Coal quality dynamic analysis method and system based on infrared detection
Patent Information
- Application Number
- CN202611250121.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-18
- Publication Date
- 2026-11-03
AI Technical Summary
Existing near-infrared coal analysis technology is susceptible to the unevenness of coal flow in industrial conveyor belt scenarios, resulting in random fluctuations in analysis results. Furthermore, the prediction model cannot cope with fluctuations in coal source and particle size, leading to information redundancy or overfitting.
By building a hardware platform to collect dynamic spectral data in real time, performing data preprocessing and weighted fusion, using a partial least squares regression model to extract characteristic bands, and establishing a dynamic adaptation scheme between the main prediction model and the verification model, real-time prediction of coal quality can be achieved.
It improves the signal-to-noise ratio of data and the representativeness of analysis results, adapts to changes in coal quality, reduces analysis interference, ensures the stability and adaptability of the model, and avoids production shocks.
Smart Images
Figure CN123028973A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of coal quality testing technology, and in particular to a method and system for dynamic analysis of coal quality based on infrared detection. Background Technology
[0002] As an important energy source and industrial raw material, the quality parameters of coal, such as moisture, ash, volatile matter, fixed carbon, and sulfur, directly affect combustion efficiency, pollutant emission control, and the economics of subsequent processing technologies. Therefore, accurate and timely testing of coal quality is crucial in all stages of coal mining, washing, transportation, and sales.
[0003] Existing near-infrared coal analysis techniques, whether offline or side-by-side, generally employ single or multiple average measurements of uniform powder within a static sample cup. However, in real industrial conveyor belt scenarios, because coal flow is a dynamic, macroscopically non-uniform aggregate of particles, a single spectral acquisition is highly susceptible to the influence of local large gangue cavities or uneven particle size distribution, leading to significant random fluctuations in the analysis results. Furthermore, the feature extraction of existing techniques largely relies on the full spectrum or empirically fixed bands, failing to fully consider the sensitivity differences of different coal quality indicators to various spectral regions, which can easily lead to information redundancy or overfitting.
[0004] Meanwhile, the prediction models for coal quality targets in existing technologies remain fixed for a long time once calibrated, and cannot cope with the shifts in spectral-composition relationships caused by fluctuations in coal sources and changes in particle size. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention provides a method and system for dynamic analysis of coal quality based on infrared detection, which solves the problems existing in the prior art.
[0006] This invention provides a method for dynamic analysis of coal quality based on infrared detection, the method comprising the following steps:
[0007] S1: Construct a hardware platform for coal quality testing and initialize the software environment and configure parameters; S2: Real-time acquisition of dynamic spectral data of coal on the conveyor belt; S3: Perform data preprocessing on the dynamic spectral data; S4: Perform data fusion and feature extraction on the preprocessed dynamic spectral data to obtain spectral feature information; S4 specifically includes: S4.1: Determine the dynamic analysis time window for coal analysis; S4.2: Within the dynamic analysis time window, perform weighted average fusion on the preprocessed dynamic spectral data based on the coal flow surface density parameter to obtain fused spectral data; S4.3: Extract feature bands from the fused spectral data to obtain spectral feature information; S5: Based on the spectral feature information, predict and output the coal quality target.
[0008] Preferably, in step S4.1, the time required to collect the coal flow of the length analysis unit L is calculated based on the instantaneous linear velocity of the conveyor belt collected in step S2, which is the dynamic analysis time window for coal analysis.
[0009] Preferably, in step S4.2, the preprocessed dynamic spectral data collected at all times within the dynamic analysis time window are formed into a spectral sequence, and a fusion weight is calculated for each frame of spectral data within the dynamic analysis time window. The fusion weight is proportional to the coal flow surface density at the time the coal in that frame of spectrum is collected. The final value of the fused spectral data is the sum of all frame spectra within the dynamic analysis time window after weighting them by their respective weights, divided by the total weight.
[0010] Preferably, in step S4.3, a partial least squares regression model is used to extract feature bands from the fused spectral data.
[0011] Preferably, in step S1, the hardware platform includes a spectral acquisition module, a light source module, an optical probe and optical path design, and auxiliary sensors.
[0012] Preferably, the spectral acquisition module is a near-infrared spectrometer, and the core parameters of the near-infrared spectrometer are: spectral range covering 900nm to 2500nm; spectral resolution better than 8nm; and equipped with an integrating sphere accessory with adjustable aperture or a fiber optic diffuse reflection probe. The light source module is connected to the spectrum acquisition module. The light source module uses two quartz halogen lamps as a stable broadband light source. The continuous light emitted by the quartz halogen lamps is collimated by a collimating lens to form a parallel beam. The continuous light is then modulated into an alternating signal of a specific frequency by an electrically adjustable mechanical chopper. The optical probe is connected to the spectral acquisition module. The optical probe is an active diffuse reflection probe. The active diffuse reflection probe integrates a light source incident fiber and a signal receiving fiber, which are arranged at a preset angle and are aligned with the coal flow surface on the conveyor belt. The optical probe is mounted on a three-dimensionally adjustable bracket.
[0013] Preferably, the software environment initialization and parameter configuration specifically involve: starting the data acquisition software and loading the pre-calibrated system configuration file; setting core analysis parameters, including the integration time, averaging times, smoothing algorithm, chopper frequency, data storage path, and alarm threshold of the spectral acquisition module; performing a dark current calibration: acquiring a set of spectral data as a dark spectrum in a non-luminescent state for subtracting electronic noise in subsequent spectral preprocessing; and performing a whiteboard calibration: placing a standard reflective whiteboard in the detection area.
[0014] Preferably, in step S3, the data preprocessing includes dark current subtraction and spectral smoothing filtering.
[0015] Preferably, the dark current subtraction is specifically performed by subtracting wavelength points from each newly acquired frame of raw dynamic spectral data during software environment initialization and parameter configuration, thereby achieving dark current subtraction.
[0016] The spectral smoothing filtering specifically involves applying a Savitzky-Golay filter to achieve spectral smoothing filtering. The window size of the Savitzky-Golay filter is 11 points, and the polynomial order is a second-order polynomial.
[0017] According to another aspect of the present invention, a dynamic coal quality analysis system based on infrared detection is provided. The system employs the aforementioned dynamic coal quality analysis method based on infrared detection, and the system includes: The build and initialization module is used to build a hardware platform for coal quality testing and to initialize the software environment and configure parameters. The data acquisition module is used to collect dynamic spectral data of coal on the conveyor belt in real time. The data preprocessing module is used to perform data preprocessing operations on the dynamic spectral data; The feature extraction module is used to perform data fusion and feature extraction on the preprocessed dynamic spectral data to obtain spectral feature information; The prediction and output module is used to predict and output coal quality targets based on the spectral feature information.
[0018] The present invention has the following technical effects: This invention, through a weighted fusion method based on areal density, effectively performs a statistically significant compaction and homogenization of a dynamically changing coal flow cross-section, synthesizing a frame of spectral data that best represents the overall condition of the analysis unit. This method significantly improves the signal-to-noise ratio and representativeness of the data, effectively suppressing analytical interference caused by the randomness of coal flow distribution; thus, it makes coal quality detection applicable to conveyor belt conditions.
[0019] A dynamic verification period is set. During this period, the prediction results of the master prediction model and the verification model are compared for the same batch of fused spectra. If the average deviation between the two exceeds a preset threshold, it is determined that the master prediction model has drifted, triggering the master prediction model update process. During the update, the master prediction model is fine-tuned by incremental learning instead of full retraining, which avoids the impact of model mutation on production and ensures the model's adaptability to new working conditions. Attached Figure Description
[0020] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.
[0021] Figure 1 This is a flowchart of a dynamic coal quality analysis method based on infrared detection provided in an embodiment of the present invention; Figure 2 This is a flowchart provided by an embodiment of the present invention for performing data fusion and feature extraction on preprocessed dynamic spectral data to obtain spectral feature information. Detailed Implementation
[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.
[0023] Example 1, Figure 1 A flowchart of a dynamic analysis method for coal quality based on infrared detection is shown, such as... Figure 1 As shown, a dynamic analysis method for coal quality based on infrared detection includes the following steps: S1: Construct a hardware platform for coal quality testing and initialize the software environment and configure parameters; The hardware platform includes a spectrum acquisition module, a light source module, an optical probe and optical path design, and auxiliary sensors.
[0024] The spectral acquisition module is a near-infrared spectrometer. The core parameters of the near-infrared spectrometer are: a spectral range covering 900nm to 2500nm, preferably in the short-wave near-infrared region of 900-1700nm to balance cost and high correlation with moisture and hydrogen-containing groups; a spectral resolution better than 8nm; a high-sensitivity InGaAs detector to adapt to weak light signals; and an integrating sphere accessory with an adjustable aperture or a fiber optic diffuse reflection probe to achieve stable diffuse reflection signal acquisition.
[0025] The light source module is connected to the spectrum acquisition module. The light source module uses two quartz halogen lamps as a stable broadband light source. The continuous light emitted by the quartz halogen lamps is collimated into a parallel beam through a collimating lens. The continuous light is modulated into an alternating signal of a specific frequency by an electrically adjustable mechanical chopper, which facilitates cooperation with the lock-in amplifier circuit of the spectrum acquisition module to effectively suppress DC interference from ambient light.
[0026] The optical probe is connected to the spectral acquisition module. It is an active diffuse reflection probe, integrating an incident optical fiber for the light source and a receiving optical fiber for the signal. These are arranged at a 45° angle, aligned with the coal flow surface on the conveyor belt. This design effectively collects diffuse reflected light from the coal particle surface and reduces the influence of specular reflection. A key design feature of the optical probe is its geometric relationship with the measured coal flow: the probe is mounted on a three-dimensionally adjustable bracket to ensure its focal spot is precisely projected above the centerline of the conveyor belt. The distance H perpendicular to the conveyor belt surface and the spot diameter D are precisely calculated. For example, for a belt with a bandwidth of B = 1.2m, H can be set to 40cm, and D can be calculated to be approximately 1 / 3 of B, ensuring that each sampling covers a representative cross-sectional area of the coal flow, rather than areas with severe edge effects. The optical probe is externally fitted with a detachable, purgeable sapphire or quartz window to prevent coal dust contamination of the fiber end face, and is equipped with a miniature air pump and air path for periodic positive pressure purging.
[0027] The auxiliary sensors are integrated at key locations on the conveyor belt. For example, a non-contact laser velocimeter and a high-precision weighing sensor are installed directly below the optical probe. The laser velocimeter is used to measure the instantaneous operating speed of the conveyor belt in real time and accurately. The weighing sensor is installed on the idler group that carries the coal flow, or measures the weight of coal per unit length of belt through an independent load-bearing beam, thereby calculating the areal density of the coal flow.
[0028] The software environment initialization and parameter configuration are specifically performed as follows: Start the data acquisition software and load the pre-calibrated system configuration file; set core analysis parameters, including the integration time, averaging times, smoothing algorithm, chopper frequency, data storage path, and alarm threshold of the spectral acquisition module; perform a dark current calibration: in a non-emitting state, acquire a set of spectral data as a dark spectrum for subsequent spectral preprocessing to subtract electronic noise; perform a whiteboard calibration: place a standard reflective whiteboard in the detection area and acquire a set of spectral data as a reference spectrum for subsequent conversion of the original spectrum into absorbance or reflectance, achieving absolute intensity normalization of the spectrum.
[0029] S2: Real-time acquisition of dynamic spectral data of coal on the conveyor belt; A laser velocimeter installed next to the conveyor belt continuously monitors the instantaneous linear velocity of the conveyor belt. When the speed of the conveyor belt is detected to be greater than the preset minimum effective working threshold within a preset time and not exceeding the safe operating limit of the conveyor belt, the conveyor belt is determined to have entered an effective working state. Then, the dynamic spectral data acquisition process is automatically activated to collect the dynamic spectral data of the coal in real time.
[0030] Simultaneously, during the spectral data acquisition process, it is necessary to collect the timestamp information of the sampling time and the weight of the coal flow per unit length for subsequent feature extraction. The weight of the coal flow per unit length is used as the coal surface density.
[0031] S3: Perform data preprocessing on the dynamic spectral data; The raw dynamic spectral data contains various noises and non-coal sample information, which must be cleaned immediately before online analysis; otherwise, it will seriously affect the accuracy and stability of subsequent models. In this step, the data preprocessing includes dark current subtraction and spectral smoothing filtering.
[0032] The dark current subtraction is specifically achieved by: during software environment initialization and parameter configuration, one or more dark spectral data have been collected, and a subtraction operation is performed on each wavelength point of each newly collected original dynamic spectral data to achieve dark current subtraction.
[0033] The spectral smoothing filtering specifically involves applying a Savitzky-Golay filter. The Savitzky-Golay filter is a smoothing method that performs polynomial least-squares fitting within a sliding window, effectively preserving spectral peak shape, width, and height while removing noise. The window size and polynomial order of the Savitzky-Golay filter are key parameters and need to be optimized based on the signal-to-noise ratio. In this step, the window size of the Savitzky-Golay filter is 11 points, and the polynomial order is second-order.
[0034] S4: Perform data fusion and feature extraction on the preprocessed dynamic spectral data to obtain spectral feature information; In this scheme, spectral data with precise time information is collected. Existing near-infrared coal analysis techniques, whether offline or linear, generally employ single or several average measurements of uniform powder in a static sample cup. However, in real industrial conveyor belt scenarios, because coal flow is a dynamic and macroscopically non-uniform aggregate of particles, a single spectral acquisition is easily affected by local large gangue cavities or uneven particle size distribution, resulting in significant random fluctuations in the analysis results. In addition, the feature extraction of existing technologies mostly relies on the full spectrum or empirical fixed bands, failing to fully consider the sensitivity differences of different coal quality indicators to different spectral regions, which can easily lead to information redundancy or overfitting.
[0035] Therefore, as Figure 2 As shown, S4 specifically includes: S4.1: Determine the dynamic analysis time window for coal analysis; In this step, a length analysis unit L with clear physical meaning is first defined. This unit represents a basic quality unit that can be processed and controlled by subsequent processes such as coal blending and combustion. Specifically, the time required to collect the coal flow from the length analysis unit L is calculated based on the instantaneous linear velocity of the conveyor belt collected in step S2; this time window is the dynamic analysis time window for coal analysis. Through this setting, regardless of whether the conveyor belt is accelerating, decelerating, or moving at a constant speed, the physical coal sample quantity represented by each data analysis is constant and known. This setting ensures the stability and comparability of the analysis results.
[0036] S4.2: Within the dynamic analysis time window, the preprocessed dynamic spectral data are weighted and averaged to obtain fused spectral data based on the coal flow surface density parameter. The preprocessed dynamic spectral data collected at all times within the dynamic analysis time window are formed into a spectral sequence. A fusion weight is calculated for each frame of spectral data within the dynamic analysis time window. This weight is proportional to the coal flow surface density at the time the spectrum was collected. That is, spectra from areas with dense coal flow and thick material layers will have a greater influence in the final fused spectrum; conversely, spectra from voids or sparse areas will have a weaker influence. The final value of the fused spectral data is the sum of all frame spectra within the dynamic analysis time window, weighted by their respective weights, divided by the total weight.
[0037] In this step, this density-based weighted fusion effectively performs a statistically significant compaction and homogenization of a dynamically changing coal flow cross-section, synthesizing a frame of spectral data that best represents the overall condition of the analysis unit. This method significantly improves the signal-to-noise ratio and representativeness of the data, effectively suppressing analytical interference caused by the randomness of coal flow distribution.
[0038] S4.3: Extract feature bands from the fused spectral data to obtain spectral feature information; The partial least squares regression (PLSR) model was used to extract the feature bands from the fused spectral data.
[0039] S5: Based on the spectral feature information, predict and output the coal quality target.
[0040] Existing coal quality target prediction models remain fixed once calibrated, failing to address shifts in spectral-composition relationships caused by fluctuations in coal source and particle size. This step addresses this by constructing a dual-model dynamic adaptation scheme: an online main model for prediction and a background verification model for auxiliary models. This ensures the model always matches the current working conditions of the coal quality analysis. Specifically, S5 involves configuring a main prediction model and a verification model to predict coal quality. The main prediction model uses Support Vector Machine Regression (SVR), while the verification model uses an Artificial Neural Network (ANN) model. Both the main prediction model and the verification model receive the spectral feature information as input. The verification model runs in the background but does not directly participate in real-time output; instead, it periodically verifies the prediction bias of the main prediction model. A dynamic verification period is set. During this period, the prediction results of the master prediction model and the verification model are compared for the same batch of fused spectra. If the average deviation between the two exceeds a preset threshold, it is determined that the master prediction model has drifted, triggering the master prediction model update process. During the update, the master prediction model is fine-tuned by incremental learning instead of full retraining, which avoids the impact of model mutation on production and ensures the model's adaptability to new working conditions.
[0041] Example 2: This invention also provides a dynamic coal quality analysis system based on infrared detection. The system employs a dynamic coal quality analysis method based on infrared detection as described in Example 1. The system includes: The build and initialization module is used to build a hardware platform for coal quality testing and to initialize the software environment and configure parameters. The data acquisition module is used to collect dynamic spectral data of coal on the conveyor belt in real time. The data preprocessing module is used to perform data preprocessing operations on the dynamic spectral data; The feature extraction module is used to perform data fusion and feature extraction on the preprocessed dynamic spectral data to obtain spectral feature information; The prediction and output module is used to predict and output coal quality targets based on the spectral feature information.
[0042] Example 3: The present invention also provides an electronic device, including one or more processors and a memory.
[0043] A processor can be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and can control other components in an electronic device to perform desired functions.
[0044] The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the program instructions to implement the infrared detection-based dynamic analysis method for coal quality described in any embodiment of this application above, and / or other desired functions. Various contents such as initial extrinsic parameters and thresholds may also be stored in the computer-readable storage medium.
[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.
Claims
1. A method for dynamic analysis of coal quality based on infrared detection, characterized in that, The method includes the following steps: S1: Construct a hardware platform for coal quality testing and initialize the software environment and configure parameters; S2: Real-time acquisition of dynamic spectral data of coal on the conveyor belt; S3: Perform data preprocessing on the dynamic spectral data; S4: Perform data fusion and feature extraction on the preprocessed dynamic spectral data to obtain spectral feature information; S4 specifically includes: S4.1: Determine the dynamic analysis time window for coal analysis; S4.2: Within the dynamic analysis time window, perform weighted average fusion on the preprocessed dynamic spectral data based on the coal flow surface density parameter to obtain fused spectral data; S4.3: Extract feature bands from the fused spectral data to obtain spectral feature information; S5: Based on the spectral feature information, predict and output the coal quality target.
2. The method for dynamic analysis of coal quality based on infrared detection according to claim 1, characterized in that, In step S4.1, the time required for the coal flow of the acquisition length analysis unit L is calculated based on the instantaneous linear velocity of the conveyor belt collected in step S2, which is the dynamic analysis time window for coal analysis.
3. The method for dynamic analysis of coal quality based on infrared detection according to claim 2, characterized in that, In step S4.2, the preprocessed dynamic spectral data collected at all times within the dynamic analysis time window are formed into a spectral sequence. A fusion weight is calculated for each frame of spectral data within the dynamic analysis time window. The fusion weight is proportional to the coal flow surface density at the time the coal in that frame of spectrum is collected. The final value of the fused spectral data is the sum of all frame spectra within the dynamic analysis time window after weighting them by their respective weights, divided by the total weight.
4. The method for dynamic analysis of coal quality based on infrared detection according to claim 3, characterized in that, In step S4.3, a partial least squares regression model is used to extract feature bands from the fused spectral data.
5. The method for dynamic analysis of coal quality based on infrared detection according to claim 1, characterized in that, In S1, the hardware platform includes a spectrum acquisition module, a light source module, an optical probe and optical path design, and auxiliary sensors.
6. The method for dynamic analysis of coal quality based on infrared detection according to claim 5, characterized in that, The spectral acquisition module is a near-infrared spectrometer. The core parameters of the near-infrared spectrometer are: spectral range covering 900nm to 2500nm; spectral resolution better than 8nm; and equipped with an integrating sphere accessory with adjustable aperture or a fiber optic diffuse reflection probe. The light source module is connected to the spectrum acquisition module. The light source module uses two quartz halogen lamps as a stable broadband light source. The continuous light emitted by the quartz halogen lamps is collimated by a collimating lens to form a parallel beam. The continuous light is then modulated into an alternating signal of a specific frequency by an electrically adjustable mechanical chopper. The optical probe is connected to the spectral acquisition module. The optical probe is an active diffuse reflection probe. The active diffuse reflection probe integrates a light source incident fiber and a signal receiving fiber, which are arranged at a preset angle and are aligned with the coal flow surface on the conveyor belt. The optical probe is mounted on a three-dimensionally adjustable bracket.
7. The method for dynamic analysis of coal quality based on infrared detection according to claim 1, characterized in that, The software environment initialization and parameter configuration are specifically performed as follows: start the data acquisition software and load the pre-calibrated system configuration file; set the core analysis parameters, including the integration time, averaging times, smoothing algorithm, chopper frequency, data storage path, and alarm threshold of the spectral acquisition module; perform a dark current calibration: in a non-luminescent state, acquire a set of spectral data as a dark spectrum for use in subsequent spectral preprocessing to subtract electronic noise; and perform a whiteboard calibration: place a standard reflective whiteboard in the detection area.
8. The method for dynamic analysis of coal quality based on infrared detection according to claim 1, characterized in that, In step S3, the data preprocessing includes dark current subtraction and spectral smoothing filtering.
9. The method for dynamic analysis of coal quality based on infrared detection according to claim 8, characterized in that, The dark current subtraction is specifically achieved by performing a subtraction operation on each wavelength point of the newly acquired original dynamic spectral data for each frame during software environment initialization and parameter configuration, thereby realizing dark current subtraction. The spectral smoothing filtering specifically involves applying a Savitzky-Golay filter to achieve spectral smoothing filtering. The window size of the Savitzky-Golay filter is 11 points, and the polynomial order is a second-order polynomial.
10. A dynamic coal quality analysis system based on infrared detection, characterized in that, The system employs a dynamic coal quality analysis method based on infrared detection as described in any one of claims 1-9, and the system comprises: The build and initialization module is used to build a hardware platform for coal quality testing and to initialize the software environment and configure parameters. The data acquisition module is used to collect dynamic spectral data of coal on the conveyor belt in real time. The data preprocessing module is used to perform data preprocessing operations on the dynamic spectral data; The feature extraction module is used to perform data fusion and feature extraction on the preprocessed dynamic spectral data to obtain spectral feature information; The prediction and output module is used to predict and output coal quality targets based on the spectral feature information.