dividing the coal field mesh model to obtain the internal and boundary temperature field spaces, the distribution characteristics of the temperature field can be analyzed in more detail, improving the modeling accuracy of the temperature field. Combining the equivalence principle of velocity gradient and temperature gradient, the dynamically filtered associated temperature parameters are allocated to the internal and boundary temperature field spaces, which can more accurately reflect the dynamic changes of the temperature field. By integrating the internal and boundary temperature field spaces, the three-dimensional temperature field of the field is obtained, which can comprehensively simulate the temperature distribution of the field and improve the integrity and accuracy of the temperature field modeling. S1043, establish a nonlinear relationship with the associated temperature parameter, the adaptive time period coal quality multi-channel spectral information as the output variable, and the derived enhancement feature as the output variable. The nonlinear relationship includes the temperature correlation coefficient and the spectral correlation coefficient. S1044, based on the particle swarm optimization algorithm, adjusts the weight coefficients of temperature correlation coefficient and spectral correlation coefficient. By introducing Gaussian perturbation and amplitude change rate, the optimization process can avoid getting trapped in local optima and improve the globality of optimization. Based on the particle swarm optimization algorithm, Gaussian perturbation and amplitude change rate are combined with the weight coefficients of temperature correlation coefficient and spectral correlation coefficient to optimize. S1045, using iterative computation, the Bayesian information criterion of the nonlinear relationship is minimized. Based on the minimization of the Bayesian information criterion, the combination of temperature correlation coefficients and spectral correlation coefficients that meet the convergence condition is determined to obtain derived enhancement features. The derived enhancement features, the on-site three-dimensional temperature field, and the adaptive time period are merged to obtain the signal enhancement set. It should be noted that derived enhancement features refer to a set of new features generated by processing and analyzing the original data. These features can better reflect the inherent structure and pattern of the data, thereby improving the performance and predictive ability of the model. In this embodiment, the derived enhancement features can be temperature-compensated spectral features, temperature anti-interference features, multi-sensor collaborative features, and adversarial generation features.

[0042] In this embodiment of the invention, by coupling the three-dimensional temperature field of the site with an adaptive time period, the influence of environmental factors on the spectral signal can be fully considered, thereby improving the accuracy and reliability of signal processing. Furthermore, by accurately establishing a grid model of the coal site, the modeling accuracy of the three-dimensional temperature field is improved. Combining the equivalence principle of velocity gradient and temperature gradient, the accurate allocation of associated temperature parameters is achieved, ultimately comprehensively simulating the temperature distribution of the site, reducing the interference of temperature changes on the spectral signal, and enhancing the stability of the signal. Moreover, by minimizing the determined temperature correlation coefficient and spectral correlation coefficient combination through the Bayesian information criterion, derived enhancement features are generated, further enhancing the feature expression capability of the signal, providing richer feature information for subsequent signal enhancement and coal quality index detection, and improving the accuracy and reliability of detection.

[0043] In this embodiment, when pre-constructing the indicator recognition model based on support vector machines, the indicator recognition model is based on the support vector machine model. The indicator recognition model also includes an input layer and an output layer. A spatiotemporal feature extraction operator layer is set between the input layer and the support vector machine. The spatiotemporal feature extraction operator layer introduces a singular value decomposition (SVD) algorithm. This spatiotemporal feature extraction operator layer, combined with the SVD algorithm, extracts features from the signal enhancement set, obtaining a feature combination set including at least one set of indicator features. For each indicator feature within the feature combination set, the mutual information between it and other indicator features is calculated. The mutual information between indicator features is used as the distance between the indicator features, determining the M nearest indicator features. The feature weight update probability of the M indicator features is determined based on the Relief algorithm. The weights of the M indicator features are updated based on the feature weight update probability, obtaining the initial feature set after weight update. An improved random forest algorithm is introduced into the support vector machine model. This improved random forest algorithm introduces a Chebyshev chaotic mapping network and a maximum information coefficient on the basis of the random forest algorithm. The random forest algorithm uses a Chebyshev chaotic mapping network for the initial feature set. The Chebyshev chaotic mapping network maps the initial feature set to weights of 0 to 1 through an approximate impulse function. Based on a preset weight selection threshold, redundant indicator features are eliminated. Indicator features are extracted from the initial feature set after eliminating redundant indicator features, and at least one indicator decision tree is constructed. The indicator decision tree is weighted based on its error value and maximum information coefficient, and the weighted value of the indicator decision tree is used as the indicator feature importance. A multi-channel weighted feature map combining feature importance is constructed. The support vector machine model sets spectral channel RBF kernel function, environmental channel linear kernel function, and derived channel polynomial kernel function for the multi-channel weighted feature map. The spectral channel RBF kernel function, environmental channel linear kernel function, and derived channel polynomial kernel function are weighted and fused based on an attention mechanism. The industrial analysis indicators of the multi-channel weighted feature map are predicted by a parallel SVM prediction head. The industrial analysis indicators are used as the initial indicator set output, where the industrial analysis indicators are coal ash content, volatile matter, and sulfur content.

[0044] The indicator recognition model training method includes: S201, Generate at least one set of modeling samples based on PSASP simulation software, and perturb the modeling samples to obtain a perturbation-processed modeling sample set; S202, Load the pre-built indicator recognition model, preset the training rounds and hyperparameters of the indicator recognition model, and perform ablation processing on the pre-built indicator recognition model; S203, perform signal enhancement preprocessing on the modeling sample set, and train the indicator recognition model using the modeling sample set to obtain the training output results; S204, calculate the mean squared error between the training output and the actual result; S205, determine whether the mean square error is less than the preset difference threshold; S206, If the mean square error is less than the preset difference threshold, output the converged index identification model. If the mean squared error is not less than the preset difference threshold, the hyperparameters of the index identification model are adjusted by the Adam optimizer, and the process returns to S203 to continue iterative training of the model.

[0045] This invention provides a method for outputting an initial index set by combining an index recognition model with a multi-channel weighted feature map. Figure 6 This diagram illustrates the implementation process of a method for outputting an initial indicator set using an indicator recognition model combined with a multi-channel weighted feature map. Specifically, this method includes: S301, extract features from the signal enhancement set to obtain a feature combination set including at least one set of index features, and calculate the mutual information between the index features in the feature combination set and other index features. S302, using the mutual information between indicator features as the distance between indicator features, determine the M indicator features that are closest to the indicator features, determine the feature weight update probability of the M indicator features based on the Relief algorithm, update the weights of the M indicator features based on the feature weight update probability, and obtain the initial feature set after weight update. S303 uses a Chebyshev chaotic mapping network on the initial feature set based on the improved random forest algorithm. The Chebyshev chaotic mapping network maps the initial feature set to weights from 0 to 1 through an approximate impulse function, and eliminates redundant index features based on the preset weights and screening thresholds. S304, extract indicator features from the initial feature set after eliminating redundant indicator features, and construct at least one indicator decision tree. Based on the error value and maximum information coefficient of the indicator decision tree, weight the indicator feature and use the error value and maximum information coefficient of the indicator decision tree as the indicator feature importance to construct a multi-channel weighted feature map that combines feature importance. S305 sets up RBF kernel functions for spectral channels, linear kernel functions for environmental channels, and polynomial kernel functions for derived channels for multi-channel weighted feature maps. Based on the attention mechanism, it weights and fuses the RBF kernel functions for spectral channels, linear kernel functions for environmental channels, and polynomial kernel functions for derived channels. It then predicts industrial analysis indicators of the multi-channel weighted feature maps through a parallel SVM prediction head, and outputs the industrial analysis indicators as the initial indicator set.

[0046] In this embodiment of the invention, when the index recognition model outputs the initial index set by combining the multi-channel weighted feature map, the index recognition model sets a spatiotemporal feature extraction operator layer between the input layer and the support vector machine. This can extract the temporal and spatial features in the signal, enhancing the model's ability to process complex data. Calculating the mutual information between features and using it as the distance between features can more accurately assess the correlation between features. Based on the Relief algorithm, the feature weight update probability is determined, which can dynamically adjust the feature weights and update the features, enhancing the contribution of important features and reducing the impact of redundant features. Finally, the support vector machine model sets a spectral channel RBF kernel function, an environmental channel linear kernel function, and a derived channel polynomial kernel function for the multi-channel weighted feature map, which can better handle different types of features. Furthermore, based on the attention mechanism, the kernel functions of different channels are weighted and fused, which can dynamically adjust the contribution of each channel, improving the model's adaptability and predictive ability.

[0047] In this embodiment of the invention, the method for collaborative correction of indicators based on the drift risk of the initial indicator set and the initial indicator set includes: S401, based on the multi-head attention mechanism combined with the exponential decay model, calculates the event decay weights of industrial analysis indicators in the initial indicator set; S402 constructs a time-space-environment attention architecture through a multi-head attention mechanism, integrates the event decay weights of industrial analysis indicators with the time-space-environment attention architecture, and calculates the detection confidence of industrial analysis indicators through attention weight aggregation. S403 uses the detection confidence level of industrial analysis indicators as a constraint to construct a partial least squares regression function based on a drift compensation strategy. The partial least squares regression function introduces a GAN architecture to generate drift risk compensation values. Based on the smoothing correction method, the initial indicator set is collaboratively corrected in conjunction with the drift risk.

[0048] On the other hand, embodiments of the present invention also provide a multi-channel online coal quality detection system based on signal enhancement. Figure 7 A schematic diagram of the architecture of a multi-channel online coal quality detection system based on signal enhancement is shown. The system specifically includes: The signal enhancement module 100 collects multi-channel spectral information of coal quality and related environmental information in real time based on the coal quality detection equipment, merges the multi-channel spectral information of coal quality and related environmental information into a real-time information set, performs signal enhancement preprocessing on the real-time information set, and outputs a signal enhancement set. The signal enhancement module 100 includes: The signal cleaning unit 110 is used to load the real-time information set, clean and remove outliers from the real-time information set based on the boundary mixing sampling strategy, obtain the real-time information set after outlier cleaning and removal, and perform normalization processing on the real-time information set. The adaptive truncation unit 120 is used to acquire the normalized real-time information set, identify the signal bands of the real-time information set based on the spectral analysis method, determine the peak characteristics of the signal bands, and adaptively truncate the coal quality multi-channel spectral information in the real-time information set according to the peak characteristics of the signal bands. The time stamp alignment method is used to align the associated environmental information with multiple sets of the adaptive time periods to obtain the adaptive time period set. The dynamic filtering unit 130 dynamically filters the adaptive time period set based on the BEADS algorithm combined with the SG filtering method, and outputs the dynamically filtered adaptive time period set. The signal coupling unit 140 is used to load an adaptive time period set, construct a three-dimensional temperature field based on the adaptive time period set, couple the three-dimensional temperature field with the adaptive time period to obtain derived enhancement features, and merge the derived enhancement features, the three-dimensional temperature field, and the adaptive time period to obtain a signal enhancement set.

[0049] The adaptive truncation unit 120 includes: The spectral analysis module 121 is used to acquire the normalized real-time information set, identify the signal bands of the real-time information set based on the spectral analysis method, and determine the peak characteristics of the signal bands. The time period extraction module 122 is used to adaptively extract the coal quality multi-channel spectral information from the real-time information set by combining the peak characteristics of the signal band. The time period alignment module 123 uses the timestamp alignment method to align the associated environmental information with multiple sets of adaptive time periods to obtain an adaptive time period set.

[0050] The multi-channel analysis module 200 is used to pre-build an index recognition model based on support vector machine, iteratively train the signal enhancement set until convergence, load the signal enhancement set, perform initial feature selection on the signal enhancement set based on the index recognition model, determine the feature importance through an improved random forest algorithm, construct a multi-channel weighted feature map combining feature importance, and the index recognition model outputs the initial index set by combining the multi-channel weighted feature map. The index correction module 300 is used to obtain an initial index set, determine the detection confidence of the initial index set based on a multi-head attention mechanism, dynamically evaluate the drift risk of the initial index set, perform collaborative index correction based on the drift risk of the initial index set and the initial index set, and output the coal quality index detection results after collaborative correction.

[0051] In summary, this invention provides a multi-channel online coal quality detection system and method based on signal enhancement. In the embodiments of this invention, signal enhancement preprocessing is performed on the real-time acquired multi-channel spectral information of coal quality and associated environmental information. The interference of environmental factors on the spectral signal is considered during the signal enhancement preprocessing. By coupling the on-site three-dimensional temperature field with an adaptive time period, derived enhancement features are obtained, which can effectively compensate for the interference of environmental factors on the signal. The signal after environmental factor compensation is more stable, reducing signal drift and distortion caused by environmental changes, thereby improving the stability and accuracy of the signal.

[0052] It should be noted that, for the sake of simplicity, the foregoing embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to the present invention. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0053] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art can still combine, add, delete, or otherwise adjust the features of the various embodiments of the present invention according to the circumstances without conflict or creative effort, thereby obtaining different technical solutions that do not fundamentally depart from the concept of the present invention. These technical solutions also fall within the scope of protection of the present invention.

Claims

1. A multi-channel online coal quality detection method based on signal enhancement, characterized in that, The method includes: Based on the real-time acquisition of multi-channel spectral information and associated environmental information of coal by coal quality testing equipment, the multi-channel spectral information and associated environmental information of coal quality are merged into a real-time information set, and the real-time information set is preprocessed for signal enhancement to output a signal enhancement set; A support vector machine-based index recognition model is pre-built, and the signal enhancement set is iteratively trained until convergence. The signal enhancement set is loaded, and initial feature selection is performed on the signal enhancement set based on the index recognition model. The feature importance is determined by improving the random forest algorithm, and a multi-channel weighted feature map combining the feature importance is constructed. The index recognition model outputs the initial index set by combining the multi-channel weighted feature map. An initial indicator set is obtained, and the detection confidence of the initial indicator set is determined based on a multi-head attention mechanism. The drift risk of the initial indicator set is dynamically evaluated, and the indicators are collaboratively corrected based on the drift risk of the initial indicator set and the initial indicator set. The coal quality indicator detection results after collaborative correction are output.

2. The multi-channel online coal quality detection method based on signal enhancement as described in claim 1, characterized in that: The method for signal enhancement preprocessing of real-time information sets includes: Load the real-time information set, perform outlier cleaning and removal on the real-time information set based on the boundary hybrid sampling strategy, obtain the real-time information set after outlier cleaning and removal, and perform normalization processing on the real-time information set. The normalized real-time information set is obtained, the signal bands of the real-time information set are identified based on the spectral analysis method, the peak characteristics of the signal bands are determined, and the coal quality multi-channel spectral information in the real-time information set is adaptively extracted based on the peak characteristics of the signal bands. The time stamp alignment method is used to align the associated environmental information with multiple sets of the adaptive time periods to obtain the adaptive time period set. The adaptive time period set is dynamically filtered based on the BEADS algorithm combined with the SG filtering method, and the dynamically filtered adaptive time period set is output. Load the adaptive time period set, construct the on-site three-dimensional temperature field based on the adaptive time period set, couple the on-site three-dimensional temperature field with the adaptive time period to obtain the derived enhancement features, and merge the derived enhancement features, the on-site three-dimensional temperature field and the adaptive time period to obtain the signal enhancement set.

3. The multi-channel online coal quality detection method based on signal enhancement as described in claim 2, characterized in that: The method for outlier cleaning and removal from real-time information sets based on a boundary hybrid sampling strategy includes: Acquire a real-time information set and use a boundary point detection algorithm to determine the coefficient of variation of data points in the real-time information set; Based on the coefficient of variation of data points, a variation threshold is set for the real-time information set. Data points with a coefficient of variation greater than the variation threshold in the real-time information set are classified as boundary data points, and the remaining data points are classified as center data points. Load boundary data points, preset the oversampling factor based on the weight of the boundary data points, and use synthetic minority oversampling technology to copy the boundary data points to obtain new boundary data points after copying. Load the center data points, preset the undersampling rate based on the density of the center data points, traverse the center data points, and use the Tomek-Links method combined with the undersampling rate to remove and clean the center data points to obtain the reconstructed center data. The copied boundary data points and the reconstructed center data are obtained and integrated into a real-time information set after outlier cleaning and removal. Load the real-time information set after outlier cleaning and removal, and normalize the real-time information set.

4. The multi-channel online coal quality detection method based on signal enhancement as described in claim 3, characterized in that: The adaptive time-time set dynamic filtering method based on the BEADS algorithm combined with the SG filtering method includes: Load the adaptive time period set, initialize the parameters of the BEADS algorithm, and set the number of iterations; The differential operator of coal quality multi-channel spectral information in the adaptive time period is calculated based on the BEADS algorithm and the spectral information estimate is updated. The baseline component and noise component of coal quality multi-channel spectral information in the adaptive time period are determined after a preset number of iterations. The estimated baseline component and noise component are subtracted from the original signal to obtain the denoised adaptive time period; Based on the smoothness of the adaptive time period, the window size and polynomial order of the SG filter corresponding to different adaptive time periods are determined. The SG filtering method achieves smoothing of coal quality multi-channel spectral information in the adaptive time period through local polynomial fitting, and outputs the dynamically filtered adaptive time period set.

5. The multi-channel online coal quality detection method based on signal enhancement as described in claim 4, characterized in that: The method for coupling the on-site three-dimensional temperature field with the adaptive time period includes: Load the adaptive time period set, extract the associated environmental information from the adaptive time period set, establish a coal field grid model based on the associated environmental information, divide the coal field grid model into sub-grids, and obtain the internal temperature field space and boundary temperature field space of the sub-grids of the coal field grid model. The internal temperature field space and boundary temperature field space of the sub-grid of the coal field grid model are obtained. Based on the equivalent principle of velocity gradient and temperature gradient, the dynamically filtered associated temperature parameters are allocated to the internal temperature field space and boundary temperature field space. The internal temperature field space and boundary temperature field space are integrated to obtain the three-dimensional temperature field of the field. A nonlinear relationship is established with the associated temperature parameter, the adaptive time period coal quality multi-channel spectral information as the output variable, and the derived enhancement feature as the output variable. The nonlinear relationship includes the temperature correlation coefficient and the spectral correlation coefficient. The weight coefficients of temperature correlation coefficient and spectral correlation coefficient are adjusted based on particle swarm optimization algorithm, and Gaussian perturbation and amplitude change rate are introduced on the basis of particle swarm optimization algorithm to optimize the weight coefficients of temperature correlation coefficient and spectral correlation coefficient. Iterative computation is used to minimize the Bayesian information criterion of the nonlinear relationship. Based on the minimization of the Bayesian information criterion, the combination of temperature correlation coefficient and spectral correlation coefficient that meet the convergence condition is determined to obtain the derived enhancement features. The derived enhancement features, the on-site three-dimensional temperature field, and the adaptive time period are merged to obtain the signal enhancement set.

6. The multi-channel online coal quality detection method based on signal enhancement as described in claim 1, characterized in that: When pre-constructing an indicator recognition model based on support vector machines, the indicator recognition model is based on the support vector machine model. The indicator recognition model also includes an input layer and an output layer. A spatiotemporal feature extraction operator layer is set between the input layer and the support vector machine. The spatiotemporal feature extraction operator layer introduces the singular value decomposition algorithm. An improved random forest algorithm is introduced into the support vector machine model. The industrial analysis indicators are predicted by the parallel SVM prediction head using multi-channel weighted feature maps. The industrial analysis indicators are used as the initial indicator set output.

7. The multi-channel online coal quality detection method based on signal enhancement as described in claim 6, characterized in that: The method for outputting an initial indicator set by combining the indicator recognition model with multi-channel weighted feature maps includes: Feature extraction is performed on the signal enhancement set to obtain a feature combination set including at least one set of index features. For the index features in the feature combination set, the mutual information between them and other index features is calculated. Using the mutual information between indicator features as the distance between indicator features, the M indicator features closest to the indicator features are determined. The feature weight update probability of the M indicator features is determined based on the Relief algorithm. The weights of the M indicator features are updated based on the feature weight update probability to obtain the initial feature set after weight update. Based on the improved random forest algorithm, a Chebyshev chaotic mapping network is used for the initial feature set. The Chebyshev chaotic mapping network maps the initial feature set to weights of 0 to 1 through an approximate impulse function, and eliminates redundant index features based on the preset weights and screening thresholds. The initial feature set, after eliminating redundant indicator features, is used to extract indicator features and construct at least one indicator decision tree. Based on the error value and maximum information coefficient of the indicator decision tree, a weighted feature map combining the indicator feature importance is constructed. For multi-channel weighted feature maps, spectral channel RBF kernel function, environmental channel linear kernel function, and derived channel polynomial kernel function are set. Based on the attention mechanism, the spectral channel RBF kernel function, environmental channel linear kernel function, and derived channel polynomial kernel function are weighted and fused. Then, the industrial analysis index of the multi-channel weighted feature map is predicted by the parallel SVM prediction head, and the industrial analysis index is used as the initial index set output.

8. A multi-channel online coal quality detection system based on signal enhancement, used to implement the multi-channel online coal quality detection method based on signal enhancement as described in any one of claims 1-7, characterized in that: The signal enhancement-based multi-channel online coal quality detection system includes: The signal enhancement module is based on the real-time acquisition of multi-channel spectral information of coal quality and related environmental information by coal quality detection equipment. It merges the multi-channel spectral information of coal quality and related environmental information into a real-time information set, performs signal enhancement preprocessing on the real-time information set, and outputs the signal enhancement set. The multi-channel analysis module is used to pre-build an index recognition model based on support vector machine, iteratively train the signal enhancement set until convergence, load the signal enhancement set, perform initial feature selection on the signal enhancement set based on the index recognition model, determine the feature importance through an improved random forest algorithm, construct a multi-channel weighted feature map combining feature importance, and the index recognition model outputs the initial index set by combining the multi-channel weighted feature map. The index correction module is used to obtain an initial index set, determine the detection confidence of the initial index set based on a multi-head attention mechanism, dynamically evaluate the drift risk of the initial index set, perform collaborative index correction based on the drift risk of the initial index set and the initial index set, and output the coal quality index detection results after collaborative correction.

9. The multi-channel online coal quality detection system based on signal enhancement as described in claim 8, characterized in that: The signal enhancement module includes: The signal cleaning unit is used to load the real-time information set, clean and remove outliers from the real-time information set based on the boundary mixing sampling strategy, obtain the real-time information set after outlier cleaning and removal, and perform normalization processing on the real-time information set. An adaptive truncation unit is used to acquire a normalized real-time information set, identify the signal bands of the real-time information set based on spectral analysis, determine the peak characteristics of the signal bands, and adaptively truncate the coal quality multi-channel spectral information in the real-time information set based on the peak characteristics of the signal bands. The unit also uses a timestamp alignment method to align the associated environmental information with multiple sets of adaptive time periods to obtain an adaptive time period set. The dynamic filtering unit uses the BEADS algorithm combined with the SG filtering method to dynamically filter the adaptive time period set and outputs the dynamically filtered adaptive time period set. The signal coupling unit is used to load the adaptive time period set, construct the on-site three-dimensional temperature field based on the adaptive time period set, couple the on-site three-dimensional temperature field with the adaptive time period to obtain the derived enhancement features, and merge the derived enhancement features, the on-site three-dimensional temperature field and the adaptive time period to obtain the signal enhancement set.

10. The multi-channel online coal quality detection system based on signal enhancement as described in claim 9, characterized in that: The adaptive truncation unit includes: The spectral analysis module is used to acquire the normalized real-time information set, identify the signal bands of the real-time information set based on spectral analysis, and determine the peak characteristics of the signal bands. The time segmentation module is used to adaptively segment the multi-channel spectral information of coal quality in real-time information set by combining the peak characteristics of the signal band. The time period alignment module uses a timestamp alignment method to align the associated environmental information with multiple sets of adaptive time periods to obtain an adaptive time period set.

Images