A rotary kiln flue gas component detection method, device and system

Abstract

The present application relates to the technical field of industrial monitoring, and in particular to a rotary kiln flue gas composition detection method, device and system, which solves the technical problem of low accuracy of rotary kiln flue gas composition detection in the prior art. The method comprises: obtaining a measured light path vector of a rotary kiln flue cross section at a current time, a concentration matrix at a historical time, and spectral data; the measured light path vector is used to represent the absorption measurement value of each laser ray in the rotary kiln flue cross section at the current time; the concentration matrix at the historical time at least includes a concentration matrix at a previous time and a concentration matrix at two previous times; the concentration matrix is used to represent the gas concentration distribution of each spatial grid in the rotary kiln flue cross section at the corresponding time; iterative concentration field reconstruction is performed according to the measured light path vector at the current time, the concentration matrix at the historical time, and the spectral data to determine the concentration matrix at the current time; and the rotary kiln flue gas composition detection process is monitored according to the concentration matrix at the current time.

Description

Technical Field

[0001] This invention relates to the field of industrial monitoring technology, specifically to a method, apparatus, and system for detecting the composition of rotary kiln flue gas. Background Technology

[0002] Rotary kilns are widely used in industrial thermal processes such as cement clinker production, lime roasting, metallurgical pellet roasting, and hazardous waste and municipal solid waste incineration, generating large amounts of high-temperature flue gas during operation. Online monitoring of gaseous components such as oxygen and carbon monoxide in the flue gas is a crucial technical means for combustion status assessment, process optimization control, and safe operation. With the increasing automation of industrial processes, industrial kiln systems place higher demands on the real-time performance and spatial characterization capabilities of flue gas component monitoring. Therefore, the continuous detection of flue gas components by arranging gas monitoring devices at key locations such as the rotary kiln preheater outlet, decomposer outlet, secondary combustion chamber outlet, and kiln tail flue gas chamber has become an important component of modern industrial kiln operation monitoring systems.

[0003] In existing technologies, large-section industrial flues such as rotary kiln preheaters, decomposition furnaces, and secondary combustion chamber outlets are prone to probe scaling and physical corrosion due to high temperatures, extremely high dust loads, and severe stratification of the gas-solid two-phase flow field. Furthermore, data from single local measurement points have spatial limitations and cannot accurately represent the non-uniform gas concentration distribution gradient within a wide cross-section, thus reducing the accuracy of rotary kiln flue gas composition detection. Summary of the Invention

[0004] To address the technical problem of low accuracy in rotary kiln flue gas composition detection, the present invention aims to provide a method, apparatus, and system for detecting rotary kiln flue gas composition. The specific technical solution adopted is as follows: This application provides a method for detecting the composition of rotary kiln flue gas, including: The measured optical path vector, historical concentration matrix, and spectral data of the rotary kiln flue cross section at the current moment are obtained. The measured optical path vector is used to characterize the absorption measurement value of each laser ray within the rotary kiln flue cross section at the current moment. The historical concentration matrix includes at least the concentration matrix of the previous moment and the concentration matrices of the two moments before that. The concentration matrix is ​​used to characterize the gas concentration distribution of each spatial grid within the rotary kiln flue cross section at the corresponding moment. The spectral data is used to characterize the spectral intensity values ​​of each frequency within the rotary kiln flue cross section. The concentration field is reconstructed iteratively based on the measured optical path vector at the current moment, the concentration matrix at historical moments, and the spectral data to determine the concentration matrix at the current moment. The rotary kiln flue gas composition detection process is monitored based on the concentration matrix at the current moment.

[0005] In one possible implementation, the method includes: The concentration matrix of the previous time step is determined as the concentration matrix of the zeroth iteration at the current time step, and the dust concentration factor is determined by analyzing the spectral intensity values ​​of each frequency in the spectral data. Temporal inertia analysis is performed based on the concentration matrix at historical moments to determine the inertial concentration matrix; the inertial concentration matrix is ​​used to characterize the concentration change inertia of each spatial grid over time. In each iteration, multi-source constraint analysis is performed based on the measured optical path vector, inertial concentration matrix, dust concentration factor at the current moment, and concentration matrix of the previous iteration to determine the corrected concentration change matrix for the current iteration. The corrected concentration change matrix is ​​used to characterize the amount of corrected concentration change of each spatial grid in the rotary kiln flue section in the current iteration relative to the previous iteration. The concentration matrix of the previous iteration is updated based on the modified concentration change matrix to determine the concentration matrix of the current iteration. The concentration matrix of the iteration corresponding to the preset iteration termination condition is used as the concentration matrix at the current moment.

[0006] In one possible implementation, the method includes: The inertial concentration matrix is ​​determined by performing time-series difference calculations based on the concentration matrix of the previous time step and the concentration matrices of the two time steps before.

[0007] In one possible implementation, the method includes: In each iteration, the optical path is inverted based on the measured optical path vector at the current moment and the concentration matrix of the previous iteration to determine the inferred concentration change matrix. The inferred concentration change matrix is ​​used to characterize the inferred concentration change of each spatial grid in the rotary kiln flue section at the current moment relative to the previous iteration. Spatial smoothing analysis is performed based on the concentration matrix of the previous iteration to determine the concentration smoothing matrix; the concentration smoothing matrix is ​​used to characterize the concentration diffusion equilibrium trend between each spatial grid and its adjacent grids in the current iteration relative to the previous iteration. The predicted concentration change matrix, concentration smoothing matrix, and inertial concentration matrix are weighted and calculated based on the dust concentration factor at the current moment to obtain the corrected concentration change matrix for the current iteration.

[0008] In one possible implementation, the method includes: Obtain the spatial distance matrix of the rotary kiln flue section; the spatial distance matrix is ​​used to characterize the physical length distribution of each laser ray passing through each spatial grid. The residual gradient is calculated based on the spatial distance matrix, the concentration matrix from the previous iteration, and the measured optical path vector at the current moment to determine the inferred concentration change matrix.

[0009] In one possible implementation, the method includes: Obtain the Laplacian smoothing operator corresponding to the cross-section of the rotary kiln flue; the Laplacian smoothing operator is used to characterize the spatial coupling relationship between each spatial grid and its adjacent grids; The diffusion trend is calculated based on the Laplace smoothing operator and the concentration matrix from the previous iteration to determine the concentration smoothing matrix.

[0010] In one possible implementation, the method includes: The inference weight, smoothing weight, and inertia weight are determined based on the dust concentration factor at the current moment. The predicted concentration change matrix, the concentration smoothing matrix, and the inertial concentration matrix are weighted according to the prediction weight, the smoothing weight, and the inertial weight to obtain the corrected concentration change matrix for the current iteration.

[0011] In one possible implementation, the method includes: Spatial interpolation is performed based on the concentration matrix at the current moment to reconstruct the continuous concentration distribution field; the continuous concentration distribution field is used to characterize the continuous topological distribution of gas concentration within the cross-section of the rotary kiln flue. Flow field features are extracted based on the continuous concentration distribution field to determine key flow field characteristic parameters; key flow field characteristic parameters include local concentration extreme value coordinates and cross-regional deflection index. The rotary kiln flue gas composition detection process is monitored based on key flow field characteristic parameters.

[0012] This application provides a rotary kiln flue gas composition detection device, including: a processor and a communication interface; the communication interface and the processor are coupled, and the processor is used to run computer programs or instructions to implement the method of any of the above embodiments.

[0013] This application provides a rotary kiln flue gas composition detection system, including: The data acquisition unit is used to acquire the measured optical path vector of the rotary kiln flue cross section at the current moment, the concentration matrix at historical moments, and the spectral data. The measured optical path vector is used to characterize the absorption measurement value of each laser ray within the rotary kiln flue cross section at the current moment. The concentration matrix at historical moments includes at least the concentration matrix of the previous moment and the concentration matrices of the previous two moments. The concentration matrix is ​​used to characterize the gas concentration distribution of each spatial grid within the rotary kiln flue cross section at the corresponding moment. The spectral data is used to characterize the spectral intensity values ​​of each frequency within the rotary kiln flue cross section. The concentration field reconstruction unit is used to iteratively reconstruct the concentration field based on the measured optical path vector at the current moment, the concentration matrix at historical moments, and spectral data, and to determine the concentration matrix at the current moment. The monitoring unit is used to monitor the rotary kiln flue gas composition detection process based on the concentration matrix at the current moment.

[0014] The present invention has the following beneficial effects: Based on the above technical solution, this application obtains the measured optical path vector of the rotary kiln flue cross-section at the current moment, the concentration matrix at historical moments, and spectral data. This yields the projected measurement information, temporal evolution information, and environmental interference information of the gas concentration within the flue cross-section. On this basis, the concentration matrix at the current moment is determined through iterative concentration field reconstruction, realizing the inversion of high-resolution concentration spatial distribution from underdetermined measurement data. Furthermore, the rotary kiln flue gas composition detection process is monitored based on the current concentration matrix, enabling real-time monitoring and early warning of combustion status and safe operation. This solution effectively solves the problems of insufficient spatial characterization capability and decreased detection accuracy under high dust conditions in traditional single-point monitoring methods, improving the accuracy and reliability of rotary kiln flue gas composition detection and providing strong technical support for the safe operation and process optimization control of industrial thermal processes. Attached Figure Description

[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic flowchart of a method for detecting the composition of rotary kiln flue gas according to an embodiment of the present invention. Figure 2 This is a system architecture diagram of a rotary kiln flue gas composition detection system provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of the hardware structure of a rotary kiln flue gas composition detection device provided in one embodiment of the present invention. Detailed Implementation

[0017] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a rotary kiln flue gas composition detection method, apparatus, and system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0019] The following description, in conjunction with the accompanying drawings, details the specific scheme of the rotary kiln flue gas composition detection method, apparatus, and system provided by the present invention.

[0020] Please see Figure 1 The diagram illustrates a flowchart of a method for detecting the composition of rotary kiln flue gas according to an embodiment of the present invention. The method includes the following steps: Step 101: Obtain the measured optical path vector of the rotary kiln flue section at the current moment, the concentration matrix at historical moments, and the spectral data.

[0021] The measured optical path vector is used to characterize the absorption measurement value of each laser ray within the rotary kiln flue section at the current moment. This absorption measurement value is normalized by factors such as line intensity and pressure, and transformed into the concentration-path length integral value of each laser ray at the current moment. That is, the dimensions of the measured optical path vector are consistent with the concentration-path length integral. The concentration matrix at historical moments includes at least the concentration matrix of the previous moment and the concentration matrices of the two moments before that. The concentration matrix is ​​used to characterize the gas concentration distribution of each spatial grid within the rotary kiln flue section at the corresponding moment. The spectral data is used to characterize the spectral intensity values ​​of each frequency within the rotary kiln flue section.

[0022] Because the spatial distribution of gas concentration within the cross-section of a rotary kiln flue is non-uniform, and influenced by the stratification of the gas-solid two-phase flow field, temperature gradient, and changes in combustion conditions, the gas concentration at different spatial locations will vary significantly. To obtain information on the spatial distribution of gas concentration within the flue cross-section, this application employs a measurement method based on tunable semiconductor laser absorption spectroscopy. Laser emitters and photoelectric receivers are arranged around the outer wall of the flue at the same horizontal height. The laser beam forms a cross-grid inside the flue. By measuring the absorption of each laser beam as it passes through the flue cross-section, the original integrated absorbance of each optical path is obtained. Subsequently, this original integrated absorbance is preprocessed to convert it into a concentration-path length integral value, yielding the measured optical path vector. The preprocessing includes: First, based on the real-time temperature measurement within the flue section, the corresponding temperature-corrected linear intensity parameter is retrieved from the standard spectral database. Simultaneously, based on the on-site pressure measurement, the corresponding pressure broadening coefficient is retrieved from the standard spectral database. Then, the original integrated absorbance is divided by the linear intensity parameter to eliminate the influence of temperature on the absorption linear intensity. The calculated result is then divided by the pressure broadening coefficient to eliminate the modulation effect of pressure broadening on the absorption signal, thus achieving pressure influence correction. After the above two division operations, the result is the concentration-path length integral value of each laser beam at the current moment. Since the physical path length of the laser beam passing through different spatial grids is different, the concentration-path length integral value of each laser beam reflects the integral effect of gas concentration in all spatial grids traversed by the optical path. Therefore, the measured optical path vector contains the projected information of the gas concentration distribution within the flue section.

[0023] The laser emitter and photoelectric detector receiver can be retracted inside the flange sleeve. The laser is emitted inward through the optical window, and the blowing air forms a high-pressure air curtain that is continuously blown into the flue inside the sleeve. This is used to block the high-temperature and high-dust gas from contacting the optical lens, thereby protecting the optical components from corrosion and dust adhesion, extending the service life of the sensor and ensuring the stability of the measurement data.

[0024] Furthermore, the aforementioned concentration matrix reflects the spatial distribution of gas concentration within the flue section. The rotary kiln flue section is discretized into multiple spatial grids, each corresponding to an element in the concentration matrix. The value of this element represents the gas concentration at a specific moment for that spatial grid. The historical concentration matrix records the temporal evolution of the gas concentration distribution within the flue section, providing temporal prior constraints for subsequent iterative concentration field reconstruction. The spectral data includes spectral intensity values ​​at various frequencies, used not only to obtain absorption measurements to calculate the measured optical path vector but also to analyze environmental interference factors such as dust concentration factors, providing a basis for weight allocation during the iterative concentration field reconstruction process.

[0025] For example, in the application scenario of the kiln tail flue of the metallurgical pelletizing and alumina roasting system, multiple laser emitters can be installed on the adjacent kiln walls on both sides of the flue cross section, and photoelectric detection receivers can be installed on the opposite kiln walls. The laser beam forms a vertical intersecting grid inside the flue. The data acquisition frequency can be set to 100Hz to meet the requirements of high-frequency sampling for time-series continuity analysis. Before the rotary kiln starts running, the sensors are turned on to continuously acquire data. Since there is no interference from high-concentration dust at this time, the composition of the air is stable and uniform. The integral measurement vector of the non-interference absorption line of each detection optical path can be obtained. According to the Beer-Lambert law, the integral absorption of a single optical path is divided by the known effective physical length of the optical path to obtain the single concentration value under that optical path. Then, the single concentration values ​​obtained from all array optical paths are weighted and averaged to obtain the reference concentration scalar. Finally, the reference concentration scalar is synchronously assigned to each independent spatial grid of the cross section to construct the initial concentration matrix, which serves as the initial condition for iterative concentration field reconstruction after hot ignition.

[0026] Step 102: Reconstruct the concentration field iteratively based on the measured optical path vector at the current moment, the concentration matrix at historical moments, and the spectral data to determine the concentration matrix at the current moment.

[0027] This application utilizes multiple rounds of iterative calculations to invert the gas concentration distribution of each spatial grid within the flue section at the current moment from measured optical path vectors, historical concentration matrices, and spectral data. Since the number of laser optical paths is typically less than the number of spatial grids, the inversion system constitutes an underdetermined system of equations, making it impossible to obtain a unique solution directly. Therefore, this application introduces spatial regularization constraints and temporal prior constraints, combined with dust concentration factor analysis to fuse multi-source information, gradually approximating the true concentration distribution during the iterative process.

[0028] It should be noted that, under extremely high-frequency, millisecond-level sampling periods, the diffusion and mass transfer mixing process of the target gas components within the kiln exhibits an objectively gradual evolution characteristic. The actual physical flow field at the current moment is a small, continuous shift from the flow field at the previous moment. Furthermore, within a real industrial thermal flue, the concentration field of the target gas, influenced by molecular diffusion and turbulent mixing, exhibits a gradual, smooth, and continuous transition in its spatial distribution, eliminating the possibility of abrupt changes between adjacent physical grids.

[0029] Therefore, based on the above physical laws, this application utilizes the optical path inversion information, spatial smoothing constraints, and temporal inertial constraints of the measured data simultaneously during the iterative concentration field reconstruction process. By adaptively adjusting the weights of each constraint term through the dust concentration factor, the measured data is fully trusted when the dust concentration factor is low, and the weights of spatial smoothing and temporal inertial constraints are enhanced when the dust concentration factor is high, thereby improving the accuracy of concentration field reconstruction under harsh high-dust conditions.

[0030] Step 103: Monitor the rotary kiln flue gas composition detection process based on the concentration matrix at the current moment.

[0031] The monitoring refers to the visualization, flow field feature extraction, and process parameter adjustment of the spatial distribution information of gas concentration in the flue section represented by the concentration matrix at the current moment, so as to realize real-time monitoring and early warning of the combustion status and safe operation of the rotary kiln.

[0032] The aforementioned concentration matrix represents the gas concentration values ​​of each spatial grid within the flue section in the form of a discrete grid. In order to present the spatial topological distribution of gas concentration more intuitively, this application can further perform spatial interpolation reconstruction on the discrete grid matrix to generate a high-resolution continuous concentration distribution field.

[0033] Based on this, key flow field characteristic parameters such as local concentration extreme value coordinates and cross-regional flow deviation index can be extracted. These parameters can reflect key process information such as whether combustion is uniform, whether there are local oxygen-deficient or oxygen-rich areas, and whether there is flow deviation. For example, when the local concentration of key components exceeds the safety threshold or the distribution shows significant flow deviation, a safety warning can be automatically triggered, and the combustion air-coal ratio and the opening of the corresponding process damper can be adjusted in conjunction, thereby achieving closed-loop monitoring of the rotary kiln flue gas composition.

[0034] Based on the above technical solution, this application obtains the measured optical path vector of the rotary kiln flue cross-section at the current moment, the concentration matrix at historical moments, and spectral data. This yields the projected measurement information, temporal evolution information, and environmental interference information of the gas concentration within the flue cross-section. On this basis, the concentration matrix at the current moment is determined through iterative concentration field reconstruction, realizing the inversion of high-resolution concentration spatial distribution from underdetermined measurement data. Furthermore, the rotary kiln flue gas composition detection process is monitored based on the current concentration matrix, enabling real-time monitoring and early warning of combustion status and safe operation. This solution effectively solves the problems of insufficient spatial characterization capability and decreased detection accuracy under high dust conditions in traditional single-point monitoring methods, improving the accuracy and reliability of rotary kiln flue gas composition detection and providing strong technical support for the safe operation and process optimization control of industrial thermal processes.

[0035] As a possible embodiment of this application, step 102 above can be implemented through the following steps: Step 201: Determine the concentration matrix of the previous time step as the concentration matrix of the zeroth iteration at the current time step, and perform dust concentration factor analysis based on the spectral intensity values ​​of each frequency in the spectral data to determine the dust concentration factor at the current time step.

[0036] The dust concentration factor characterizes the degree of interference between dust in the flue and laser measurement at the current moment. The higher the dust concentration factor, the more severe the scattering and absorption of laser light by dust particles during propagation, resulting in greater spectral intensity attenuation and lower reliability of the measured optical path vector.

[0037] During rotary kiln operation, the dust concentration factor is the main interfering factor affecting the accuracy of laser absorption spectroscopy measurements. In the absence of high-concentration dust interference, the intensity values ​​of each frequency in the laser spectrum remain stable, and the absorption lines are clearly distinguishable. When the dust concentration factor increases, the scattering and absorption of laser light by dust particles intensifies, leading to an overall rise in the spectral baseline, a shallower absorption line depth, and a decrease in the intensity values ​​of each frequency. Therefore, by analyzing the degree of attenuation of the spectral intensity values ​​of each frequency relative to historical normal conditions, the dust concentration factor at the current moment can be quantitatively assessed.

[0038] For example, the dust concentration factor at the current moment satisfies the following formula: in, The dust concentration factor at the current moment. The number of frequencies in the spectral data. For the first in the spectral data The average intensity value of a frequency over a historical period (e.g., the most recent hour). For the first in the spectral data The intensity value of each frequency at the current moment. The function normalizes the average intensity values ​​of each frequency in spectral data over a historical period to a range of 0 to 1, with the sum of the values ​​being 1. The function assigns higher weights to frequencies with more pronounced spectral attenuation (i.e., higher average intensity over historical periods), meaning these frequencies are more sensitive to dust scattering. This is a safety parameter used to correct for denominators of 0; its specific value can be determined based on... The value of the value determines the outcome, such as .

[0039] For each frequency, this application can calculate its average intensity value over a historical period. As a baseline, the intensity value of that frequency at the current moment is also obtained. The process involves calculating the attenuation ratio of the spectral intensity value at each frequency at the current moment relative to the historical average intensity of that frequency, and then taking a weighted average to obtain the overall residual attenuation rate at the current moment. This yields the dust concentration factor. The dust concentration factor ranges from 0 to 1; a value closer to 1 indicates more severe dust interference, while a value closer to 0 indicates less severe dust interference. The dust concentration factor refers to the overall attenuation of the measured spectral intensity at the current moment compared to the historical average intensity due to the scattering and absorption of laser light by dust particles.

[0040] It should be noted that, in the context of this application, As the average intensity of the spectrum, its value directly reflects the signal-to-noise ratio level of that frequency in historical measurements. enter The function can adaptively suppress the interference of frequencies with excessively low historical intensity (poor signal-to-noise ratio) on the calculation of dust concentration factor. Simultaneously, through... The arithmetic mean of the residual attenuation rates at each frequency is used to obtain a comprehensive average residual attenuation rate. The dust concentration factor ultimately characterizes the overall spectral intensity attenuation caused by dust. Under actual rotary kiln operating conditions, the spectral attenuation caused by changes in dust concentration is significant. Even after arithmetic averaging, the resulting dust concentration factor still possesses sufficient dynamic range to distinguish different levels of dust interference.

[0041] Step 202: Perform time-series inertial analysis based on the concentration matrix at historical moments to determine the inertial concentration matrix.

[0042] Among them, the inertial concentration matrix is ​​used to characterize the concentration change inertia of each spatial grid over time, that is, the concentration change trend of each spatial grid at the current moment inferred based on the historical concentration evolution trend.

[0043] Under extremely high frequency sampling conditions, the temporal changes in gas concentration exhibit physical inertia. That is, the concentration state at the current moment is derived from the concentration state at the previous moment, and changes continuously according to certain evolutionary rules, without any instantaneous disordered jumps. This temporal inertia originates from the objective physical laws governing the diffusion and mass transfer mixing processes of gas molecules. The macroscopic mixing and diffusion of a large-volume gas phase flow field requires a certain characteristic time, which is much longer than the time interval of high-frequency sampling.

[0044] Therefore, by analyzing the concentration matrix at historical moments and extracting the temporal trend information of concentration changes, reliable prior constraints can be provided for the concentration prediction at the current moment. Especially when the measured data is severely affected by dust interference and the reliability of optical path inversion decreases, the temporal inertial constraint can effectively maintain the continuity and stability of the concentration field reconstruction.

[0045] In some embodiments, this application can determine the inertial concentration matrix by performing time-series difference calculations based on the concentration matrix of the previous time step and the concentration matrices of the previous two time steps. Each element of the inertial concentration matrix represents the amount of concentration change inferred from the historical evolution trend of the corresponding spatial grid.

[0046] It should be noted that by calculating the change in the concentration matrix at the previous time step relative to the concentration matrices at the previous two time steps, the concentration change rate of each spatial grid within the most recent sampling period can be obtained. Since the high-frequency sampling period is much shorter than the characteristic time of macroscopic mixing and diffusion in a large-volume gas-phase flow field, this rate of change can be approximately constant within the current sampling period. Therefore, the concentration change at the previous time step can be directly used as the inertial prediction value of the concentration change at the current time step. This fully utilizes the physical continuity of gas concentration changes, avoids complex time-series model fitting, and reduces computational complexity while ensuring prediction accuracy.

[0047] For example, the inertial concentration matrix satisfies the following formula: in, The inertial concentration matrix at the current moment. This is the concentration matrix from the previous time step. This is the concentration matrix for the previous two time steps. Positive elements in the inertial concentration matrix indicate an increasing concentration trend in the corresponding spatial grid, while negative elements indicate a decreasing concentration trend. The absolute value reflects the rate of concentration change. This inertial concentration matrix remains unchanged throughout all iterations at the current time step, serving as a fixed input for the temporal inertial constraint in multi-source constraint analysis.

[0048] Step 203: In each iteration, perform multi-source constraint analysis based on the measured optical path vector, inertial concentration matrix, dust concentration factor at the current moment, and concentration matrix of the previous iteration to determine the corrected concentration change matrix for the current iteration.

[0049] The modified concentration change matrix is ​​used to characterize the modified concentration change of each spatial grid in the rotary kiln flue section in the current iteration relative to the previous iteration.

[0050] In some embodiments, multi-source constraint analysis includes constraint dimensions such as measured data constraints, spatial smoothing constraints, and temporal inertia constraints. Through adaptive weighted fusion of multi-source constraints, this application can calculate the optimal correction direction of the concentration matrix.

[0051] For example, in the iterative concentration field reconstruction process, the measured optical path vector provides direct measurement information from the sensor, the spatial smoothing constraint utilizes the continuity of the spatial distribution of gas concentration, and the temporal inertia constraint utilizes the temporal continuity of concentration changes.

[0052] The three types of constraint information mentioned above have different reliability levels under different dust concentration factors: when the dust concentration factor is low, the measured optical path vector is less affected by interference and has the highest reliability, so it should be given a larger weight. When the dust concentration factor is high, the measured optical path vector is severely affected by dust scattering and absorption, and its reliability decreases. In this case, the weights of spatial smoothing constraints and temporal inertia constraints should be increased, relying on prior knowledge of physical laws to fill the monitoring blind spots. By adaptively adjusting the weight ratio of the three constraints according to the dust concentration factor at the current moment, optimal fusion of multi-source information can be achieved, improving the robustness and accuracy of concentration field reconstruction under various operating conditions.

[0053] Step 204: Update the concentration matrix of the previous iteration based on the corrected concentration change matrix to determine the concentration matrix of the current iteration.

[0054] Concentration update refers to correcting the concentration matrix from the previous iteration using a modified concentration change matrix, making the concentration matrix approximate the true concentration distribution. The modified concentration change matrix contains information on the direction and magnitude of concentration adjustment obtained based on multi-source constraint analysis. By subtracting this correction from the concentration matrix of the previous iteration, the updated concentration matrix for the current iteration can be obtained.

[0055] In each iteration, the corrected concentration change matrix integrates the results of three constraints: measured data, spatial smoothing, and temporal inertia. It represents the optimal corrected estimate of the concentration matrix for the current iteration. By repeatedly performing multi-source constraint analysis and concentration updates, the concentration matrix moves closer to the true concentration distribution in each iteration, gradually eliminating the deviation between the initial guess and the true value. The convergence of the iterative process benefits from the synergistic effect of the multi-source constraints: optical path inversion ensures that the solution direction is consistent with the measured data, spatial smoothing constraints prevent unreasonable oscillations in the solution in space, and temporal inertia constraints maintain the temporal continuity of the solution.

[0056] For example, the concentration matrix satisfies the following formula: in, For the current moment, the first Concentration matrix of round iteration, For the current moment, the first Concentration matrix of round iteration, For the current moment, the first The corrected concentration change matrix for each iteration.

[0057] It should be noted that the above iterative update formula This is a gradient descent iteration performed under multi-source constraints. Specifically, it involves correcting the concentration change matrix. Based on the inferred concentration change matrix Concentration smoothing matrix and inertial concentration matrix The weighted composition inherently incorporates constraints based on measured data, spatial continuity, and temporal inertia. Furthermore, each concentration update ensures that the reconstruction results remain within the physically feasible region.

[0058] Step 205: Use the concentration matrix of the iteration corresponding to the preset iteration termination condition as the concentration matrix at the current moment.

[0059] The preset iteration termination condition is used to determine whether the iteration process has converged to the desired target. Conditions may include, for example, reaching a preset maximum number of iterations, or the change in the concentration matrix between two consecutive iterations being less than a preset threshold. When the preset iteration termination condition is met, it indicates that the concentration matrix of the current iteration is sufficiently close to the true concentration distribution, and the iteration can be stopped and output as the final concentration matrix for the current moment.

[0060] The setting of the iteration termination condition needs to strike a balance between computational efficiency and reconstruction accuracy. Since the concentration matrix of the zeroth iteration at the current moment uses the concentration matrix that has converged in the previous moment, and the gas concentration change under high-frequency sampling has a continuous and gradual physical characteristic, the above iterative update process usually converges after multiple iterations. Too few iterations may lead to insufficient reconstruction of the concentration field and fail to effectively eliminate the bias of the initial guess; too many iterations will increase the computational cost and may introduce the risk of overfitting. For example, this application can set the maximum number of iterations to 5. Under most operating conditions, 5 iterations are sufficient for the concentration matrix to converge to the expected accuracy level. At the same time, a convergence threshold (e.g., 0.1) can also be set. When the difference between the concentration matrices of two consecutive iterations is less than the convergence threshold, the iteration is terminated early to save computational resources.

[0061] Based on the above technical solution, this application can determine the concentration matrix of the previous time step as the concentration matrix of the zeroth iteration at the current time step and perform dust concentration analysis, establish the iteration starting point and assess the degree of environmental interference. Furthermore, it determines the inertial concentration matrix through temporal inertial analysis, thereby utilizing the temporal continuity of concentration changes to provide prior constraints. Subsequently, this application can perform multiple iterations, determine the corrected concentration change matrix through multi-source constraint analysis, gradually approximate the true concentration distribution through concentration updates, and control the convergence of the iteration process through preset iteration termination conditions. This scheme effectively improves the accuracy and robustness of concentration field reconstruction under underdetermined measurement conditions, especially under harsh high-dust conditions. Through an adaptive weight allocation mechanism, it can maintain reconstruction accuracy based on prior knowledge of physical laws when the reliability of measured data decreases, providing a reliable concentration distribution data foundation for rotary kiln flue gas composition detection.

[0062] As a possible embodiment of this application, step 203 above can be implemented through the following steps: Step 301: In each iteration, perform optical path inversion based on the measured optical path vector at the current moment and the concentration matrix of the previous iteration to determine the predicted concentration change matrix.

[0063] The inferred concentration change matrix is ​​used to characterize the inferred concentration change of each spatial grid within the rotary kiln flue section at the current moment relative to the previous iteration. Optical path inversion refers to using the residual information between the measured optical path vector and the concentration matrix from the previous iteration to calculate the correction direction and magnitude of the concentration matrix.

[0064] In laser absorption spectroscopy measurements, the absorption measurement of each laser beam is equal to the sum of the products of the gas concentration in all spatial grids traversed by the optical path and the physical length of the optical path within the corresponding grid. Therefore, the residual between the measured optical path vector and the predicted optical path vector calculated based on the concentration matrix from the previous iteration reflects the deviation between the current concentration estimate and the true value. By inverting the residual information onto each spatial grid, the concentration correction direction for each spatial grid can be obtained, i.e., the inferred concentration change matrix. This process essentially solves a minimization problem, aiming to minimize the error between the predicted optical path vector calculated based on the current concentration estimate and the measured optical path vector.

[0065] In one possible implementation, this application can obtain the spatial distance matrix of the rotary kiln flue section, and then perform residual gradient calculation based on the spatial distance matrix, the concentration matrix of the previous iteration, and the measured optical path vector at the current moment to determine the inferred concentration change matrix.

[0066] The spatial distance matrix is ​​used to characterize the physical length distribution of each laser ray as it passes through each spatial grid.

[0067] In laser absorption spectroscopy measurements, each laser beam passes through multiple spatial grids as it crosses the flue cross-section. The physical path length within each grid differs, and therefore, the contribution of each grid to the overall absorption measurement is proportional to its physical path length. The rows of the spatial distance matrix correspond to each laser beam, and the columns to each spatial grid. The values ​​of the matrix elements represent the physical path length of the corresponding laser beam within its respective spatial grid (i.e., the intersection length of the corresponding laser beam within its corresponding spatial grid). This spatial distance matrix can be calculated once the sensor layout is determined and remains unchanged throughout the monitoring process; it is an inherent geometric parameter of the system.

[0068] In some embodiments, the spatial distance matrix can be determined based on the geometry of the flue, the arrangement of the laser emitter and receiver, and the method of spatial grid division. For a rectangular flue, the laser beam forms a straight propagation path between the emitter on the adjacent two sides of the kiln wall and the receiver on the opposite side. The intersection length of each laser ray with each spatial grid can be calculated using geometric optics methods. The spatial grid division should balance spatial resolution and computational efficiency; a uniform grid division method is typically used, dividing the flue cross-section into rectangular grids of equal size.

[0069] This application can calculate the residual between the measured optical path vector and the optical path vector predicted based on the concentration matrix and spatial distance matrix of the previous iteration, and then use gradient descent to invert this residual information to each spatial grid to obtain the correction amount of the concentration matrix.

[0070] Based on the Beer-Lambert law, the absorption measurement of each laser beam can be expressed as the sum of the products of the concentration of each spatial grid and the physical length of the optical path within the corresponding grid. In other words, the predicted optical path vector equals the product of the spatial distance matrix and the concentration matrix. Therefore, the residual between the predicted and measured optical path vectors reflects the degree of deviation in the current concentration estimate. To minimize this residual, the gradient descent method can be used to calculate the gradient of the residual with respect to the concentration matrix; the direction of this gradient is the optimal correction direction for the concentration matrix.

[0071] For example, the inferred concentration change matrix satisfies the following formula: in, For the current moment, the first The inferred concentration change matrix from rounds of iterations. The normalized matrix corresponding to each spatial grid (the dimension of each element in the matrix is ​​[1 / length²]) is the reciprocal of the sum of the squares of the path lengths of all laser rays passing through the corresponding spatial grid within that spatial grid. This represents element-wise multiplication. It is a spatial distance matrix. For the current moment, the first Concentration matrix of round iteration, This is the measured optical path vector at the current moment. The operator is used to flatten a two-dimensional concentration matrix into a one-dimensional column vector by columns or rows.

[0072] Indicates the first The concentration matrix from each iteration is flattened into a one-dimensional column vector and multiplied by the spatial distance matrix to obtain the optical path prediction vector. Each element of this optical path prediction vector is the sum of the products of the concentration and path length of all spatial grids traversed by the corresponding laser ray. This represents the residual vector between the predicted and measured optical path vectors, reflecting the deviation between the current concentration estimate and the measured data. This means mapping the residual vector back to the spatial grid domain through the transpose of the spatial distance matrix, then reconstructing the one-dimensional result back to the matrix form corresponding to the original two-dimensional spatial grid, and multiplying it element-by-element with the normalized matrix to infer the dimensions of each element in the concentration change matrix to concentration, thus obtaining the concentration correction amount for each spatial grid.

[0073] The final result For the current moment, the first The inferred concentration change matrix is ​​generated in each iteration, where each element represents the concentration correction amount obtained by inverting measured data from the corresponding spatial grid. This inferred concentration change matrix guides the concentration matrix to be adjusted in the direction of reducing optical path residuals in the next iteration. Through multiple iterations, it gradually approximates the true concentration distribution, thereby minimizing the error between the optical path prediction vector calculated based on the reconstructed concentration matrix and the measured optical path vector.

[0074] Step 302: Perform spatial smoothing analysis based on the concentration matrix from the previous iteration to determine the concentration smoothing matrix.

[0075] The concentration smoothing matrix is ​​used to characterize the concentration diffusion equilibrium trend between each spatial grid and its neighboring grids in the current iteration relative to the previous iteration.

[0076] In real industrial thermal flues, the concentration field of the target gas is affected by molecular diffusion and turbulent mixing, resulting in a gradual and smooth continuous transition in its spatial distribution. There should be no drastic abrupt changes in gas concentration between adjacent spatial grids; high-concentration gases will naturally diffuse towards lower-concentration regions until a local equilibrium is reached. Based on this physical law, this application can calculate the difference in concentration between each spatial grid and the average concentration of its surrounding adjacent grids, assessing the degree of concentration deviation of that grid in space. If the concentration of a grid is significantly higher than the average concentration of its surrounding adjacent grids, that grid is more likely to experience a concentration decrease due to diffusion at the current moment; conversely, if the concentration is significantly lower than the surrounding average concentration, it is more likely to be filled up due to diffusion.

[0077] In one possible implementation, this application can obtain the Laplace smoothing operator corresponding to the cross section of the rotary kiln flue, calculate the diffusion trend based on the Laplace smoothing operator and the concentration matrix of the previous iteration, and determine the concentration smoothing matrix.

[0078] The Laplace smoothing operator characterizes the spatial coupling relationship between each spatial grid and its neighboring grids. The Laplace smoothing operator is a differential operator used to measure the second-order rate of change of a function in space. On a discrete grid, it can be represented as a weighted sum of the differences between the current grid value and the average value of neighboring grids, measuring the smoothness of the function in space. The Laplace smoothing operator is typically represented as a sparse matrix, where the non-zero elements correspond to the adjacency relationships between spatial grids, and the element values ​​reflect the coupling strength between adjacent grids. This Laplace smoothing operator can be constructed after the spatial grid partitioning scheme is determined and remains unchanged throughout the monitoring process.

[0079] For example, in a two-dimensional rectangular mesh partitioning, each internal mesh typically has four adjacent meshes (top, bottom, left, and right), while the boundary meshes have fewer adjacent meshes. The Laplace smoothing operator can be defined as follows: for each mesh, the element value in the corresponding row of the Laplace smoothing operator in the column of that mesh itself is the number of adjacent meshes, the element value in the corresponding column of the adjacent meshes is -1, and the remaining elements are 0.

[0080] This application can transform the concentration matrix of the previous iteration using the Laplace smoothing operator to obtain a concentration smoothing matrix that characterizes the concentration diffusion equilibrium trend of each spatial grid. Based on the physical laws of gas molecule diffusion, gas in high-concentration regions will diffuse to low-concentration regions until a local concentration equilibrium is reached.

[0081] In this model, each element of the concentration smoothing matrix represents the concentration correction for the corresponding spatial grid based on the spatial smoothing assumption. If the concentration of a grid is significantly higher than the average concentration of its surrounding grids, the concentration smoothing matrix at that grid has a positive and large value, indicating that the grid is more likely to experience a concentration decrease due to diffusion at the current moment. Conversely, if the concentration of a grid is significantly lower than the average concentration of its surrounding grids, the concentration smoothing matrix at that grid has a negative value and a large absolute value, indicating that the grid is more likely to be filled up due to diffusion. By incorporating the concentration smoothing matrix into multi-source constraint analysis, unreasonable oscillations and abrupt changes in the spatial distribution of the concentration field reconstruction results can be effectively suppressed, ensuring that the reconstruction results conform to the physical laws of gas diffusion.

[0082] For example, the concentration smoothing matrix satisfies the following formula: in, For the current moment, the first Concentration smoothing matrix of round iteration, This is the quadratic form of the Laplace smoothing operator. For the current moment, the first Concentration matrix of round iteration.

[0083] Indicates the first The concentration matrix of each iteration is smoothed using the Laplace smoothing operator to calculate the deviation of the average concentration of each spatial grid from that of its neighboring grids. A positive value indicates that the concentration of that grid is higher than the average concentration of its surroundings, while a negative value indicates that the concentration of that grid is lower than the average concentration of its surroundings. This means that the transpose of the Laplace smoothing operator is applied again to the above deviation results for further smoothing and mapping back to the spatial grid domain, resulting in a concentration correction based on the assumption of spatial diffusion equilibrium. This allows the concentration matrix to be adjusted towards a smoother spatial direction in the next iteration, suppressing abrupt concentration jumps between adjacent grids. This makes the reconstructed concentration distribution conform to the gradual, smooth, and continuous transition characteristics caused by gas molecule diffusion and turbulent mixing, thereby improving the spatial rationality of the concentration field reconstruction results.

[0084] Step 303: Based on the dust concentration factor at the current moment, perform weighted calculations on the predicted concentration change matrix, the concentration smoothing matrix, and the inertial concentration matrix to obtain the corrected concentration change matrix for the current iteration.

[0085] The inferred concentration change matrix, concentration smoothing matrix, and inertial concentration matrix represent the possible trends in gas concentration from three different dimensions: measured data, spatial continuity, and temporal inertia, respectively. Under conditions of low dust concentration, the measured optical path vector is less affected by interference, and the inferred concentration change matrix obtained through optical path inversion has high reliability and should be given a larger weight. Under conditions of high dust concentration, the measured data is severely affected by dust scattering and absorption, and the reliability of the inferred concentration change matrix decreases. In this case, the weights of the concentration smoothing matrix and the inertial concentration matrix should be increased, relying on prior knowledge of physical laws to predict concentration changes.

[0086] In one possible implementation, this application can determine the inference weight, smoothing weight, and inertia weight based on the dust concentration factor at the current moment. Then, the inference concentration change matrix, the concentration smoothing matrix, and the inertia concentration matrix are weighted and calculated according to the inference weight, smoothing weight, and inertia weight to obtain the corrected concentration change matrix for the current iteration.

[0087] Among them, the inference weight is used to characterize the confidence contribution of the inferred concentration change matrix in the fusion process, the smoothing weight is used to characterize the spatial diffusion contribution of the concentration smoothing matrix in the fusion process, and the inertia weight is used to characterize the temporal continuation contribution of the inertial concentration matrix in the fusion process.

[0088] For example, the inferred weights satisfy the following formula: in, To infer the weights, This represents the dust concentration factor at the current moment.

[0089] The smoothing weights satisfy the following formula: in, To smooth out the weights, The dust concentration factor at the current moment. The distribution coefficient has a value range of (0,1).

[0090] The inertia weight satisfies the following formula: in, For inertial weights, The dust concentration factor at the current moment; The distribution coefficient has a value range of (0,1).

[0091] The allocation coefficient can be set based on prior experience of dust diffusion velocity in the historical operating conditions of rotary kiln. For example, it can be 0.3, that is, the smoothing weight of the concentration smoothing matrix is ​​less than the inertial weight of the inertial concentration matrix (the high-frequency sampling period is much smaller than the characteristic time of macroscopic mixing and diffusion of large volume gas phase flow field, which ensures that the historical state matrix has a very high physical confidence. In addition, for the severe working conditions where raw material collapse often causes local spatial blockage, the heavy reliance on the spatial smoothing of adjacent grids is very likely to cause the cross-propagation of errors and the excessive smoothing of the true concentration topology).

[0092] The corrected concentration change matrix satisfies the following formula: in, For the current moment, the first The corrected concentration change matrix for each iteration. To infer the weights, To smooth out the weights, For inertial weights, For the current moment, the first The inferred concentration change matrix from rounds of iterations. For the current moment, the first Concentration smoothing matrix of round iteration, This is the inertial concentration matrix. (The above...) , and The three terms are calculated from the optical path residual gradient, the spatial smoothing operator, and the temporal difference, respectively, and are related to the concentration matrix. They have different physical dimensions or scales; however, by introducing iterative weights, , and These represent the dimensionless concentration change direction under their respective constraints. Furthermore, since subsequent iterative updates are essentially gradient descent processes with prior constraints, the weight coefficients before the update implicitly include a scalable step size factor, thus ensuring that the corrected concentration change matrix obtained after weighted combination... With concentration matrix Maintain uniformity of dimensions.

[0093] Based on the above technical solution, this application can determine the inferred concentration change matrix through optical path inversion in each iteration, determine the concentration smoothing matrix through spatial smoothing analysis, and then perform weighted calculations on the inferred concentration change matrix, concentration smoothing matrix, and inertial concentration matrix according to the dust concentration factor at the current moment, thereby realizing the adaptive fusion of multi-source constraint information. This scheme effectively solves the problem of non-uniqueness in solving underdetermined equations. Through the synergistic effect of multi-source constraints, the concentration field reconstruction results conform to both measured data and physical laws. In particular, under high dust conditions, the reconstruction accuracy is maintained through a weight adaptive adjustment mechanism, improving the robustness and accuracy of rotary kiln flue gas composition detection under various operating conditions.

[0094] As a possible embodiment of this application, step 103 above can be implemented through the following steps: Step 401: Perform spatial interpolation reconstruction based on the concentration matrix at the current moment to determine the continuous concentration distribution field.

[0095] The continuous concentration distribution field is used to characterize the continuous topological distribution of gas concentration within the cross-section of the rotary kiln flue. This application can extend the discrete concentration matrix into a continuous concentration distribution field through an interpolation algorithm, thereby more intuitively presenting the spatial topological distribution of gas concentration and providing a high-resolution data foundation for subsequent flow field feature extraction.

[0096] The concentration matrix stores the gas concentration values ​​of each spatial grid in the form of a discrete grid. However, this discrete representation is difficult to intuitively reflect the continuous variation characteristics of the concentration distribution and is not conducive to extracting accurate flow field characteristic parameters. Through spatial interpolation reconstruction, intermediate values ​​can be inserted between discrete grids to generate a high-resolution continuous concentration distribution field, making the contour lines, gradient directions, and other features of the concentration distribution clearer and more discernible.

[0097] For example, spatial interpolation reconstruction can employ various interpolation algorithms, such as bilinear interpolation and bicubic interpolation, and the specific choice can be made by comprehensively considering computational complexity and reconstruction accuracy.

[0098] In some embodiments, a high-order spatial interpolation algorithm can be used to reconstruct the continuous concentration distribution field. High-order interpolation algorithms can better maintain the smoothness and continuity of the concentration distribution and reduce spurious oscillations introduced during the interpolation process. By reconstructing the discrete grid matrix into a high-resolution continuous concentration thermogram, the topological distribution of gaseous substances in the cross-section can be intuitively mapped, facilitating operators to quickly identify areas of abnormal concentration and uneven distribution.

[0099] Step 402: Extract flow field features based on the continuous concentration distribution field and determine key flow field feature parameters.

[0100] Key flow field characteristic parameters include local concentration extreme value coordinates and cross-regional flow deviation index. Local concentration extreme value coordinates are used to identify the spatial locations of the highest and lowest gas concentrations within the flue section, while the cross-regional flow deviation index is used to quantify the degree of non-uniformity in gas concentration distribution and the intensity of airflow deviation.

[0101] Local concentration extreme value coordinates can reveal hot and cold spots in the combustion process. High concentration areas may correspond to areas of complete combustion, while low concentration areas may correspond to areas with insufficient oxygen supply or incomplete combustion. By monitoring the position and value changes of local concentration extreme value coordinates, combustion anomalies and process deviations can be detected in a timely manner. The cross-regional flow deviation index reflects the uniformity of gas concentration distribution within the flue cross-section. A larger flow deviation index indicates a more uneven concentration distribution, and the airflow may exhibit flow deviation, which can affect combustion efficiency and the stability of the thermal process.

[0102] In some embodiments, the coordinates of local concentration extrema can be determined by searching for the locations of local maximum and minimum values ​​in a continuous concentration distribution field. The cross-regional flow deviation index can be quantified by calculating the difference between the average concentrations of different regions. For example, the flue section can be divided into multiple sub-regions, the average concentration of each sub-region can be calculated, and then the standard deviation or maximum difference of the average concentration of the sub-regions can be calculated as a measure of the flow deviation index. The extraction of key flow field characteristic parameters provides a quantitative basis for subsequent safety warnings and process control.

[0103] Step 403: Monitor the corresponding operating conditions of the rotary kiln flue gas composition detection process based on key flow field characteristic parameters.

[0104] This application can monitor the operating status of a rotary kiln in real time based on key flow field characteristic parameters. When abnormal operating conditions are detected, corresponding safety warnings or automatic control actions are triggered to ensure the safe operation of the rotary kiln and process optimization.

[0105] The operational safety of a rotary kiln is closely related to the distribution of flue gas components. When the local oxygen concentration is too low, incomplete combustion may occur, producing harmful gases such as carbon monoxide, and even posing an explosion risk. Conversely, when the local oxygen concentration is too high, overburning may occur, damaging refractory materials. Significant flow deviation in concentration distribution indicates an unreasonable airflow organization, which may lead to localized overheating or decreased thermal efficiency. Therefore, by setting safety thresholds for key flow field characteristic parameters, automatic identification and early warning of abnormal operating conditions can be achieved.

[0106] In some embodiments, this application can incorporate key flow field characteristic parameters as feedforward parameters into the rotary kiln distributed control system. When the local concentration of key components exceeds the safety threshold or the distribution shows significant deviation, an explosion-proof and anti-scaling safety warning is automatically triggered, and the combustion air-coal ratio and the opening of the corresponding process damper are adjusted accordingly, thereby achieving closed-loop monitoring of the rotary kiln flue gas composition. By linking the concentration field reconstruction results with the process control system, an automated closed loop from detection to control can be achieved, improving the safety and economy of rotary kiln operation.

[0107] Based on the above technical solution, this application transforms discrete grid data into an intuitive continuous concentration distribution field by spatial interpolation reconstruction based on the concentration matrix at the current moment. Flow field features are extracted based on this continuous concentration distribution field, determining key flow field characteristic parameters such as local concentration extremum coordinates and cross-regional flow deviation index. Finally, the corresponding operating conditions of the rotary kiln flue gas composition detection process are monitored based on these key flow field characteristic parameters, achieving real-time assessment and early warning of combustion status and safe operation. This solution effectively enhances the practical value of rotary kiln flue gas composition detection, transforming high-resolution concentration distribution information into operable process control decisions, and providing a complete technical solution for the safe operation and optimized control of industrial thermal processes.

[0108] Please see Figure 2 The diagram illustrates a system architecture of a rotary kiln flue gas composition detection system according to an embodiment of the present invention. The rotary kiln flue gas composition detection system 20 includes: The data acquisition unit 21 is used to acquire the measured optical path vector of the rotary kiln flue cross section at the current moment, the concentration matrix at historical moments, and the spectral data. The measured optical path vector is used to characterize the absorption measurement value of each laser ray in the rotary kiln flue cross section at the current moment. The concentration matrix at historical moments includes at least the concentration matrix of the previous moment and the concentration matrices of the previous two moments. The concentration matrix is ​​used to characterize the gas concentration distribution of each spatial grid in the rotary kiln flue cross section at the corresponding moment. The spectral data is used to characterize the spectral intensity value of each frequency in the rotary kiln flue cross section. The concentration field reconstruction unit 22 is used to perform iterative concentration field reconstruction based on the measured optical path vector at the current moment, the concentration matrix at historical moments, and spectral data, and to determine the concentration matrix at the current moment. The monitoring unit 23 is used to monitor the rotary kiln flue gas composition detection process based on the concentration matrix at the current moment.

[0109] It should be noted that the various embodiments of this application can be referenced or learned from each other. For example, the same or similar steps, method embodiments, system embodiments and device embodiments can be referenced from each other without limitation.

[0110] This application also provides a hardware structure diagram of a rotary kiln flue gas composition detection device (denoted as rotary kiln flue gas composition detection device 30), see [link to diagram]. Figure 3 The rotary kiln flue gas composition detection device 30 includes a processor 31, and optionally, a memory 32 connected to the processor 31.

[0111] In the first possible implementation, see Figure 3 The rotary kiln flue gas composition detection device 30 also includes a communication interface 33. The processor 31, memory 32, and communication interface 33 are connected via a bus. The communication interface 33 is used to communicate with other devices or communication networks. Optionally, the communication interface 33 may include a transmitter and a receiver. The device in the communication interface 33 used to implement the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of this application. The device in the communication interface 33 used to implement the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of this application.

[0112] Based on the first possible implementation method Figure 3 The schematic diagram shown can be used to illustrate the structure of the rotary kiln flue gas composition detection device involved in the above embodiments.

[0113] in, Figure 3 The diagram can also illustrate the system chip in the rotary kiln flue gas composition detection device. In this case, the actions performed by the aforementioned rotary kiln flue gas composition detection device can be implemented by this system chip; the specific actions performed can be found above and will not be repeated here.

[0114] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0115] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0116] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for detecting the composition of rotary kiln flue gas, characterized in that, include: The measured optical path vector, historical concentration matrix, and spectral data of the rotary kiln flue cross-section at the current moment are obtained. The measured optical path vector is used to characterize the absorption measurement value of each laser ray within the rotary kiln flue cross-section at the current moment. The historical concentration matrix includes at least the concentration matrix of the previous moment and the concentration matrices of the previous two moments. The concentration matrix is ​​used to characterize the gas concentration distribution of each spatial grid within the rotary kiln flue cross-section at the corresponding moment. The spectral data is used to characterize the spectral intensity value of each frequency within the rotary kiln flue cross-section. Based on the measured optical path vector at the current moment, the concentration matrix at historical moments, and the spectral data, the concentration field is reconstructed iteratively to determine the concentration matrix at the current moment. The rotary kiln flue gas composition detection process is monitored based on the concentration matrix at the current moment.

2. The method for detecting the composition of rotary kiln flue gas according to claim 1, characterized in that, Based on the measured optical path vector at the current moment, the concentration matrix at historical moments, and the spectral data, an iterative concentration field reconstruction is performed to determine the concentration matrix at the current moment, including: The concentration matrix of the previous time step is determined as the concentration matrix of the zeroth iteration at the current time step, and the dust concentration factor is determined by performing dust concentration factor analysis based on the spectral intensity values ​​of each frequency in the spectral data. A temporal inertial analysis is performed based on the concentration matrix at the historical time point to determine the inertial concentration matrix; the inertial concentration matrix is ​​used to characterize the temporal concentration change inertia of each spatial grid. In each iteration, multi-source constraint analysis is performed based on the measured optical path vector, inertial concentration matrix, dust concentration factor at the current moment, and concentration matrix of the previous iteration to determine the corrected concentration change matrix for the current iteration. The corrected concentration change matrix is ​​used to characterize the amount of corrected concentration change of each spatial grid in the rotary kiln flue section in the current iteration relative to the previous iteration. The concentration matrix of the previous iteration is updated based on the modified concentration change matrix to determine the concentration matrix of the current iteration. The concentration matrix of the iteration corresponding to the preset iteration termination condition is taken as the concentration matrix at the current moment.

3. The method for detecting the composition of rotary kiln flue gas according to claim 2, characterized in that, Based on the concentration matrix at the historical time points, a time-series inertial analysis is performed to determine the inertial concentration matrix, including: The inertial concentration matrix is ​​determined by performing time-series difference calculation based on the concentration matrix of the previous time step and the concentration matrices of the two previous time steps.

4. The method for detecting the composition of rotary kiln flue gas according to claim 2, characterized in that, In each iteration, multi-source constraint analysis is performed based on the measured optical path vector, inertial concentration matrix, dust concentration factor at the current moment, and concentration matrix from the previous iteration to determine the corrected concentration change matrix for the current iteration, including: In each iteration, the optical path is inverted based on the measured optical path vector at the current moment and the concentration matrix of the previous iteration to determine the inferred concentration change matrix; the inferred concentration change matrix is ​​used to characterize the inferred concentration change of each spatial grid in the rotary kiln flue section at the current moment relative to the previous iteration. Spatial smoothing analysis is performed based on the concentration matrix from the previous iteration to determine the concentration smoothing matrix; the concentration smoothing matrix is ​​used to characterize the concentration diffusion equilibrium trend between each spatial grid and its adjacent grids in the current iteration relative to the previous iteration. The predicted concentration change matrix, the concentration smoothing matrix, and the inertial concentration matrix are weighted and calculated based on the dust concentration factor at the current moment to obtain the corrected concentration change matrix for the current iteration.

5. The method for detecting the composition of rotary kiln flue gas according to claim 4, characterized in that, In each iteration, optical path inversion is performed based on the measured optical path vector at the current moment and the concentration matrix from the previous iteration to determine the inferred concentration change matrix, including: Obtain the spatial distance matrix of the rotary kiln flue section; the spatial distance matrix is ​​used to characterize the physical length distribution of each laser ray passing through each spatial grid. The residual gradient is calculated based on the spatial distance matrix, the concentration matrix of the previous iteration, and the measured optical path vector at the current moment to determine the inferred concentration change matrix.

6. The method for detecting the composition of rotary kiln flue gas according to claim 4, characterized in that, Spatial smoothing analysis is performed based on the concentration matrix from the previous iteration to determine the concentration smoothing matrix, including: Obtain the Laplace smoothing operator corresponding to the cross-section of the rotary kiln flue; the Laplace smoothing operator is used to characterize the spatial coupling relationship between each spatial grid and its adjacent grids; The diffusion trend is calculated based on the Laplace smoothing operator and the concentration matrix of the previous iteration to determine the concentration smoothing matrix.

7. The method for detecting the composition of rotary kiln flue gas according to claim 4, characterized in that, The predicted concentration change matrix, the concentration smoothing matrix, and the inertial concentration matrix are weighted and calculated based on the dust concentration factor at the current moment to obtain the corrected concentration change matrix for the current iteration, including: The inference weight, smoothing weight, and inertia weight are determined based on the dust concentration factor at the current moment. The predicted concentration change matrix, the concentration smoothing matrix, and the inertia concentration matrix are weighted and calculated based on the predicted weight, the smoothing weight, and the inertia weight to obtain the corrected concentration change matrix for the current iteration.

8. The method for detecting the composition of rotary kiln flue gas according to claim 1, characterized in that, The rotary kiln flue gas composition detection process is monitored based on the concentration matrix at the current moment, including: Spatial interpolation is performed based on the concentration matrix at the current moment to reconstruct a continuous concentration distribution field; the continuous concentration distribution field is used to characterize the continuous topological distribution of gas concentration within the cross-section of the rotary kiln flue. Based on the continuous concentration distribution field, flow field features are extracted to determine key flow field feature parameters; the key flow field feature parameters include local concentration extreme value coordinates and cross-regional deflection index. The corresponding operating conditions of the rotary kiln flue gas composition detection process are monitored based on the key flow field characteristic parameters.

9. A rotary kiln flue gas composition detection device, characterized in that, include: A processor and a communication interface; the communication interface is coupled to the processor, the processor being used to run computer programs or instructions to implement the rotary kiln flue gas composition detection method as described in any one of claims 1-8.

10. A rotary kiln flue gas composition detection system, characterized in that, include: The data acquisition unit is used to acquire the measured optical path vector, historical concentration matrix, and spectral data of the rotary kiln flue cross section at the current moment; the measured optical path vector is used to characterize the absorption measurement value of each laser ray within the rotary kiln flue cross section at the current moment; the historical concentration matrix includes at least the concentration matrix of the previous moment and the concentration matrices of the previous two moments; the concentration matrix is ​​used to characterize the gas concentration distribution of each spatial grid within the rotary kiln flue cross section at the corresponding moment; the spectral data is used to characterize the spectral intensity value of each frequency within the rotary kiln flue cross section. The concentration field reconstruction unit is used to perform iterative concentration field reconstruction based on the measured optical path vector at the current moment, the concentration matrix at historical moments, and spectral data to determine the concentration matrix at the current moment. The monitoring unit is used to monitor the rotary kiln flue gas composition detection process based on the concentration matrix at the current moment.

Images