A method for imaging concentration of combustion temperature components by fusing absorption spectroscopy and induced fluorescence

Abstract

The application provides a combustion temperature component concentration imaging method combining absorption spectrum and induced fluorescence, and the specific steps comprise the following steps: fan-shaped beam laser emitted by a high-speed tunable laser passes through a combustion field and is received by a detection array; sheet laser emitted by a low-speed dye laser covers the same area; a plurality of light path absorption spectrums and high-resolution fluorescence images are synchronously acquired; a mapping from a fluorescence intensity histogram to a temperature concentration path histogram is established; a histogram matrix containing real light paths and virtual light paths is constructed; a temperature distribution and a fluorescence molecule / free radical group distribution of a to-be-measured area are calculated; the fluorescence image is corrected and the temperature distribution is updated; and the temperature and component concentration distributions of the combustion field are solved; the high spatial resolution information of laser-induced fluorescence is used to compensate for the sparseness of single-angle laser absorption spectrum measurement; the number of constraint equations is greatly increased by using virtual light paths; the ill-posed problem of combustion field tomography under few angles is effectively solved; and the image spatial resolution and accuracy are significantly improved.

Description

Technical Field

[0001] This invention proposes a combustion temperature component concentration imaging method that integrates absorption spectroscopy and induced fluorescence, belonging to the field of combustion field parameter imaging, and involving two measurement techniques: laser absorption spectroscopy and planar laser-induced fluorescence. Background Technology

[0003] The temperature and component concentration distribution of the combustion field are key parameters reflecting combustion efficiency, pollutant emissions, and chemical reaction mechanisms. Temperature reflects the distribution and trend of thermal energy in the combustion field, while concentration reveals the distribution of particles involved in combustion. Accurate measurement of these parameters is of great significance for aero-engine combustor design, industrial furnace optimization, and fundamental combustion science research. Therefore, measurement techniques for these two parameters have attracted much attention. Current measurement methods can be divided into contact measurement methods and non-contact measurement methods.

[0004] Contact measurement methods (such as thermocouples and gas sampling analysis) have significant advantages, including mature technology, simple system structure, low cost, and intuitive data reading, and are widely used in traditional combustion diagnostics and industrial applications. However, as combustion research moves towards extreme conditions, these methods inevitably reveal limitations. As invasive sensors, thermocouples inevitably interfere with the original state of the flow and temperature fields. Even in mature industrial applications, thermocouples still face severe challenges in extreme high-temperature environments. The paper "A high-precision temperature measurement system based on noise cancellation," published by Zhu et al. in IEEE Transactions on Instrumentation and Measurement, Volume 72, Article No. 9508808 in 2023, points out that traditional contact temperature measurement methods require direct contact with the object being measured, thus interfering with the target's temperature field and affecting the accuracy of the temperature measurement. Traditional thermocouples are usually made of specific materials, and their operating temperature is limited by the melting point of the materials used, making it difficult to measure at higher temperatures. In ultra-high temperature regions, the thermoelectric performance of thin-film thermocouples is unstable, resulting in signal drift and inaccurate measurement results, even leading to equipment failure. While gas sampling analysis can provide detailed component information for component concentration measurement, its "sampling-analysis" process suffers from significant lag and error. Hanraths et al., in their 2021 paper "Unsteady Effects on NOx Measurements in Pulse Detonation Combustion" published in *Flow Turbulence and Combustion*, Volume 107, Issue 3, pp. 781-809, pointed out that the geometry of the sampling probe and the extraction process systematically affect the measurement results, causing the measured NOx concentration to vary significantly with sampling parameters, making it difficult to accurately reflect the instantaneous component distribution during unsteady combustion.

[0005] In contrast, non-contact measurement methods based on optical technology do not come into contact with the measured medium and do not disrupt the measured flow field. Representative techniques include coherent anti-stokes Raman scattering spectrometry (CARS), planar laser induced fluorescence (PLIF), and tunable diode laser absorption spectroscopy (TDLAS).

[0006] CARS (Coherent Anti-Stokes Raman Spectroscopy) technology utilizes three laser beams—pump, Stokes, and probe—to generate a nonlinear four-wave mixing effect in a medium, producing a strong coherent anti-Stokes signal. Temperature and concentration are then retrieved by analyzing the spectral characteristics of this signal. While CARS is unaffected by background fluorescence and boasts extremely high measurement accuracy, it demands stringent phase matching, has an extremely complex optical path structure, and requires expensive equipment. This typically limits it to point or single-line measurements, making it difficult to extend to high spatial resolution two-dimensional imaging. Roy et al.'s 2021 paper, "Recent advances in coherent anti-Stokes Raman spectroscopy: Fundamental developments and applications in reacting flows," published in *Progress in Energy and Combustion Science*, Volume 36, Issue 2, points out that the complex optical path and computational processes of CARS, coupled with expensive equipment, make it unsuitable for high-resolution image reconstruction.

[0007] Planar laser-induced fluorescence (PLIF) technology can obtain two-dimensional high-resolution distribution images of specific components (such as OH and CH radicals) in a combustion field, clearly revealing the flame topology and reaction zone location. However, PLIF technology faces significant challenges in quantitative measurement. On the one hand, the fluorescence signal intensity is nonlinearly affected by temperature, pressure, and collisional quenching effects, making it difficult to directly convert into accurate concentration values. The paper "Improved Method for Quantitative Measurement of OH Radicals Based on Absorption Spectroscopy" published by Yang et al. in *Molecules*, Volume 31, Issue 1, 2026 (Article No. 118) points out that OH-PLIF quantitative measurement is limited by both temperature sensitivity and poor applicability of calibration constants. The paper "PLIF Flame Study on the Qualitative and Quantitative Measurement of OH Species for Conventional and Alternative Jet Fuels; Experimental and Theoretical Investigations," published in *Combustion Science and Technology*, Volume 197, Issue 11, pp. 2768-2782 in 2025, by Saraee et al., points out that in heavy liquid hydrocarbon fuel flames, directly calibrating the original relative OH concentration using PLIF is very difficult and susceptible to errors. Especially under stoichiometric and fuel-rich conditions, residual hydrocarbon collision bodies in the mixture exhibit significant quenching cross-sections, causing fluorescence signals to attenuate due to strong collisional quenching effects, thus requiring higher corrections when converting absolute concentrations to quantitative values. In high-temperature combustion environments, traditional calibration models often neglect the influence of temperature changes and collisional quenching effects on fluorescence signals. On the other hand, the excitation laser is absorbed by the target molecule when it passes through the combustion field, causing the laser energy to gradually decay along the optical path, resulting in a non-uniform distortion of the fluorescence image, which is "stronger at the beginning and weaker at the end".The paper "Quantification of NO in the post-flame region of laminar-premixed ammonia / hydrogen / nitrogen-air flames using laser-induced fluorescence" published by Richter et al. in Volume 277 of *Combustion and Flame* (Article No. 114139) in 2025 points out that the NO-LIF signal is affected by both laser absorption and fluorescence reabsorption. When the laser passes through a NO-containing gas, its intensity is attenuated due to absorption, leading to signal distortion downstream of the measurement location. Simultaneously, the emitted fluorescent photons may also be reabsorbed by other NO molecules along the path. PLIF technology for temperature measurement requires simultaneous excitation by two lasers, necessitating two dye laser systems, resulting in complex optical path arrangements and extremely high equipment costs. In their 2024 paper, "Quantitatively OH-PLIF measurements in laminar diffusion flames of n-heptane at elevated pressures," published in *Fuel*, Volume 357 (Article No. 129943), Li et al. pointed out that using PLIF technology for two-dimensional quantitative measurements of component concentration is limited by both impact quenching effects and stringent optical system requirements. In saturated LIF technology, extremely high laser output pulse energy is required to reach the saturation threshold and eliminate the quenching effect, which is extremely difficult to achieve in practical experiments. Furthermore, two-color LIF technology not only has a complex optical path configuration but also imposes extremely stringent requirements on the spatial alignment of the probe beam and the signal-to-noise ratio of the LIF signal, making the application of these techniques in complex two-dimensional quantitative measurements a significant challenge.

[0008] Tunable Diode Laser Absorption Spectroscopy (TDLAS) utilizes the absorption characteristics of molecules to specific wavelengths of laser light to achieve quantitative measurements of path-averaged temperature and concentration. Combined with computed tomography (CT) algorithms, TDLAS can reconstruct the two-dimensional distribution of a combustion field. However, traditional TDLAS techniques require CT imaging to obtain the two-dimensional distribution, typically relying on multi-angle, high-density projection optical paths to meet the data requirements of the reconstruction algorithm, resulting in complex optical path arrangements and low grid resolution. Zhou et al.'s 2024 paper, "Quantification of NO in the post-flame region of laminar premixed ammonia / hydrogen / nitrogen-air flames using laser-induced fluorescence," published in IEEE Transactions on Instrumentation and Measurement, Volume 73, Article No. 4508910, pointed out that limited optical space leads to insufficient projections, resulting in low image reconstruction resolution and inherent ill-posed problems. Liang et al.'s 2025 paper, "Reconstruction of absorption tomography from limited data based on attention mechanism and U²net structure," published in Instrumentation Science & Technology, Volume 53, pp. 760-777, also pointed out that limited optical space leads to insufficient projections, resulting in low image reconstruction resolution and inherent ill-posed problems. In practical experiments such as those involving aero-engines, the limited space and optical observation windows often make it difficult to arrange multi-angle annular optical path arrays. This makes traditional absorption tomography prone to severe reconstruction artifacts, resulting in a significant decrease in spatial resolution. This paper proposes an absorption tomography reconstruction method that combines an attention mechanism and a U²-Net structure. The attention mechanism effectively extracts key features from limited LAS projection data, enabling high-precision reconstruction of high-quality two-dimensional images of the flow field's interior with only a very small number of projection angles (limited data).The paper "Unsupervised neural-implicit laser absorption tomography for quantitative imaging of unsteady flames," published by Molnar et al. in Volume 279 of *Combustion and Flame* in 2025 (Article No. 114298), proposes a novel unsupervised neural-implicit reconstruction method to overcome the spatial resolution limitations imposed by small-angle and sparse beams. This method utilizes a coordinate neural network to directly represent thermochemical state variables as continuous spatiotemporal functions, and combines this with a differentiable observation operator based on a standard spectral database. The network optimization relies entirely on a limited set of actual acquired LAS measurement data for unsupervised network optimization. This innovative strategy effectively eliminates the physical dependence on high-density projection optical paths, successfully achieving high-fidelity, quantitative spatiotemporal imaging of unsteady flames even under extremely limited optical observation conditions. The paper "Application of UVAS and TDLAS-based multi-combustion-parameter diagnosis using computerized tomography" published by Zhou et al. in Optics and Lasers in Engineering, Volume 178, No. 108255 in 2024, addresses the limitation of insufficient LAS projections by employing a modified Tikhonov regularization method based on the regularization coefficient matrix in the reconstruction algorithm. This effectively suppresses reconstruction artifacts caused by limited optical paths and insufficient multi-angle projections, and successfully achieves stable and high-precision two-dimensional imaging of multiple key combustion parameters under limited optical windows.

[0009] In combustion experiments, LAS technology is limited by narrow space and limited optical observation windows, making it difficult to arrange multi-angle ring optical path arrays. This makes tomography a serious pathological problem, resulting in severe artifacts in the reconstructed image or making it impossible to solve.

[0010] In summary, a single optical measurement technique cannot simultaneously achieve both "high-precision quantification" and "high spatial resolution." In exploring the integration of multiple technologies, combining high-precision CARS with other techniques faces engineering obstacles such as extremely complex optical paths and uncontrollable costs. Therefore, this invention chooses to combine the quantitative advantages of LAS with the high spatial resolution advantages of PLIF. This is not only because the two are complementary in their measurement principles—LAS provides quantitative physical constraints, while PLIF provides high-resolution fluorescence images—but also because these two technologies have high compatibility in their optical systems, facilitating integration within confined spaces.

[0011] To address the limitations of existing LAS and PLIF fusion techniques, which still rely on multi-angle optical paths and where some methods neglect the correction of absorption effects in PLIF, this invention proposes a combustion temperature and component concentration imaging method that integrates absorption spectroscopy and induced fluorescence. This invention utilizes only single-angle LAS data and constructs a virtual optical path through a neural network to increase tomographic constraints, effectively solving the ill-conditioned reconstruction problem under limited angles. Simultaneously, iterative correction of laser absorption loss in the PLIF image is achieved using the reconstructed high-precision temperature and concentration fields, thus realizing high-resolution, high-accuracy quantitative imaging of the combustion field with a minimally simplistic optical path configuration. Summary of the Invention

[0013] This invention proposes a method for imaging combustion temperature and component concentration by fusing absorption spectroscopy and laser-induced fluorescence, achieving the fusion of laser absorption spectroscopy and laser-induced fluorescence spectroscopy. This invention compensates for the sparsity of single-angle laser absorption spectroscopy measurements by utilizing the high spatial resolution information of laser-induced fluorescence, increases the number of constraint equations using a virtual optical path, corrects the fluorescence image based on temperature and concentration distribution, establishes an iterative process between the fluorescence image and the temperature and concentration images, and solves for the temperature and component concentration distribution results of the combustion field. The specific implementation steps are as follows:

[0014] Step 1: Based on the absorption spectrum data, calculate the temperature-concentration path histogram obtained from the laser absorption spectrum inversion, and obtain the fluorescence intensity histogram at the position that coincides with the laser absorption spectrum light.

[0015] Set the temperature reconstruction range and concentration reconstruction range, and select based on prior knowledge of combustion field characteristics. Temperature value and Each concentration value is used to construct... Discrete temperature concentration pairs For a given pressure ,temperature and concentration , wave number Absorption value per unit absorption distance It can be calculated from the HITRAN database.

[0016]

[0017] In spectral scanning At each wavenumber point, The absorbance per unit absorption distance at each temperature and concentration is used to form a unit absorption matrix. According to Beer-Lambert's law, at each wavenumber... The total uptake of the entire path is the sum of the uptakes of all path segments, expressed as:

[0018]

[0019] Each wave number The equations at each point can be established simultaneously. Therefore, the absorption spectrum along the entire path... yes The sum of the absorption spectra of each gas component can be expressed as:

[0020]

[0021] In the formula, the absorption spectrum This is the absorption spectrum of the measured flow field distribution. A finite number of discrete temperature and concentration values ​​are selected, and the unit absorption matrix is ​​calculated. Then, an iterative algorithm is used to obtain the temperature-concentration histogram from the linear model. The single-path temperature-concentration histogram model is a linear inversion model based on absorption spectra, which can effectively utilize multispectral data for gas parameter inversion.

[0022] Measurement values ​​of absorption spectra at K wavenumber points on M projected rays Establish a linear imaging model based on histograms.

[0023]

[0024] Where A is the sensitivity matrix, and Y is... A binary matrix with all elements taking values ​​of 0 or 1, used to represent the normalized distribution of the values ​​for each temperature-concentration pair. Each column represents a set of temperature-concentration pairs, and each row represents a grid cell in the region of interest. N is the number of grid cells in the sensitivity matrix. This is a unit absorption rate matrix. and It is a column vector with all elements being 1, used to add path length constraints.

[0025] The product of the sensitivity matrix A and the binary matrix Y is defined as the histogram matrix H, i.e. ,for A 3D matrix. The m-th row in H represents the path length of each temperature-concentration pair on the m-th projected ray, i.e., the temperature-concentration histogram; based on the measured absorptivity matrix... The temperature-concentration histogram can be retrieved for each projected ray:

[0026]

[0027] In the formula, This represents the m-th row of the histogram matrix H. Absorption rate matrix The m-th line, It is the absorptivity matrix In the m-th row, using reconstruction algorithms such as SART, the histogram matrix H can be obtained by sequentially inverting the M rays.

[0028] Extract the fluorescence intensity distribution that coincides with the absorption spectrum optical path from a single-frame fluorescence image, and calculate the fluorescence intensity histogram along the path. .

[0029] Step 2: Establish a nonlinear mapping model for the temperature-concentration path histogram and fluorescence intensity histogram using a hybrid model based on Transformer and fully connected neural networks (Transformer-FNN). The fluorescence intensity histogram... As input to the Transformer-FNN model, the temperature-concentration path length histogram H obtained from absorption spectral inversion is used as the output of the Transformer-FNN model. The optical path projection data is used as the training set to learn the network weight parameters in the Transformer-FNN model through training. The Transformer-FNN model specifically includes an input embedding layer, a position encoding layer, a Transformer encoder, and a fully connected neural network. Its specific mapping process is as follows: the input histogram data... After adding the sequence dimension, it is linearly mapped to the feature dimension through the input embedding layer. The feature vector is then scaled and its features are added to the positional encoding layer. The feature vector is then input into the Transformer encoder for self-attention feature extraction. The Transformer encoder consists of multiple stacked encoding layers with multi-head attention mechanisms. The terminal feature vector of the output feature sequence of the Transformer encoder is extracted and input into the fully connected neural network. The fully connected neural network contains multiple alternating linear layers, ReLU nonlinear activation functions, and Dropout layers. Finally, the terminal linear layer of the fully connected neural network outputs the predicted histogram H.

[0030] A mapping model is used to transform the data into a virtual temperature-concentration histogram. A histogram matrix containing both the real and virtual optical paths is constructed to calculate the two-dimensional temperature and concentration distribution within the test area. The absorption distance of a given temperature-concentration pair across all laser paths can be considered as the projection of the laser beam through this 0-1 distribution. Therefore, the 0-1 distribution, i.e., the 0-1 binary matrix Y, can be reconstructed using tomographic imaging.

[0031]

[0032] In the formula, column vector The first of the binary matrix Y The column corresponds to a total of The spatial distribution of temperature-concentration pairs within the region of interest, represented by a column vector. The histogram matrix H is the first... Similarly, using image reconstruction algorithms such as SART, the binary matrix Y is inverted and processed. Each temperature-concentration pair is inverted sequentially to obtain a binary matrix Y. From Y and the specific values ​​of temperature and component concentration in each temperature-concentration pair, the spatial distribution of temperature and component concentration can be recovered, enabling two-dimensional or three-dimensional imaging of temperature and component concentration distributions.

[0033]

[0034] Step 3: Correct the fluorescence image, establish an iterative process between the fluorescence image and the temperature and component concentration images, and solve for the temperature and component concentration distributions in the combustion field. Obtain the fluorescence image. Temperature images from LAS tomography Then, the concentration of fluorescent molecules can be calculated.

[0035]

[0036] In the formula, A constant that can be calibrated using a standard flame produced by a standard burner. This indicates the proportion of fluorescent molecules at low energy levels.

[0037] The wavenumber of the fluorescent molecule pair can be obtained as follows: Laser absorption rate and compensate fluorescence images

[0038]

[0039] For discrete pixel data, the absorptivity is accumulated pixel by pixel. q represents the th pixel along the PLIF laser direction. The laser intensity at this location was received from the preceding... The absorption of each column requires its own absorption rate to be included in the calculation. In the formula... Indicates the preceding number The pixel corresponds to the region, and the length of laser propagation.

[0040] get Subsequently, the fluorescence molecule concentration image can be further updated, and the new concentration distribution is recorded as follows.

[0041]

[0042] Image of concentration of new fluorescent molecular groups In the process, some of the laser attenuation errors have been compensated for. The new concentration image has not been fully transferred. Decision made. The new temperature will then be calculated based on the fluorescence model.

[0043]

[0044] First round of temperature updates completed:

[0045]

[0046] Using the temperature distribution updated in the first round as the current temperature distribution, the following process is repeated: The concentration of fluorescent molecules and laser absorption are recalculated based on the current temperature distribution; fluorescence intensity attenuation is corrected to obtain a new actual fluorescence intensity distribution; and a new fluorescent molecule concentration is calculated based on this distribution. The new temperature update distribution is then obtained from the new fluorescent molecule concentration through a neural network. The temperature is updated using the matrix Newton iteration method. First, the residual is calculated.

[0047]

[0048] In the formula, Let be the iteration number. The adaptive calculation of the residual-driven step size is used to obtain the _th_ iteration. Temperature distribution in the next iteration:

[0049]

[0050] In the formula, To fix the relaxation factor, If the value is a local minimum, the fluorescence compensation solution for this round is completed after the temperature update. The solution for the new absorbance distribution can then be obtained to compensate the fluorescence image again.

[0051] This process is repeated. To check whether the iteration should be terminated, this criterion can be used. This criterion requires that after p iterations, the relative changes in temperature and fluorescent molecule concentration, when summed pixel by pixel, are less than a threshold.

[0052]

[0053] In the formula, This represents the threshold for ending the iteration. According to simulation studies, this threshold can be set to one percent of the number of image pixels. Once this condition is met, fluorescence correction ends, and the corresponding temperature distribution and fluorescent molecule concentration distribution are obtained. Attached Figure Description

[0054] Figure 1 This is a flowchart of the present invention for reconstructing high-resolution temperature and concentration distribution.

[0055] Figure 2 This is the temperature distribution obtained by inverting the laser absorption spectrum histogram in the example, with a resolution of 300 × 300.

[0056] Figure 3 The image shown is a fluorescence image captured by a high-speed camera in the example, with a resolution of 1000 × 1000.

[0057] Figure 4 It is a high-resolution temperature distribution obtained through iteration in the instance.

[0058] Figure 5 It is the high-resolution OH molecule concentration distribution obtained through iteration in the example. Detailed Implementation

[0060] The invention will be further illustrated below with reference to examples. The laser absorption spectroscopy measurement object used in the examples is water molecules at 7185.58 cm⁻¹. -1 and 7444.43 cm -1 Two spectral lines were observed. The absorption spectroscopy single-sided sensor has a side length of 15 cm, with 12 photoelectric sensors arranged along the side. A laser emitted from a collimating lens illuminates the 12 sensors on the opposite side. A laser-induced fluorescence system measures the fluorescence spectral line of the OH molecule at 283 nm. The laser absorption spectroscopy reconstructed image has a resolution of 50 × 50, and the laser-induced fluorescence image has a resolution of 1000 × 1000. A circular combustion region with a diameter of 6 cm is located in the center of the pentagonal sensor; the fuel is methane, forming a diffuse flame in the air.

[0061] Step 1: Obtain the temperature-concentration path histogram and fluorescence intensity histogram derived from the laser absorption spectrum inversion. Calculate the water molecule concentration path at 7185.58 cm⁻¹ from the light intensity signal received by the detector. -1 and 7444.43 cm -1 The measured absorption spectra of the two spectral lines, combined with a set discrete temperature and concentration grid, were used to calculate the temperature-concentration path length histogram H under the actual measurement optical path using a histogram-based linear imaging model. Based on this low-resolution histogram data, the temperature distribution within the grid can be preliminarily obtained (results are attached). Figure 2 (As shown). Meanwhile, in a single-frame high-resolution fluorescence image (as attached) Figure 3 As shown, the two-dimensional fluorescence distribution signal at the position that completely overlaps with the above absorption spectroscopy measurement optical path is extracted, and the fluorescence intensity histogram on the corresponding path is calculated. .

[0062] Step 2: Calculate the initial two-dimensional temperature and component concentration distribution of the combustion field using a histogram mapping model. A hybrid model based on Transformer and fully connected neural networks (Transformer-FNN) is used to establish the mapping relationship. Specifically, the fluorescence intensity histogram obtained in Step 1 is used... The input embedding layer of the Transformer-FNN model, after feature scaling and positional encoding, is fed into a multi-layer Transformer encoder for self-attention feature extraction. Finally, a fully connected neural network at the ends outputs the predicted temperature-concentration path histogram H. After training the model using actual optical path data, several virtual optical paths without physical probes are selected on the fluorescence image. Their fluorescence intensity histograms are extracted and input into the trained Transformer-FNN model to predict the virtual temperature-concentration histogram. The histogram matrices of the real and virtual optical paths are combined to construct a complete sensitivity matrix, which is then solved using the Joint Algebraic Reconstruction Technique (SART) to obtain a high-resolution initial temperature image. and initial OH concentration image .

[0063] Step 3: Correct the fluorescence image and solve for the final temperature and component concentration distribution of the combustion field. (This step involves modifying the initial temperature image.) Image with OH concentration Substitute the components into the laser absorption model to establish component correlations. The concentration distribution of water molecules can then be calculated. Subsequently, absorption compensation was performed on the original fluorescence signal based on the calculated temperature and component concentration distribution. According to Beer-Lambert's law, the pixel-by-pixel laser absorptivity along the PLIF laser propagation direction was calculated, and the accumulated laser attenuation rate in the preceding region was added to the calculation to obtain a new fluorescence image. Based on the compensation and updated water molecule concentration The concentration of new fluorescent molecules was recalculated based on the fluorescence model. and new temperature distribution .

[0064] The temperature distribution obtained in the previous round and fluorescent molecule concentration Substitute these parameters back into the fluorescence image compensation formula to calculate the fluorescence image. Substitute into the fluorescence model to update the concentration of fluorescent molecules. and temperature distribution The fluorescence image correction and parameter update process described above is repeated. The iteration terminates when the relative change rate of temperature and OH molecule concentration between two consecutive iterations, after pixel-by-pixel summation, is less than a set threshold (one percent of the image pixel count). After iteration, the final high-resolution combustion field parameter distribution is output, including the high-resolution temperature distribution shown in the attached figure. Figure 4 As shown in the attached figure, the concentration distribution of OH molecules is as follows. Figure 5 As shown.

Claims

1. A method for imaging combustion temperature and component concentration by fusing absorption spectroscopy and induced fluorescence, characterized in that... The process involves obtaining the temperature-concentration path histogram of the laser absorption spectrum and the fluorescence intensity histogram at the overlapping positions, establishing a mapping from the fluorescence intensity histogram to the temperature-concentration path histogram, calculating the histogram matrix of the virtual optical path, solving for the initial two-dimensional temperature and component concentration distribution of the combustion field, compensating for the laser absorption loss of induced fluorescence and correcting the fluorescence image, and solving for the high-resolution temperature and component concentration distribution of the combustion field. Specifically, this includes the following steps: Step 1: Obtain the temperature-concentration path histogram of the laser absorption spectrum, and acquire the fluorescence intensity histogram at the location coinciding with the laser absorption spectrum. A high-speed laser absorption spectroscopy system acquires absorption spectral signals from multiple optical paths passing through the test area and covering different absorption path lengths. Simultaneously, a low-speed planar laser-induced fluorescence system acquires a high spatial resolution two-dimensional fluorescence distribution image of the test area, where the laser sheet plane for inducing fluorescence coincides with the laser absorption spectrum optical path plane. The temperature reconstruction range and concentration reconstruction range are set, and based on prior knowledge of the combustion field characteristics, the appropriate parameters are selected. Temperature value and Each concentration value is used to construct... Discrete temperature concentration pairs Based on the principle of laser absorption spectroscopy, the absorption spectral lines at K wavenumber points on M projected rays are measured. Establish a linear imaging model based on histograms. Where A is the sensitivity matrix, and Y is... A binary matrix of dimension N, where all elements are either 0 or 1, represents the normalized distribution of the values ​​for each temperature-concentration pair. Each column represents a set of temperature-concentration pairs, and each row represents a grid cell in the region of interest. N is the number of grid cells in the sensitivity matrix. This is a unit absorption rate matrix. and It is a column vector with all elements being 1, used to add path length constraints; The product of the sensitivity matrix A and the binary matrix Y is defined as the histogram matrix H, i.e. ,for In the dimensional matrix H, the Mth row represents the path length occupied by each temperature-concentration pair on the Mth projection ray, i.e., the temperature-concentration histogram. The fluorescence intensity distribution at the location coinciding with the absorption spectrum path is extracted from a single-frame fluorescence image, and the fluorescence intensity histogram along that path is calculated. ; Step 2: Calculate the two-dimensional temperature and absorbing molecule concentration distribution of the combustion field using temperature-concentration histograms and fluorescence intensity histograms; establish a fluorescence intensity histogram using a neural network. The mapping relationship model between the fluorescence intensity histogram and the temperature-concentration path length histogram H is established using a neural network based on Transformer and fully connected neural network (Transformer-FNN). Specifically, the configuration involves mapping the fluorescence intensity histogram... As input to the Transformer-FNN model, the temperature-concentration path length histogram H obtained from absorption spectral inversion is used as the output of the Transformer-FNN model. Optical path projection data is used as the training set to train the network weight parameters. Using the trained neural network mapping model, the fluorescence intensity histogram of the virtual optical path is transformed into a virtual temperature-concentration path length histogram, thus constructing a virtual absorption spectral measurement dataset covering the area to be measured. Combining the actual measurement histogram of the physical optical path with the predicted histogram of the virtual optical path, a sensitivity matrix A is constructed. The binary matrix Y is solved column by column to obtain the values ​​of each temperature-concentration pair within the grid of the area to be measured, ultimately obtaining the two-dimensional temperature distribution of the area to be measured. and component concentration distribution; Step 3: Correct the fluorescence image and solve for the temperature and component concentration distribution of the combustion field. Using the two-dimensional temperature distribution obtained from the absorbing molecules, the fluorescence image is decoupled from the fluorescence molecule concentration to obtain the fluorescence molecule concentration distribution. The absorption rate distribution of the fluorescence molecules to the induced laser is then calculated to compensate for the fluorescence image. In the compensated fluorescence image, the temperature and the concentrations of fluorescence molecules and absorbing molecules can be decoupled again. The iterative process of fluorescence image decoupling and compensation forms the image solving process, ultimately yielding a high spatial resolution and accurate quantitative temperature and component concentration image.

2. A combustion temperature component concentration imaging method that integrates absorption spectroscopy and induced fluorescence, characterized in that... The specific process for solving the temperature and component concentration distribution of the combustion field by correcting the fluorescence image in step three of claim 1 is as follows: absorption compensation is performed on the original fluorescence signal intensity along the path to eliminate the nonlinear attenuation of the fluorescence signal caused by laser absorption, thereby obtaining the corrected fluorescence intensity distribution, and then the temperature and component concentration distribution are calculated accordingly. The process is as follows: First, fluorescence image correction and parameter updates are performed based on the input temperature and fluorescent molecule concentration: According to Beer-Lambert law, the wavenumber of the fluorescent molecule pair is... laser absorption rate for In the formula, Indicates pressure, Indicates the concentration of fluorescent molecules. Indicates the temperature of the fluorescent molecule. Indicates spectral line intensity. Representing a linear function, using the initial temperature obtained in step two of claim 1. and fluorescent molecule concentration distribution Calculate the absorption rate distribution This allows for the compensation of the fluorescence image, resulting in a new fluorescence image. for In the formula, Represents the first fluorescence image along the PLIF laser direction The column, because the laser was in front To calculate the absorption of fluorescent molecules, their absorbances need to be included in the calculation. Indicates the preceding number The length of laser propagation in the region corresponding to the pixel; This yields... Then, the concentration image of fluorescent molecules can be further updated, and the new concentration distribution of fluorescent molecules is recorded as follows. ,have Obtain new fluorescence images and new concentration Then, based on the fluorescence model, an equation was constructed to solve for the temperature of the fluorescent molecules. The equation is This equation has a unique solution within the measured temperature range, allowing for the calculation of temperature distribution. The first round of temperature updates has been completed. Secondly, iterative calculations of the combustion field parameters are performed: Let the current iteration number be... The temperature distribution of fluorescent molecules after the last round of updates and concentration distribution As input, the absorption rate distribution of fluorescent molecules was recalculated. : Obtain the corrected fluorescence image : Update # Concentration of fluorescent molecules in the wheel : Based on the concentration of new fluorescent molecules and fluorescence images Substitute into the equation Solve for the intermediate temperature of this cycle. : Calculate the residual with respect to the previous temperature distribution: The adaptive calculation of the residual-driven step size is used to update the th Temperature distribution in the wheel: In the formula, To fix the relaxation factor, To find the minimum value, perform a convergence criterion for this round. The calculation terminates when the relative change rate of temperature and fluorescent molecule concentration between two consecutive iterations, after summing pixel by pixel, is less than a set threshold. The convergence criterion is: In the formula, This represents the threshold for ending the iteration. The threshold is set to one percent of the number of image pixels. Once this condition is met, fluorescence compensation ends; otherwise, it proceeds to the next iteration. Round iteration, the temperature distribution obtained in this round and fluorescent molecule concentration distribution As input, recalculate the first... Distribution of fluorescence molecular absorption rates : And update the corrected fluorescence image. : Update fluorescent molecule concentration : Solve for the intermediate temperature of this cycle. : Recalculate the residuals: Update temperature distribution : And perform convergence judgment; and so on, we can obtain and , and , and Repeat the above parameter update process until the calculated result is obtained. and And satisfy the set convergence conditions or reach the preset maximum number of iterations; Finally, the temperature distribution and fluorescent molecule concentration distribution are obtained. The absorption molecule concentration distribution is obtained by mapping the temperature and concentration, and images of the distribution of parameters such as temperature, component concentration and fluorescent molecule concentration are output.

Images