Coal detection system, method and storage medium

Abstract

The embodiment of the present invention provides a coal detection system, method and storage medium, which belongs to the field of coal detection technology. The system includes: a detection unit, which is used to collect spectral data of coal samples for detection; wherein the detection unit includes a near-infrared spectrum acquisition module and an X-ray fluorescence spectrum acquisition module; a training unit, which is used to construct a neural network based on greedy search in each neural network dimension through the spectral data, and obtain a coal sample detection model based on simulated sample training after sample augmentation; an analysis unit, which is used to perform fusion processing on the spectral data to obtain target spectral data, perform target spectral data inference based on the coal sample detection model, and obtain coal detection results. The scheme of the present invention performs coal detection based on dual-spectrum fusion information, and performs adaptive search of neural network structure based on the collected spectral data, thereby ensuring the accuracy of coal detection.

Description

Technical Field

[0001] The present invention relates to the technical field of coal detection, and in particular to a coal detection system, a coal detection method and a storage medium. Background Art

[0002] Online coal inspection is an important means of coal inspection. It has the significant advantages of real-time high efficiency, continuous operation around the clock, and assisting digital management. The main difficulty of online coal inspection is how to simultaneously achieve high-precision detection of multiple indicators such as coal calorific value, ash content, sulfur content and moisture content under complex and changeable working conditions. Existing online coal component analysis methods include near-infrared spectroscopy (NIRS), laser-induced breakdown spectroscopy, X-ray fluorescence spectrometry (XRF), neutron activation technology, machine vision imaging, etc. The above technologies can only achieve effective detection of some or a single indicator. To achieve comprehensive coal inspection, it is necessary to perform various tests at a single speed, which not only leads to the need for a large number of repeated operations, such as sample preparation, but also leads to the problem of low accuracy of test results due to differences in test standards between various devices.

[0003] There is a common situation in the current coal monitoring scheme, that is, in terms of coal sample preparation, transportation and inspection schemes, as well as coal sample index monitoring schemes, each link is carried out independently. This decentralized execution method leads to inconsistency in sample status and makes it impossible to achieve unified standard control. Especially in online coal detection, factors such as coal particle size and surface flatness have a significant impact on the stability of the detection system. This independently executed monitoring scheme leads to the phenomenon of information islands in the monitoring process. The executors of each link only focus on the completion of their own tasks and lack comprehensive control and coordination of the overall monitoring process. Due to the lack of unified standard control, there may be execution deviations in different links, which affects the reliability and consistency of the monitoring results. In online coal detection, factors such as coal particle size and surface flatness are crucial to the stability of the detection system. The size of coal particles not only affects the combustion characteristics and calorific value, but also directly affects the accuracy of the detection equipment. Surface flatness affects the detection effect of optical sensors. Uneven surfaces may cause uneven light reflection and affect the accuracy of data collection. Therefore, understanding the characteristics and surface flatness of coal particles is crucial to optimizing the monitoring system. Only by fully understanding the physical properties and surface morphology of coal samples can we effectively adjust the parameters and methods of monitoring equipment and improve the stability and accuracy of the monitoring system. At the same time, establishing a unified standard management and control system to ensure coordination and information sharing in all links is the key to improving the efficiency and accuracy of coal monitoring.

[0004] In the existing scheme, patent CN114112976A "XRF-NIRS combined coal calorific value high repeatability detection method" discloses XRF NIRS combined coal calorific value high repeatability detection method, but the scheme does not mention the solution to overcome the coal particles and surface flatness in the online coal detection, which will have a great impact on the stability of the XRF system and NIRS system, and it is still a single indicator (coal calorific value) test scheme, and it is still unable to achieve comprehensive integrated detection of multiple indicators. It can be seen that this mutually isolated detection method is bound to cause the problem of inability to guarantee detection accuracy and low detection efficiency. To address this problem, a new coal detection scheme needs to be proposed. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a coal detection system, method and storage medium to at least solve the problems of the existing coal detection schemes that the detection accuracy cannot be guaranteed and the detection efficiency is low.

[0006] In order to achieve the above-mentioned purpose, the first aspect of the present invention provides a coal detection system, which includes: a detection unit, which is used to collect spectral data for detecting coal samples; wherein the detection unit includes a near-infrared spectrum acquisition module and an X-ray fluorescence spectrum acquisition module; a training unit, which is used to construct a neural network based on greedy search in each neural network dimension through the spectral data, and obtain a coal sample detection model based on simulated sample training after sample augmentation; an analysis unit, which is used to perform fusion processing on the spectral data to obtain target spectral data, and perform target spectral data inference based on the coal sample detection model to obtain coal detection results.

[0007] Optionally, the system further comprises a sample preparation unit for collecting coal samples and performing sample preparation processing on the coal samples to obtain test coal samples.

[0008] Optionally, the sample preparation unit includes: a sampling module, used to randomly collect raw coal samples during the transportation or storage of raw coal, or to receive raw coal samples deposited by users; a crushing module, used to crush the raw coal samples to obtain crushed coal samples; a processing module, used to pre-process the crushed coal samples to obtain basic coal samples; wherein the pre-processing of the crushed coal samples includes: drying, grinding and screening; a shaping module, used to shape the basic coal samples to obtain test coal samples.

[0009] Optionally, the sampling module, the crushing module, the processing module and the shaping module are connected based on a transfer conveyor belt; the transfer conveyor belt is triggered and controlled based on a servo system.

[0010] Optionally, the servo system is configured to: start timing based on the trigger signal of the position trigger at the preset position of each module of the sample preparation unit, and control the power servo motor of the transfer conveyor belt to start, until it runs for a predetermined time, and then turn off the power servo motor of the transfer conveyor belt; or start timing in response to the switch trigger signal corresponding to each module of the sample preparation unit, and control the power servo motor of the transfer conveyor belt to start, until it runs for a predetermined time, and then turn off the power servo motor of the transfer conveyor belt.

[0011] Optionally, the shaping module includes: a coal conveying component, including a conveying frame and a conveyor belt rotatably arranged on the conveying frame, the conveying frame being provided with an inlet for detecting coal samples; a first pretreatment component, including a limit frame and a scraper, the limit frame is supported above the conveyor belt by the conveying frame, and the limit frame is located below the inlet for detecting coal samples, the scraper is arranged at the discharge end of the limit frame and forms a first gap with the conveyor belt; a second pretreatment component, including at least one pressure roller, the pressure roller is rotatably arranged on the conveying frame, and the pressure roller is located at the outlet end of the limit frame, a second gap connected to the first gap is formed between the pressure roller and the conveyor belt, and the pressure roller is provided with a limit member for limiting the width of the second gap.

[0012] Optionally, the detection unit also includes: a sampling hood for setting the near-infrared spectrum acquisition module, the X-ray fluorescence spectrum acquisition module and the visual analysis module; a lighting module, an optical fiber and a photoelectric sensor are also arranged inside the sampling hood; an aperture extending from the inner wall of the sampling hood is arranged between the lighting module and the optical fiber; the photoelectric sensor is arranged in the light path direction of the lighting module, and is used to monitor the light intensity of the corresponding lighting module.

[0013] Optionally, the lighting module includes a plurality of symmetrically arranged light sources; the optical fiber is arranged at a central position on the top of the sampling cover; and each lighting module is provided with at least one corresponding photoelectric sensor, which is arranged on the inner wall of the sampling cover.

[0014] Optionally, the analysis unit is also used to monitor the status of the corresponding lighting module based on the light intensity collected by the photoelectric sensor, including: performing preprocessing including filtering and denoising on the light data collected by the photoelectric sensor; performing interpolation or fitting calculations based on the voltage value or digital value of the electrical signal corresponding to the preprocessed light data and a calibration curve between the light intensity to obtain the light intensity at the detection position; performing light intensity fitting of the corresponding lighting module based on the light intensity at the detection position and the positional relationship between the light point sensor and the corresponding lighting module to obtain the detected light intensity of the corresponding lighting module; comparing the detected light intensity with a preset light intensity threshold, and outputting an alarm message if the detected light intensity is less than the preset light intensity threshold.

[0015] Optionally, the detection unit also includes: a Mylar film replacement device, including a sample conveyor belt, a test coal sample is placed on the sample conveyor belt, a mounting frame is provided above the test coal sample, a driving wheel and a driven wheel are provided on the mounting frame, a Mylar film transmission channel is formed on the mounting frame, the Mylar film can pass through the Mylar film transmission channel and abut against the lower edge of the outer circumferential surface of the driving wheel and the driven wheel, and move under the drive of the driving wheel.

[0016] Optionally, the system also includes a visual analysis module, which includes one or more image acquisition modules, each image acquisition module is arranged at each preset direction corresponding to the detection position of the detected coal sample, and is used to collect image information of the detected coal sample at a viewing angle; the analysis unit is also used to evaluate the appearance of the detected coal sample based on the image information at each viewing angle.

[0017] Optionally, the appearance evaluation of the detected coal sample based on the image information at each viewing angle includes: performing denoising, grayscale and edge detection processing on the image information at each viewing angle in sequence, and calibrating the detection coal sample area; fitting the appearance dimensions of the detected coal sample based on the image detection coal sample area at each viewing angle; collecting surface flatness indicators of the image detection coal sample area at each viewing angle, and fitting the surface flatness of the detected coal sample based on the surface flatness indicators; and evaluating the appearance of the detected coal sample based on the fitted appearance dimensions and the fitted surface flatness.

[0018] Optionally, the appearance evaluation of the detected coal sample is performed based on the fitted outer dimensions and the fitted surface flatness, including: calculating the Euclidean distance between the fitted outer dimensions and the preset outer dimensions as the first deviation value; calculating the Euclidean distance between the fitted surface flatness and the preset surface flatness as the second deviation value; performing normalization processing on the first deviation value and the second deviation value, and performing arithmetic mean calculation on the normalized first deviation value and the second deviation value to obtain a deviation coefficient; and assigning a preset correction coefficient in the coal sample detection model based on the deviation coefficient.

[0019] Optionally, the spectral data includes near-infrared spectral signals and X-ray fluorescence spectral data; performing fusion processing on the spectral data to obtain target spectral data includes: performing fusion processing on the near-infrared spectral signals collected by the near-infrared spectral acquisition module and the X-ray fluorescence spectral data collected by the X-ray fluorescence spectral acquisition module to obtain target spectral data.

[0020] Optionally, a fusion process is performed on the near-infrared spectrum signal collected by the near-infrared spectrum acquisition module and the X-ray fluorescence spectrum data collected by the X-ray fluorescence spectrum acquisition module to obtain target spectrum data, including: preprocessing the near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively; downsampling the preprocessed near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively, and scaling the downsampled near-infrared spectrum signal and the X-ray fluorescence spectrum data to a normal distribution; and splicing the normally distributed near-infrared spectrum signal and the normally distributed X-ray fluorescence spectrum data to obtain the target spectrum data.

[0021] Optionally, the preprocessing of the near-infrared spectral signal and the X-ray fluorescence spectral data respectively includes: performing SG convolution smoothing processing on the near-infrared spectral signal and the X-ray fluorescence spectral data respectively to obtain the near-infrared spectral signal after SG convolution smoothing processing and the X-ray fluorescence spectral data after SG convolution smoothing processing; performing area normalization processing on the near-infrared spectral signal after SG convolution smoothing processing.

[0022] Optionally, the SG convolution smoothing processing is performed on the near-infrared spectral signal and the X-ray fluorescence spectral data respectively, including: calculating the coefficients of the SG convolution kernel based on a preset window size and polynomial order; performing weighted averaging of adjacent data points in each spectral signal based on the coefficients of the SG convolution kernel for the near-infrared spectral signal and the X-ray fluorescence spectral data respectively to obtain smoothed data points; performing symmetric expansion or zero filling processing on the boundaries of each spectral signal; and obtaining the spectral signal after SG convolution smoothing processing based on the smoothed data points and the processed boundaries.

[0023] Optionally, the downsampling processing is performed on the preprocessed near-infrared spectral signal and X-ray fluorescence spectral data respectively, including: downsampling processing is performed on the X-ray fluorescence spectral data after SG convolution smoothing processing and the near-infrared spectral signal after area normalization processing respectively, including: filtering processing is performed on each spectral signal respectively, and in the spectral signal after filtering processing, one sampling point is reserved at a fixed interval to obtain multiple sampling points; or, the spectral signal after filtering processing is intercepted into multiple segments, and averaging processing is performed on each sampling point in each segment, and one sampling point is obtained in each segment to obtain multiple sampling points; signal reconstruction is performed based on each sampling point to obtain the spectral signal after downsampling processing.

[0024] Optionally, the downsampled near-infrared spectral signals and X-ray fluorescence spectral data are scaled to a normal distribution, including: calculating the statistical characteristics of each spectral signal respectively; wherein the statistical characteristics are variance and / or standard deviation; based on the statistical characteristics of each spectral signal, performing standardization or normalization processing on each spectral signal, scaling the value of each spectral signal to a preset value range, and obtaining the scaled value of each spectral signal; based on a preset transformation algorithm and the scaled value of each spectral signal, converting each spectral signal into a normal distribution.

[0025] Optionally, a splicing process is performed on a normally distributed near-infrared spectral signal and a normally distributed X-ray fluorescence spectral data to obtain target spectral data, including: performing an alignment operation on the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data; after completing the alignment of the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data, splicing the two spectral signals based on weighted average to obtain initial target spectral data; verifying the initial target spectral data based on the normally distributed near-infrared spectral signal and / or the normally distributed X-ray fluorescence spectral data, and using the verified initial target spectral data as the target spectral data.

[0026] Optionally, the pre-trained coal sample detection model is:

[0027] θ=argmax θ L 2 (f(X|θ),Y);

[0028] Among them, f(·|θ) is a deep neural network with θ as parameter; is the target spectrum data; The component of coal.

[0029] Optionally, the system also includes a training unit for performing coal sample detection model training, including: collecting historical near-infrared spectral signals and historical X-ray fluorescence spectral data, and constructing corresponding historical target spectral data based on the historical near-infrared spectral signals and the historical X-ray fluorescence spectral data; using the historical target spectral data as training data to perform PLS model parameter initialization; generating simulated samples based on the initialized PLS model, performing model training in a pre-built neural network based on the simulated samples, and obtaining an initial model for coal sample detection; and verifying the initial model for coal sample detection based on the reserved historical target spectral data to obtain a coal sample detection model.

[0030] Optionally, the method of generating simulated samples based on the initialized PLS model includes: adaptively determining a signal type and signal parameters that match the target spectral data, and generating a basic signal based on the signal type and the model parameters; adaptively selecting an augmentation scheme, and performing a data augmentation operation on the basic signal based on the selected augmentation scheme to obtain an augmented signal set; screening out deviation data in the augmented signal set, and using the screened-out augmented signal set as the simulated sample.

[0031] Optionally, the method of generating simulated samples based on the initialized PLS model includes: adaptively determining a signal type and signal parameters that match the target spectral data, and generating a basic signal based on the signal type and the model parameters; adaptively selecting an augmentation scheme, and performing a data augmentation operation on the basic signal based on the selected augmentation scheme to obtain an augmented signal set; screening out deviation data in the augmented signal set, and using the screened-out augmented signal set as the simulated sample.

[0032] Optionally, performing model training based on the simulated samples to obtain an initial model for coal sample detection includes: annotating the simulated samples with pseudo labels based on the PLS model after parameter initialization, and using the annotated simulated samples as training samples; based on the training samples, performing model training in a neural network based on the target spectral data search to obtain an initial model for coal sample detection.

[0033] Optionally, the coal sample detection initial model is verified based on the reserved historical target spectral data to obtain the coal sample detection model, including: constructing a verification set based on the reserved historical target spectral data, and inputting the verification set into the component monitoring initial model to obtain a corresponding prediction result; performing a performance evaluation of the component monitoring initial model based on the prediction result and the actual result of the corresponding reserved historical target spectral data; if the performance evaluation of the component monitoring initial model fails, adjusting the hyperparameters, optimizing the model structure and / or adding regularization operations to obtain an updated component monitoring initial model; re-performing a performance evaluation on the updated component monitoring initial model based on the reserved historical target spectral data until a component monitoring initial model that meets the preset performance requirements is obtained as the coal sample detection model.

[0034] Optionally, the performance evaluation of the initial component monitoring model is performed based on the predicted results and the actual results of the corresponding reserved historical target spectral data, including: comparing the predicted results with the actual results, and evaluating the accuracy and recall of the model based on their deviations; if any evaluation result of the accuracy and recall fails, the performance evaluation of the initial component monitoring model fails.

[0035] Optionally, the training unit is also used to perform a neural network search based on the spectral data, including: dividing a plurality of optimization dimensions based on the structural parameters of the neural network; in each optimization dimension, adaptively adjusting the parameters in the neural network structure corresponding to the optimization dimension, and calculating the MAE index after each adjustment; comparing the MAE index corresponding to each parameter, and determining the parameter combination with the highest MAE index as the optimization result in the corresponding optimization dimension; performing a greedy search between the optimization dimensions to obtain the optimization results of each optimization dimension; based on the optimization results of each optimization dimension, determining the structural parameters corresponding to each neural network structure, and constructing a neural network corresponding to the search result.

[0036] Optionally, the neural network dimensions include: a basic operator dimension of a neural network layer, a resolution dimension of input data, a network depth dimension, and a network width dimension.

[0037] Optionally, within the basic operator dimension of the neural network layer, the corresponding optimization result obtaining rules include: traversing the convolution type of the linear layer, and traversing the activation function type of the nonlinear layer, adaptively combining the convolution type of the fully connected layer and the activation function type of the nonlinear layer, and calculating the MAE index after each combination; comparing the MAE index after each combination, selecting the combination corresponding to the maximum MAE index, and determining the convolution type of the fully connected layer and the activation function type of the nonlinear layer corresponding to the combination as the optimization result within the basic operator dimension of the neural network layer.

[0038] Optionally, the convolution type of the linear layer is: a fully connected layer, a 1D convolution layer with a kernel size of 5, a 1D convolution layer with a kernel size of 10, or a 1D convolution layer with a kernel size of 15; the activation function type of the nonlinear layer is: a TanH function, an ELU function, a Sigmoid activation function, or a Softmax function.

[0039] Optionally, the calculation rule of the MAE indicator is:

[0040]

[0041] in, is the true value of the coal composition corresponding to the i-th training data; y i is the model prediction value corresponding to the i-th training data; m is the size of the data set.

[0042] Optionally, the system further includes an output unit for visually outputting the coal detection results.

[0043] Optionally, the test results of the coal sample include: one or more of ash composition, ash content, volatile matter, hydrocarbons, ash melting point, total water, total sulfur and calorific value.

[0044] Optionally, the visual output of the coal detection results includes: determining a corresponding detection result object in response to a user data query instruction; selecting a preset data visualization scheme based on the corresponding detection result object, and pushing the visualized data to the user end.

[0045] A second aspect of the present invention provides a coal detection method, which is applied to the above-mentioned coal detection system, and the method includes: collecting spectral data of coal samples for detection; constructing a neural network based on greedy search in each neural network dimension through the spectral data, and obtaining a coal sample detection model based on simulated sample training after sample augmentation; performing fusion processing on the spectral data to obtain target spectral data, and performing target spectral data inference based on the coal sample detection model to obtain coal detection results.

[0046] Optionally, before executing the spectral data collection of the test coal sample, the method further includes: collecting a raw coal sample, and performing sample preparation processing on the raw coal sample to obtain a test coal sample.

[0047] Optionally, the sample preparation process of the raw coal sample to obtain the test coal sample includes: randomly collecting raw coal samples during the transportation or storage of raw coal, or receiving raw coal samples deposited by users; crushing the raw coal sample to obtain a crushed coal sample; pre-processing the crushed coal sample to obtain a basic coal sample; wherein the pre-processing of the crushed coal sample includes: drying, grinding and screening; shaping the basic coal sample to obtain a test coal sample.

[0048] Optionally, the method also includes: monitoring the status of the corresponding lighting module, including: performing preprocessing including filtering and denoising on the illumination data collected by the photoelectric sensor; wherein each illumination module is provided with at least one photoelectric sensor corresponding to the photoelectric sensor, which is arranged on the inner wall of the sampling cover; interpolating or fitting calculations are performed based on the voltage value or digital value of the electrical signal corresponding to the preprocessed illumination data and a calibration curve between the illumination intensity to obtain the illumination intensity at the detection position; fitting the illumination intensity of the corresponding illumination module based on the illumination intensity at the detection position and the positional relationship between the light point sensor and the corresponding illumination module to obtain the detected illumination intensity of the corresponding illumination module; comparing the detected illumination intensity with a preset illumination intensity threshold, and if the detected illumination intensity is less than the preset illumination intensity threshold, outputting an alarm message.

[0049] Optionally, the method further includes collecting image data of the detected coal sample, and evaluating the appearance of the detected coal sample based on the image data of the detected coal sample.

[0050] Optionally, the appearance evaluation of the detected coal sample based on the image data of the detected coal sample includes: performing denoising, grayscale and edge detection processing on the image information at each viewing angle in turn, and calibrating the detection coal sample area; wherein the image information at each viewing angle is acquired based on image acquisition modules set in each preset direction corresponding to the detection position of the detected coal sample; based on the image detection coal sample area at each viewing angle, the appearance size of the detected coal sample is fitted; surface flatness index is collected in the image detection coal sample area at each viewing angle, and surface flatness fitting of the detected coal sample is performed based on the surface flatness index; and the appearance of the detected coal sample is evaluated based on the fitted appearance size and the fitted surface flatness.

[0051] Optionally, the appearance evaluation of the detected coal sample is performed based on the fitted outer dimensions and the fitted surface flatness, including: calculating the Euclidean distance between the fitted outer dimensions and the preset outer dimensions as the first deviation value; calculating the Euclidean distance between the fitted surface flatness and the preset surface flatness as the second deviation value; performing normalization processing on the first deviation value and the second deviation value, and performing arithmetic mean calculation on the normalized first deviation value and the second deviation value to obtain a deviation coefficient; and assigning a preset correction coefficient in the coal sample detection model based on the deviation coefficient.

[0052] Optionally, the spectral data includes: near-infrared spectral signals and X-ray fluorescence spectral data; the spectral data reasoning is performed based on the pre-trained coal sample detection model to obtain the reasoning result, including: performing fusion processing on the near-infrared spectral signals collected by the near-infrared spectral acquisition module and the X-ray fluorescence spectral data collected by the X-ray fluorescence spectral acquisition module to obtain target spectral data; and performing target spectral data reasoning based on the assigned coal sample detection model to obtain the reasoning result.

[0053] Optionally, a fusion process is performed on the near-infrared spectrum signal collected by the near-infrared spectrum acquisition module and the X-ray fluorescence spectrum data collected by the X-ray fluorescence spectrum acquisition module to obtain target spectrum data, including: preprocessing the near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively; downsampling the preprocessed near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively, and scaling the downsampled near-infrared spectrum signal and the X-ray fluorescence spectrum data to a normal distribution; and splicing the normally distributed near-infrared spectrum signal and the normally distributed X-ray fluorescence spectrum data to obtain the target spectrum data.

[0054] Optionally, the method also includes: training a coal sample detection model, including: collecting historical near-infrared spectral signals and historical X-ray fluorescence spectral data, and constructing corresponding historical target spectral data based on the historical near-infrared spectral signals and the historical X-ray fluorescence spectral data; using the historical target spectral data as training data to perform PLS model parameter initialization; generating simulated samples based on the initialized PLS model, performing model training in a pre-constructed neural network based on the simulated samples, and obtaining an initial model for coal sample detection; and verifying the initial model for coal sample detection based on the reserved historical target spectral data to obtain a coal sample detection model.

[0055] Optionally, the pre-constructed rules of the neural network include: dividing multiple optimization dimensions based on the structural parameters of the neural network; in each optimization dimension, adaptively adjusting the parameters in the neural network structure corresponding to the optimization dimension, and calculating the MAE index after each adjustment based on the target spectral data; comparing the MAE index corresponding to each parameter, and determining the parameter combination with the highest MAE index as the optimization result in the corresponding optimization dimension; performing greedy search between the optimization dimensions to obtain the optimization results of each optimization dimension; based on the optimization results of each optimization dimension, determining the structural parameters corresponding to each neural network structure, and constructing a neural network corresponding to the search result.

[0056] Optionally, one or more of ash composition, ash content, volatile matter, hydrocarbons, ash melting point, total water, total sulfur and calorific value.

[0057] Optionally, the method also includes: visually outputting the coal detection results, including: determining a corresponding detection result object in response to a user data query instruction; selecting a preset data visualization scheme based on the corresponding detection result object, and pushing the visualized data to the user end.

[0058] On the other hand, the present invention provides a computer-readable storage medium, on which instructions are stored, which, when executed on a computer, enable the computer to execute the above-mentioned coal detection method.

[0059] Through the above technical solution, the solution of the present invention has the following beneficial effects:

[0060] 1. Improved detection accuracy: By using multiple detection technologies such as near-infrared spectrum acquisition module, X-ray fluorescence spectrum acquisition module and visual analysis module, the system can detect coal samples comprehensively and from multiple angles, improving the detection accuracy and precision.

[0061] 2. Improved detection efficiency: Through automated sample preparation and data collection, the system can quickly and efficiently complete the coal sample detection process, saving time and labor costs and improving detection efficiency.

[0062] 3. Training model optimization: The analysis unit performs data training based on the pre-trained coal sample detection model, which can continuously optimize the model and improve the detection capability and adaptability of the system.

[0063] 4. Result information output: The output unit determines the test result information of the coal sample according to the reasoning results, so that users can obtain and analyze the test data in time, providing strong support for decision-making.

[0064] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0066] Figure 1 is a system structure diagram of a coal detection system provided by an embodiment of the present invention;

[0067] Figure 2 It is a system structure diagram of a coal detection system including a sample preparation module provided by one embodiment of the present invention;

[0068] Figure 3 It is a structural schematic diagram of a specific embodiment of a shaping module provided by an embodiment of the present invention;

[0069] Figure 4 is a schematic structural diagram of a second pretreatment component provided by an embodiment of the present invention;

[0070] Figure 5 is a schematic structural diagram of a pressure roller provided in one embodiment of the present invention;

[0071] Figure 6 is a structural schematic diagram of a specific embodiment of a detection unit provided by an embodiment of the present invention;

[0072] Figure 7 It is a structural schematic diagram of a specific embodiment of a Mylar film replacement device provided by one embodiment of the present invention;

[0073] Figure 8 is a structural schematic diagram of a specific embodiment of a buffer adjustment structure provided by an embodiment of the present invention;

[0074] Fig. 9 is a system structure diagram of a coal detection system including an output module provided in one embodiment of the present invention;

[0075] Fig.10is a flow chart of steps of a coal detection method provided by one embodiment of the present invention;

[0076] Fig.11 It is a flow chart of the steps of a coal detection method including a sample preparation step provided by one embodiment of the present invention;

[0077] Fig.12 It is a schematic diagram of a sample preparation process for testing coal samples provided by an embodiment of the present invention;

[0078] Fig.13 is a flowchart of the steps of the spectral signal fusion process provided by one embodiment of the present invention;

[0079] Fig.14 It is a step flow chart of a coal detection method including an output step provided in one embodiment of the present invention.

[0080] Description of Reference Numerals

[0081] 1. Coal conveying member; 11. Conveying frame; 112. Mounting frame; 1121. Baffle; 12. Conveyor belt; 13. Feed hopper; 14. Mounting position; 2. First pre-treatment member; 21. Limiting frame; 211. Limiting plate; 22. Scraper; 3. Second pre-treatment member; 31. Pressing roller; 311. Rotating shaft; 3111. Sprocket; 3113. Stop wheel; 32. Limiting member; 321. Limiting wheel; 33. First driving member; 35. Support; 4. Sampling cover; 41. Lighting module; 4 2. Optical fiber; 43. Photoelectric sensor; 44. Aperture; 5. Sample conveyor belt; 6. Detection of coal sample; 7. Mounting bracket; 73. Mylar film transmission channel; 71. Driving wheel; 72. Driven wheel; 8. Mylar film; 9. Buffer adjustment structure; 901. Fixed shaft; 902. Longitudinal threaded shaft; 903. Lower fixed sleeve; 904. Upper threaded sleeve; 905. Compression spring; 101. Positive pressure blowing device; 102. X-ray generating device; 103. Scanning device; 104. Energy receiving device. DETAILED DESCRIPTION

[0082] The specific implementation of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the present invention, and is not used to limit the present invention.

[0083] Figure 1 1 is a system structure diagram of a coal detection system provided by an embodiment of the present invention. Figure 1As shown, an embodiment of the present invention provides a coal detection system, which includes: a detection unit, which is used to collect spectral data of coal samples for detection; wherein the detection unit includes a near-infrared spectrum acquisition module and an X-ray fluorescence spectrum acquisition module; a training unit, which is used to construct a neural network based on greedy search in each neural network dimension through the spectral data, and obtain a coal sample detection model based on simulated sample training after sample augmentation; an analysis unit, which is used to perform fusion processing on the spectral data to obtain target spectral data, and perform target spectral data inference based on the coal sample detection model to obtain coal detection results.

[0084] Preferably, Figure 2 The system also includes a sample preparation unit, which is used to collect coal samples and perform sample preparation on the coal samples to obtain test coal samples. The sample preparation unit includes: a sampling module, which is used to randomly collect raw coal samples during the transportation or storage of raw coal, or to receive raw coal samples deposited by users; a crushing module, which is used to perform crushing on the raw coal samples to obtain crushed coal samples; a processing module, which is used to perform pre-processing on the crushed coal samples to obtain basic coal samples; wherein the pre-processing of the crushed coal samples includes: drying, grinding and screening; a shaping module, which is used to perform shaping on the basic coal samples to obtain test coal samples.

[0085] In a possible implementation, in the coal industry, sampling is a crucial link that directly affects the accuracy and reliability of subsequent coal quality analysis and utilization. The present invention proposes an intelligent sampling device that combines sensing technology and automatic control system to realize intelligent identification and collection of raw coal samples. By carrying high-precision sensors and image recognition technology, it can monitor the flow of coal in the process of raw coal transportation or storage in real time, and perform random sampling according to preset algorithms and rules. By using data analysis and simulation technology, the selection of sampling positions is optimized to ensure the representativeness and comprehensiveness of the sampling points. By analyzing the coal flow path and speed, the optimal sampling position is determined to avoid sampling bias and local problems and improve the reliability of sampling. A real-time monitoring and feedback mechanism is introduced to monitor key parameters in the sampling process, such as sampling volume, sampling frequency, etc., and the sampling strategy is adjusted in time. Through data collection and analysis, real-time monitoring and quality control of the sampling process are realized to ensure the accuracy and reliability of the sampling results. An automated sampling system is designed to realize the automated collection and processing of raw coal samples. In combination with machine learning and artificial intelligence technology, the sampling algorithm and process are optimized to improve sampling efficiency and automation, and reduce human intervention and errors. The design of the intelligent sampling device and optimized sampling position of the present invention can improve the accuracy and representativeness of sampling, reduce sampling errors and deviations, and improve the reliability of sampling results. The introduction of real-time monitoring and feedback mechanism can optimize the sampling process, improve sampling efficiency and speed, and save manpower and time costs. The design of the automated sampling system can reduce human intervention, reduce operational risks and errors, and improve the controllability and stability of the sampling process.

[0086] In a possible implementation, during the coal processing, crushing is an essential link, but it also generates a lot of dust pollution. Integrate efficient dust removal equipment in the crushing module, such as bag dust collector or electrostatic precipitator, to capture and filter dust particles generated during the crushing process. Through equipment integration, timely removal and treatment of dust during the crushing process can be achieved to ensure the cleanliness of the production environment and the health of employees. Adopt wet dust removal technology, through a spray system or a wet scrubber, the dust is humidified during the crushing process to reduce the diffusion and flying of dust. The wet dust removal technology can effectively control dust emissions and reduce the impact on air quality. Design the negative pressure closed structure of the crushing module, control the flow of dust in the closed space through the negative pressure system, reduce dust leakage and diffusion, and the negative pressure closed design can effectively prevent dust pollution and protect the surrounding environment and equipment. Introduce an online monitoring and control system to monitor the dust emission concentration and particle size distribution in real time, adjust the operating parameters and cleaning cycle of the dust removal equipment according to the monitoring data, and ensure the stability and efficiency of the dust removal effect through real-time monitoring and control. The scheme of the present invention effectively controls dust emission, reduces environmental pollution, and protects the surrounding ecological environment through the integration of dust removal equipment and the application of wet dust removal technology. The negative pressure closed design and the introduction of the online monitoring system of the scheme of the present invention can protect the health of employees, reduce the harm of dust to the respiratory tract and skin, and improve the comfort of the working environment.

[0087] In a possible implementation, the scheme of the present invention introduces a sample grading and screening device in the processing module, and performs fine screening and grading treatment on the samples according to the particle size and shape characteristics of the crushed coal samples. Through grading and screening, basic coal samples of different particle size ranges can be obtained, providing more accurate and representative samples for subsequent detection. The scheme of the present invention designs a chemical treatment reaction tank, which is used to chemically treat and react the crushed coal samples, extract target components or remove interfering substances. The chemical treatment reaction tank can change the chemical properties of the coal samples, making them more suitable for subsequent detection and analysis, and improving the accuracy of the detection results. The scheme of the present invention introduces a magnetic separation device, which performs magnetic separation treatment on the magnetic impurities or magnetic minerals in the crushed coal samples, and separates the magnetic substances from the basic coal samples. The magnetic separation device can effectively remove interfering substances, purify the basic coal samples, and improve the accuracy and reliability of the detection. A drying and dehumidification system is designed to dry and dehumidify the treated basic coal samples to ensure the dry state and stability of the samples. The drying and dehumidification system can avoid the influence of moisture on the detection results and ensure the accuracy and reliability of the detection data.

[0088] Preferably, the sampling module, the crushing module, the processing module and the shaping module are connected based on a transfer conveyor belt; the transfer conveyor belt is triggered and controlled based on a servo system.

[0089] Preferably, the servo system is configured to: start timing based on the trigger signal of the position trigger at the preset position of each module of the sample preparation unit, and control the power servo motor of the transfer conveyor belt to start, until it runs for a predetermined time, and then turn off the power servo motor of the transfer conveyor belt; or start timing in response to the switch trigger signal corresponding to each module of the sample preparation unit, and control the power servo motor of the transfer conveyor belt to start, until it runs for a predetermined time, and then turn off the power servo motor of the transfer conveyor belt.

[0090] Specifically, an intelligent sensor such as an infrared sensor is installed at the preset position of each module of the sample preparation unit to detect the position of the sample and trigger the corresponding operation. The sensor can monitor the sample position in real time, send a signal to the servo system, and start the corresponding operation. According to the preset position trigger or switch trigger signal of each module of the sample preparation unit, the servo system starts timing and controls the power servo motor of the transfer conveyor belt to start, and performs corresponding operations such as sampling, crushing, processing or shaping according to the predetermined time, and then turns off the power servo motor of the conveyor belt. The control logic of the servo system is optimized so that it can flexibly adjust the flow speed and residence time of the conveyor belt according to the operation requirements of different modules to ensure the smooth progress and efficient completion of each operation link. The scheme of the present invention realizes the automatic flow and operation execution of samples between modules, reduces manual intervention, and improves the efficiency and consistency of sample processing. The servo system accurately controls the start and stop of the conveyor belt according to the preset position trigger signal to ensure the accurate operation and flow of samples in each module. By optimizing the control logic, the flexible adjustment of the conveyor belt speed and residence time is realized to adapt to different operation requirements and improve the applicability and flexibility of the system.

[0091] Preferably, Figure 3 The shaping module includes: a coal conveying component 1, a first pre-processing component 2 and a second pre-processing component 3.

[0092] The coal conveying member 1 is composed of a conveying frame 11 and a rotating conveyor belt 12 installed thereon. The conveying frame 11 is provided with a detection coal sample inlet and adopts a rectangular parallelepiped structure. The conveyor belt 12 transports the detection coal sample along the length direction of the conveying frame 11, and a plurality of legs are provided at the bottom to ensure the support effect. The detection coal sample enters the conveying frame 11 through the detection coal sample inlet and directly falls into the conveyor belt 12, and then is transported toward the exit direction. In addition, a second driving member is also provided on the conveying frame 11, which is used to drive the conveyor belt 12 to rotate. The specific rotation method belongs to conventional technology, so it will not be described in detail.

[0093] The first pretreatment component 2 is composed of a limit frame 21 and a scraper 22. The limit frame 21 is supported above the conveyor belt 12 by the conveyor frame 11 and is located below the entrance of the test coal sample, so that the test coal sample can fall into it smoothly. The scraper 22 is located between the discharge end of the limit frame 21 and the conveyor belt 12 to form a first gap. Two mounting plates are provided on the conveyor frame 11, and the two ends of the limit frame 21 are connected to the two mounting plates to support the limit frame 21 above the conveyor belt 12. The limit frame 21 and the conveyor belt 12 are spaced in the vertical direction, and the gap between them is small, which prevents the test coal sample from flowing out of the limit frame 21 from the side, and does not affect the rotation of the conveyor belt 12, ensuring the smooth transportation of the test coal sample. The limit frame 21 adopts a rectangular frame structure, including two side plates extending along the direction of the conveyor belt 12. The gap between the side plates can be designed as needed to meet the initial shaping requirements of the width of the test coal sample. The scraper 22 is arranged above the discharge end of the two side plates, and the gap between the scraper 22 and the conveyor belt 12 also meets the initial shaping requirements.

[0094] By setting the limit frame 21 and the conveyor belt 12, the limit frame 21 is located just below the detection coal sample inlet of the conveyor frame 11, so that the detection coal sample can fall onto the conveyor belt 12 after being poured into the conveyor frame 11, thereby limiting the coal sample within the limit frame 21, and the rotating conveyor belt 12 can convey the coal sample within the limit frame 21. The scraper 22 is set on the limit frame 21, and when the conveyor belt 12 drives the detection coal sample located in the limit frame 21 to contact the scraper 22, the scraper 22 can scrape the detection coal sample flat on the conveyor belt 12, the thickness of the detection coal sample is the distance between the scraper 22 and the conveyor belt 12, and the width of the coal sample is equal to the width inside the limit frame 21.

[0095] The second pretreatment component 3 includes at least one pressing roller 31, which is extended and arranged on the conveying frame 11 along the conveying direction perpendicular to the conveying belt 12, and is located at the outlet end of the limit frame 21. A second gap is formed between the pressing roller 31 and the conveying belt 12, which is connected to the first gap, so that the test coal sample preliminarily shaped by the limit frame 21 and the scraper 22 can enter between the pressing roller 31 and the conveying frame 11, thereby limiting the thickness of the test coal sample within a preset range. A limiter 32 is provided on the pressing roller 31 to limit the width of the second gap to ensure that the width of the test coal sample is within a predetermined range.

[0096] The coal pre-detection pretreatment device provided by the present invention can perform preliminary shaping of the coal sample through the first pretreatment component 2 during the coal sample transportation process to limit the width and height of the coal sample within a certain range, and then perform secondary shaping on the preliminary shaped coal sample through the second pretreatment component 3 to roll the coal sample so that the height and height of the coal sample after rolling reach a preset range, and make the surface of the coal sample smooth, so that the coal sample meets the detection requirements, improves the stability of the coal sample spectral data, and improves the detection accuracy.

[0097] In some embodiments, in combination Figure 5 The limiting member 32 includes two limiting wheels 321 coaxially arranged on the pressure roller 31 , and the two limiting wheels 321 are spaced apart along the extending direction of the pressure roller 31 to form a second gap between the two limiting wheels 321 .

[0098] Specifically, two limiting wheels 321 are respectively arranged on the outer sides of the two side plates, so that the primary shaped material flowing out through the first gap can enter the second gap, and then the material is secondary shaped by the pressing roller 31. The two limiting wheels 321 can rotate synchronously with the pressing roller 31, and the distance between the two limiting wheels 321 can be designed according to actual needs to meet the rolling and shaping requirements of the material.

[0099] The design of the limiting member 32 is simple in structure and enables the material processed by the pressing roller 31 to meet the preset requirements.

[0100] For further optimization, a slope surface is provided on the inner side of the limiting wheel 321 .

[0101] Specifically, the middle part of the limiting wheel 321 bulges inwardly, so that the inner side of the limiting wheel 321 forms a truncated cone, and the side wall of the truncated cone forms a slope surface. The limiting wheel 321 under this design can guide the material entering the second gap, so that the material can enter the second gap.

[0102] Combination Figure 4 Two mounting frames 112 are arranged on the conveying frame 11 at intervals along the conveying direction perpendicular to the conveyor belt 12. A rotating shaft 311 is passed through the pressure roller 31. Both ends of the rotating shaft 311 are rotatably arranged on the two mounting frames 112 respectively. A first driving member 33 for driving the rotating shaft 311 to rotate is provided on the conveying frame 11.

[0103] Specifically, the rotating shaft 311 can drive the pressure roller 31 to rotate coaxially, and the two ends of the rotating shaft 311 are respectively installed on the two mounting frames 112 through the support 35, and the ends of the rotating shaft 311 are connected to the support 35 through the bearing to ensure the smoothness of the rotation of the rotating shaft 311. Among them, the support 35 includes a fixed frame arranged on the mounting frame 112 and a bearing seat arranged on the fixed frame, and the bearing is arranged on the bearing seat. By setting the fixed frame, the installation position of the bearing seat can be positioned, so that the mounting hole on the bearing seat can correspond to the mounting hole on the mounting frame 112, and then the external bolts can be accurately screwed into the mounting holes of the bearing seat and the mounting frame 112, thereby increasing the convenience of installation.

[0104] The first driving member 33 may be a driving motor disposed on the conveying frame 11, and the output shaft of the driving motor is connected to the rotating shaft 311, so that the rotating shaft 311 is driven to rotate by the driving motor, thereby driving the pressing roller 31 and the above-mentioned limiting wheel 321 to rotate, so that the device operates stably. Among them, the driving motor is a variable frequency motor, and it is understandable that the driving motor may also be other forms of motors, which can be designed according to actual needs.

[0105] In some embodiments, in combination Figure 4 As shown, when there are multiple pressure rollers 31, a sprocket 3111 is provided at one end of the rotating shaft 311 of the multiple pressure rollers 31 away from the first driving member 33, and the multiple sprockets 3111 are connected by a chain. The first driving member 33 is used to drive one of the rotating shafts 311 to rotate.

[0106] Specifically, a plurality of pressure rollers 31 are arranged at intervals along the conveying direction of the conveyor belt 12, one end of the rotating shaft 311 away from the first driving member 33 extends out of the support 35, and a sprocket 3111 is connected to the extended end of the rotating shaft 311, and the sprocket 3111 rotates synchronously with the rotating shaft 311. The plurality of sprockets 3111 are connected by a chain, so that when one of the rotating shafts 311 rotates, the other rotating shafts 311 can be driven to rotate synchronously through the cooperation of the sprocket 3111 and the chain, thereby reducing the number of driving devices and thus reducing the structural cost. Figure 4 Schematically shows a design in which there are two pressing rollers 31 , the protruding ends of the two rotating shafts 311 are connected with sprockets 3111 , and the two sprockets 3111 are connected by a chain, that is, the chain is meshed with the sprockets 3111 .

[0107] Preferably, there are two pressure rollers 31, and the two pressure rollers 31 are arranged at intervals along the conveying direction of the conveyor belt 12. For the convenience of description, the pressure roller 31 arranged close to the limit frame 21 is called the first pressure roller 31, and the pressure roller 31 arranged away from the limit frame 21 is called the second pressure roller 31. The height of the second gap between the first pressure roller 31 and the conveyor belt 12 is greater than the height of the second gap between the second pressure roller 31 and the conveyor belt 12.

[0108] As a feasible implementation, the height of the second gap between the first pressure roller 31 and the conveyor belt 12 is 3.5 cm, and the width of the second gap (the distance between the two limiting wheels 321) is 10 cm. The height of the second gap between the second pressure roller 31 and the conveyor belt 12 is 3 cm, and the width of the second gap (the distance between the two limiting wheels 321) is 10 cm. This design can compact the material to a thickness of 3 cm through the two pressure rollers 31 and limit the width of the material to 10 cm.

[0109] Under this design, the first pressing roller 31 is used to receive the preliminary shaped material conveyed from the limit frame 21. Since the incoming material quantity is relatively large, the thickness of the first gap between the first pressing roller 31 and the conveyor belt 12 is relatively large, so as to press the preliminary shaped material into a coal sample with a thickness of 3.5 cm and a width of 10 cm; and then the second pressing roller 31 is used to press the material into a coal sample with a thickness of 3 cm and a width of 10 cm, thereby ensuring the compaction effect of the material.

[0110] Figure 4 The pressure roller 31 which is not connected to the first driving member 33 is shown. Figure 4 The pressure roller 31 connected to the first driving member 33 is schematically shown, and specifically, a synchronous shaft is provided at the end of the rotating shaft 311 of the pressure roller 31 connected to the first driving member 33, and the synchronous shaft is used to connect with the output shaft of the first driving member 33, so that the output shaft of the first driving member 33 can drive the rotating shaft 311 to rotate synchronously through the synchronous shaft. In order to ensure that the synchronous shaft can rotate synchronously with the output shaft of the first driving member 33, a spline can be inserted at the connection between the synchronous shaft and the output shaft of the first driving member 33.

[0111] Combination Figure 4 As shown, in some embodiments, two blocking wheels 3113 are further provided on the rotating shaft 311 , and baffle plates 1121 are provided on the two mounting frames 112 . The two blocking wheels 3113 are supported on the inner sides of the two baffle plates 1121 , respectively.

[0112] Specifically, the two blocking wheels 3113 on the rotating shaft 311 are respectively arranged on the outside of the two limiting wheels 321, and the blocking wheels 3113 can rotate synchronously with the rotating shaft 311, or the blocking wheels 3113 and the rotating shaft 311 can rotate in coordination, which can be designed according to actual needs. Figure 4 and Figure 5 As shown, one end of the mounting frame 112 is connected to the conveying frame 11 to support the other end of the mounting frame 112 above the conveyor belt 12, and the mounting frame 112 is spaced apart from the conveyor belt 12, so that the position of the conveyor belt 12 in the vertical direction can be limited by the mounting frame 112, so as to avoid the phenomenon of warping or arching of the conveyor belt 12, and ensure the conveying effect of the conveyor belt 12. The baffle 1121 is arranged on the mounting frame 112, and the inner side of the baffle 1121 contacts and cooperates with the outer side of the stop wheel 3113, so that the movement of the rotating shaft 311 along its axial direction is limited by the two baffles 1121 and the two stop wheels 3113, thereby limiting the position of the pressure roller 31 perpendicular to the conveying direction of the conveyor belt 12, so that the pressure roller 31 can stably receive and press the material.

[0113] In some embodiments, Figure 4As shown, the output ends of the two sides of the limiting frame 21 are provided with limiting plates 211 , which extend in an arc shape. The limiting plates 211 are arranged opposite to the pressing roller 31 , and the curvature of the limiting plates 211 matches the curvature of the pressing roller 31 .

[0114] Specifically, the ends of the two side plates of the limit frame 21 are respectively provided with limit plates 211, which extend in an arc shape to the bottom of the pressure roller 31 to limit the material flowing out of the limit frame 21, ensuring that the material can be conveyed to the bottom of the pressure roller 31, and the curvature of the limit plate 211 matches the curvature of the pressure roller 31, so that the gap between the arc plate and the pressure roller 31 is smaller, thereby preventing the material from escaping from the second gap and meeting the material conveying requirements.

[0115] like Figure 3 As shown, in some embodiments, a protective cover is provided above the second pretreatment member 3. Specifically, when there are two pressing rollers 31, the first pressing roller 31 and the second pressing roller 31 are both provided in the protective cover, and the protective cover is located between the first driving member 33 and the chain. By providing the protective cover, it is not easy for external dust to contaminate the materials pressed by the first pressing roller 31 and the second pressing roller 31, and the dust generated by the first pressing roller 31 and the second pressing roller 31 when pressing the materials is not easy to spread in the environment around the device.

[0116] Combination Figure 3 and Figure 4 As shown, in some embodiments, a feed hopper 13 is provided above the material inlet, wherein the top of the feed hopper 13 extends upward from the conveyor frame 11, and the inlet of the feed hopper 13 can be connected to the outlet of the equipment in the previous process, so that the material can enter the conveyor frame 11 through the feed hopper 13. In order to increase the sealing performance of the connection between the feed hopper 13 and the outlet of the equipment in the previous process, a flange plate can be provided at the top of the feed hopper 13.

[0117] The conveyor frame 11 is provided with a mounting position 14 for mounting a detection module, and the mounting position 14 is located on the side of the second pretreatment component 3 away from the first pretreatment component 2. Among them, the detection module can be a coal rapid detection module, and the coal rapid detection module is installed on the mounting position 14. The mounting position 14 is provided with a plurality of mounting holes and detection ports, and the mounting holes are used to install the coal rapid detection module. When working, the coal rapid detection module can detect a coal sample with a width of 10 cm and a height of 3 cm on the conveyor belt 12 through the detection port. The coal rapid detection module is a conventional technology in the field of coal detection, so its structure and working principle are not described in detail here.

[0118] During operation, the coal rapid detection module is installed at the installation position 14, and the coal sample enters the conveyor frame 11 through the feed hopper 13 and falls onto the part of the conveyor belt 12 opposite to the limit frame 21, so that the coal sample is transported by the conveyor belt 12, and the coal sample during the transportation process is scraped flat by the scraper 22 to obtain a coal sample of a certain width and thickness. The scraped coal sample is sequentially conveyed to two pressing rollers 31, so that the coal sample can be pressed into a preset size (the height of the coal sample is 3 cm and the width is 10 cm). The surface of the obtained coal sample is smooth, so that the coal sample meets the detection requirements of the coal rapid detection module, improves the stability of the coal sample spectral data, and improves the detection accuracy.

[0119] Preferably, Figure 5 The detection unit also includes: a sampling cover 4 for setting the near-infrared spectrum acquisition module, the X-ray fluorescence spectrum acquisition module and the visual analysis module; an illumination module 41, an optical fiber 42 and a photoelectric sensor 43 are also arranged inside the sampling cover; an aperture 44 extending from the inner wall of the sampling cover is arranged between the illumination module 41 and the optical fiber 42; the photoelectric sensor 43 is arranged in the light path direction of the illumination module, and is used to monitor the light intensity of the corresponding illumination module 41.

[0120] In an embodiment of the present invention, the scheme of the present invention will also test the reflectivity of the coal, that is, a corresponding optical fiber is set to recover the reflected light of the sample. Generally, the reflectivity test of coal uses a halogen lamp as the light source. The light source will decay if it is lit for a long time. At the same time, the reflectivity of coal is low, and the accuracy of the reflectivity test is easily affected by factors such as stray light. In order to improve the test accuracy and reduce the influence of stray light, the scheme of the present invention proposes a corresponding detection module improvement. By setting an aperture between the lighting module and the optical fiber, stray light can be blocked from entering the optical fiber.

[0121] Preferably, the lighting module 41 includes a plurality of symmetrically arranged light sources; the optical fiber 4 is arranged at the center position of the top of the sampling cover; each lighting module 41 is provided with at least one corresponding photoelectric sensor 43, which is arranged on the inner wall of the sampling cover.

[0122] Furthermore, the analysis unit is also used to monitor the status of the corresponding lighting module based on the light intensity collected by the photoelectric sensor, including: performing preprocessing including filtering and denoising on the light data collected by the photoelectric sensor; performing interpolation or fitting calculations based on the voltage value or digital value of the electrical signal corresponding to the preprocessed light data and a calibration curve between the light intensity to obtain the light intensity at the detection position; performing light intensity fitting of the corresponding lighting module based on the light intensity at the detection position and the positional relationship between the light point sensor and the corresponding lighting module to obtain the detected light intensity of the corresponding lighting module; comparing the detected light intensity with a preset light intensity threshold, and outputting an alarm message if the detected light intensity is less than the preset light intensity threshold.

[0123] In the embodiments of the present invention, it has been explained above that the reflectivity test of coal generally uses a halogen lamp as the light source. The light source will decay when it is lit for a long time. Under the existing scheme, relevant personnel are often required to independently judge the life of the light source and replace the light source when it needs to be replaced. In this case, it is very dependent on the subjective experience of the personnel. Based on this, the scheme of the present invention proposes a corresponding light intensity monitoring scheme, which monitors the light intensity of the light source in real time through a photoelectric sensor, and outputs an alarm message when the light intensity of the light source does not meet the requirements, reminding relevant personnel to replace the light source.

[0124] Preferably, Figure 6 The detection unit also includes: a Mylar film replacement device, including a sample conveyor belt 5, on which a detection coal sample 6 is placed, and a mounting frame 7 is provided above the detection coal sample 6, on which a driving wheel 71 and a driven wheel 72 are provided, and a Mylar film transmission channel 73 is formed on the mounting frame 7, and the Mylar film 8 can pass through the Mylar film transmission channel 73 and abut against the lower edge of the outer circumferential surface of the driving wheel 71 and the driven wheel 72, and move under the drive of the driving wheel 71.

[0125] In the embodiment of the present invention, in the existing online coal composition analysis process, the Mylar film of the X-ray fluorescence spectrum acquisition module (RF signal light transmission module) is very easy to be contaminated during the transmission and detection process. During the use of the existing Mylar film, the degree of dirtiness is usually judged manually to determine whether the Mylar film needs to be replaced, which has low work efficiency and high replacement cost. Based on this, the solution of the present invention proposes a corresponding Mylar film replacement device.

[0126] Specifically, a conveyor belt driving wheel and a conveyor belt driven wheel are respectively provided at both ends of the sample conveyor belt 5. The sample conveyor belt 5 conveys the test coal sample 6 through the conveyor belt driving wheel. The conveyor belt driving wheel can be connected to a motor to provide power for the conveyor belt driving wheel. Of course, since the sample conveyor belt 5 is relatively soft, a support plate can be provided under the sample conveyor belt 5 to support the sample conveyor belt 5, so as to avoid placing multiple test coal samples 6 on the sample conveyor belt 5, causing deformation of the sample conveyor belt 5.

[0127] Furthermore, in the present invention, a driving wheel 71 and a driven wheel 72 are provided on the mounting frame 7, and a motor is connected to the driving wheel 71 to provide power for the driving wheel 71. A Mylar film transmission channel 71 is provided on the mounting frame 7, which is suitable for the Mylar film 8 to pass through. It is conceivable that a protective structure may be provided in the Mylar film transmission channel 71 to prevent the Mylar film 8 from being damaged. The Mylar film 8 is attached to the driving wheel 71 and the driven wheel 72, and the driven wheel 72 can flatten the Mylar film 8 and provide a certain tension.

[0128] In a possible implementation, Figure 8 , a buffer adjustment structure 9 is provided on the driving wheel 71, and a buffer adjustment structure 9 is also provided on the driven wheel 72. The structures of the two buffer adjustment structures 9 can be set to be exactly the same, thereby reducing the production cost. A single buffer adjustment structure 9 includes a fixed shaft 901 and a longitudinal threaded shaft 902 connected to the fixed shaft 901 or integrally formed with the fixed shaft 901. An external thread is formed on the outer peripheral surface of the longitudinal threaded shaft 902. A lower fixed sleeve 903 and an upper threaded sleeve 904 are sleeved on the longitudinal threaded shaft 902. A gap is formed between the lower fixed sleeve 903 and the longitudinal threaded shaft 902, and a threaded connection is formed between the upper threaded sleeve 904 and the longitudinal threaded shaft 902. Of course, a locking nut can also be provided above the upper threaded sleeve 904 to prevent the upper threaded sleeve 904 from loosening during use. Of course, other anti-loosening measures can also be adopted, for example, thread glue can also be used. It is conceivable that a compression spring 905 is provided between the lower fixing sleeve 903 and the upper threaded sleeve 904 , thereby providing a buffer force for the driving wheel 71 and the driven wheel 72 , and also providing downward pressure for the driving wheel 71 and the driven wheel 72 .

[0129] In a possible implementation manner, bearings are provided between the fixed shaft 901 and the driving wheel 71 , and between the fixed shaft 901 and the driven wheel 72 .

[0130] In a specific embodiment of the present invention, bearings are provided at both ends of the fixed shaft 901, and the outer ring of the bearing is against the inner circumference of the inner hole of the driving wheel 71 or the driven wheel 72. Specifically, steps are provided at both ends of the fixed shaft 901, and one side of the step is against one side of the inner ring of the bearing. A shaft elastic circlip is also provided on the fixed shaft 901, and the shaft elastic circlip is against the outer side of the inner ring of the bearing to prevent the bearing from moving. Of course, other structural forms can also be used to prevent the bearing from moving left and right, which also belongs to the protection scope of the present invention.

[0131] As a preferred embodiment of the present invention, the central area of ​​the lower fixing sleeve 903 is formed as a through hole, and the central area of ​​the upper threaded sleeve 904 is formed with a threaded inner hole, which is adapted to the longitudinal threaded shaft 902 .

[0132] In a specific embodiment of the present invention, the central area of ​​the lower fixing sleeve 903 is formed as a light hole, so that the longitudinal threaded shaft 902 can move freely in the lower fixing sleeve 903. The premise that the longitudinal threaded shaft 902 can move freely in the lower fixing sleeve 903 is that there is a fixed support on the lower fixing sleeve 903. During the up and down movement of the longitudinal threaded shaft 902, the lower fixing sleeve 903 is always in a fixed state and will not move with the up and down movement of the longitudinal threaded shaft 902. Therefore, as a preferred embodiment of the present invention, the lower fixing sleeve 903 is fixedly connected to the mounting frame 7.

[0133] As a preferred embodiment of the present invention, a silicone protective layer is provided on the outer circumference of the driving wheel 71 and the driven wheel 72 .

[0134] As a specific embodiment of the present invention, the driving wheel 71 includes a driving wheel core shaft and a silicone protective layer arranged on the outer peripheral surface of the driving wheel core shaft. The central area of ​​the driving wheel core shaft is provided with an installation through hole for installing the bearing and the fixed shaft 901.

[0135] As a preferred embodiment of the present invention, the automatic Mylar film replacement device also includes a positive pressure blowing device 101, an X-ray generating device 102, a scanning device 103 and an energy receiving device 104. The positive pressure blowing device 101 is arranged above the operating window of the mounting frame 7, and the X-ray generating device 102, the scanning device 103 and the energy receiving device 104 are arranged above the positive pressure blowing device 101.

[0136] In a specific embodiment of the present invention, a control system is also included, and the control system is respectively connected to the positive pressure blowing device 101, the X-ray generating device 102, the scanning device 103, the energy receiving device 104, the motor of the driving wheel 71 and other signals.

[0137] As a preferred embodiment of the present invention, the positive pressure blowing device 101, the X-ray generating device 102, the scanning device 103 and the energy receiving device 104 are all fixedly connected to the mounting frame 7 via a mounting bracket.

[0138] As a preferred embodiment of the present invention, the surface of the sample conveyor belt 5 is provided with an anti-slip layer or an anti-slip structure.

[0139] In a preferred embodiment of the present invention, the sample conveyor belt 5 is provided with a test coal sample 6, a mounting frame 7 is provided above the test coal sample 6, a driving wheel 71 and a driven wheel 72 are provided on the mounting frame 7, a motor is connected to the driving wheel 71, a Mylar film transmission channel 71 and an operation window that penetrates the upper and lower thicknesses of the mounting frame 7 are formed on the mounting frame 7, the Mylar film 8 can pass through the Mylar film transmission channel 71, the operation window is provided corresponding to the test coal sample 6, and a buffer adjustment is provided on the driving wheel 71 and the driven wheel 72. The buffer adjustment structure 9 includes a fixed shaft 901 and a longitudinal threaded shaft 902 connected to the fixed shaft 901, a lower fixed sleeve 903 and an upper threaded sleeve 904 are provided on the longitudinal threaded shaft 902, a compression spring 905 is provided between the lower fixed sleeve 903 and the upper threaded sleeve 904, bearings are provided between the fixed shaft 901 and the driving wheel 71, and between the fixed shaft 901 and the driven wheel 72, elastic retaining rings for the shaft are also provided at both ends of the fixed shaft 901, and a silicone protective layer is provided on the outer circumference of the driving wheel 71 and the driven wheel 72.

[0140] Preferably, the system also includes a visual analysis module, which includes one or more image acquisition modules, each of which is arranged at each preset direction corresponding to the detection position of the detected coal sample, and is used to collect image information of the detected coal sample at a viewing angle; the analysis unit is also used to evaluate the appearance of the detected coal sample based on the image information at each viewing angle, including: performing denoising, grayscale and edge detection processing on the image information at each viewing angle in turn, and calibrating the detection coal sample area; fitting the outer dimensions of the detected coal sample based on the image detection coal sample area at each viewing angle; collecting surface flatness indicators in the image detection coal sample area at each viewing angle, and fitting the surface flatness of the detected coal sample based on the surface flatness indicators; and evaluating the appearance of the detected coal sample based on the fitted outer dimensions and the fitted surface flatness.

[0141] In the embodiment of the present invention, the appearance of the coal sample may have the following effects on the results of coal composition monitoring based on spectral signals:

[0142] 1) Influence of light path scattering: The irregular shape or rough surface of coal samples may cause light path scattering, causing the light signal to be deflected or scattered during transmission, affecting the collection and accuracy of spectral signals.

[0143] 2) Differences in light absorption: The different shapes of coal samples may lead to different light absorption capabilities in different parts, which in turn affects the intensity and characteristics of the spectral signal. For example, the differences in particle size, shape and density on the surface of coal will affect the transmission and absorption of light.

[0144] 3) Surface reflection effect: The light reflection characteristics of the coal sample surface will affect the collection and analysis of spectral signals. Coal samples of different shapes may have different reflectivities, resulting in different reflection and diffuse reflection of light signals on the sample surface, affecting the quality and accuracy of spectral signals.

[0145] 4) Spectral signal interference: The uneven appearance of coal samples may cause interference with spectral signals. For example, the presence of impurities, oxides or moisture on the sample surface will affect the purity and stability of the spectral signal, thereby affecting the accuracy of component monitoring.

[0146] 5) Difference in optical path length: The different sizes and shapes of coal samples may lead to differences in optical path length, causing different degrees of attenuation and absorption of the light signal during transmission inside the sample, affecting the intensity and clarity of the spectral signal.

[0147] In the embodiment of the present invention, the appearance of the coal sample has a certain influence on the results of coal composition monitoring based on spectral signals. Therefore, when performing composition monitoring, it is necessary to consider and perform appropriate correction and processing on the appearance factor to improve the accuracy and reliability of the monitoring results.

[0148] Although the scheme of the present invention proposes a corresponding scheme for limiting the shape of the coal sample, in order to ensure that the shape shaping meets expectations, the shaping result needs to be tested. While collecting spectral signals, the scheme of the present invention will also collect visual signals, such as image information, and perform target detection coal sample shape recognition based on the image information to determine whether it meets expectations. If it does not meet expectations, it is necessary to output an alarm instruction to remind relevant personnel to replace the sample to avoid obtaining erroneous detection results.

[0149] Preferably, the surface flatness identification of the coal sample to be detected based on the visual image data includes: performing denoising, graying and edge detection processing on the visual image in sequence, calibrating the coal sample detection area; extracting features related to surface flatness in the coal sample detection area; performing surface flatness index identification on the features related to surface flatness, and determining the surface flatness of the coal sample to be detected based on the identification result. The features related to surface flatness include: one or more of texture features, color features and edge features. The surface flatness index includes: average grayscale and / or texture uniformity.

[0150] Specifically, the following steps are included:

[0151] 1) Data acquisition and preprocessing: Collect visual image data of coal samples to ensure image clarity and consistency. De-noise the visual images to eliminate the impact of noise on subsequent processing. Convert the processed images to grayscale images to simplify the processing and highlight surface features.

[0152] 2) Surface flatness identification processing: Perform edge detection processing to accurately calibrate the boundaries and contours of the test coal sample. Calibrate the test coal sample area to ensure that subsequent processing is focused on the sample surface. Extract features related to surface flatness in the test coal sample area, including texture features, color features, and edge features. Feature extraction and surface flatness index identification:

[0153] 3) Analyze and process the extracted features to determine features related to surface flatness. Calculate surface flatness indicators such as average grayscale and texture uniformity based on one or more of texture features, color features, and edge features.

[0154] 4) Surface flatness identification and result determination: According to the identification results, the surface flatness level of the tested coal sample is determined. Combined with the surface flatness index and characteristic analysis results, the surface flatness of the tested coal sample is evaluated and classified.

[0155] In an embodiment of the present invention, the surface flatness recognition of the coal sample based on visual image data can be effectively applied to quality control and production process monitoring in the coal industry to improve production efficiency and product quality management level.

[0156] Specifically, the appearance evaluation of the detected coal sample based on the fitted outer dimensions and the fitted surface flatness includes: calculating the Euclidean distance between the fitted outer dimensions and the preset outer dimensions as the first deviation value; calculating the Euclidean distance between the fitted surface flatness and the preset surface flatness as the second deviation value; performing normalization processing on the first deviation value and the second deviation value, and performing arithmetic mean calculation on the normalized first deviation value and the second deviation value to obtain a deviation coefficient; and assigning a preset correction coefficient in the coal sample detection model based on the deviation coefficient.

[0157] In an embodiment of the present invention, the scheme of the present invention can also integrate the detected coal samples with certain differences in appearance into the subsequent detection model by means of a deviation correction coefficient. This ensures that even if there are certain differences in the detected coal samples, unified standard processing can be achieved by correcting the deviation coefficient.

[0158] Preferably, the analysis unit is configured to: execute the spectral data reasoning based on the assigned coal sample detection model to obtain the reasoning result, including: performing fusion processing on the near-infrared spectral signal collected by the near-infrared spectral acquisition module and the X-ray fluorescence spectral data collected by the X-ray fluorescence spectral acquisition module to obtain the target spectral data; execute the target spectral data reasoning based on the assigned coal sample detection model to obtain the reasoning result.

[0159] Specifically, preprocessing is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively; downsampling is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data after preprocessing respectively, and the downsampled near-infrared spectrum signal and the X-ray fluorescence spectrum data are scaled to normal distribution; splicing is performed on the near-infrared spectrum signal with normal distribution and the X-ray fluorescence spectrum data with normal distribution to obtain the target spectrum data. Specifically, the following steps are included:

[0160] Step 1: Preprocess the near-infrared spectral signal and X-ray fluorescence spectral data respectively.

[0161] Specifically, SG convolution smoothing processing is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively; and area normalization processing is performed on the near-infrared spectrum signal after the convolution smoothing processing.

[0162] Preferably, the SG convolution smoothing processing is performed on the near-infrared spectral signal and the X-ray fluorescence spectral data respectively, including: calculating the coefficients of the SG convolution kernel based on a preset window size and polynomial order; in the near-infrared spectral signal and the X-ray fluorescence spectral data, respectively, weighted averaging the coefficients in the convolution kernel with adjacent data points in each spectral signal to obtain smoothed data points; performing symmetric expansion or zero filling processing on the boundary of each spectral signal; and obtaining the spectral signal subjected to SG convolution smoothing processing based on the smoothed data points and the processed boundaries.

[0163] In an embodiment of the present invention, SG convolution smoothing is a linear smoothing method. By locally fitting the data, the data curve can be better smoothed, the spikes and fluctuations in the data can be eliminated, and the data can be made more stable and continuous. SG convolution smoothing can effectively smooth the noise part of the data while retaining the characteristics and trends of the signal, which helps to reduce the interference of high-frequency noise in the data on the signal and improve the readability and analysis accuracy of the data. Compared with other smoothing methods, SG convolution smoothing can better maintain the overall shape and trend of the data while smoothing the data, will not cause distortion or offset of the data shape, and maintains the original characteristics of the data. SG convolution smoothing is a simple and efficient data smoothing method with a fast calculation speed. It is suitable for the processing of large-scale data sets and can complete data smoothing operations in a shorter time. It is suitable for large-scale coal sample detection, ensuring detection accuracy while improving detection efficiency.

[0164] Furthermore, the area normalization processing of the near-infrared spectral signal after convolution smoothing includes: determining a signal set of the near-infrared spectral signal after convolution smoothing; the signal set includes multiple signal samples, and the reflectivity of each signal sample under the corresponding spectral type at each preset wavelength point; calculating the signal after area normalization processing based on the signal set of the near-infrared spectral signal.

[0165] In an embodiment of the present invention, baseline drift is prone to occur in near-infrared spectral signals, so the scheme of the present invention will also eliminate the baseline drift in the near-infrared spectral signal by means of area normalization. Baseline drift is a signal deviation caused by factors such as instrument drift, environmental changes or sample differences, which affects the accuracy and stability of the spectral signal. Through area normalization, the signal can be moved up or down as a whole, so that the baseline level is more stable. The optical path difference in the near-infrared spectral signal will lead to differences in signal intensity, affecting the comparison and analysis of the signal. The area normalization process can normalize the overall intensity of the signal, reduce the impact of the optical path difference on the signal, and make the comparison between different samples more accurate. The area normalization process can highlight the characteristic peaks or troughs in the near-infrared spectral signal, making the characteristics of the signal more obvious and prominent, which helps to more accurately identify and analyze specific components or features in the spectrum.

[0166] Specifically, the calculation rule of the signal after area normalization is:

[0167]

[0168] in, Indicates the NIR reflectance corresponding to the mth wavelength point of the nth sample; is the signal after area normalization.

[0169] Step 2: Downsampling is performed on the near-infrared spectrum signal and X-ray fluorescence spectrum data after preprocessing.

[0170] Specifically, filtering processing is performed on each spectral signal respectively, and in the spectral signal after filtering processing, one sampling point is retained at a fixed interval to obtain multiple sampling points; or the spectral signal after filtering processing is cut into multiple segments, and averaging processing is performed on each sampling point in each segment, and one sampling point is obtained in each segment to obtain multiple sampling points; signal reconstruction is performed based on each sampling point to obtain a spectral signal after downsampling processing.

[0171] In an embodiment of the present invention, downsampling processing can reduce the amount of data, that is, reduce the sampling rate or the number of sampling points, thereby saving storage space and computing resources. In particular, for large-scale data sets or data sampled at high frequencies, downsampling can effectively reduce the volume of data, making it easier to store and process. Downsampling can simplify the data analysis process, reduce the complexity and dimension of the data, and make the data easier to understand and process. By reducing the resolution of the data, some detailed information can be removed, the main features of the data can be highlighted, and the model establishment and analysis process can be simplified. Downsampling processing can help remove noise and interference in the data, smooth the data curve, and improve the quality and stability of the data. By reducing the sampling rate or averaging the sampling points, the volatility of the data can be reduced, making the data clearer and more reliable. The scheme of the present invention can reduce the complexity and dimension of the data, reduce the amount of data, and make the data easier to process and analyze by retaining the interval sampling points and multi-segment averaging. And reduce the noise and interference in the signal, the interval sampling point retention and averaging processing can further smooth the signal curve, reduce the volatility of the data, and improve the signal quality and stability.

[0172] Step 3: Scale the downsampled near-infrared spectral signal and X-ray fluorescence spectral data to a normal distribution.

[0173] Specifically, the statistical characteristics of each spectral signal are calculated respectively; wherein the statistical characteristics are variance and / or standard deviation; based on the statistical characteristics of each spectral signal, each spectral signal is subjected to standardization or normalization processing, and the value of each spectral signal is scaled to a preset value range to obtain the scaled value of each spectral signal; based on a preset transformation algorithm and the scaled value of each spectral signal, each spectral signal is converted into a normal distribution.

[0174] In an embodiment of the present invention, the scheme of the present invention converts the signal into a normal distribution to simplify the data analysis process, making the data easier to understand and process. By converting the signal into a normal distribution, the fitting effect and prediction accuracy of the model can be improved, and converting the signal into a normal distribution can better meet the requirements of hypothesis testing and ensure the validity of the test results. And the normal distribution has the characteristic of standardization, that is, the mean is 0 and the standard deviation is 1. Converting the signal into a normal distribution can achieve data standardization, so that different signals are comparable, which is convenient for subsequent splicing of two spectral signals.

[0175] Step 4: Perform splicing processing on the normally distributed near-infrared spectrum signal and the normally distributed X-ray fluorescence spectrum data to obtain the target spectrum data.

[0176] Specifically, an alignment operation of a normally distributed near-infrared spectral signal and a normally distributed X-ray fluorescence spectral data is performed; after completing the alignment of the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data, the two spectral signals are spliced ​​based on weighted average to obtain initial target spectral data; the initial target spectral data is verified based on the normally distributed near-infrared spectral signal and / or the normally distributed X-ray fluorescence spectral data, and the verified initial target spectral data is used as the target spectral data.

[0177] Furthermore, the operation of performing the alignment of the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data includes: selecting the same reference point in the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data, and respectively calculating the time difference between the corresponding reference point and other data points in each spectral signal to obtain a time difference sequence; and performing a time axis adjustment of one of the spectral signals based on a random interpolation method and the corresponding time difference sequence until the time axes of the two spectral signals are aligned.

[0178] Furthermore, the random interpolation method is:

[0179] X aug =X 1 +(1-)X 2 ,α∈[0,1]

[0180] y aug = PLS ( aug )

[0181] Among them, f PLS represents the PLS mapping function; X aug and aug denote the synthesized NIRS-XRF signal and the corresponding pseudo-label respectively; α denotes the interpolation weight randomly sampled in the range [0,1]; X 1 and X 2Signal samples randomly selected from historical NIRS-XRF signals of coal samples.

[0182] In an embodiment of the present invention, the scheme of the present invention performs two spectral signal splicing based on the random interpolation method, which can retain the characteristics and information of the original data during the signal splicing process and avoid data loss or distortion. Splicing the signals by interpolation can better maintain the integrity of the data. The random interpolation method can achieve a smooth transition in the transition area of ​​the signal splicing to avoid mutations or discontinuities, which helps to improve the continuity and stability of the signal splicing. The random interpolation method can adjust the interpolation parameters as needed, such as the number of interpolation points, the selection of interpolation functions, etc., so as to flexibly control the effect of signal splicing, which makes the signal splicing process more customizable and can be customized based on the user's test requirements. The random interpolation method can effectively reduce the artifacts and distortion that may occur during the signal splicing process and improve the quality and accuracy of the signal splicing. By reasonably selecting the interpolation method and parameters, the error introduced by the signal splicing can be reduced, thereby improving the subsequent detection accuracy.

[0183] Preferably, the pre-trained coal sample detection model is:

[0184] θ=argmax θ L 2 (f(X|θ),Y);

[0185] Among them, f(·|θ) is a deep neural network with θ as parameter; is the target spectrum data; In one possible implementation, the present invention uses a deep neural network to construct a prediction model for coal composition prediction, and establishes a mapping between the input NIRS-XRF fusion spectrum and the corresponding real component label through an end-to-end training method. represents the input signal after preprocessing, represents the true value of the corresponding coal composition. n ∈R D is the concatenation of the NIRS and XRF spectra after preprocessing in step S20, y n is a non-negative scalar label. N and D are the number of training set samples and the input resolution, respectively. Note that since X nIt is possible to downsample from the original data, so D can be smaller than the resolution of the original data. f(·|θ) is a deep neural network with θ as a parameter, which can consist of multiple linear and nonlinear layers. The nonlinear layer here is also called an activation layer. A linear layer is usually followed by a nonlinear layer, and the combination of the two is called a module. In this paper, the network consists of multiple identical modules stacked in sequence and ends with a fully connected layer, outputting a scalar value y. Specifically, the base operator of the linear layer can be a fully connected layer or a 1D convolutional layer with different kernel sizes, while the base operator of the nonlinear layer can be a hyperbolic tangent function (TanH), an exponential linear unit (ELU), or a sigmoid function, etc.

[0186] Preferably, the method also includes: pre-training a coal sample detection model, including: collecting historical near-infrared spectral signals and historical X-ray fluorescence spectral data, and constructing corresponding historical target spectral data based on the historical near-infrared spectral signals and the historical X-ray fluorescence spectral data; using the historical target spectral data as training data to perform PLS model parameter initialization; generating simulated samples based on the initialized PLS model, performing model training based on the simulated samples, and obtaining an initial model for coal sample detection; and verifying the initial model for coal sample detection based on the reserved historical target spectral data to obtain a coal sample detection model.

[0187] In the embodiment of the present invention, in order to obtain an accurate detection model, a large amount of historical data is actually required as training samples to perform model training. However, it is actually difficult to obtain comprehensive historical data. On the one hand, there is the problem of historical data retention, and on the other hand, there is the problem of data interoperability. Therefore, it is actually difficult to train and obtain an accurate detection model based on the existing historical data. Based on this, the present invention proposes a solution for performing pre-training based on the PLS model.

[0188] In an embodiment of the present invention, historical target spectral data is used as training data to initialize the parameters of the partial least squares regression (PLS) model, laying the foundation for subsequent model training. Based on the initialized PLS model, simulated sample data is generated for model training, the training data set is expanded, and the generalization ability of the model is improved. The model is trained using simulated sample data, and the initial model for coal sample detection is established by learning the patterns and characteristics of historical data.

[0189] Preferably, the PLS model parameters include: initial weight, learning rate, activation function, regularization parameter, initialization bias term, and optimizer type.

[0190] Preferably, the generation of simulated samples based on the initialized PLS model includes: adaptively determining the signal type and signal parameters that match the target spectral data, and generating a basic signal based on the signal type and the model parameters; adaptively selecting an augmentation scheme, and performing a data augmentation operation on the basic signal based on the selected augmentation scheme to obtain an augmented signal set; screening out deviation data in the augmented signal set, and using the screened-out augmented signal set as a simulated sample.

[0191] In an embodiment of the present invention, adaptive algorithms and machine learning techniques are used to determine the best signal type and parameter combination to match the characteristics and changes of different spectral signals. Based on the determined signal type and parameters, a basic signal is generated as the basis for the augmented signal set. According to the data characteristics and model requirements, an augmentation scheme is adaptively selected, including data interpolation, noise reduction processing, etc., to improve data quality and model performance. The selected augmentation scheme, such as data interpolation, noise addition and other operations, is executed on the basic signal to generate a diversified augmented signal set. After completing the data augmentation operation, an augmented signal set containing diversified signals is obtained, which is used to enrich the training data and improve the generalization ability and robustness of the model. Deviation data screening is performed in the augmented signal set to identify and eliminate data points that may introduce errors, thereby ensuring the accuracy and reliability of the training data. Model training and verification improve the model's adaptability to uncertainty and noise.

[0192] In an embodiment of the present invention, the adaptively selected augmentation scheme includes: randomly selecting one or more schemes from noise addition, translation, scaling, rotation, shearing and transformation as pre-selected schemes; randomly adjusting parameters within a preset adjustable parameter range of each pre-selected scheme to obtain a pre-selected scheme determined by the parameters; if there is only one pre-selected scheme determined by the parameters, directly processing the basic signal based on the scheme to obtain an augmented signal; if there are multiple pre-selected schemes determined by the parameters, executing each scheme in turn to obtain an augmented signal.

[0193] Furthermore, the method of screening out deviation data in the augmented signal set and using the screened augmented signal set as a simulation sample includes: respectively calculating the Euclidean distance between each augmented signal in each augmented signal set and the basic signal; screening out augmented signals whose Euclidean distance is greater than a preset Euclidean distance threshold, and using the screened augmented signal set as a simulation sample.

[0194] In an embodiment of the present invention, the scheme of the present invention adaptively determines the augmentation scheme, and can randomly generate a large number of simulation signals to simulate the detection signals under various coal conditions, thereby ensuring the comprehensiveness of the data and making up for the problem of being unable to train an accurate detection model due to insufficient historical retained data.

[0195] Furthermore, the model training is performed based on the simulated samples to obtain an initial model for coal sample detection, including: pseudo-labeling the simulated samples based on the PLS model after parameter initialization, and using the labeled simulated samples as training samples; based on the training samples, model training is performed in a neural network based on the target spectral data search to obtain an initial model for coal sample detection.

[0196] In an embodiment of the present invention, after obtaining the augmented data set, it is also necessary to annotate the coal components corresponding to the data set, so as to facilitate model training based on the coal components corresponding to the simulated signal. Based on this, it is necessary to annotate the coal components of the augmented signals in the augmented signal set, so that each augmented signal corresponds to a simulated coal component prediction result. After the annotation is completed, the corresponding training samples are obtained, and the corresponding detection model can be obtained by performing model training based on the training samples. However, in order to ensure the accuracy of the model, the model needs to be verified.

[0197] Furthermore, the coal sample detection initial model is verified based on the reserved historical target spectral data to obtain the coal sample detection model, including: constructing a verification set based on the reserved historical target spectral data, and inputting the verification set into the component monitoring initial model to obtain the corresponding prediction result; performing performance evaluation of the component monitoring initial model based on the prediction result and the actual result of the corresponding reserved historical target spectral data; if the performance evaluation of the component monitoring initial model fails, adjusting the hyperparameters, optimizing the model structure and / or adding regularization operations to obtain an updated component monitoring initial model; re-performing a performance evaluation on the updated component monitoring initial model based on the reserved historical target spectral data until a component monitoring initial model that meets the preset performance requirements is obtained as the coal sample detection model.

[0198] In an embodiment of the present invention, the validation set is input into the initial model of coal sample detection to obtain the predicted results, and the prediction results are compared with the actual results to perform performance evaluation and error analysis. According to the performance evaluation results, if the model does not pass the preset performance requirements, the hyperparameters are adjusted, the model structure is optimized, and the regularization operation is added to improve the accuracy and stability of the model. According to the adjusted model, the performance evaluation is re-performed until the initial model of coal sample detection that meets the preset performance requirements is obtained as the updated coal sample detection model. The scheme of the present invention realizes continuous optimization and iteration of the coal sample detection model through repeated verification, evaluation and adjustment, thereby improving the performance and accuracy of the model. Based on the validation set of the reserved historical target spectral data, the performance and generalization ability of the model can be more accurately evaluated, and the reliability of the model evaluation can be improved. According to the performance evaluation results, the hyperparameters and model structure are adjusted in time to make the model better adapt to the data characteristics and task requirements, thereby improving the accuracy of coal sample detection. By adding regularization operations and other means, the stability and generalization ability of the model are improved, the risk of overfitting is reduced, and the robustness of the model is enhanced.

[0199] Preferably, the performance evaluation of the initial component monitoring model is performed based on the prediction results and the actual results of the corresponding reserved historical target spectral data, including: comparing the prediction results and the actual results of the corresponding reserved historical target spectral data, and evaluating the accuracy and recall of the model based on the deviation between the two; if any evaluation result of the accuracy and recall fails, the performance evaluation of the initial component monitoring model fails.

[0200] Preferably, a neural network search is performed based on the target spectral data, including: dividing a plurality of optimization dimensions based on neural network structure parameters; in each optimization dimension, adaptively adjusting the parameters in the neural network structure corresponding to the optimization dimension, and calculating the MAE index after each adjustment; comparing the MAE index corresponding to each parameter, and determining the parameter combination with the highest MAE index as the optimization result in the corresponding optimization dimension; performing a greedy search between the optimization dimensions to obtain the optimization results of each optimization dimension; based on the optimization results of each optimization dimension, determining the structural parameters corresponding to each neural network structure, and constructing a neural network corresponding to the search result.

[0201] Furthermore, the optimization dimensions include: the basic operator dimension of the neural network layer, the resolution dimension of the input data, the network depth dimension, and the network width dimension.

[0202] In the embodiment of the present invention, the neural network dimensions that affect the fusion of NIRS-XRF dual spectra include the basic operators of the linear and nonlinear layers, the resolution of the input data, the network depth (the number of stacked modules) and the network width (the output channels of the linear layer). The scheme of the present invention decomposes the search process of these four dimensions into four stages, each stage searches a single dimension and performs greedy search between the dimensions.

[0203] In an embodiment of the present invention, within the basic operator dimension of the neural network layer, the corresponding optimization result obtaining rules include: traversing the convolution type of the linear layer, and traversing the activation function type of the nonlinear layer, adaptively combining the convolution type of the fully connected layer and the activation function type of the nonlinear layer, and calculating the MAE index after each combination; comparing the MAE index after each combination, selecting the combination corresponding to the maximum MAE index, and determining the convolution type of the fully connected layer and the activation function type of the nonlinear layer corresponding to the combination as the optimization result within the basic operator dimension of the neural network layer.

[0204] Furthermore, the convolution type of the linear layer is: a fully connected layer, a 1D convolution layer with a kernel size of 5, a 1D convolution layer with a kernel size of 10, or a 1D convolution layer with a kernel size of 15.

[0205] Furthermore, the activation function type of the nonlinear layer is: TanH function, ELU function, Sigmoid activation function or Softmax function.

[0206] Furthermore, the calculation rule of the MAE indicator is:

[0207]

[0208] in, is the true value of the coal composition corresponding to the i-th training data; y i is the model prediction value corresponding to the i-th training data; m is the size of the data set.

[0209] In the embodiment of the present invention, the scheme of the present invention repeatedly generates N (take N=10) different initialization network configurations to repeat the search process multiple times, and these configurations have different input resolutions, network depths and network widths. The cross-validation MAE is calculated for different random network initialization configurations, and the N results are used to vote on the best settings of the basic operators in the first stage to determine the final search results.

[0210] Furthermore, the scheme of the present invention freezes the linear and nonlinear basic operator settings in several other dimensions. A similar process as the first stage is also used in the search of the remaining three dimensions. Finally, a network configuration with the best basic operator, input resolution, network depth and network width settings is generated to model NIRS-XRF dual-spectrum signals for the prediction of coal composition.

[0211] In an embodiment of the present invention, a neural network structure automatic search strategy is proposed to automatically search for the optimal configuration for predicting coal composition from NIRS-XRF dual-spectrum signals. This method avoids manual network design, so that the optimal neural network structure can be output without too much knowledge of the professional characteristics of NIRS-XRF dual-spectrum signals. The neural network structure can fully capture the complex features in NIRS and XRF data and effectively integrate the information from these two different sources. It is worth noting that the neural network automatic search strategy is also highly efficient. Assuming that the basic operator, input resolution, network depth and network width have No, Nr, Nd and Nw candidates respectively, the complexity of the exhaustive search method through enumeration is Our proposed greedy search method can reduce this complexity to This significantly improves the search efficiency and is beneficial to model updating and deployment in practical applications.

[0212] Preferably, optionally, the test results of the coal sample include: one or more of ash composition, ash content, volatile matter, hydrocarbons, ash melting point, total water, total sulfur and calorific value.

[0213] In the embodiment of the present invention, the present invention can realize the following coal detection target detection:

[0214] 1) Ash composition analysis: Coal is a complex organic matter containing various elements and compounds, among which ash is the inorganic matter left over when coal is burned, including minerals, soil, metal oxides, etc. The ash content has an important influence on the combustion characteristics and utilization value of coal.

[0215] 2) Ash analysis: Ash content is the content of non-combustible matter in coal, which can be inferred by specific characteristic peaks in the spectral signal. Accurate determination of ash content is crucial to evaluating the purity and combustion characteristics of coal.

[0216] 3) Volatile matter analysis: The volatile matter content is the content of gas and liquid volatilized in the coal during the heating process. The volatile matter content in the coal sample can be inferred by the change of the spectral signal, which is crucial for understanding the combustion characteristics and stability of coal.

[0217] 4) Carbon-hydrogen analysis: Carbon-hydrogen analysis of coal refers to the process of quantitatively analyzing the content of carbon (C) and hydrogen (H) elements in coal samples. Through carbon-hydrogen analysis, we can understand the content ratio of carbon and hydrogen elements in coal, and then infer the calorific value, combustion characteristics and chemical properties of coal.

[0218] 5) Ash melting point analysis: The ash melting point of coal refers to the temperature at which the ash in the coal melts during the heating process at high temperature. The ash melting point can reflect the fusibility of the ash in the coal. Coal with a low ash melting temperature is prone to form slag during the combustion process, which has an adverse effect on the combustion equipment and the environment; while coal with a higher ash melting temperature is more suitable for combustion, reducing the ash slag problem during the combustion process. Ash melting point analysis can help researchers and engineers select the appropriate type of coal, optimize the combustion process, reduce environmental pollution, and improve energy efficiency.

[0219] 6) Total moisture analysis: Total moisture analysis of coal refers to the process of analyzing and measuring the content of all moisture in a coal sample. The results of total moisture analysis of coal can help researchers and engineers understand the moisture content in coal, which in turn affects the combustion performance, combustion efficiency and temperature control of coal during combustion. Coal with high moisture content will consume more heat to evaporate moisture during combustion, reducing combustion efficiency and increasing smoke emissions during combustion.

[0220] 7) Total sulfur analysis: Total sulfur analysis of coal refers to the process of analyzing and measuring the content of all sulfur elements in coal samples. The results of total sulfur analysis of coal can help researchers and engineers understand the sulfur content in coal, and then evaluate the combustion characteristics of coal, the sulfur emissions in the combustion products, and the possible environmental impacts during the combustion process. High-sulfur coal will produce sulfur oxides and sulfuric acid mist during the combustion process, which will have a negative impact on the environment and health.

[0221] 8) Calorific value analysis: Calorific value analysis of coal refers to the process of measuring and analyzing the heat released when coal is burned. The calorific value of coal is one of the important indicators for evaluating coal combustion performance and energy efficiency. Coal with high calorific value usually has higher combustion efficiency, can provide more heat energy, and reduce energy consumption costs. Therefore, calorific value analysis of coal is of great significance for selecting suitable fuels, optimizing combustion processes, and evaluating energy efficiency.

[0222] Preferably, Fig. 9 The system further includes an output unit for visually outputting the coal detection results. The visual outputting of the coal detection results includes: determining the corresponding detection result object in response to a user data query instruction; selecting a preset data visualization scheme based on the corresponding detection result object, and pushing the visualized data to the user end.

[0223] Furthermore, for different test result objects, the system can select a preset data visualization scheme to display the test results in the form of intuitive charts, curves or images, which is convenient for users to understand and analyze. After data visualization processing, the system pushes the visualized data to the user end, and users can intuitively view the test results of coal samples through the interface to help them make decisions and evaluations.

[0224] Fig.10 FIG. 1 is a flow chart of the steps of a coal detection method provided by an embodiment of the present invention. Fig.10 As shown, an embodiment of the present invention provides a coal detection method, the system comprising:

[0225] Step S10: Execute the spectral data collection of the detected coal sample.

[0226] Specifically, near infrared spectroscopy (NIRS) is a valuable non-destructive analytical technique for detecting organic materials, which has been widely used and recognized in multiple research works. Since coal is mainly composed of organic matter, containing various functional groups and minerals, NIRS is very suitable for coal quality analysis. Specifically, the spectral range of NIRS includes a broad absorption band that is sensitive to hydrogen-containing organic functional groups, which helps to accurately and reliably analyze coal quality. In contrast, X-ray fluorescence spectrometry (XRF) is usually used for elemental analysis. In this process, high-energy X-rays are used to irradiate materials, excite atoms and induce energy absorption. When the excited atoms undergo electronic transitions back to the ground state, they release energy in the form of fluorescence radiation. This fluorescence radiation can be used to detect inorganic ash-forming elements such as aluminum (Al), silicon (Si), calcium (Ca) and other similar elements. By combining NIRS and XRF techniques, we can have a more comprehensive understanding of the organic and inorganic composition of coal, thereby more accurately evaluating its quality. This integrated approach will help improve the efficiency and accuracy of coal quality detection and promote the sustainable development of the coal industry.

[0227] Recently, there have been some studies based on NIRS-XRF dual spectrum fusion to do coal sample detection work. These studies usually use simple network structures and are not accurately adapted to the coal sample detection task. Near infrared spectroscopy signals (NIRS) perform well in detecting molecular-level information, while X-ray fluorescence spectroscopy (XRF) is good at extracting atomic-level information. Combining NIRS and XRF sensors with different material perception capabilities poses a major challenge to linear-based techniques (such as partial least squares regression) and simple network structures.

[0228] In order to solve this problem and ensure the accuracy of dual-spectrum signals in coal sample detection, the present invention proposes a corresponding dual-spectrum signal fusion scheme and an adaptive neural network training scheme, which achieves the technical effect of accurately predicting coal composition based on target spectral data. Based on this, the present invention needs to collect dual-spectrum information of coal samples for detection.

[0229] Furthermore, the solution of the present invention will also test the reflectivity of the coal, that is, a corresponding optical fiber is set to recover the reflected light of the sample.

[0230] In a possible implementation, Fig.11 Before step S10, the method further includes: collecting raw coal samples, and performing sample preparation processing on the raw coal samples to obtain test coal samples.

[0231] Specifically, the raw coal sample preparation process includes sampling, crushing, processing and shaping, such as Fig.12 , specifically including:

[0232] Step S101: Execute raw coal sampling.

[0233] Specifically, in the coal industry, sampling is a crucial link, which directly affects the accuracy and reliability of subsequent coal quality analysis and utilization. The scheme of the present invention proposes an intelligent sampling device, which combines sensing technology and automatic control system to realize intelligent identification and collection of raw coal samples. By carrying high-precision sensors and image recognition technology, it can monitor the flow of coal in the process of raw coal transportation or storage in real time, and perform random sampling according to preset algorithms and rules. By using data analysis and simulation technology, the selection of sampling positions is optimized to ensure the representativeness and comprehensiveness of the sampling points. By analyzing the coal flow path and speed, the optimal sampling position is determined to avoid sampling bias and local problems and improve the reliability of sampling. A real-time monitoring and feedback mechanism is introduced to monitor key parameters in the sampling process, such as sampling volume, sampling frequency, etc., and the sampling strategy is adjusted in time. Through data collection and analysis, real-time monitoring and quality control of the sampling process are realized to ensure the accuracy and reliability of the sampling results. An automated sampling system is designed to realize the automated collection and processing of raw coal samples. Combined with machine learning and artificial intelligence technology, the sampling algorithm and process are optimized to improve sampling efficiency and automation, and reduce human intervention and errors. The design of the intelligent sampling device and optimized sampling position of the present invention can improve the accuracy and representativeness of sampling, reduce sampling errors and deviations, and improve the reliability of sampling results. The introduction of real-time monitoring and feedback mechanism can optimize the sampling process, improve sampling efficiency and speed, and save manpower and time costs. The design of the automated sampling system can reduce human intervention, reduce operational risks and errors, and improve the controllability and stability of the sampling process. Fig. 9 , including the following steps:

[0234] Step S102: Execute raw coal pulverization processing.

[0235] Specifically, in the process of coal processing, crushing is an indispensable link, but it also generates a lot of dust pollution. Integrate efficient dust removal equipment in the crushing module, such as bag dust collector or electrostatic precipitator, to capture and filter dust particles generated during the crushing process. Through equipment integration, timely removal and treatment of dust during the crushing process can be achieved to ensure the cleanliness of the production environment and the health of employees. Adopt wet dust removal technology, through the spray system or wet scrubber, humidify the dust during the crushing process to reduce the diffusion and flying of dust. Wet dust removal technology can effectively control dust emissions and reduce the impact on air quality. Design the negative pressure closed structure of the crushing module, control the flow of dust in the closed space through the negative pressure system, reduce dust leakage and diffusion, and the negative pressure closed design can effectively prevent dust pollution and protect the surrounding environment and equipment. Introduce an online monitoring and control system to monitor the dust emission concentration and particle size distribution in real time, adjust the operating parameters and cleaning cycle of the dust removal equipment according to the monitoring data, and ensure the stability and efficiency of the dust removal effect through real-time monitoring and control. The scheme of the present invention effectively controls dust emission, reduces environmental pollution, and protects the surrounding ecological environment through the integration of dust removal equipment and the application of wet dust removal technology. The negative pressure closed design and the introduction of the online monitoring system of the scheme of the present invention can protect the health of employees, reduce the harm of dust to the respiratory tract and skin, and improve the comfort of the working environment.

[0236] Step S103: Processing the raw coal into test coal samples.

[0237] Specifically, the scheme of the present invention introduces a sample grading and screening device in the processing module, and performs fine screening and grading treatment on the samples according to the particle size and shape characteristics of the crushed coal samples. Through grading and screening, basic coal samples of different particle size ranges can be obtained, providing more accurate and representative samples for subsequent detection. The scheme of the present invention designs a chemical treatment reaction tank, which is used to chemically treat and react the crushed coal samples, extract target components or remove interfering substances. The chemical treatment reaction tank can change the chemical properties of the coal samples, making them more suitable for subsequent detection and analysis, and improving the accuracy of the detection results. The scheme of the present invention introduces a magnetic separation device, which performs magnetic separation treatment on the magnetic impurities or magnetic minerals in the crushed coal samples, and separates the magnetic substances from the basic coal samples. The magnetic separation device can effectively remove interfering substances, purify the basic coal samples, and improve the accuracy and reliability of the detection. A drying and dehumidification system is designed to dry and dehumidify the treated basic coal samples to ensure the dry state and stability of the samples. The drying and dehumidification system can avoid the influence of moisture on the detection results and ensure the accuracy and reliability of the detection data.

[0238] Step S104: performing shaping processing on the detected coal sample.

[0239] Specifically, the appearance of the coal sample may have the following effects on the results of coal composition monitoring based on spectral signals:

[0240] 1) Influence of light path scattering: The irregular shape or rough surface of coal samples may cause light path scattering, causing the light signal to be deflected or scattered during transmission, affecting the collection and accuracy of spectral signals.

[0241] 2) Differences in light absorption: The different shapes of coal samples may lead to different light absorption capabilities in different parts, which in turn affects the intensity and characteristics of the spectral signal. For example, the differences in particle size, shape and density on the surface of coal will affect the transmission and absorption of light.

[0242] 3) Surface reflection effect: The light reflection characteristics of the coal sample surface will affect the collection and analysis of spectral signals. Coal samples of different shapes may have different reflectivities, resulting in different reflection and diffuse reflection of light signals on the sample surface, affecting the quality and accuracy of spectral signals.

[0243] 4) Spectral signal interference: The uneven appearance of coal samples may cause interference with spectral signals. For example, the presence of impurities, oxides or moisture on the sample surface will affect the purity and stability of the spectral signal, thereby affecting the accuracy of component monitoring.

[0244] 5) Difference in optical path length: The different sizes and shapes of coal samples may lead to differences in optical path length, causing different degrees of attenuation and absorption of the light signal during transmission inside the sample, affecting the intensity and clarity of the spectral signal.

[0245] In the embodiment of the present invention, the appearance of the coal sample has a certain influence on the results of coal composition monitoring based on spectral signals. Therefore, when performing composition monitoring, it is necessary to consider and perform appropriate correction and processing on the appearance factor to improve the accuracy and reliability of the monitoring results.

[0246] Furthermore, although the scheme of the present invention proposes a corresponding scheme for limiting the shape of the coal sample, in order to ensure that the shape shaping meets expectations, the shaping result needs to be tested. While collecting spectral signals, the scheme of the present invention will also collect visual signals, such as image information, and perform target detection coal sample shape recognition based on the image information to determine whether it meets expectations. If it does not meet expectations, it is necessary to output an alarm instruction to remind relevant personnel to replace the sample to avoid obtaining erroneous detection results.

[0247] Preferably, the surface flatness identification of the coal sample to be detected based on the visual image data includes: performing denoising, graying and edge detection processing on the visual image in sequence, calibrating the coal sample detection area; extracting features related to surface flatness in the coal sample detection area; performing surface flatness index identification on the features related to surface flatness, and determining the surface flatness of the coal sample to be detected based on the identification result. The features related to surface flatness include: one or more of texture features, color features and edge features. The surface flatness index includes: average grayscale and / or texture uniformity.

[0248] Specifically, the following steps are included:

[0249] 1) Data acquisition and preprocessing: Collect visual image data of coal samples to ensure image clarity and consistency. De-noise the visual images to eliminate the impact of noise on subsequent processing. Convert the processed images to grayscale images to simplify the processing and highlight surface features.

[0250] 2) Surface flatness identification processing: Perform edge detection processing to accurately calibrate the boundaries and contours of the test coal sample. Calibrate the test coal sample area to ensure that subsequent processing is focused on the sample surface. Extract features related to surface flatness in the test coal sample area, including texture features, color features, and edge features. Feature extraction and surface flatness index identification:

[0251] 3) Analyze and process the extracted features to determine features related to surface flatness. Calculate surface flatness indicators such as average grayscale and texture uniformity based on one or more of texture features, color features, and edge features.

[0252] 4) Surface flatness identification and result determination: According to the identification results, the surface flatness level of the tested coal sample is determined. Combined with the surface flatness index and characteristic analysis results, the surface flatness of the tested coal sample is evaluated and classified.

[0253] In an embodiment of the present invention, the surface flatness recognition of the coal sample based on visual image data can be effectively applied to quality control and production process monitoring in the coal industry to improve production efficiency and product quality management level.

[0254] Specifically, the appearance evaluation of the detected coal sample based on the fitted outer dimensions and the fitted surface flatness includes: calculating the Euclidean distance between the fitted outer dimensions and the preset outer dimensions as the first deviation value; calculating the Euclidean distance between the fitted surface flatness and the preset surface flatness as the second deviation value; performing normalization processing on the first deviation value and the second deviation value, and performing arithmetic mean calculation on the normalized first deviation value and the second deviation value to obtain a deviation coefficient; and assigning a preset correction coefficient in the coal sample detection model based on the deviation coefficient.

[0255] In an embodiment of the present invention, the scheme of the present invention can also integrate the detected coal samples with certain differences in appearance into the subsequent detection model by means of a deviation correction coefficient. This ensures that even if there are certain differences in the detected coal samples, unified standard processing can be achieved by correcting the deviation coefficient.

[0256] Step S20: Execute the spectral data reasoning based on the pre-trained coal sample detection model to obtain the reasoning result.

[0257] Specifically, a fusion process is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data in the detection data to obtain target spectrum data, and model training is performed based on the fusion signal. Preprocessing is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively; downsampling is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data after preprocessing respectively, and the downsampled near-infrared spectrum signal and the X-ray fluorescence spectrum data are scaled to a normal distribution; splicing is performed on the normally distributed near-infrared spectrum signal and the normally distributed X-ray fluorescence spectrum data to obtain target spectrum data. Specifically, as Fig.13 , including the following steps:

[0258] Step S201: Preprocessing the near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively.

[0259] Specifically, SG convolution smoothing processing is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data respectively; and area normalization processing is performed on the near-infrared spectrum signal after the convolution smoothing processing.

[0260] Preferably, the SG convolution smoothing processing is performed on the near-infrared spectral signal and the X-ray fluorescence spectral data respectively, including: calculating the coefficients of the SG convolution kernel based on a preset window size and polynomial order; in the near-infrared spectral signal and the X-ray fluorescence spectral data, respectively, weighted averaging the coefficients in the convolution kernel with adjacent data points in each spectral signal to obtain smoothed data points; performing symmetric expansion or zero filling processing on the boundary of each spectral signal; and obtaining the spectral signal subjected to SG convolution smoothing processing based on the smoothed data points and the processed boundaries.

[0261] In an embodiment of the present invention, SG convolution smoothing is a linear smoothing method. By locally fitting the data, the data curve can be better smoothed, the spikes and fluctuations in the data can be eliminated, and the data can be made more stable and continuous. SG convolution smoothing can effectively smooth the noise part of the data while retaining the characteristics and trends of the signal, which helps to reduce the interference of high-frequency noise in the data on the signal and improve the readability and analysis accuracy of the data. Compared with other smoothing methods, SG convolution smoothing can better maintain the overall shape and trend of the data while smoothing the data, will not cause distortion or offset of the data shape, and maintains the original characteristics of the data. SG convolution smoothing is a simple and efficient data smoothing method with a fast calculation speed. It is suitable for the processing of large-scale data sets and can complete data smoothing operations in a shorter time. It is suitable for large-scale coal sample detection, ensuring detection accuracy while improving detection efficiency.

[0262] Furthermore, the area normalization processing of the near-infrared spectral signal after convolution smoothing includes: determining a signal set of the near-infrared spectral signal after convolution smoothing; the signal set includes multiple signal samples, and the reflectivity of each signal sample under the corresponding spectral type at each preset wavelength point; calculating the signal after area normalization processing based on the signal set of the near-infrared spectral signal.

[0263] In an embodiment of the present invention, baseline drift is prone to occur in near-infrared spectral signals, so the scheme of the present invention will also eliminate the baseline drift in the near-infrared spectral signal by means of area normalization. Baseline drift is a signal deviation caused by factors such as instrument drift, environmental changes or sample differences, which affects the accuracy and stability of the spectral signal. Through area normalization, the signal can be moved up or down as a whole, so that the baseline level is more stable. The optical path difference in the near-infrared spectral signal will lead to differences in signal intensity, affecting the comparison and analysis of the signal. The area normalization process can normalize the overall intensity of the signal, reduce the impact of the optical path difference on the signal, and make the comparison between different samples more accurate. The area normalization process can highlight the characteristic peaks or troughs in the near-infrared spectral signal, making the characteristics of the signal more obvious and prominent, which helps to more accurately identify and analyze specific components or features in the spectrum.

[0264] Specifically, the calculation rule of the signal after area normalization is:

[0265]

[0266] in, Indicates the NIR reflectance corresponding to the mth wavelength point of the nth sample; is the signal after area normalization.

[0267] Step S202: performing downsampling processing on the near-infrared spectrum signal and the X-ray fluorescence spectrum data after preprocessing respectively.

[0268] Specifically, filtering processing is performed on each spectral signal respectively, and in the spectral signal after filtering processing, one sampling point is retained at a fixed interval to obtain multiple sampling points; or the spectral signal after filtering processing is cut into multiple segments, and averaging processing is performed on each sampling point in each segment, and one sampling point is obtained in each segment to obtain multiple sampling points; signal reconstruction is performed based on each sampling point to obtain a spectral signal after downsampling processing.

[0269] In an embodiment of the present invention, downsampling processing can reduce the amount of data, that is, reduce the sampling rate or the number of sampling points, thereby saving storage space and computing resources. In particular, for large-scale data sets or data sampled at high frequencies, downsampling can effectively reduce the volume of data, making it easier to store and process. Downsampling can simplify the data analysis process, reduce the complexity and dimension of the data, and make the data easier to understand and process. By reducing the resolution of the data, some detailed information can be removed, the main features of the data can be highlighted, and the model establishment and analysis process can be simplified. Downsampling processing can help remove noise and interference in the data, smooth the data curve, and improve the quality and stability of the data. By reducing the sampling rate or averaging the sampling points, the volatility of the data can be reduced, making the data clearer and more reliable. The scheme of the present invention can reduce the complexity and dimension of the data, reduce the amount of data, and make the data easier to process and analyze by retaining the interval sampling points and multi-segment averaging. And reduce the noise and interference in the signal, the interval sampling point retention and averaging processing can further smooth the signal curve, reduce the volatility of the data, and improve the signal quality and stability.

[0270] Step S203: scaling the downsampled near-infrared spectrum signal and X-ray fluorescence spectrum data to a normal distribution.

[0271] Specifically, the statistical characteristics of each spectral signal are calculated respectively; wherein the statistical characteristics are variance and / or standard deviation; based on the statistical characteristics of each spectral signal, each spectral signal is subjected to standardization or normalization processing, and the value of each spectral signal is scaled to a preset value range to obtain the scaled value of each spectral signal; based on a preset transformation algorithm and the scaled value of each spectral signal, each spectral signal is converted into a normal distribution.

[0272] In an embodiment of the present invention, the scheme of the present invention converts the signal into a normal distribution to simplify the data analysis process, making the data easier to understand and process. By converting the signal into a normal distribution, the fitting effect and prediction accuracy of the model can be improved, and converting the signal into a normal distribution can better meet the requirements of hypothesis testing and ensure the validity of the test results. And the normal distribution has the characteristic of standardization, that is, the mean is 0 and the standard deviation is 1. Converting the signal into a normal distribution can achieve data standardization, so that different signals are comparable, which is convenient for subsequent splicing of two spectral signals.

[0273] Step S204: performing splicing processing on the normally distributed near-infrared spectrum signal and the normally distributed X-ray fluorescence spectrum data to obtain target spectrum data.

[0274] Specifically, an alignment operation of a normally distributed near-infrared spectral signal and a normally distributed X-ray fluorescence spectral data is performed; after completing the alignment of the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data, the two spectral signals are spliced ​​based on weighted average to obtain initial target spectral data; the initial target spectral data is verified based on the normally distributed near-infrared spectral signal and / or the normally distributed X-ray fluorescence spectral data, and the verified initial target spectral data is used as the target spectral data.

[0275] Furthermore, the operation of performing the alignment of the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data includes: selecting the same reference point in the normally distributed near-infrared spectral signal and the normally distributed X-ray fluorescence spectral data, and respectively calculating the time difference between the corresponding reference point and other data points in each spectral signal to obtain a time difference sequence; and performing a time axis adjustment of one of the spectral signals based on a random interpolation method and the corresponding time difference sequence until the time axes of the two spectral signals are aligned.

[0276] Furthermore, the random interpolation method is:

[0277] X aug =αX 1 +(1-α)X 2 ,α∈[0,1]

[0278] y aug =f PLS (X aug )

[0279] Among them, f PLS represents the PLS mapping function; X aug and aug denote the synthesized NIRS-XRF signal and the corresponding pseudo-label respectively; α denotes the interpolation weight randomly sampled in the range [0,1]; X 1 and X2 Signal samples randomly selected from historical NIRS-XRF signals of coal samples.

[0280] In an embodiment of the present invention, the scheme of the present invention performs two spectral signal splicing based on the random interpolation method, which can retain the characteristics and information of the original data during the signal splicing process and avoid data loss or distortion. Splicing the signals by interpolation can better maintain the integrity of the data. The random interpolation method can achieve a smooth transition in the transition area of ​​the signal splicing to avoid mutations or discontinuities, which helps to improve the continuity and stability of the signal splicing. The random interpolation method can adjust the interpolation parameters as needed, such as the number of interpolation points, the selection of interpolation functions, etc., so as to flexibly control the effect of signal splicing, which makes the signal splicing process more customizable and can be customized based on the user's test requirements. The random interpolation method can effectively reduce the artifacts and distortion that may occur during the signal splicing process and improve the quality and accuracy of the signal splicing. By reasonably selecting the interpolation method and parameters, the error introduced by the signal splicing can be reduced, thereby improving the subsequent detection accuracy.

[0281] Furthermore, the scheme of the present invention uses a deep neural network to construct a prediction model for coal composition prediction, and establishes a mapping between the input NIRS-XRF fusion spectrum and the corresponding real component label through an end-to-end training method. represents the input signal after preprocessing, represents the true value of the corresponding coal composition. n ∈R D is the concatenation of the NIRS and XRF spectra after preprocessing in step S20, y n is a non-negative scalar label. N and D are the number of training set samples and the input resolution, respectively. Note that since X n It is possible to downsample from the original data, so D can be smaller than the resolution of the original data. f(·|θ) is a deep neural network with θ as a parameter, which can consist of multiple linear and nonlinear layers. The nonlinear layer here is also called an activation layer. A linear layer is usually followed by a nonlinear layer, and the combination of the two is called a module. In this paper, the network consists of multiple identical modules stacked in sequence and ends with a fully connected layer, outputting a scalar value y. Specifically, the base operator of the linear layer can be a fully connected layer or a 1D convolutional layer with different kernel sizes, while the base operator of the nonlinear layer can be a hyperbolic tangent function (TanH), an exponential linear unit (ELU), or a sigmoid function, etc.

[0282] Preferably, the method also includes: pre-training a coal sample detection model, including: collecting historical near-infrared spectral signals and historical X-ray fluorescence spectral data, and constructing corresponding historical target spectral data based on the historical near-infrared spectral signals and the historical X-ray fluorescence spectral data; using the historical target spectral data as training data to perform PLS model parameter initialization; generating simulated samples based on the initialized PLS model, performing model training based on the simulated samples, and obtaining an initial model for coal sample detection; and verifying the initial model for coal sample detection based on the reserved historical target spectral data to obtain a coal sample detection model.

[0283] In the embodiment of the present invention, in order to obtain an accurate detection model, a large amount of historical data is actually required as training samples to perform model training. However, it is actually difficult to obtain comprehensive historical data. On the one hand, there is the problem of historical data retention, and on the other hand, there is the problem of data interoperability. Therefore, it is actually difficult to train and obtain an accurate detection model based on the existing historical data. Based on this, the present invention proposes a solution for performing pre-training based on the PLS model.

[0284] In an embodiment of the present invention, historical target spectral data is used as training data to initialize the parameters of the partial least squares regression (PLS) model, laying the foundation for subsequent model training. Based on the initialized PLS model, simulated sample data is generated for model training, the training data set is expanded, and the generalization ability of the model is improved. The model is trained using simulated sample data, and the initial model for coal sample detection is established by learning the patterns and characteristics of historical data.

[0285] Preferably, the PLS model parameters include: initial weight, learning rate, activation function, regularization parameter, initialization bias term, and optimizer type.

[0286] Preferably, the generation of simulated samples based on the initialized PLS model includes: adaptively determining the signal type and signal parameters that match the target spectral data, and generating a basic signal based on the signal type and the model parameters; adaptively selecting an augmentation scheme, and performing a data augmentation operation on the basic signal based on the selected augmentation scheme to obtain an augmented signal set; screening out deviation data in the augmented signal set, and using the screened-out augmented signal set as a simulated sample.

[0287] In an embodiment of the present invention, adaptive algorithms and machine learning techniques are used to determine the best signal type and parameter combination to match the characteristics and changes of different spectral signals. Based on the determined signal type and parameters, a basic signal is generated as the basis for the augmented signal set. According to the data characteristics and model requirements, an augmentation scheme is adaptively selected, including data interpolation, noise reduction processing, etc., to improve data quality and model performance. The selected augmentation scheme, such as data interpolation, noise addition and other operations, is executed on the basic signal to generate a diversified augmented signal set. After completing the data augmentation operation, an augmented signal set containing diversified signals is obtained, which is used to enrich the training data and improve the generalization ability and robustness of the model. Deviation data screening is performed in the augmented signal set to identify and eliminate data points that may introduce errors, thereby ensuring the accuracy and reliability of the training data. Model training and verification improve the model's adaptability to uncertainty and noise.

[0288] In an embodiment of the present invention, the adaptively selected augmentation scheme includes: randomly selecting one or more schemes from noise addition, translation, scaling, rotation, shearing and transformation as pre-selected schemes; randomly adjusting parameters within a preset adjustable parameter range of each pre-selected scheme to obtain a pre-selected scheme determined by the parameters; if there is only one pre-selected scheme determined by the parameters, directly processing the basic signal based on the scheme to obtain an augmented signal; if there are multiple pre-selected schemes determined by the parameters, executing each scheme in turn to obtain an augmented signal.

[0289] Furthermore, the method of screening out deviation data in the augmented signal set and using the screened augmented signal set as a simulation sample includes: respectively calculating the Euclidean distance between each augmented signal in each augmented signal set and the basic signal; screening out augmented signals whose Euclidean distance is greater than a preset Euclidean distance threshold, and using the screened augmented signal set as a simulation sample.

[0290] In an embodiment of the present invention, the scheme of the present invention adaptively determines the augmentation scheme, and can randomly generate a large number of simulation signals to simulate the detection signals under various coal conditions, thereby ensuring the comprehensiveness of the data and making up for the problem of being unable to train an accurate detection model due to insufficient historical retained data.

[0291] Furthermore, the model training is performed based on the simulated samples to obtain an initial model for coal sample detection, including: pseudo-labeling the simulated samples based on the PLS model after parameter initialization, and using the labeled simulated samples as training samples; based on the training samples, model training is performed in a neural network based on the target spectral data search to obtain an initial model for coal sample detection.

[0292] In an embodiment of the present invention, after obtaining the augmented data set, it is also necessary to annotate the coal components corresponding to the data set, so as to facilitate model training based on the coal components corresponding to the simulated signal. Based on this, it is necessary to annotate the coal components of the augmented signals in the augmented signal set, so that each augmented signal corresponds to a simulated coal component prediction result. After the annotation is completed, the corresponding training samples are obtained, and the corresponding detection model can be obtained by performing model training based on the training samples. However, in order to ensure the accuracy of the model, the model needs to be verified.

[0293] Furthermore, the coal sample detection initial model is verified based on the reserved historical target spectral data to obtain the coal sample detection model, including: constructing a verification set based on the reserved historical target spectral data, and inputting the verification set into the component monitoring initial model to obtain the corresponding prediction result; performing performance evaluation of the component monitoring initial model based on the prediction result and the actual result of the corresponding reserved historical target spectral data; if the performance evaluation of the component monitoring initial model fails, adjusting the hyperparameters, optimizing the model structure and / or adding regularization operations to obtain an updated component monitoring initial model; re-performing a performance evaluation on the updated component monitoring initial model based on the reserved historical target spectral data until a component monitoring initial model that meets the preset performance requirements is obtained as the coal sample detection model.

[0294] In an embodiment of the present invention, the validation set is input into the initial model of coal sample detection to obtain the predicted results, and the prediction results are compared with the actual results to perform performance evaluation and error analysis. According to the performance evaluation results, if the model does not pass the preset performance requirements, the hyperparameters are adjusted, the model structure is optimized, and the regularization operation is added to improve the accuracy and stability of the model. According to the adjusted model, the performance evaluation is re-performed until the initial model of coal sample detection that meets the preset performance requirements is obtained as the updated coal sample detection model. The scheme of the present invention realizes continuous optimization and iteration of the coal sample detection model through repeated verification, evaluation and adjustment, thereby improving the performance and accuracy of the model. Based on the validation set of the reserved historical target spectral data, the performance and generalization ability of the model can be more accurately evaluated, and the reliability of the model evaluation can be improved. According to the performance evaluation results, the hyperparameters and model structure are adjusted in time to make the model better adapt to the data characteristics and task requirements, thereby improving the accuracy of coal sample detection. By adding regularization operations and other means, the stability and generalization ability of the model are improved, the risk of overfitting is reduced, and the robustness of the model is enhanced.

[0295] Preferably, the performance evaluation of the initial component monitoring model is performed based on the prediction results and the actual results of the corresponding reserved historical target spectral data, including: comparing the prediction results and the actual results of the corresponding reserved historical target spectral data, and evaluating the accuracy and recall of the model based on the deviation between the two; if any evaluation result of the accuracy and recall fails, the performance evaluation of the initial component monitoring model fails.

[0296] Preferably, a neural network search is performed based on the target spectral data, including: dividing a plurality of optimization dimensions based on neural network structure parameters; in each optimization dimension, adaptively adjusting the parameters in the neural network structure corresponding to the optimization dimension, and calculating the MAE index after each adjustment; comparing the MAE index corresponding to each parameter, and determining the parameter combination with the highest MAE index as the optimization result in the corresponding optimization dimension; performing a greedy search between the optimization dimensions to obtain the optimization results of each optimization dimension; based on the optimization results of each optimization dimension, determining the structural parameters corresponding to each neural network structure, and constructing a neural network corresponding to the search result.

[0297] Furthermore, the optimization dimensions include: the basic operator dimension of the neural network layer, the resolution dimension of the input data, the network depth dimension, and the network width dimension.

[0298] In the embodiment of the present invention, the neural network dimensions that affect the fusion of NIRS-XRF dual spectra include the basic operators of the linear and nonlinear layers, the resolution of the input data, the network depth (the number of stacked modules) and the network width (the output channels of the linear layer). The scheme of the present invention decomposes the search process of these four dimensions into four stages, each stage searches a single dimension and performs greedy search between the dimensions.

[0299] In an embodiment of the present invention, within the basic operator dimension of the neural network layer, the corresponding optimization result obtaining rules include: traversing the convolution type of the linear layer, and traversing the activation function type of the nonlinear layer, adaptively combining the convolution type of the fully connected layer and the activation function type of the nonlinear layer, and calculating the MAE index after each combination; comparing the MAE index after each combination, selecting the combination corresponding to the maximum MAE index, and determining the convolution type of the fully connected layer and the activation function type of the nonlinear layer corresponding to the combination as the optimization result within the basic operator dimension of the neural network layer.

[0300] Furthermore, the convolution type of the linear layer is: a fully connected layer, a 1D convolution layer with a kernel size of 5, a 1D convolution layer with a kernel size of 10, or a 1D convolution layer with a kernel size of 15.

[0301] Furthermore, the activation function type of the nonlinear layer is: TanH function, ELU function, Sigmoid activation function or Softmax function.

[0302] Furthermore, the calculation rule of the MAE indicator is:

[0303]

[0304] in, is the true value of the coal composition corresponding to the i-th training data; y i is the model prediction value corresponding to the i-th training data; m is the size of the data set.

[0305] In the embodiment of the present invention, the scheme of the present invention repeatedly generates N (take N=10) different initialization network configurations to repeat the search process multiple times, and these configurations have different input resolutions, network depths and network widths. The cross-validation MAE is calculated for different random network initialization configurations, and the N results are used to vote on the best settings of the basic operators in the first stage to determine the final search results.

[0306] Furthermore, the scheme of the present invention freezes the linear and nonlinear basic operator settings in several other dimensions. A similar process as the first stage is also used in the search of the remaining three dimensions. Finally, a network configuration with the best basic operator, input resolution, network depth and network width settings is generated to model NIRS-XRF dual-spectrum signals for the prediction of coal composition.

[0307] In an embodiment of the present invention, a neural network structure automatic search strategy is proposed to automatically search for the optimal configuration for predicting coal composition from NIRS-XRF dual-spectrum signals. This method avoids manual network design, so that the optimal neural network structure can be output without too much knowledge of the professional characteristics of NIRS-XRF dual-spectrum signals. The neural network structure can fully capture the complex features in NIRS and XRF data and effectively integrate the information from these two different sources. It is worth noting that the neural network automatic search strategy is also highly efficient. Assuming that the basic operator, input resolution, network depth and network width have No, Nr, Nd and Nw candidates respectively, the complexity of the exhaustive search method through enumeration is Our proposed greedy search method can reduce this complexity to This significantly improves the search efficiency and is beneficial to model updating and deployment in practical applications.

[0308] In another possible implementation, Fig.14 The method also includes visually outputting the coal detection results.

[0309] Specifically, the test results of the coal sample include: one or more of ash composition, ash content, volatile matter, carbon and hydrogen, ash melting point, total water, total sulfur and calorific value.

[0310] In the embodiment of the present invention, the present invention can realize the following coal detection target detection:

[0311] 1) Ash composition analysis: Coal is a complex organic matter containing various elements and compounds, among which ash is the inorganic matter left over when coal is burned, including minerals, soil, metal oxides, etc. The ash content has an important influence on the combustion characteristics and utilization value of coal.

[0312] 2) Ash analysis: Ash content is the content of non-combustible matter in coal, which can be inferred by specific characteristic peaks in the spectral signal. Accurate determination of ash content is crucial to evaluating the purity and combustion characteristics of coal.

[0313] 3) Volatile matter analysis: The volatile matter content is the content of gas and liquid volatilized in the coal during the heating process. The volatile matter content in the coal sample can be inferred by the change of the spectral signal, which is crucial for understanding the combustion characteristics and stability of coal.

[0314] 4) Carbon-hydrogen analysis: Carbon-hydrogen analysis of coal refers to the process of quantitatively analyzing the content of carbon (C) and hydrogen (H) elements in coal samples. Through carbon-hydrogen analysis, we can understand the content ratio of carbon and hydrogen elements in coal, and then infer the calorific value, combustion characteristics and chemical properties of coal.

[0315] 5) Ash melting point analysis: The ash melting point of coal refers to the temperature at which the ash in the coal melts during the heating process at high temperature. The ash melting point can reflect the fusibility of the ash in the coal. Coal with a low ash melting temperature is prone to form slag during the combustion process, which has an adverse effect on the combustion equipment and the environment; while coal with a higher ash melting temperature is more suitable for combustion, reducing the ash slag problem during the combustion process. Ash melting point analysis can help researchers and engineers select the appropriate type of coal, optimize the combustion process, reduce environmental pollution, and improve energy efficiency.

[0316] 6) Total moisture analysis: Total moisture analysis of coal refers to the process of analyzing and measuring the content of all moisture in a coal sample. The results of total moisture analysis of coal can help researchers and engineers understand the moisture content in coal, which in turn affects the combustion performance, combustion efficiency and temperature control of coal during combustion. Coal with high moisture content will consume more heat to evaporate moisture during combustion, reducing combustion efficiency and increasing smoke emissions during combustion.

[0317] 7) Total sulfur analysis: Total sulfur analysis of coal refers to the process of analyzing and measuring the content of all sulfur elements in coal samples. The results of total sulfur analysis of coal can help researchers and engineers understand the sulfur content in coal, and then evaluate the combustion characteristics of coal, the sulfur emissions in the combustion products, and the possible environmental impacts during the combustion process. High-sulfur coal will produce sulfur oxides and sulfuric acid mist during the combustion process, which will have a negative impact on the environment and health.

[0318] 8) Calorific value analysis: Calorific value analysis of coal refers to the process of measuring and analyzing the heat released when coal is burned. The calorific value of coal is one of the important indicators for evaluating coal combustion performance and energy efficiency. Coal with high calorific value usually has higher combustion efficiency, can provide more heat energy, and reduce energy consumption costs. Therefore, calorific value analysis of coal is of great significance for selecting suitable fuels, optimizing combustion processes, and evaluating energy efficiency.

[0319] The method further includes: visually outputting the coal detection results. Visualizing the coal detection results includes: determining the corresponding detection result object in response to a user data query instruction; selecting a preset data visualization scheme based on the corresponding detection result object, and pushing the visualized data to the user end.

[0320] Based on the inference results, the scheme of the present invention organizes and stores the coal sample test results output by the model for subsequent display and analysis. Using data visualization technology, the coal sample test results are displayed in the form of charts, images, etc., and the component information of the coal samples is intuitively presented. A result display interface is designed, and users can input coal sample information through the interface to view the corresponding coal sample test results, thereby realizing personalized display and query of the results. The real-time update function of the results is realized. When new coal samples are tested, the displayed coal sample test results are updated in time to maintain the timeliness and accuracy of the data. The result interpretation and analysis function is provided to explain the content and significance of each component, so as to help users better understand the composition and characteristics of coal samples.

[0321] In an embodiment of the present invention, the coal sample test results are displayed through data visualization, making complex data intuitive and easy to understand, thereby improving the user's understanding and analysis capabilities of the coal sample components. A friendly result display interface is designed so that users can easily and quickly query and view the coal sample test results, thereby improving user experience and operational efficiency. The displayed results are updated in real time so that users can keep abreast of the latest coal sample test conditions and provide timely feedback and support for decision-making. Through the result interpretation and analysis functions, users are helped to gain an in-depth understanding of the content and influencing factors of coal components, thereby providing a scientific basis and guidance for coal production and utilization. The scheme of the present invention can realize the output and display of coal sample test results, provide the coal industry with more intelligent and convenient coal sample detection services, and promote the effective management and utilization of coal resources.

[0322] The embodiment of the present invention further provides a computer-readable storage medium, on which instructions are stored, which, when executed on a computer, enable the computer to execute the above-mentioned coal detection method.

[0323] Those skilled in the art can understand that all or part of the steps in the method for implementing the above-mentioned embodiments can be completed by instructing the relevant hardware through a program, and the program is stored in a storage medium, including several instructions for making a single-chip microcomputer, a chip or a processor perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory ROM, random access memory RAM, random access memory, disk or optical disk and other media that can store program codes.

[0324] The optional embodiments of the present invention are described in detail above in conjunction with the accompanying drawings. However, the embodiments of the present invention are not limited to the specific details in the above embodiments. Within the technical concept of the embodiments of the present invention, the technical scheme of the embodiments of the present invention can be subjected to a variety of simple modifications, and these simple modifications all belong to the protection scope of the embodiments of the present invention. It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the embodiments of the present invention will not further describe various possible combinations.

[0325] In addition, various embodiments of the present invention may be arbitrarily combined, and as long as they do not violate the concept of the embodiments of the present invention, they should also be regarded as the contents disclosed in the embodiments of the present invention.

Claims

1. A coal detection system, characterized in that: The system comprises: The detection unit is used to collect spectral data of the coal sample; wherein, The detection unit includes a near-infrared spectrum acquisition module and an X-ray fluorescence spectrum acquisition module; A visual analysis module, wherein the visual analysis module includes one or more image acquisition modules, each of which is arranged at each preset direction corresponding to the detection position of the detection coal sample, and is used to acquire image information of the detection coal sample at a viewing angle; The analysis unit is also used to evaluate the appearance of the detected coal sample based on the image information at each viewing angle; A training unit is used to construct a neural network based on greedy search in each neural network dimension through the spectral data, and obtain a coal sample detection model based on simulated sample training after sample augmentation; wherein, The training unit is specifically configured as follows: Collecting historical near-infrared spectrum signals and historical X-ray fluorescence spectrum data, and constructing corresponding historical target spectrum data based on the historical near-infrared spectrum signals and the historical X-ray fluorescence spectrum data; Using the historical target spectral data as training data to perform PLS model parameter initialization; Generate simulated samples based on the initialized PLS model, perform model training based on the simulated samples, and obtain an initial model for coal sample detection; Performing verification on the coal sample detection initial model based on the reserved historical target spectrum data to obtain a coal sample detection model; The analysis unit is used to perform fusion processing on the spectral data to obtain target spectral data, and perform target spectral data reasoning based on the coal sample detection model to obtain coal detection results.

2. The system according to claim 1, characterized in that The system also includes a sample preparation unit, which is used to collect coal samples and perform sample preparation processing on the coal samples to obtain test coal samples.

3. The system according to claim 2, characterized in that The sample preparation unit comprises: Sampling module, used to randomly collect raw coal samples during the transportation or storage of raw coal, or to receive raw coal samples deposited by users; A pulverizing module, used for pulverizing the raw coal sample to obtain a pulverized coal sample; A processing module is used to perform preprocessing on the pulverized coal sample to obtain a basic coal sample; wherein, Pre-processing the pulverized coal sample includes: Drying, grinding and screening; A shaping module is used to perform shaping processing on the basic coal sample to obtain a test coal sample.

4. The system according to claim 3, characterized in that The sampling module, the crushing module, the processing module and the shaping module are connected based on a transfer conveyor belt; The transfer conveyor is triggered and controlled based on a servo system.

5. The system according to claim 4, characterized in that The servo system is configured as follows: Based on the trigger signal of the position trigger at the preset position of each module of the sample preparation unit, start timing and control the power servo motor of the transfer conveyor belt to start until the predetermined running time, and then turn off the power servo motor of the transfer conveyor belt; or In response to the switch trigger signals corresponding to the modules of the sample preparation unit, the timing is started, and the power servo motor of the transfer conveyor belt is controlled to start until the predetermined running time is reached, and then the power servo motor of the transfer conveyor belt is turned off.

6. The system according to claim 3, characterized in that The shaping module comprises: A coal conveying component (1) comprises a conveying frame (11) and a conveying belt (12) rotatably arranged on the conveying frame (11); a coal sample detection inlet is provided on the conveying frame (11); A first pretreatment component (2), comprising a limit frame (21) and a scraper (22), wherein the limit frame (21) is supported above the conveyor belt (12) by the conveyor frame (11), and the limit frame (21) is located below the detection coal sample inlet, and the scraper (22) is arranged at the discharge end of the limit frame (21) and forms a first gap with the conveyor belt (12); The second pretreatment component (3) comprises at least one pressure roller (31), wherein the pressure roller (31) is rotatably arranged on the conveying frame (11), and the pressure roller (31) is located at the outlet end of the limit frame (21), a second gap connected to the first gap is formed between the pressure roller (31) and the conveying belt (12), and a limit member (32) for limiting the width of the second gap is provided on the pressure roller (31).

7. The system according to claim 1, characterized in that The detection unit also includes: A sampling cover (4) for arranging the near-infrared spectrum acquisition module and the X-ray fluorescence spectrum acquisition module; The sampling cover is also provided with an illumination module (41), an optical fiber (42) and a photoelectric sensor (43); A diaphragm (44) extending from the inner wall of the sampling cover is provided between the lighting module (41) and the optical fiber (42); The photoelectric sensor (43) is arranged in the light path direction of the lighting module and is used to monitor the light intensity of the corresponding lighting module (41).

8. The system according to claim 7, characterized in that The lighting module (41) comprises a plurality of symmetrically arranged light sources; The optical fiber (42) is arranged at the center position of the top of the sampling cover (4); Each lighting module (41) is provided with at least one corresponding photoelectric sensor (43), which is arranged on the inner wall of the sampling cover.

9. The system according to claim 8, characterized in that The analysis unit is further used to monitor the state of the corresponding lighting module based on the light intensity collected by the photoelectric sensor, including: Performing pre-processing including filtering and denoising on the illumination data collected by the photoelectric sensor; According to the voltage value or digital value of the electrical signal corresponding to the pre-processed illumination data, an interpolation or fitting calculation is performed on the calibration curve between the illumination intensity to obtain the illumination intensity at the detection position; Based on the light intensity at the detection position and the positional relationship between the light point sensor and the corresponding lighting module, performing light intensity fitting of the corresponding lighting module to obtain the detection light intensity of the corresponding lighting module; The detected light intensity is compared with a preset light intensity threshold, and if the detected light intensity is less than the preset light intensity threshold, an alarm message is output.

10. The system according to claim 1, characterized in that The detection unit also includes: A Mylar film replacement device comprises a sample conveyor belt (5), a test coal sample (6) is placed on the sample conveyor belt (5), a mounting frame (7) is provided above the test coal sample (6), a driving wheel (71) and a driven wheel (72) are provided on the mounting frame (7), a Mylar film transmission channel (73) is formed on the mounting frame (7), and a Mylar film (8) can pass through the Mylar film transmission channel (73) and abut against the lower edge of the outer peripheral surface of the driving wheel (71) and the driven wheel (72), and move under the drive of the driving wheel (71).

11. The system according to claim 10, characterized in that The coal sample appearance evaluation based on the image information at each viewing angle includes: De-noising, graying and edge detection are performed on the image information at each viewing angle in sequence, and the coal sample area is calibrated and detected; Detect the coal sample area based on the images at each viewing angle, and fit the outer dimensions of the detected coal sample; The surface flatness index is collected in the coal sample area detected by the image at each viewing angle, and the surface flatness fitting of the detected coal sample is performed based on the surface flatness index; Based on the fitted outer dimensions and fitted surface flatness, the appearance of the tested coal sample is evaluated.

12. The system according to claim 11, characterized in that The process of evaluating the appearance of the coal sample based on the fitted appearance size and the fitted surface flatness includes: Calculating the Euclidean distance between the fitted outer dimensions and the preset outer dimensions as a first deviation value; Calculating the Euclidean distance between the fitted surface flatness and the preset surface flatness as a second deviation value; Performing normalization processing on the first deviation value and the second deviation value, and performing arithmetic mean calculation on the normalized first deviation value and the second deviation value to obtain a deviation coefficient; The preset correction coefficient in the coal sample detection model is assigned a value based on the deviation coefficient.

13. The system according to claim 1, characterized in that The spectral data includes near-infrared spectral data and X-ray fluorescence spectral data; The performing fusion processing on the spectral data to obtain target spectral data comprises: A fusion process is performed on the near-infrared spectrum signal and the X-ray fluorescence spectrum data to obtain target spectrum data.

14. The system according to claim 13, characterized in that Performing fusion processing on the near-infrared spectrum data and the X-ray fluorescence spectrum data to obtain target spectrum data includes: Preprocessing the near infrared spectrum data and the X-ray fluorescence spectrum data respectively; Downsampling is performed on the preprocessed near infrared spectrum data and X-ray fluorescence spectrum data respectively, and the downsampled near infrared spectrum data and X-ray fluorescence spectrum data are scaled to a normal distribution; The normally distributed near-infrared spectrum data and the normally distributed X-ray fluorescence spectrum data are spliced ​​to obtain target spectrum data.

15. The system according to claim 14, characterized in that The preprocessing of the near infrared spectrum data and the X-ray fluorescence spectrum data respectively includes: Performing SG convolution smoothing processing on the near-infrared spectrum data and the X-ray fluorescence spectrum data respectively to obtain the near-infrared spectrum data after SG convolution smoothing processing and the X-ray fluorescence spectrum data after SG convolution smoothing processing; The near-infrared spectral data after SG convolution smoothing were subjected to area normalization.

16. The system according to claim 15, characterized in that The performing SG convolution smoothing processing on the near infrared spectrum data and the X-ray fluorescence spectrum data respectively comprises: Based on the preset window size and polynomial order, the coefficients of the SG convolution kernel are calculated; For near-infrared spectral data and X-ray fluorescence spectral data, weighted averages of adjacent data points in each spectral data are performed based on the coefficients of the SG convolution kernel to obtain smoothed data points; Performing symmetric expansion or zero filling processing on the boundaries of each spectral data; The spectral data after SG convolution smoothing is obtained based on the smoothed data points and processed boundaries.

17. The system according to claim 14, characterized in that The downsampling process is performed on the pre-processed near infrared spectrum data and X-ray fluorescence spectrum data respectively, including: Downsampling is performed on the X-ray fluorescence spectrum data after SG convolution smoothing and the near infrared spectrum data after area normalization, including: Perform filtering processing on each spectral data respectively, and retain a sampling point at a fixed interval in the spectral data after filtering processing to obtain multiple sampling points; or, cut multiple segments of the spectral data after filtering processing, and perform averaging processing on each sampling point in each segment, and obtain a sampling point in each segment to obtain multiple sampling points; Signal reconstruction is performed based on each sampling point to obtain spectral data after downsampling processing.

18. The system according to claim 14, characterized in that The downsampled near-infrared spectral data and X-ray fluorescence spectral data are scaled to a normal distribution, including: Calculating statistical features of each spectral data respectively; wherein the statistical features are variance and / or standard deviation; Based on the statistical characteristics of each spectral data, performing standardization or normalization processing on each spectral data, scaling the value of each spectral data to a preset value range, and obtaining a scaled value of each spectral data; Based on a preset transformation algorithm and a scaling value of each spectral data, each spectral data is converted into a normal distribution.

19. The system according to claim 14, characterized in that Perform splicing processing on the normally distributed near-infrared spectrum data and the normally distributed X-ray fluorescence spectrum data to obtain target spectrum data, including: Performing an alignment operation of normally distributed near-infrared spectral data and normally distributed X-ray fluorescence spectral data; After completing the alignment of the normally distributed near-infrared spectrum signal and the normally distributed X-ray fluorescence spectrum data, the two spectrum signals are spliced ​​based on weighted average to obtain the initial target spectrum data; The initial target spectrum data is verified based on the normally distributed near-infrared spectrum signal and / or the normally distributed X-ray fluorescence spectrum data, and the verified initial target spectrum data is used as the target spectrum data.

20. The system according to claim 1, characterized in that The coal sample detection model is: θ=argmax θ L2(f(X|θ),Y); Among them, f(·|θ) is a deep neural network with θ as parameter; is the target spectrum data; The component of coal.

21. The system according to claim 1, characterized in that The generating of simulation samples based on the initialized PLS model includes: Adaptively determine the signal type and signal parameters that match the target spectral data, and generate a basic signal based on the signal type and the model parameters; Adaptively selecting an augmentation scheme, and performing a data augmentation operation on the basic signal based on the selected augmentation scheme to obtain an augmented signal set; Deviation data is screened out in the augmented signal set, and the augmented signal set after the screening is used as a simulation sample.

22. The system according to claim 21, characterized in that The generating of simulation samples based on the initialized PLS model includes: Adaptively determine the signal type and signal parameters that match the target spectral data, and generate a basic signal based on the signal type and the model parameters; Adaptively selecting an augmentation scheme, and performing a data augmentation operation on the basic signal based on the selected augmentation scheme to obtain an augmented signal set; Deviation data is screened out in the augmented signal set, and the augmented signal set after the screening is used as a simulation sample.

23. The system according to claim 1, characterized in that The performing model training based on the simulated sample to obtain an initial model for coal sample detection includes: Annotate the simulated samples with pseudo labels based on the PLS model after parameter initialization, and use the annotated simulated samples as training samples; Based on the training samples, model training is performed in a neural network based on the target spectral data search to obtain an initial model for coal sample detection.

24. The system according to claim 1, characterized in that The method of verifying the initial coal sample detection model based on the reserved historical target spectrum data to obtain the coal sample detection model includes: Building a validation set based on the reserved historical target spectral data, and inputting the validation set into the component monitoring initial model to obtain a corresponding prediction result; Based on the prediction results and the actual results of the corresponding reserved historical target spectral data, the performance of the initial component monitoring model is evaluated; If the performance evaluation of the initial component monitoring model fails, hyperparameters are adjusted, model structure is optimized, and / or regularization operations are added to obtain an updated initial component monitoring model; The performance of the updated initial component monitoring model is re-evaluated based on the reserved historical target spectral data until an initial component monitoring model that meets the preset performance requirements is obtained as the coal sample detection model.

25. The system according to claim 24, characterized in that The performance evaluation of the initial component monitoring model based on the prediction results and the actual results of the corresponding reserved historical target spectral data includes: Comparing the predicted result with the actual result, and evaluating the accuracy and recall of the model based on the deviation between the two; If any of the evaluation results of precision and recall fails, the performance evaluation of the initial model for component monitoring fails.

26. The system according to claim 1, characterized in that The training unit is also used to perform a neural network search based on the spectral data, including: Divide multiple optimization dimensions based on the structural parameters of the neural network; In each optimization dimension, adaptively adjust the parameters in the neural network structure corresponding to the optimization dimension, and calculate the MAE index after each adjustment; Compare the MAE indicators corresponding to each parameter, and determine the parameter combination with the highest MAE indicator as the optimization result in the corresponding optimization dimension; Perform greedy search among the optimization dimensions to obtain the optimization results of each optimization dimension; Based on the optimization results of each optimization dimension, the structural parameters corresponding to each neural network structure are determined, and the neural network corresponding to the search result is constructed.

27. The system according to claim 26, characterized in that The neural network dimensions include: The basic operator dimension of the neural network layer, the resolution dimension of the input data, the network depth dimension, and the network width dimension.

28. The system according to claim 27, characterized in that In the basic operator dimension of the neural network layer, the corresponding optimization result obtaining rules include: Traverse the convolution type of the linear layer and the activation function type of the nonlinear layer, adaptively combine the convolution type of the fully connected layer and the activation function type of the nonlinear layer, and calculate the MAE index after each combination; Compare the MAE indicators after each combination, select the combination with the maximum MAE indicator, and determine the convolution type of the fully connected layer and the activation function type of the nonlinear layer corresponding to the combination as the optimization result within the basic operator dimension of the neural network layer.

29. The system according to claim 28, characterized in that The convolution type of the linear layer is: Fully connected layer, 1D convolutional layer with kernel size 5, 1D convolutional layer with kernel size 10, or 1D convolutional layer with kernel size 15; The activation function type of the nonlinear layer is: TanH function, ELU function, Sigmoid activation function or Softmax function.

30. The system according to claim 28, characterized in that The calculation rules of the MAE indicator are: in, is the true value of the coal composition corresponding to the i-th training data; y i is the model prediction value corresponding to the i-th training data; m is the size of the dataset.

31. The system according to claim 1, characterized in that The system also includes an output unit for visually outputting the coal detection results.

32. The system according to claim 1 or 31, characterized in that The coal test results include: One or more of ash composition, ash content, volatile matter, hydrocarbons, ash melting point, total water, total sulfur and calorific value.

33. The system according to claim 31, characterized in that The visual output of the coal detection result comprises: In response to a user data query instruction, determining a corresponding detection result object; A preset data visualization scheme is selected based on the corresponding detection result object, and the visualized data is pushed to the user end.

34. A coal detection method, characterized in that: The method is applied to the coal detection system according to any one of claims 1 to 33, and the method comprises: Collect spectral data of coal samples; A neural network is constructed based on greedy search in each neural network dimension through the spectral data, and a coal sample detection model is obtained based on simulated sample training after sample augmentation; The spectral data is subjected to fusion processing to obtain target spectral data, and the target spectral data is inferred based on the coal sample detection model to obtain a coal detection result.

35. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions which, when executed on a computer, enable the computer to execute the coal detection method of claim 34.

Images