A method and system for rapid detection of coal quality based on space-time fusion spectrum

Abstract

The application belongs to the technical field of coal quality detection, and discloses a coal quality rapid detection method and system based on space-time fusion spectrum, which comprises the following steps: acquiring characteristic spectral lines corresponding to multiple key elements in each to-be-detected coal sample based on full-band spectrum of multiple to-be-detected samples; constructing a space-time spectral image based on the characteristic spectral lines; acquiring morphological features of the space-time spectral image, and acquiring evolution trends of the characteristic spectral lines based on the morphological features; constructing a quantitative model based on the evolution trends, and detecting the quality of the to-be-detected sample based on the quantitative model. The application can improve the quantitative detection precision of coal quality.

Description

Technical Field

[0001] This invention belongs to the field of coal quality testing and analysis technology, and more specifically, relates to a rapid detection method and system for coal quality based on spatiotemporal fusion spectroscopy. Background Technology

[0002] Industrial analysis and elemental analysis are used in coal production to determine its chemical composition. Industrial analysis requires placing the sample in a furnace for heating and testing, but this method is time-consuming and cannot perform elemental analysis. Currently available elemental analysis methods include X-ray fluorescence (XRF) and procedural gamma neutron activation analysis (PGNAA). These methods offer the advantage of non-destructive testing, but suffer from low accuracy, high cost, and radiation risks, making them unsuitable for the practical needs of the coal industry.

[0003] When laser-induced breakdown spectroscopy (LIBS) is used for coal detection, in addition to the inherent heterogeneity of coal, it also contains volatile components. These volatile components are released during laser ablation of the sample, increasing the matrix effect and reducing the quantitative accuracy of coal analysis. In 2021, Researcher Dong Meirong from South China University of Technology investigated the impact of volatile matter in coal on plasma and concluded that the content of volatile components has a significant influence on the spatiotemporal characteristics of elements in coal plasma. Currently, commonly used improvement methods include sample preprocessing, data processing, and combined techniques. In 2022, Lu et al. proposed a hybrid model based on wavelet threshold denoising and cross-validation recursive feature elimination. By extracting effective features from the denoised spectrum, this model improves the accuracy and precision of LIBS quantitative analysis of coal.

[0004] Currently, Chinese patent document CN102313722A discloses a coal quality industrial analysis method based on multiple linear regression, which mainly extracts relevant elements that significantly contribute to the spectrum and calculates industrial analysis indicators using multiple regression equations; CN111044504A discloses a coal quality analysis method considering the uncertainty of laser-induced breakdown spectroscopy, which mainly divides multiple spectra of each sample into multiple groups, then obtains the average spectrum of each group, and uses these average spectra to predict industrial indicators. Both of these methods use single-scale spectroscopy, only acquiring information on elements in the current state, resulting in limited spectral information and low detection accuracy. Therefore, there is an urgent need for a detection method that can improve the accuracy of coal quality detection using LIBS. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a rapid detection method and system for coal quality based on spatiotemporal fusion spectroscopy. By effectively acquiring plasma evolution information and leveraging the rich dynamic characteristics such as element migration and plasma parameter decay contained in the spatiotemporal evolution of plasma, the invention deeply explores the characteristics of plasma evolution and further improves the quantitative detection accuracy of coal quality.

[0006] To achieve the above objectives, according to one aspect of the present invention, a rapid detection method for coal quality based on spatiotemporal fusion spectroscopy is provided, comprising the following steps:

[0007] S1 obtains the characteristic spectral lines corresponding to multiple key elements in each coal sample based on the full-band spectrum of multiple test samples.

[0008] S2 constructs a spatiotemporal spectral image based on the characteristic spectral lines;

[0009] S3 acquires the morphological features of the spatiotemporal spectral image, and acquires the evolution trend of the feature spectral lines based on the morphological features;

[0010] S4 constructs a quantitative model based on the evolution trend and detects the quality of the sample to be tested based on the quantitative model.

[0011] Furthermore, before step S1, the full-band spectrum is acquired. The acquisition method includes: setting multiple different delays and using an orthogonal experimental method to acquire the spectra corresponding to multiple different defocus amounts under each delay.

[0012] Furthermore, the delay is in the range of 0 μs to 15 μs.

[0013] Furthermore, the defocusing amount is within -5mm to 5mm.

[0014] Furthermore, in step S1, preliminary feature spectral lines are first selected manually, and then the feature selection method is used to obtain the feature spectral lines from the preliminary feature spectral lines.

[0015] Furthermore, in step S3, the method for obtaining the morphological features includes at least the gray-level co-occurrence matrix method and the oriented gradient histogram method.

[0016] Furthermore, in step S2, the spatiotemporal spectral image is reduced to 1-dimensional image features.

[0017] Furthermore, the method for constructing a quantitative model based on the aforementioned evolution trend is as follows: the partial least squares method is used to fit the evolution trend with the test quality index of the coal sample to be tested into a linear fitting relationship.

[0018] Furthermore, each coal sample to be tested contains at least the following key elements: C, H, O, N, Si, Al, Fe, Ca, K, Mg, and Ti.

[0019] According to another aspect of the present invention, a rapid coal quality detection system based on spatiotemporal fusion spectroscopy is also provided, comprising:

[0020] The feature spectral line acquisition module is used to acquire the feature spectral lines corresponding to multiple key elements in each coal sample based on the full-band spectrum of multiple samples to be tested.

[0021] A spatiotemporal spectral image construction module is used to construct a spatiotemporal spectral image based on the feature spectral lines;

[0022] The evolution trend acquisition module is used to acquire the morphological features of the spatiotemporal spectral image and acquire the evolution trend of the feature spectral lines based on the morphological features.

[0023] The quantitative model construction and detection module is used to construct a quantitative model based on the evolution trend and to detect the quality of the sample to be tested based on the quantitative model.

[0024] Compared with the prior art, the above technical solutions conceived by this invention have the following main advantages:

[0025] 1. This invention obtains the characteristic spectral lines corresponding to multiple key elements in each coal sample by acquiring the full-band spectrum of multiple test samples. By acquiring the multidimensional fingerprint spectral information of the same sample to form an evolution spectrum, the differences between coal samples are increased, the comprehensiveness of the spectral information of coal samples is improved, and the abundance of spectral signals is significantly improved, thereby further improving the prediction accuracy of quantitative analysis of coal samples.

[0026] 2. This invention acquires multi-dimensional plasma spectral signals with different delays and spatial locations, expanding transient spectral signals into dynamic spectral signals with plasma evolution information. This provides a more detailed description of plasma dynamics. The obtained spatiotemporal spectral images reveal the evolution trends of elemental spectral lines and elemental distributions in the plasma, as well as the correlations of various elemental spectral lines in both time and space dimensions. This allows for a more accurate analysis of the influence of matrix effects on plasma spectra. Furthermore, the extracted plasma evolution factor (i.e., evolution trend) compensates for and corrects the influence of plasma spectral matrix effects, thereby improving the accuracy of coal quality detection.

[0027] 3. The present invention combines laser-induced breakdown spectroscopy technology to enable excellent analytical performance in coal quality detection. Based on the quantitative analysis results of coal, it can guide the adjustment of coal combustion process parameters and further improve coal combustion efficiency. Attached Figure Description

[0028] Figure 1 This is a flowchart of a rapid coal quality detection method based on spatiotemporal fusion spectroscopy provided by an embodiment of the present invention;

[0029] Figure 2 Typical LIBS spectra of coal obtained under different delays and different decoking amounts are provided in the embodiments of the present invention.

[0030] Figure 3 Typical LIBS spatiotemporal spectral images of coal samples provided in embodiments of the present invention;

[0031] Figure 4 A comparison chart of the predicted and actual calorific value concentrations of coal samples before and after correction according to the present invention, provided in an embodiment of the invention.

[0032] Figure 5 This is a comparison chart of the predicted and actual carbon concentrations of coal samples before and after correction according to the present invention, provided in an embodiment of the invention. Detailed Implementation

[0033] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0034] According to an embodiment of the present invention, a rapid detection method for coal quality based on spatiotemporal fusion spectroscopy is provided, the method comprising the following steps:

[0035] S1 obtains the characteristic spectral lines corresponding to multiple key elements in each coal sample based on the full-band spectrum of multiple test samples.

[0036] S2 constructs spatiotemporal spectral images based on feature spectral lines, following temporal and spatial order.

[0037] S3 acquires the morphological features of the spatiotemporal spectral image and obtains the evolution trend of the feature spectral lines based on the morphological features; the morphological features here include, but are not limited to, texture features, shape features, etc.

[0038] S4 constructs a quantitative model based on the evolution trend and uses the quantitative model to detect the quality of the sample to be tested.

[0039] In a preferred embodiment, a full-band spectrum is acquired before step S1. The acquisition method includes: setting multiple different delays, all of which are located within a specific continuous delay period; using an orthogonal experimental method to acquire multiple spectra corresponding to different defocus amounts under each delay, i.e., repeatedly acquiring spectra multiple times under a single delay; and adjusting the distance between the laser focusing lens and the coal sample to be tested before each acquisition so that the defocus amount is different in each acquisition.

[0040] In a more preferred embodiment, the delay is within the range of 0 μs to 15 μs; specifically, the continuous delay period can be selected as 0 μs to 10 μs, 0.5 μs to 15 μs, 0.6 μs to 9 μs, 0.7 μs to 8 μs, 0.5 μs to 7 μs, 0.8 μs to 6 μs, 0.9 μs to 9 μs, 1 μs to 15 μs, 2 μs to 14 μs, 2.5 μs to 13 μs, 3 μs to 15 μs, 4 μs to 11 μs, 5 μs, etc. Within each continuous delay period, such as 0.5μs to 10μs, multiple delays can be selected. For example, within the range of 0.5μs to 10μs, 0.5μs, 0.8μs, 1μs, 1.5μs, 2μs, and 2.5μs can be selected. Within the range of 0μs to 15μs, 3.5μs, 4μs, 4.5μs, 5.5μs, 6μs, and 7μs can be selected. It should be understood that the aforementioned specific continuous delay period and each delay include, but are not limited to, the values ​​listed above.

[0041] In a more preferred embodiment, the defocus amount is within -5mm to 5mm; specifically, the defocus amount is the relative distance between the laser focusing lens and the sample. Changing the defocus amount changes the spatial position of the coal sample to be tested. For example, the defocus amount corresponding to each delay is selected as 1mm, 0.5mm, 0mm, -0.5mm and -1mm, or the multiple defocus amounts corresponding to different delays are different, such as 5mm, 3mm, 0mm, -3mm and -5mm corresponding to a 0.5μs delay, and 0.9mm, 0.6mm, 0.1mm, -0.5mm and -0.8mm corresponding to a 2.5μs delay, etc. The value of the defocus amount includes, but is not limited to, the values ​​listed above.

[0042] In a more preferred embodiment, in step S1, preliminary feature spectral lines are first selected manually, and then feature spectral lines are obtained from the preliminary feature spectral lines using a feature selection method. This provides a more accurate spectral basis for the establishment of a quantitative model. Specifically, relevant key elements are first manually determined, and elemental spectral lines with better quality corresponding to each key element are selected. Then, a feature selection algorithm is used to select spectral lines with greater contributions from the manually selected elemental spectral lines. More specifically, the aforementioned feature selection method includes, but is not limited to, the random forest algorithm. When using the random forest algorithm, multiple decision trees are constructed and the importance of features in the decision trees is evaluated. The average contribution of features in the decision trees is taken as the feature importance. Then, the feature importance is ranked, and the top-ranked features are selected for analysis. The manual selection method involves manually selecting spectral lines of multiple elements that are highly correlated with the sample matrix. For example, key elements such as C, H, O, N, Si, Al, Fe, Ca, K, Mg, and Ti are manually determined, and the spectral lines corresponding to these elements are selected.

[0043] In a preferred embodiment, in step S3, the method for obtaining morphological features includes at least: the gray-level co-occurrence matrix method and / or the histogram of directional gradients method; specifically, when using the histogram of directional gradients to collect texture information such as spectral image gradients, the sliding window size is optimized to [2, 4], which has strong edge detection capability and photometric invariance capability; when using the gray-level co-occurrence matrix method, texture features based on the co-occurrence matrix are extracted from the gray-level image, the image is directly read in, and features such as contrast and correlation are calculated.

[0044] In a preferred embodiment, step S2 further reduces the spatiotemporal spectral image to 1-dimensional image features.

[0045] In a preferred embodiment, the method for constructing a quantitative model based on the evolution trend is as follows: the partial least squares method is used to fit the evolution trend and the test quality index of the coal sample to be tested into a linear fitting relationship, and the quantitative model is obtained based on the linear fitting relationship.

[0046] In a preferred embodiment, each coal sample to be tested contains at least the following key elements: C, H, O, N, Si, Al, Fe, Ca, K, Mg, and Ti.

[0047] According to another aspect of the present invention, a system is also provided that can implement the rapid coal quality detection method based on spatiotemporal fusion spectroscopy provided in any of the foregoing embodiments, the system comprising:

[0048] The feature spectral line acquisition module is used to acquire the feature spectral lines corresponding to multiple key elements in each coal sample based on the full-band spectrum of multiple samples to be tested.

[0049] Spatiotemporal spectral image construction module, used to construct spatiotemporal spectral images based on feature spectral lines;

[0050] The evolution trend acquisition module is used to acquire the morphological features of spatiotemporal spectral images and acquire the evolution trend of feature spectral lines based on the morphological features.

[0051] The quantitative model construction and detection module is used to construct quantitative models based on evolution trends and to detect the quality of the samples to be tested based on the quantitative models.

[0052] To better illustrate the implementation details of the present invention, the following embodiments are provided to further illustrate the present invention. It should be understood that the following embodiments are only preferred implementation methods and are not intended to limit the scope of protection of the present invention in any way.

[0053] Example 1

[0054] In this embodiment, a group of coal samples with known industrial analytical index values ​​and elemental contents were selected to calibrate their calorific value and carbon content (C). The specific steps of the quantitative method are as follows:

[0055] The calorific value and carbon content of the selected coal samples are shown in Table 1, where each number represents the number of the coal sample to be tested:

[0056] Table 1. Calorific value and carbon content of coal samples

[0057]

[0058]

[0059] After performing routine pretreatment on each of the coal samples in the table above, they are placed on the worktable of the spectral acquisition device. The spectral acquisition device includes a control unit, a laser, and a spectrometer, which are respectively connected to the timing controller. The laser is used to emit laser light onto the coal sample on the worktable, the timing controller is used to adjust the timing of laser emission and spectral acquisition, and the spectrometer is used to acquire the plasma spectrum of the coal sample excited by the laser.

[0060] Before laser emission, the spectral acquisition delay was set to 0.5μs, 1μs, 1.5μs, 2μs, and 2.5μs, with defocusing values ​​of 1mm, 0.5mm, 0mm, -0.5mm, and -1mm for each delay. Then, an orthogonal experimental method was used to acquire the spatiotemporal evolution spectrum. Specifically, at 0.5μs, plasma spectra with defocusing values ​​of 1mm, 0.5mm, 0mm, -0.5mm, and -1mm were acquired, and at 1μs, defocusing values ​​of 1mm, 0.5mm, and 0mm were acquired. Plasma spectra at defocusing depths of 1 mm, -0.5 mm, and -1 mm were collected at 1.5 μs, 2 μs, and 2.5 μs. The final collected plasma spectra are shown below. Figure 2 As shown.

[0061] Due to the influence of the matrix effect, there are differences in the evolution of plasmas. That is, the spectrum under a single parameter is greatly affected by the matrix effect, so the spectral line intensity and element concentration are no longer linearly related. Therefore, the plasma evolution law is used to correct the spectrum.

[0062] Specifically, spectral lines corresponding to important elements related to coal quality, such as C, H, O, N, Si, Al, Fe, Ca, K, Mg, and Ti, were screened from an atomic spectral database. These manually selected spectral lines were then fed into a genetic algorithm model to further refine the feature lines that contribute significantly. Finally, spatiotemporal spectral images of each coal sample were constructed based on these extracted feature lines, transforming the one-dimensional spectrum into a two-dimensional spectral image. The constructed spatiotemporal spectral images are shown below. Figure 3 As shown.

[0063] To extract information on the temporal and spatial variation trends of elemental spectral intensity from spatiotemporal spectral images, it is necessary to extract image features from these images. Specifically, the evolutionary patterns are determined by extracting morphological features. In this embodiment, a histogram of oriented gradients (HOG) is used to describe the features of the spectral image. Specifically, texture features are constructed by calculating and counting the directional gradients of local regions in the image. This method has strong edge detection capabilities and photometric invariance capabilities.

[0064] Subsequently, based on PLSR (Partial Least Squares Regression), regression equations were established between image features and the calorific value and carbon content of coal samples, respectively. The predicted values ​​of calorific value and carbon content were then obtained through these regression equations. Figure 4 (a) and Figure 5 In the middle (a), the predicted calorific value and carbon content are shown respectively based on the traditional PLSR. Figure 4 (b) and Figure 5 In section (b), the predicted calorific value and carbon content based on the spatiotemporal spectral-PLSR provided by this invention are shown respectively. Figure 4 and Figure 5 The evaluation parameters of the quantitative model for calorific value and carbon content are summarized in Table 2:

[0065] Table 2 Comparison of Evaluation Parameters for Quantitative Results of Calorific Value and Carbon Content

[0066]

[0067] As can be seen from Table 2, the coefficient of determination R of the calorific value prediction set is... 2 The root mean square error (RMSEp) increased from 0.9320 to 0.9900, the mean square error (RMSEp) decreased from 0.7269 to 0.3670, and the mean relative error decreased from 0.0194 to 0.0107; the coefficient of determination (R²) for the carbon content prediction set increased. 2 The accuracy improved from 0.8692 to 0.9857, the root mean square error (RMSEp) decreased from 2.3574 to 0.7194, and the average relative error of the prediction set decreased from 0.0232 to 0.0090. Combining the above figures and tables, it is clear that the spatiotemporal fusion-based LIBS detection method significantly improves the evaluation indicators of the calibration results compared to the traditional LIBS spectral quantitative detection method. The influence of matrix effects is clearly corrected, and the accuracy and precision of quantitative analysis are greatly enhanced.

[0068] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A rapid detection method for coal quality based on spatiotemporal fusion spectroscopy, characterized in that, Includes the following steps: S1 acquires the full-band spectrum of multiple samples to be tested. The acquisition method includes: setting multiple different delays and using an orthogonal experimental method to acquire the spectra corresponding to multiple different defocus amounts at each delay. S2 obtains the characteristic spectral lines corresponding to multiple key elements in each sample based on the full-band spectrum of multiple test samples. S3 constructs spatiotemporal spectral images based on feature spectral lines, following temporal and spatial order. S4 acquires the morphological features of the spatiotemporal spectral image, and obtains the evolution trend of the feature spectral lines based on the morphological features; S5 constructs a quantitative model based on the evolution trend and detects the quality of the sample to be tested based on the quantitative model.

2. The rapid detection method for coal quality based on spatiotemporal fusion spectroscopy as described in claim 1, characterized in that, The delay is in the range of 0 μs to 15 μs.

3. The rapid detection method for coal quality based on spatiotemporal fusion spectroscopy as described in claim 1, characterized in that, The defocusing amount is within -5mm to 5mm.

4. The rapid detection method for coal quality based on spatiotemporal fusion spectroscopy as described in claim 1, characterized in that, In step S2, preliminary feature spectral lines are first selected manually, and then the feature selection method is used to obtain the feature spectral lines from the preliminary feature spectral lines.

5. The rapid detection method for coal quality based on spatiotemporal fusion spectroscopy as described in claim 1, characterized in that, In step S4, the methods for obtaining the morphological features include at least the gray-level co-occurrence matrix method and the oriented gradient histogram method.

6. The rapid detection method for coal quality based on spatiotemporal fusion spectroscopy as described in claim 1, characterized in that, The method for constructing a quantitative model based on the aforementioned evolution trend is as follows: the partial least squares method is used to fit the evolution trend with the test quality index of the coal sample to be tested into a linear fit relationship.

7. The rapid detection method for coal quality based on spatiotemporal fusion spectroscopy as described in claim 1, characterized in that, Each coal sample to be tested contains at least the following key elements: C, H, O, N, Si, Al, Fe, Ca, K, Mg, and Ti.

8. A system capable of implementing the rapid coal quality detection method as described in any one of claims 1-7, characterized in that, include: The feature spectral line acquisition module is used to acquire the feature spectral lines corresponding to multiple key elements in each coal sample based on the full-band spectrum of multiple samples to be tested. A spatiotemporal spectral image construction module is used to construct a spatiotemporal spectral image based on the feature spectral lines; The evolution trend acquisition module is used to acquire the morphological features of the spatiotemporal spectral image and acquire the evolution trend of the feature spectral lines based on the morphological features. The quantitative model construction and detection module is used to construct a quantitative model based on the evolution trend and to detect the quality of the sample to be tested based on the quantitative model.

Images