Rapid online coal quality detection and analysis method based on multi-modal spectroscopy information

Abstract

The invention discloses a quick online coal quality detection and analysis method based on multi-modal spectroscopy information, and belongs to the technical field of coal quality detection. The implementation method comprises the following steps of: 1, synchronously acquiring spectrum, near infrared spectrum and sound spectrum signals on the surface of the coal flow by respectively utilizing a time-resolved laser-induced plasma spectrum, a diffuse reflection probe, an acousto-optic adjustable light filter and a microphone; 2, the collected TRLIBS spectrum is sequentially subjected to anomaly elimination, background deduction, normalization and peak position translation completion, the collected near infrared spectrum is subjected to dark background deduction, sound spectrum characteristics are extracted, and corrected spectrum data are corrected; 3, performing multi-modal weighted fusion on the NIRS coal quality parameter prediction value and the TRLIBS coal quality parameter prediction value which are obtained respectively to obtain a coal quality industrial index prediction result; compared with the prior art, the invention solves the technical problems of insufficient accuracy of single LIBS detection of moisture index in coal quality detection and spectrum fluctuation caused by coal flow surface fluctuation.

Description

Technical Field

[0001] This invention relates to a rapid online detection and analysis method for coal quality based on multimodal spectroscopic information, belonging to the field of coal quality detection technology, and is applicable to the rapid and accurate online detection of coal quality industrial indicators. Background Technology

[0002] Energy is the foundation of social development. As a major energy consumer, China relies heavily on fossil fuels, especially coal. Coal is the primary source of China's electricity supply. Although clean energy is developing, coal-fired power generation will continue to play a crucial role until new energy sources replace coal. The calorific value, ash content, volatile matter, and sulfur content of coal are critical to the combustion of coal in power plant boilers. Traditional manual coal quality testing methods cannot provide rapid, real-time guidance for production.

[0003] Online detection methods relying on traditional radiation sources (such as neutron activation methods, gamma-ray and X-ray radiation methods) have environmental problems. Material detection technologies based on spectroscopic information are green and safe, and are an inevitable trend for future industry development. Although single-spectrum technology can detect multiple elements and indicators in coal in real time, the current limitations of spectroscopic detection equipment and detection content make it difficult to meet the requirements for accurate determination of complex coal quality parameters.

[0004] Therefore, how to solve the problem of insufficient accuracy of LIBS detection of moisture index in coal quality testing and the spectral fluctuation caused by the surface undulation of coal flow has become an urgent problem to be solved. Summary of the Invention

[0005] The purpose of this invention is to address the technical problems of insufficient accuracy in detecting moisture content using a single LIBS spectrometer and spectral fluctuations caused by surface undulations in coal flow during coal quality testing. This invention proposes a rapid online coal quality detection and analysis method based on multimodal spectroscopic information. By integrating multiple spectroscopic information, spectral and acoustic preprocessing, and PLS modeling, the accuracy of coal moisture content detection is improved, and the fluctuation of spectral data is reduced.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] This invention discloses a rapid online detection and analysis method for coal quality based on multimodal spectroscopic information, comprising the following steps:

[0008] Step 1: Simultaneously acquire spectral, near-infrared spectral, and acoustic spectral signals from the coal flow surface using time-resolved laser-induced plasma spectroscopy, a diffuse reflection probe, an acousto-optic tunable filter, and a microphone.

[0009] Step 1.1: Use time-resolved laser-induced breakdown spectroscopy (TRLIBS) to collect spectra on the surface of the coal flow;

[0010] Step 1.1.1: When the coal flow is transported to the laser focusing point surface by the belt, the laser emits a laser with a wavelength of 1064nm. After passing through the dichroic mirror, the laser is focused onto the coal flow surface by the coaxial zoom system composed of concave and convex mirrors, generating laser-induced plasma.

[0011] Step 1.1.2: The wide-wavelength beam emitted by the laser-induced plasma is reflected by a dichroic mirror to the focusing lens after passing through a coaxial zoom system composed of concave and convex mirrors;

[0012] Step 1.1.3: The light beam passing through the focusing lens is focused onto the optical fiber and transmitted to the spectrometer to form spectral data;

[0013] Step 1.2: Near-infrared spectral data are collected from the surface of the coal flow using a diffuse reflection probe and an acousto-optic tunable filter;

[0014] Step 1.2.1: Use a diffuse reflection probe to emit detection light to illuminate the surface of the coal flow, and a near-infrared spectrometer to receive the diffuse reflection light from the surface of the coal flow;

[0015] Step 1.2.2: Use an acousto-optic tunable filter to select the wavelength of diffuse reflected light to obtain the near-infrared spectrum;

[0016] Step 1.3: Acquire the acoustic signal generated by the laser-induced plasma using a microphone;

[0017] Step 2: The acquired TRLIBS spectra are sequentially subjected to anomaly removal, background subtraction, normalization, and peak position shifting and completion. The acquired near-infrared spectra (NIRS) are subjected to dark background subtraction, the acoustic spectral features are extracted, and the corrected spectral data are corrected.

[0018] Step 2.1: In the preprocessing of spectral data, anomaly removal, background subtraction, normalization, and peak position shifting and completion are performed in sequence to obtain corrected spectral data;

[0019] Step 2.1.1: Use the Laida criterion to remove outlier spectra from the spectral data;

[0020] Step 2.1.2: Remove the spectral background of the channel using the window translation minimum method combined with polynomial fitting;

[0021] Step 2.1.3: Normalize the background intensity of the spectrum after removing background interference by channel;

[0022] Step 2.1.4: Mark the characteristic peak intensities of the normalized spectrum; shift the drift peaks; establish a multiple linear regression relationship using the peak sequence of correlation coefficients between characteristic peaks; use the multiple linear regression relationship to shift and complete the peak positions to form corrected spectral data;

[0023] Step 2.2: Dark background subtraction is performed on the near-infrared spectrum by subtracting the instrument's dark current noise from the spectral intensity;

[0024] Step 2.3: Extract the positive peak value and signal energy of the acoustic spectrum as acoustic spectrum features, and use the acoustic spectrum features to correct the modified spectral data.

[0025] Step 3: Perform multimodal weighted fusion on the NIRS and TRLIBS coal quality parameter prediction values ​​obtained separately to obtain the prediction results of coal quality industrial indicators;

[0026] Step 3.1: Extract the near-infrared spectral feature matrix from the near-infrared spectrum after dark background subtraction, and use partial least squares regression to obtain the predicted values ​​of NIRS coal quality parameters;

[0027] Step 3.2: Extract the spectral feature matrix from the calibrated spectral data and use multiple linear regression to obtain the predicted values ​​of TRLIBS coal quality parameters;

[0028] Step 3.3: After weighted fusion of the predicted values ​​of coal quality parameters from NIRS and TRLIBS, the predicted results of coal quality industrial indicators are obtained.

[0029] Compared with existing technologies, it has the following beneficial effects:

[0030] 1. High detection accuracy: By integrating multi-modal spectroscopy fusion, multi-dimensional information from atomic, molecular and acoustic signals is combined, which improves the analysis accuracy and makes the prediction results of coal quality industrial indicators more accurate compared with single spectral technology.

[0031] 2. Fast testing speed: It realizes online rapid testing, and can complete the simultaneous testing and analysis of multiple indicators in a short time to meet the needs of real-time production guidance.

[0032] 3. Green and safe: The detection technology based on spectroscopic information avoids the environmentally unfriendly problems of traditional radiation source detection methods and is in line with the industry's green development trend.

[0033] 4. Strong anti-interference capability: Through a series of spectral preprocessing algorithms, abnormal spectra are effectively eliminated, background noise and spectral fluctuations are reduced, and the anti-interference capability and measurement repeatability of the system are improved. Attached Figure Description

[0034] Figure 1 This is a flowchart of coal quality testing based on multimodal spectroscopic information. Detailed Implementation

[0035] To better illustrate the purpose and advantages of this invention, the invention will be further described below with reference to the accompanying drawings and examples. It should be noted that the implementation of this invention is not limited to the following embodiments, and any modifications or alterations made to this invention will fall within the scope of protection of this invention.

[0036] Example

[0037] like Figure 1 As shown in the figure, the specific implementation steps of the rapid online detection and analysis method for coal quality based on multimodal spectroscopic information in this embodiment are as follows:

[0038] Step 1: Simultaneously acquire spectral, near-infrared spectral, and acoustic spectral signals from the coal flow surface using time-resolved laser-induced plasma spectroscopy, a diffuse reflection probe, an acousto-optic tunable filter, and a microphone.

[0039] Step 1.1: Use time-resolved laser-induced breakdown spectroscopy (TRLIBS) to collect spectra on the surface of the coal flow;

[0040] Step 1.1.1: When the coal flow is transported to the laser focusing point surface by the belt, the laser emits a laser with a wavelength of 1064nm. After passing through the dichroic mirror, the laser is focused onto the coal flow surface by the coaxial zoom system composed of concave and convex mirrors, generating laser-induced plasma.

[0041] Step 1.1.2: The wide-wavelength beam emitted by the laser-induced plasma is reflected by a dichroic mirror to the focusing lens after passing through a coaxial zoom system composed of concave and convex mirrors;

[0042] Step 1.1.3: The light beam passing through the focusing lens is focused onto the optical fiber and transmitted to the spectrometer to form spectral data;

[0043] In this embodiment, the laser emits a 1064nm laser, which is focused onto the surface of the coal sample by a dichroic mirror and a coaxial zoom system. The resulting plasma spectrum is reflected by a concave mirror and transmitted to a spectrometer via an optical fiber. The spectrum is collected every 10ms, and after 150 consecutive collections, it is stored in a computer.

[0044] Step 1.2: Near-infrared spectral data are collected from the surface of the coal flow using a diffuse reflection probe and an acousto-optic tunable filter;

[0045] Step 1.2.1: Use a diffuse reflection probe to emit detection light to illuminate the surface of the coal flow, and a near-infrared spectrometer to receive the diffuse reflection light from the surface of the coal flow;

[0046] Step 1.2.2: Use an acousto-optic tunable filter to select the wavelength of diffuse reflected light to obtain the near-infrared spectrum;

[0047] In this embodiment, the diffuse reflection probe emits detection light, which is reflected by the coal sample and fed back to the near-infrared spectrometer. The AOTF spectrometer uses spectral technology to complete three spectral scans within one second, acquires absorbance spectral data, and transmits it to the embedded industrial computer via network cable.

[0048] Step 1.3: Acquire the acoustic signal generated by the laser-induced plasma using a microphone;

[0049] In this embodiment, a microphone (30cm away from the coal sample surface) collects the acoustic wave signal generated by the plasma, which is then converted into an electrical signal by an oscilloscope and transmitted synchronously to a computer. The acquisition time is consistent with the single acquisition cycle of TRLIBS (1.5 seconds).

[0050] Step 2: The acquired TRLIBS spectra are sequentially subjected to anomaly removal, background subtraction, normalization, and peak position shifting and completion. The acquired near-infrared spectra (NIRS) are subjected to dark background subtraction, the acoustic spectral features are extracted, and the corrected spectral data are corrected.

[0051] Step 2.1: In the preprocessing of spectral data, anomaly removal, background subtraction, normalization, and peak position shifting and completion are performed in sequence to obtain corrected spectral data;

[0052] Step 2.1.1: Use the Laida criterion to remove outlier spectra from the spectral data;

[0053] In the embodiment, for the removal of abnormal spectra, the Euclidean distance between each sample point and the center spectrum was calculated for 150 sets of TRLIBS spectra. After normality test, 7 sets of abnormal spectra with deviation >3σ were removed according to the 3σ criterion.

[0054] Step 2.1.2: Remove the spectral background of the channel using the window translation minimum method combined with polynomial fitting;

[0055] In the example, during baseline calibration and smoothing: the remaining 143 groups of spectra are divided into 128 parts (n=64) with an average of 8192 pixels each, and the minimum value of each group is extracted and fitted to the background spectrum through cubic spline interpolation to obtain the background-free spectrum.

[0056] Step 2.1.3: Normalize the background intensity of the spectrum after removing background interference by channel;

[0057] In the embodiment, the background intensity normalization method is used for normalization, which divides the spectral intensity by the average background intensity, thereby reducing the RSD of the sample spectrum after processing.

[0058] Step 2.1.4: Mark the characteristic peak intensities of the normalized spectrum; shift the drift peaks; establish a multiple linear regression relationship using the peak sequence of correlation coefficients between characteristic peaks; use the multiple linear regression relationship to shift and complete the peak positions to form corrected spectral data;

[0059] In the embodiment, for the correction of characteristic peaks, the peak finding algorithm is used to mark characteristic peaks such as CI 193.09nm, and the drift peaks are corrected by 1 pixel shift; for the 3 missing peaks, 8 sets of peaks with correlation coefficient R>0.8 are extracted to establish a multiple linear regression model, and after completion, the correlation of characteristic peaks is improved to 0.96.

[0060] Step 2.2: Dark background subtraction is performed on the near-infrared spectrum by subtracting the instrument's dark current noise from the spectral intensity;

[0061] Step 2.3: Extract the positive peak value and signal energy of the acoustic spectrum as acoustic spectrum features, and use the acoustic spectrum features to correct the modified spectral data.

[0062] Step 3: Perform multimodal weighted fusion on the NIRS and TRLIBS coal quality parameter prediction values ​​obtained separately to obtain the prediction results of coal quality industrial indicators;

[0063] Step 3.1: Extract the near-infrared spectral feature matrix from the near-infrared spectrum after dark background subtraction, and use partial least squares regression to obtain the predicted values ​​of NIRS coal quality parameters;

[0064] Step 3.2: Extract the spectral feature matrix from the calibrated spectral data and use multiple linear regression to obtain the predicted values ​​of TRLIBS coal quality parameters;

[0065] Step 3.3: After weighted fusion of the predicted values ​​of coal quality parameters from NIRS and TRLIBS, the predicted results of coal quality industrial indicators are obtained.

[0066] In this embodiment, a feature matrix is ​​extracted from the near-infrared spectrum, and the predicted moisture content of the received substrate (Rb) is obtained through partial least squares regression (PLSR). 2 =0.92). After extracting the TRLIBS spectral feature matrix from the spectral data corrected by the acoustic signal, multiple linear regression was used to obtain the dry-basis total sulfur prediction value (R0). 2 =0.98). A weighted fusion algorithm (NIRS weight 0.86, TRLIBS weight 0.14) was used to integrate the above results, ultimately yielding the predicted baseline moisture content R0. 2 =0.998, dry basis total sulfur prediction R 2=0.991. Experimental verification shows that the detection error of this method for coal quality industrial indicators is controlled within 5%, and the time for a single detection is less than 3 seconds, meeting the requirements for online rapid analysis.

Claims

1. A rapid online detection and analysis method for coal quality based on multimodal spectroscopic information, characterized in that: Includes the following steps, Step 1: Simultaneously acquire spectral, near-infrared spectral, and acoustic spectral signals from the coal flow surface using time-resolved laser-induced plasma spectroscopy, a diffuse reflection probe, an acousto-optic tunable filter, and a microphone. Step 2: The acquired TRLIBS spectra are sequentially subjected to anomaly removal, background subtraction, normalization, and peak position shifting and completion. The acquired near-infrared spectra are subjected to dark background subtraction, acoustic spectral features are extracted, and the corrected spectral data are corrected. Step 3: Perform multimodal weighted fusion on the NIRS and TRLIBS coal quality parameter prediction values ​​obtained separately to obtain the prediction results of coal quality industrial indicators; Step 3.1: Extract the near-infrared spectral feature matrix from the near-infrared spectrum after dark background subtraction, and use partial least squares regression to obtain the predicted values ​​of NIRS coal quality parameters; Step 3.2: Extract the spectral feature matrix from the calibrated spectral data and use multiple linear regression to obtain the predicted values ​​of TRLIBS coal quality parameters; Step 3.3: After weighted fusion of the predicted values ​​of coal quality parameters from NIRS and TRLIBS, the predicted results of coal quality industrial indicators are obtained.

2. The rapid online detection and analysis method for coal quality based on multimodal spectroscopic information as described in claim 1, characterized in that: Step 1 is implemented as follows: Step 1.1: Spectral acquisition of the coal flow surface is performed using time-resolved laser-induced plasma spectroscopy; Step 1.2: Near-infrared spectral data are collected from the surface of the coal flow using a diffuse reflection probe and an acousto-optic tunable filter; Step 1.3: Use a microphone to collect the acoustic wave signal generated by the laser-induced plasma.

3. The rapid online detection and analysis method for coal quality based on multimodal spectroscopic information as described in claim 2, characterized in that: Step 1.1 is implemented as follows: Step 1.1.1: When the coal flow is transported to the laser focusing point surface by the belt, the laser emits a laser with a wavelength of 1064nm. After passing through the dichroic mirror, the laser is focused onto the coal flow surface by the coaxial zoom system composed of concave and convex mirrors, generating laser-induced plasma. Step 1.1.2: The wide-wavelength beam emitted by the laser-induced plasma is reflected by a dichroic mirror to the focusing lens after passing through a coaxial zoom system composed of concave and convex mirrors; Step 1.1.3: The light beam passing through the focusing lens is focused onto the optical fiber and transmitted to the spectrometer to form spectral data.

4. The rapid online detection and analysis method for coal quality based on multimodal spectroscopic information as described in claim 2, characterized in that: Step 1.2 is implemented as follows: Step 1.2.1: Use a diffuse reflection probe to emit detection light to illuminate the surface of the coal flow, and a near-infrared spectrometer to receive the diffuse reflection light from the surface of the coal flow; Step 1.2.2: Use an acousto-optic tunable filter to select the wavelength of diffuse reflected light to obtain the near-infrared spectrum.

5. The rapid online detection and analysis method for coal quality based on multimodal spectroscopic information as described in claim 1, characterized in that: Step 2 is implemented as follows: Step 2.1: In the preprocessing of spectral data, anomaly removal, background subtraction, normalization, and peak position shifting and completion are performed in sequence to obtain corrected spectral data; Step 2.2: Dark background subtraction is performed on the near-infrared spectrum by subtracting the instrument's dark current noise from the spectral intensity; Step 2.3: Extract the positive peak value and signal energy of the acoustic spectrum as acoustic spectrum features, and use the acoustic spectrum features to correct the modified spectral data.

6. The rapid online detection and analysis method for coal quality based on multimodal spectroscopic information as described in claim 5, characterized in that: Step 2.1 is implemented as follows: Step 2.1.1: Use the Laida criterion to remove outlier spectra from the spectral data; Step 2.1.2: Remove the spectral background of the channel using the window translation minimum method combined with polynomial fitting; Step 2.1.3: Normalize the background intensity of the spectrum after removing background interference by channel; Step 2.1.4: Mark the characteristic peak intensities of the normalized spectrum; The drift peaks are shifted; a multiple linear regression relationship is established using the peak sequence of correlation coefficients between characteristic peaks; the peak positions are shifted and completed using the multiple linear regression relationship to form corrected spectral data.

Images