Laser ray-based real-time measurement system for coal concentration in suspension kiln
By constructing an equivalent particle scattering tensor field of pulverized coal inside the suspension kiln using laser beam technology, the problem of large measurement error in pulverized coal concentration in traditional suspension kilns was solved, enabling more accurate measurement and production control of pulverized coal concentration and improving the quality of finished lime powder.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANXI FUYUANTONG MINING CO LTD
- Filing Date
- 2026-04-02
- Publication Date
- 2026-07-24
Smart Images
Figure CN122448698A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of pulverized coal concentration measurement technology, and in particular to a real-time pulverized coal concentration metering system for suspension kilns based on laser beams. Background Technology
[0002] Currently, the suspension kiln is a new type of high-efficiency heat treatment equipment, mainly used for calcination to produce lime powder. Compared with the traditional rotary kiln, the suspension kiln feeds powdered coal and raw materials through the top of the kiln, and then completes the production of lime powder by calcination for 1.3-1.5 seconds.
[0003] In traditional methods, the concentration of pulverized coal in a suspension kiln is directly calculated by the gas flow rate. However, this method is easily affected by factors such as airflow disturbance, particle agglomeration, and pipe wear, resulting in large measurement errors of pulverized coal concentration in the suspension kiln. This has an adverse effect on the production control of the suspension kiln and reduces the qualified rate of the finished lime powder in the suspension kiln. Summary of the Invention
[0004] To address the aforementioned technical problems, this invention provides a real-time coal powder concentration metering system for suspension kilns based on laser beams. The technical solution of this invention is as follows: A real-time coal powder concentration metering system based on laser beams for suspension kilns includes: The signal acquisition module is used to emit laser rays of different wavelengths through a laser ray emitter and to acquire multi-wavelength load reflection waves through the gas-solid two-phase flow inside the suspension kiln through a receiving array. The wavelength feature recognition module is used to obtain the wavelength features of load reflection waves of different wavelengths based on multi-wavelength load reflection waves, and to generate the original spectral matrix of solid particles inside the suspension kiln based on the wavelength features of load reflection waves of all wavelengths. The decoupling module is used to decouple the original spectral matrix to obtain an overcomplete dictionary, and to extract the independent spectral components of coal powder in the overcomplete dictionary through an orthogonal matching algorithm. The inversion module is used to invert the independent spectral components of pulverized coal to obtain the equivalent particle scattering tensor field of pulverized coal inside the suspension kiln. The concentration calculation module is used to establish a mapping relationship between coal powder particle scattering and particle density based on the equivalent particle scattering tensor field of coal powder inside the suspension kiln, and to calculate the coal powder concentration inside the suspension kiln based on the mapping relationship between coal powder particle scattering and particle density.
[0005] Preferably, the wavelength feature recognition module includes: The signal processing unit is used to process multi-wavelength load reflected waves using fast Fourier transform to obtain order domain eigenvectors. The single-wavelength feature extraction unit is used to extract the core features of the order domain feature vector by single wavelength to obtain the wavelength features of the load reflection wave of each wavelength. Curve construction unit is used to map the wavelength characteristics of the load reflected wave of each wavelength onto a two-dimensional complex plane, and construct a feature reference surface based on the mapping result; The matrix generation unit is used to perform equidistant sampling on the feature reference surface according to a preset step size and a preset distance to generate the original spectral matrix of solid particles inside the suspension kiln.
[0006] Preferably, the single-wavelength feature extraction unit includes: The high-dimensional mapping subunit is used to map the feature vectors of the first-order domain to a high-dimensional space to generate a high-dimensional point cloud set; The complex construction sub-unit is used to construct a simple complex based on the Euclidean distance between every two points in the high-dimensional point cloud set. The vertices of the simple complex represent points in the high-dimensional point cloud set, and the edges of the simple complex are generated based on the Euclidean distance between two points in the high-dimensional point cloud set. The filtering subunit is used to obtain the rank of each preset wavelength threshold in the simplex based on the preset wavelength threshold corresponding to each wavelength of laser beam emitted by the laser beam emitter, sort all the ranks in descending order, and select the points corresponding to the first preset number of ranks in the high-dimensional point cloud set as core feature points. The feature generation subunit is used to locate the load reflection wave corresponding to each core feature point in the multi-wavelength load reflection wave, and to use the amplitude and initial phase of the load reflection wave corresponding to each core feature point as the wavelength feature of the load reflection wave of each wavelength.
[0007] Preferably, the decoupling module includes: The initialization unit is used to standardize the original spectral matrix to obtain a standard spectral matrix, and to obtain a preset number of column vectors randomly selected from the standard spectral matrix to construct an initial dictionary; The dictionary optimization unit is used to iteratively optimize the initial dictionary to obtain a super-complete dictionary containing multiple dictionary elements; The support solution calculation unit is used to calculate the inner product of the initial residual and each dictionary element in the overcomplete dictionary with the preset load reflection wavelength of pulverized coal as the initial residual, select the dictionary element with the smallest inner product as the support solution, add the support solution to the support solution set, and update the initial residual based on the support solution to obtain the latest residual. Support solution set construction unit is used to iterate the latest residual and add the support solution obtained in each iteration to the support solution set until the change value of the latest residual after two consecutive iterations is less than the preset residual threshold, at which point the iteration stops and the optimal support solution set is obtained; The spectral component calculation unit is used to reconstruct the optimal support solution set through the orthogonal matching algorithm to obtain the independent spectral components of coal powder in the overcomplete dictionary.
[0008] Preferably, the dictionary optimization unit includes: The sparse coding subunit is used to construct the sparse coding error matrix of each initial element based on the column index of each initial element in the initial dictionary; The iterative subunit is used to perform singular value decomposition on the sparse coding error matrix of each initial element to obtain the decomposition matrix of each initial element. The first column of the decomposition matrix of each initial element is taken as the dictionary element of each initial element, and an overcomplete dictionary is generated based on the dictionary elements of all initial elements.
[0009] Preferably, the inversion module includes: The model building unit is used to build a three-dimensional model of the suspension kiln according to the preset factory dimensions of the suspension kiln, and to divide the three-dimensional model of the suspension kiln into multiple unit models of the same volume, and to configure a position index for each unit model. Tensor function construction unit is used to construct the tensor function of each unit model based on the position index of each unit model and the preset standard parameters of coal powder, and to obtain the initial tensor of each unit model based on the tensor function of each unit model. The inversion unit is used to invert each unit model based on the initial tensor of each unit model to obtain the equivalent particle scattering tensor of each unit model. The equivalent particle scattering tensors of all unit models are then spliced together to form the equivalent particle scattering tensor field of the coal powder inside the suspension kiln.
[0010] Preferably, when the inversion unit inverts each unit model based on the initial tensor of each unit model to obtain the equivalent particle scattering tensor of each unit model, it includes: The theoretical spectral signal of each unit model is determined based on the initial tensor of each unit model. The measured spectral signal of each unit model in the multi-wavelength load reflection wave is obtained. The signal residual between the theoretical spectral signal and the measured spectral signal of each unit model is calculated. Calculate the tensor gradient of each unit model based on the initial tensor of each unit model, and update the initial tensor of each unit model based on the tensor gradient of each unit model and the signal residual to obtain the updated tensor. The updated tensor of each unit model is used as the new initial tensor, and the new theoretical spectral signal of each unit model is reacquired. The updated tensor of each unit model is iteratively updated until the signal residual of each unit model is less than a preset threshold. The updated tensor of each unit model after the iteration stops is used as the equivalent particle scattering tensor of each unit model.
[0011] Preferably, the concentration calculation module includes: The spectral decomposition unit is used to obtain the equivalent particle scattering tensor of each unit model based on the equivalent particle scattering tensor field of the coal powder inside the suspension kiln, and decompose the equivalent particle scattering tensor of each unit model into multiple particle features of each unit model according to the radiative transfer theory. The concentration conversion unit is used to determine the mapping relationship between coal powder particle scattering and particle density of each unit model based on the various particle characteristics of each unit model, and to calculate the unit coal powder concentration of each unit model based on the mapping relationship between coal powder particle scattering and particle density of each unit model. The pulverized coal concentration calculation unit is used to calculate the pulverized coal concentration inside the suspension kiln based on the pulverized coal concentration of each unit model and the number of 3D model unit models of the suspension kiln.
[0012] Preferably, the concentration conversion unit includes: The particle total coefficient calculation subunit is used to perform weighted calculation of multiple particle features of each unit model and the preset weight coefficients corresponding to each particle feature to obtain the particle total coefficient of each unit model. The mapping operator obtains sub-units and is used to calculate the mapping operator of each unit model based on the total particle coefficient of each unit model and the particle size among the various particle characteristics of each unit model. The mapping operator of each unit model represents the mapping relationship between the scattering of coal powder particles and the particle density of each unit model. The concentration calculation sub-unit is used to determine the number of particles in each unit model based on the mapping operator of each unit model and the various particle characteristics of each unit model, and to calculate the unit coal powder concentration of each unit model based on the unit model volume and the number of particles in each unit model.
[0013] All of the above-mentioned optional technical solutions can be combined arbitrarily, and the present invention will not provide a detailed description of the structure after each combination.
[0014] By means of the above solution, the beneficial effects of the present invention are as follows: This invention acquires multi-wavelength load reflection waves from the gas-solid two-phase flow inside a suspension kiln using a laser beam receiving array. The original spectral matrix is obtained through wavelength feature identification. The equivalent particle scattering tensor field of the pulverized coal inside the suspension kiln is accurately obtained through decoupling and inversion. Finally, the pulverized coal concentration inside the suspension kiln is calculated based on this tensor field. Compared to traditional methods that calculate pulverized coal concentration based on gas flow rate, the use of multiple wavelengths of laser beams can accurately match the reflected waves of the pulverized coal, is less affected by environmental factors, and, combined with the relationship between the equivalent particle scattering tensor field and particle density, reduces the measurement error of the pulverized coal concentration in the suspension kiln. The calculated pulverized coal concentration inside the suspension kiln provides an accurate data basis for the production control of the suspension kiln, thereby improving the yield of finished lime powder.
[0015] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings. Attached Figure Description
[0016] Figure 1 This is a schematic diagram of the structure of the real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams provided by the present invention. Detailed Implementation
[0017] The specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.
[0018] like Figure 1 As shown in the embodiment of the present invention, the real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams includes: The signal acquisition module is used to emit laser rays of different wavelengths through a laser ray emitter and to acquire multi-wavelength load reflection waves through the gas-solid two-phase flow inside the suspension kiln through a receiving array. The wavelength feature recognition module is used to obtain the wavelength features of load reflection waves of different wavelengths based on multi-wavelength load reflection waves, and to generate the original spectral matrix of solid particles inside the suspension kiln based on the wavelength features of load reflection waves of all wavelengths. The decoupling module is used to decouple the original spectral matrix to obtain an overcomplete dictionary, and to extract the independent spectral components of coal powder in the overcomplete dictionary through an orthogonal matching algorithm. The inversion module is used to invert the independent spectral components of pulverized coal to obtain the equivalent particle scattering tensor field of pulverized coal inside the suspension kiln. The concentration calculation module is used to establish a mapping relationship between coal powder particle scattering and particle density based on the equivalent particle scattering tensor field of coal powder inside the suspension kiln, and to calculate the coal powder concentration inside the suspension kiln based on the mapping relationship between coal powder particle scattering and particle density.
[0019] Specifically, in the signal acquisition module, the laser beam emitted by the laser beam emitter has multiple pre-selected wavelengths, including 400-700nm, 700-1000nm, and 1100-2500nm. Gas-solid two-phase flow refers to a mixed fluid of solid particles and gas inside the suspension kiln; the solid particles are a mixture of particles such as magnesium oxide, pulverized coal, and calcium oxide.
[0020] In the wavelength feature identification module, the multi-wavelength load reflected wave is composed of load reflected waves of various wavelengths. In this embodiment of the invention, the wavelength feature refers to amplitude and initial phase. The original spectral matrix contains information about the wavelength and time-varying directions of the solid particles inside the suspension kiln.
[0021] In the decoupling module, decoupling refers to the processing step of separating the complex original spectral matrix into simpler, more independent overcomplete dictionaries. An overcomplete dictionary is a dictionary that stores signals, composed of multiple dictionary elements.
[0022] In the inversion module, inversion refers to the process of deriving a deeper equivalent particle scattering tensor field from the independent spectral components of pulverized coal. The equivalent particle scattering tensor field is a mathematical representation describing the reflection of laser rays by pulverized coal particles in a suspension kiln.
[0023] This invention utilizes multi-wavelength laser reflection wave acquisition and analysis to extract independent spectral components of pulverized coal and construct an equivalent particle scattering tensor field, ultimately calculating the pulverized coal concentration. Compared to directly calculating pulverized coal concentration through gas flow rate, this method improves the accuracy and reliability of pulverized coal concentration monitoring in suspension kilns.
[0024] In one specific embodiment, the wavelength feature recognition module includes: The signal processing unit is used to process multi-wavelength load reflected waves using fast Fourier transform to obtain order domain eigenvectors. The single-wavelength feature extraction unit is used to extract the core features of the order domain feature vector by single wavelength to obtain the wavelength features of the load reflection wave of each wavelength. Curve construction unit is used to map the wavelength characteristics of the load reflected wave of each wavelength onto a two-dimensional complex plane, and construct a feature reference surface based on the mapping result; The matrix generation unit is used to perform equidistant sampling on the feature reference surface according to a preset step size and a preset distance to generate the original spectral matrix of solid particles inside the suspension kiln.
[0025] Specifically, when generating the order domain feature vector, the signal processing unit first performs a fast Fourier transform on the multi-wavelength load reflection wave to convert the time-domain multi-wavelength load reflection wave into a frequency-domain order domain signal. Then, according to the preset frequency components (the preset frequency components correspond one-to-one with each wavelength of laser beam emitted by the laser beam emitter), the order domain signal is processed to obtain multiple order signals. Finally, the frequency domain features such as amplitude, initial phase, order, and extreme values of each order signal are extracted, and the frequency domain features of all order signals constitute the order domain feature vector.
[0026] In the curve construction unit, the two-dimensional complex plane in this embodiment of the invention refers to the Gaussian plane. This embodiment constructs a two-dimensional coordinate system in the two-dimensional complex plane, with the horizontal axis representing the real part and the vertical axis representing the imaginary part (in this embodiment, the imaginary part is time-dependent), and the origin is a random point in the two-dimensional complex plane. If the wavelength characteristics of a load-reflected wave of a certain wavelength include amplitude... and initial phase Then, the wavelength characteristics of the load-reflected wave of that wavelength, mapped onto a two-dimensional coordinate system in the two-dimensional complex plane, are expressed as: ;in, Indicates the real part, The imaginary part is represented by this. After mapping the wavelength characteristics of the load reflected wave of each wavelength to a two-dimensional coordinate system in the two-dimensional complex plane, the imaginary part of the load reflected wave of each wavelength forms a time-dependent imaginary curve. Connecting the imaginary curves of the load reflected waves of all wavelengths yields the characteristic reference surface.
[0027] In the matrix generation unit, when performing equidistant sampling on the feature reference surface according to a preset step size and a preset distance, a preset reference point on the feature reference surface is first selected as the starting point. The wavelength change direction (real part direction) is sampled equidistantly along the feature reference surface according to the preset step size to obtain M wavelength change direction vectors. The time change direction (imaginary part direction) is sampled equidistantly along the preset distance to obtain N time direction vectors. The M wavelength change direction vectors and the N time direction vectors constitute the M×N original spectral matrix.
[0028] This invention uses Fast Fourier Transform to process multi-wavelength load reflection waves, extracts the wavelength features of each wavelength of the load reflection waves, and constructs a feature reference surface. This enables an accurate description of the spectral characteristics of solid particles inside the suspension kiln, and the generated original spectral matrix provides a high-quality data foundation for subsequent concentration analysis.
[0029] In one specific embodiment, the single-wavelength feature extraction unit includes: The high-dimensional mapping subunit is used to map the feature vectors of the first-order domain to a high-dimensional space to generate a high-dimensional point cloud set; The complex construction sub-unit is used to construct a simple complex based on the Euclidean distance between every two points in the high-dimensional point cloud set. The vertices of the simple complex represent points in the high-dimensional point cloud set, and the edges of the simple complex are generated based on the Euclidean distance between two points in the high-dimensional point cloud set. The filtering subunit is used to obtain the rank of each preset wavelength threshold in the simplex based on the preset wavelength threshold corresponding to each wavelength of laser beam emitted by the laser beam emitter, sort all the ranks in descending order, and select the points corresponding to the first preset number of ranks in the high-dimensional point cloud set as core feature points. The feature generation subunit is used to locate the load reflection wave corresponding to each core feature point in the multi-wavelength load reflection wave, and to use the amplitude and initial phase of the load reflection wave corresponding to each core feature point as the wavelength feature of the load reflection wave of each wavelength.
[0030] Specifically, in the high-dimensional mapping subunit, when mapping the order-domain feature vector to the high-dimensional space, a Gaussian mapping is performed on the frequency domain features of a certain order signal in the order-domain feature vector (implemented through a Gaussian kernel function in this embodiment of the invention) to obtain the points corresponding to that order signal in the high-dimensional space. The points corresponding to all order signals constitute a high-dimensional point cloud set. After mapping a signal of order 1 to the high-dimensional space, its corresponding point in the high-dimensional point cloud set can be represented as: ,in, This represents the point of order 1 with wavelength c of the reflected wave from the load. This represents the point of order 1 with wavelength J of the reflected wave from the load. This represents a point of order 1 with a wavelength of V for the reflected wave from the load.
[0031] In the complex construction subunit, when constructing a simple complex, all points in the high-dimensional point cloud set are first used as a vertex set. Then, any two points in the vertex set are obtained. If the Euclidean distance between these two points is less than a preset edge-forming threshold, the two points are connected as an edge; otherwise, they are not connected. This operation is performed on all points in the vertex set to obtain a simple complex generated from the high-dimensional point cloud set. The length of any side of the simple complex is the Euclidean distance between the two points connecting that side. In this embodiment of the invention, the preset edge-forming threshold is preferably 0.5 cm.
[0032] In the filtering subunit, when obtaining the rank of a preset wavelength threshold in a simplex, the absolute difference between the side length of each side of the simplex and the preset wavelength threshold is first obtained. Sides with an absolute difference less than the wavelength difference threshold are considered as the nearest sides corresponding to the preset wavelength threshold in the simplex. The number of nearest sides corresponding to the preset wavelength threshold in the simplex is then used as the rank of the preset wavelength threshold in the simplex. In this embodiment, the preset number is preferably 5. The preset wavelength threshold for each wavelength of laser beam is an empirical threshold determined based on historical data, and the wavelength difference threshold is a threshold based on the absolute difference obtained through expert evaluation.
[0033] In the feature generation subunit, when locating the load reflection wave corresponding to each core feature point in a multi-wavelength load reflection wave, the wavelength corresponding to each core feature point in the load reflection wave is found through the reverse index number of each core feature point. For example, a certain core feature point is... Therefore, we can determine that the wavelength of the load reflection wave corresponding to this core feature point is c.
[0034] This invention, through the construction and analysis of high-dimensional point cloud sets and the efficient extraction of core feature points using simple complexes, generates the wavelength features of each wavelength of the load reflection wave of a multi-wavelength load reflection wave, providing an accurate data foundation for the construction of the original spectral matrix.
[0035] In one specific embodiment, the decoupling module includes: The initialization unit is used to standardize the original spectral matrix to obtain a standard spectral matrix, and to obtain a preset number of column vectors randomly selected from the standard spectral matrix to construct an initial dictionary; The dictionary optimization unit is used to iteratively optimize the initial dictionary to obtain a super-complete dictionary containing multiple dictionary elements; The support solution calculation unit is used to calculate the inner product of the initial residual and each dictionary element in the overcomplete dictionary with the preset load reflection wavelength of pulverized coal as the initial residual, select the dictionary element with the smallest inner product as the support solution, add the support solution to the support solution set, and update the initial residual based on the support solution to obtain the latest residual. Support solution set construction unit is used to iterate the latest residual and add the support solution obtained in each iteration to the support solution set until the change value of the latest residual after two consecutive iterations is less than the preset residual threshold, at which point the iteration stops and the optimal support solution set is obtained; The spectral component calculation unit is used to reconstruct the optimal support solution set through the orthogonal matching algorithm to obtain the independent spectral components of coal powder in the overcomplete dictionary.
[0036] Specifically, in the initialization unit, standardization refers to standardizing each element in the original spectral matrix to the range [0,1] using a normal distribution. The preset number of columns is a number pre-set based on historical experience values. The initial dictionary can be represented as: {column 1, column 2, column 3}.
[0037] In the support solution computation unit, the initial residual is updated based on the support solution. Obtain the latest residual When, the calculation formula is: ;in, Let represent the modulus of the matrix formed by all support solutions in the support solution set, and s represent the inner product of the initial residual and the support solutions.
[0038] In the support solution set construction unit, when iterating the latest residual, the latest residual is used as the new initial residual. The inner product of the new initial residual and each dictionary element in the overcomplete dictionary is calculated. The dictionary element with the smallest inner product is selected as the support solution obtained by iteration. The support solution obtained by iteration is added to the support solution set.
[0039] In the spectral component calculation unit, when reconstructing the optimal support solution set using the orthogonal matching algorithm to obtain the independent spectral components of coal powder in the overcomplete dictionary, the optimal support solution set is fitted using the least squares method, specifically as follows: , This represents a matrix constructed from all support solutions in the optimal support solution set, where each row represents a support solution. This represents the fit coefficient vector (each element in the fit coefficient vector corresponds one-to-one with each support solution). This represents the inner product. After fitting, the fitted coefficient vector corresponding to each support solution is obtained. Each fitted coefficient in the fitted coefficient vector is multiplied by the corresponding support solution in the optimal support solution set to obtain the independent vector corresponding to each support solution. Combining the independent vectors of all support solutions yields the independent spectral components of coal powder in the overcomplete dictionary.
[0040] The embodiments of the present invention construct an ultra-complete dictionary through standardization processing and iterative optimization of the initial dictionary, and utilize iterative calculations supporting solution sets to accurately extract the independent spectral components of coal powder, thereby improving the accuracy and efficiency of coal powder spectral analysis and providing a more accurate data foundation for coal powder concentration measurement.
[0041] In one specific embodiment, the dictionary optimization unit includes: The sparse coding subunit is used to construct the sparse coding error matrix of each initial element based on the column index of each initial element in the initial dictionary; The iterative subunit is used to perform singular value decomposition on the sparse coding error matrix of each initial element to obtain the decomposition matrix of each initial element. The first column of the decomposition matrix of each initial element is taken as the dictionary element of each initial element, and an overcomplete dictionary is generated based on the dictionary elements of all initial elements.
[0042] Specifically, in the sparse coding subunit, the sparse coding error matrix of a certain initial element is constructed. When, the formula is: ;in, Represents the original spectral matrix. This represents the initial element at column index k in the initial dictionary. This represents the row with row index k in the original spectral matrix.
[0043] In the iterative sub-unit, the sparse coding error matrix for a certain initial element mentioned above Perform singular value decomposition to obtain the left singular matrix, the right singular matrix, and the diagonal matrix. Use the left singular matrix as the decomposition matrix of the initial element.
[0044] The embodiments of the present invention can accurately construct an overcomplete dictionary by constructing a sparse coding error matrix and performing singular value decomposition, providing an accurate data foundation for the subsequent extraction of independent spectral components of coal powder.
[0045] In one specific embodiment, the inversion module includes: The model building unit is used to build a three-dimensional model of the suspension kiln according to the preset factory dimensions of the suspension kiln, and to divide the three-dimensional model of the suspension kiln into multiple unit models of the same volume, and to configure a position index for each unit model. Tensor function construction unit is used to construct the tensor function of each unit model based on the position index of each unit model and the preset standard parameters of coal powder, and to obtain the initial tensor of each unit model based on the tensor function of each unit model. The inversion unit is used to invert each unit model based on the initial tensor of each unit model to obtain the equivalent particle scattering tensor of each unit model. The equivalent particle scattering tensors of all unit models are then spliced together to form the equivalent particle scattering tensor field of the coal powder inside the suspension kiln.
[0046] Specifically, in the model building unit, when splitting the 3D model of the suspension kiln, this embodiment of the invention automatically achieves this through the 3D irregular mesh generation method in Revit software. The location index is a 3D location index representing the spatial positioning relationship assigned to each unit model after the 3D irregular mesh generation.
[0047] In the tensor function construction unit, the preset standard parameters of pulverized coal include density, refractive index, absorption coefficient, and scattering coefficient. The tensor function of a given element model is constructed based on its location index and the preset standard parameters of the pulverized coal. When, it can be represented as ;in, The positions are represented by o, p, and q, which are the position indices of the unit model, respectively. This represents the second-order tensor construction function. The position index of each unit model is input into the corresponding tensor function to obtain the initial tensor of each unit model (in this embodiment, it is in the form of a second-order tensor matrix).
[0048] In the inversion unit, when inverting a certain unit model to obtain the equivalent particle scattering tensor of the unit model, the initial tensor of the unit model is input into the lattice Boltzmann forward model, and the lattice Boltzmann forward model outputs the equivalent particle scattering tensor of the unit model.
[0049] This invention, through the construction of a three-dimensional model of a suspension kiln and its fine division into multiple unit models, combined with tensor functions and inversion methods, can accurately obtain the equivalent particle scattering tensor of pulverized coal, thereby improving the ability to analyze the characteristics of pulverized coal inside the suspension kiln and providing a scientific basis for analyzing pulverized coal concentration.
[0050] In a specific embodiment, when the inversion unit inverts each unit model based on the initial tensor of each unit model to obtain the equivalent particle scattering tensor of each unit model, it includes: The theoretical spectral signal of each unit model is determined based on the initial tensor of each unit model. The measured spectral signal of each unit model in the multi-wavelength load reflection wave is obtained. The signal residual between the theoretical spectral signal and the measured spectral signal of each unit model is calculated. Calculate the tensor gradient of each unit model based on the initial tensor of each unit model, and update the initial tensor of each unit model based on the tensor gradient of each unit model and the signal residual to obtain the updated tensor. The updated tensor of each unit model is used as the new initial tensor, and the new theoretical spectral signal of each unit model is reacquired. The updated tensor of each unit model is iteratively updated until the signal residual of each unit model is less than a preset threshold. The updated tensor of each unit model after the iteration stops is used as the equivalent particle scattering tensor of each unit model.
[0051] Specifically, when determining the theoretical spectral signal of a unit model based on its initial tensor, the mean of all elements in the initial tensor is calculated as the scattering intensity of the unit model. The ratio of the sum of the diagonal elements to the sum of the off-diagonal elements in the initial tensor is calculated as the anisotropy factor of the unit model. The theoretical spectral signal is then calculated based on the scattering intensity, the anisotropy factor, and the absorption coefficient in the preset standard parameters of the pulverized coal. The formula is: ;in, The scattering intensity of the element model is represented by g, and the anisotropy factor of the element model is represented by g. Indicates the absorption coefficient. This represents the measured spectral signal corresponding to the unit model in multi-wavelength load reflection waves. To obtain the measured spectral signal corresponding to the unit model in multi-wavelength load reflection waves, the unit model is first divided into unit models according to the three-dimensional model of the suspension kiln. The center position of the unit model is then used as the measurement point. The collected multi-wavelength load reflection waves are then correlated with the position index of the measurement point of the unit model to obtain the measured spectral signal corresponding to the unit model in multi-wavelength load reflection waves.
[0052] When calculating the tensor gradient of the unit model based on the initial tensor of the unit model, the partial derivative of the initial tensor of the unit model with respect to each column is calculated to obtain the component gradient of the initial tensor with respect to each column. All component gradients form a combined vector, which is the tensor gradient of the unit model.
[0053] Furthermore, when updating the initial tensor of the unit model based on the tensor gradient and signal residual of the unit model to obtain the updated tensor, the product of the tensor gradient and the signal residual is first calculated to obtain the update amount. When the signal residual is greater than 0, the initial tensor is superimposed with the update amount to obtain the updated tensor; when the signal residual is less than 0, the initial tensor is subtracted from the update amount to obtain the updated tensor; when the signal residual is equal to 0, the initial tensor is used as the updated tensor.
[0054] This invention achieves precise optimization of the equivalent particle scattering tensor by calculating the signal residual between the theoretical spectral signal and the measured spectral signal and updating the initial tensor of each unit model. This improves the accuracy of tensor analysis of unit models in the three-dimensional model and provides a solid foundation for subsequent analysis of coal powder concentration.
[0055] In one specific embodiment, the concentration calculation module includes: The spectral decomposition unit is used to obtain the equivalent particle scattering tensor of each unit model based on the equivalent particle scattering tensor field of the coal powder inside the suspension kiln, and decompose the equivalent particle scattering tensor of each unit model into multiple particle features of each unit model according to the radiative transfer theory. The concentration conversion unit is used to determine the mapping relationship between coal powder particle scattering and particle density of each unit model based on the various particle characteristics of each unit model, and to calculate the unit coal powder concentration of each unit model based on the mapping relationship between coal powder particle scattering and particle density of each unit model. The pulverized coal concentration calculation unit is used to calculate the pulverized coal concentration inside the suspension kiln based on the pulverized coal concentration of each unit model and the number of 3D model unit models of the suspension kiln.
[0056] Specifically, in the spectral decomposition unit, particle characteristics include scattering intensity, anisotropy factor, and particle size. When decomposing the equivalent particle scattering tensor of a certain unit model into various particle characteristics of that unit model according to radiative transfer theory, the scattering intensity and anisotropy factor of the equivalent particle scattering tensor of the unit model are first calculated. The calculation method is the same as the principle for calculating the scattering intensity and anisotropy factor of the initial tensor in the above embodiment, and will not be repeated here. Then, the particle size of the equivalent particle scattering tensor of the unit model is calculated. Specifically, the particle size of the equivalent particle scattering tensor of the unit model is calculated... When, the formula is: , g represents the heterogeneity factor of the unit model, This represents the measured spectral signal corresponding to the unit model in multi-wavelength load reflection waves.
[0057] In the concentration conversion unit, the mapping relationship between coal powder particle scattering and particle density of a certain unit model is represented by calculating the mapping operator between coal powder particle scattering and particle density of that unit model.
[0058] In the coal powder concentration calculation unit, the sum of the coal powder concentrations of all unit models of the 3D model of the suspension kiln is calculated to obtain the total coal powder concentration. The total coal powder concentration is then divided by the number of unit models of the 3D model of the suspension kiln to obtain the coal powder concentration inside the suspension kiln.
[0059] This invention, through spectral decomposition and concentration conversion, accurately analyzes the particle characteristics of pulverized coal inside the suspension kiln, achieving an effective mapping between pulverized coal particle scattering and particle density, thereby calculating the unit pulverized coal concentration of each unit model and improving the accuracy of pulverized coal concentration measurement.
[0060] In one specific embodiment, the concentration conversion unit includes: The particle total coefficient calculation subunit is used to perform weighted calculation of multiple particle features of each unit model and the preset weight coefficients corresponding to each particle feature to obtain the particle total coefficient of each unit model. The mapping operator obtains sub-units and is used to calculate the mapping operator of each unit model based on the total particle coefficient of each unit model and the particle size among the various particle characteristics of each unit model. The mapping operator of each unit model represents the mapping relationship between the scattering of coal powder particles and the particle density of each unit model. The concentration calculation sub-unit is used to determine the number of particles in each unit model based on the mapping operator of each unit model and the various particle characteristics of each unit model, and to calculate the unit coal powder concentration of each unit model based on the unit model volume and the number of particles in each unit model.
[0061] Specifically, in the particle total coefficient calculation subunit, particle characteristics include scattering intensity, anisotropy factor, and particle size. The preset weighting coefficients corresponding to each particle characteristic are determined in advance based on historical experience values. In this embodiment of the invention, the preset weighting coefficient for scattering intensity is preferably 0.3, the preset weighting coefficient for anisotropy factor is preferably 0.3, and the preset weighting coefficient for particle size is preferably 0.4.
[0062] The mapping operator obtains the total particle coefficient of a certain unit model within the sub-unit. Given the particle size d among various particle characteristics of the unit model, calculate the mapping operator for the unit model. When, the formula is: ,in, This refers to the density in the preset standard parameters of pulverized coal. This represents the scattering intensity of the unit model.
[0063] In the concentration calculation subunit, the mapping operator is based on the above unit model. When determining the number of particles N in a unit model based on various particle characteristics, the formula is: Calculate the ratio of particle count to unit volume for each unit model to obtain the unit coal powder concentration for each unit model.
[0064] This invention provides an embodiment that accurately obtains the total particle coefficient and mapping operator for each unit model, thereby effectively describing the relationship between pulverized coal particle scattering and particle density, and then calculating the unit pulverized coal concentration, thus improving the accuracy of pulverized coal concentration measurement inside the suspension kiln.
[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams, characterized in that, include: The signal acquisition module is used to emit laser rays of different wavelengths through a laser ray emitter and to acquire multi-wavelength load reflection waves through the gas-solid two-phase flow inside the suspension kiln through a receiving array. The wavelength feature recognition module is used to obtain the wavelength features of load reflection waves of different wavelengths based on multi-wavelength load reflection waves, and to generate the original spectral matrix of solid particles inside the suspension kiln based on the wavelength features of load reflection waves of all wavelengths. The decoupling module is used to decouple the original spectral matrix to obtain an overcomplete dictionary, and to extract the independent spectral components of coal powder in the overcomplete dictionary through an orthogonal matching algorithm. The inversion module is used to invert the independent spectral components of pulverized coal to obtain the equivalent particle scattering tensor field of pulverized coal inside the suspension kiln. The concentration calculation module is used to establish a mapping relationship between coal powder particle scattering and particle density based on the equivalent particle scattering tensor field of coal powder inside the suspension kiln, and to calculate the coal powder concentration inside the suspension kiln based on the mapping relationship between coal powder particle scattering and particle density.
2. The real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams according to claim 1, characterized in that, The wavelength feature recognition module includes: The signal processing unit is used to process multi-wavelength load reflected waves using fast Fourier transform to obtain order domain eigenvectors. The single-wavelength feature extraction unit is used to extract the core features of the order domain feature vector by single wavelength to obtain the wavelength features of the load reflection wave of each wavelength. Curve construction unit is used to map the wavelength characteristics of the load reflected wave of each wavelength onto a two-dimensional complex plane, and construct a feature reference surface based on the mapping result; The matrix generation unit is used to perform equidistant sampling on the feature reference surface according to a preset step size and a preset distance to generate the original spectral matrix of solid particles inside the suspension kiln.
3. The real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams according to claim 2, characterized in that, The single-wavelength feature extraction unit includes: The high-dimensional mapping subunit is used to map the feature vectors of the first-order domain to a high-dimensional space to generate a high-dimensional point cloud set; The complex construction sub-unit is used to construct a simple complex based on the Euclidean distance between every two points in the high-dimensional point cloud set. The vertices of the simple complex represent points in the high-dimensional point cloud set, and the edges of the simple complex are generated based on the Euclidean distance between two points in the high-dimensional point cloud set. The filtering subunit is used to obtain the rank of each preset wavelength threshold in the simplex based on the preset wavelength threshold corresponding to each wavelength of laser beam emitted by the laser beam emitter, sort all the ranks in descending order, and select the points corresponding to the first preset number of ranks in the high-dimensional point cloud set as core feature points. The feature generation subunit is used to locate the load reflection wave corresponding to each core feature point in the multi-wavelength load reflection wave, and to use the amplitude and initial phase of the load reflection wave corresponding to each core feature point as the wavelength feature of the load reflection wave of each wavelength.
4. The real-time pulverized coal concentration metering system for suspension kilns based on laser beams according to claim 1, characterized in that, The decoupling module includes: The initialization unit is used to standardize the original spectral matrix to obtain a standard spectral matrix, and to obtain a preset number of column vectors randomly selected from the standard spectral matrix to construct an initial dictionary; The dictionary optimization unit is used to iteratively optimize the initial dictionary to obtain a super-complete dictionary containing multiple dictionary elements; The support solution calculation unit is used to calculate the inner product of the initial residual and each dictionary element in the overcomplete dictionary with the preset load reflection wavelength of pulverized coal as the initial residual, select the dictionary element with the smallest inner product as the support solution, add the support solution to the support solution set, and update the initial residual based on the support solution to obtain the latest residual. Support solution set construction unit is used to iterate the latest residual and add the support solution obtained in each iteration to the support solution set until the change value of the latest residual after two consecutive iterations is less than the preset residual threshold, at which point the iteration stops and the optimal support solution set is obtained; The spectral component calculation unit is used to reconstruct the optimal support solution set through the orthogonal matching algorithm to obtain the independent spectral components of coal powder in the overcomplete dictionary.
5. The real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams according to claim 4, characterized in that, The dictionary optimization unit includes: The sparse coding subunit is used to construct the sparse coding error matrix of each initial element based on the column index of each initial element in the initial dictionary; The iterative subunit is used to perform singular value decomposition on the sparse coding error matrix of each initial element to obtain the decomposition matrix of each initial element. The first column of the decomposition matrix of each initial element is taken as the dictionary element of each initial element, and an overcomplete dictionary is generated based on the dictionary elements of all initial elements.
6. The real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams according to claim 1, characterized in that, The inversion module includes: The model building unit is used to build a three-dimensional model of the suspension kiln according to the preset factory dimensions of the suspension kiln, and to divide the three-dimensional model of the suspension kiln into multiple unit models of the same volume, and to configure a position index for each unit model. Tensor function construction unit is used to construct the tensor function of each unit model based on the position index of each unit model and the preset standard parameters of coal powder, and to obtain the initial tensor of each unit model based on the tensor function of each unit model. The inversion unit is used to invert each unit model based on the initial tensor of each unit model to obtain the equivalent particle scattering tensor of each unit model. The equivalent particle scattering tensors of all unit models are then spliced together to form the equivalent particle scattering tensor field of the coal powder inside the suspension kiln.
7. The real-time pulverized coal concentration metering system for suspension kilns based on laser beams according to claim 6, characterized in that, The inversion unit, when inverting each unit model based on its initial tensor to obtain the equivalent particle scattering tensor of each unit model, includes: The theoretical spectral signal of each unit model is determined based on the initial tensor of each unit model. The measured spectral signal of each unit model in the multi-wavelength load reflection wave is obtained. The signal residual between the theoretical spectral signal and the measured spectral signal of each unit model is calculated. Calculate the tensor gradient of each unit model based on the initial tensor of each unit model, and update the initial tensor of each unit model based on the tensor gradient of each unit model and the signal residual to obtain the updated tensor. The updated tensor of each unit model is used as the new initial tensor, and the new theoretical spectral signal of each unit model is reacquired. The updated tensor of each unit model is iteratively updated until the signal residual of each unit model is less than a preset threshold. The updated tensor of each unit model after the iteration stops is used as the equivalent particle scattering tensor of each unit model.
8. The real-time metering system for pulverized coal concentration in a suspension kiln based on laser beams according to claim 6, characterized in that, The concentration calculation module includes: The spectral decomposition unit is used to obtain the equivalent particle scattering tensor of each unit model based on the equivalent particle scattering tensor field of the coal powder inside the suspension kiln, and decompose the equivalent particle scattering tensor of each unit model into multiple particle features of each unit model according to the radiative transfer theory. The concentration conversion unit is used to determine the mapping relationship between coal powder particle scattering and particle density of each unit model based on the various particle characteristics of each unit model, and to calculate the unit coal powder concentration of each unit model based on the mapping relationship between coal powder particle scattering and particle density of each unit model. The pulverized coal concentration calculation unit is used to calculate the pulverized coal concentration inside the suspension kiln based on the pulverized coal concentration of each unit model and the number of 3D model unit models of the suspension kiln.
9. The real-time coal powder concentration metering system based on laser beams in a suspension kiln according to claim 8, characterized in that, The concentration conversion unit includes: The particle total coefficient calculation subunit is used to perform weighted calculation of multiple particle features of each unit model and the preset weight coefficients corresponding to each particle feature to obtain the particle total coefficient of each unit model. The mapping operator obtains sub-units and is used to calculate the mapping operator of each unit model based on the total particle coefficient of each unit model and the particle size among the various particle characteristics of each unit model. The mapping operator of each unit model represents the mapping relationship between the scattering of coal powder particles and the particle density of each unit model. The concentration calculation sub-unit is used to determine the number of particles in each unit model based on the mapping operator of each unit model and the various particle characteristics of each unit model, and to calculate the unit coal powder concentration of each unit model based on the unit model volume and the number of particles in each unit model.