A method for detecting the stability of a torch flame

Abstract

The present application provides a method for detecting the stability of a torch flame, including: determining the relationship function between the hydrogen-ammonia mixing ratio and the flame length, color, and temperature according to the flame stability parameter dataset, training a flame stability prediction model through a support vector machine algorithm, and judging the stability state of the flame under different ammonia mixing ratios; performing kinetic analysis on the output results of the flame stability prediction model, simulating the combustion process of hydrogen and ammonia, obtaining the variation curve of the concentration of key intermediate products with time, and judging the propagation characteristics of the reaction chain under different ammonia mixing ratios; aiming at the occurrence conditions of the flame stability key points, simulating the flow field distribution under different burner structures, obtaining the fuel injection velocity field, measuring the temperature field and species concentration field in the combustion area, and judging the influence of the burner structure on the flame stability.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and particularly to a method for detecting the stability of a torch flame. Background Art

[0002] In a hydrogen-ammonia combustion torch system, there is a technical contradiction in the influence of the ammonia mixing ratio during the storage tank supply process on the stability of the hydrogen flame. As the ammonia mixing ratio increases, the flame length, color, and temperature will change, but this change is not a linear relationship. At low ammonia mixing ratios, the flame stability improves with the increase in the ratio, and the combustion efficiency also increases. However, when the ammonia mixing ratio exceeds a certain critical value, the flame stability begins to decline, and unstable combustion occurs. This contradiction stems from the complexity of the combustion kinetics of the hydrogen-ammonia mixture gas. The presence of ammonia changes the concentration distribution and diffusion rate of the reactants, affecting the progress of the free radical chain reaction. At the same time, the endothermic effect of ammonia decomposition competes with the exothermic effect of hydrogen combustion, resulting in uneven flame temperature distribution. In addition, the presence of ammonia also affects the flame propagation speed and combustion limit, making the flame structure exhibit different characteristics at different mixing ratios. Therefore, it is difficult to obtain ideal flame stability by simply adjusting the ammonia mixing ratio in practical applications. It is necessary to comprehensively consider multiple factors such as reaction kinetics, heat and mass transfer processes, and flame structure, and establish a more accurate mathematical model to describe this complex system. This requires the collection and analysis of a large amount of experimental data under different working conditions to find the optimal operating parameter range. Summary of the Invention

[0003] The present invention provides a method for detecting the stability of a torch flame, mainly including:

[0004] Obtain mixed fuel samples with different ratios from the storage tank according to a preset hydrogen-ammonia ratio range, adjust the mixing ratio of hydrogen and ammonia using a mass flow controller, and monitor the concentrations of hydrogen and ammonia in the mixed fuel in real time through a gas analyzer to determine whether the components of the mixed fuel meet the predetermined ratio requirements;

[0005] Conduct a flame stability test on the obtained mixed fuel samples, capture the flame morphology using a high-speed imaging system, analyze the change in flame length, measure the flame color characteristics, and determine the flame temperature distribution to obtain a dataset of flame stability parameters at different ammonia mixing ratios;

[0006] According to the dataset of flame stability parameters, determine the relationship function between the hydrogen-ammonia ratio and the flame length, color, and temperature, train a flame stability prediction model through a support vector machine algorithm, and judge the stability state of the flame at different ammonia mixing ratios;

[0007] Perform kinetic analysis on the output results of the flame stability prediction model, simulate the combustion processes of hydrogen and ammonia, obtain the curves of the concentrations of key intermediate products changing with time, and judge the propagation characteristics of the reaction chain at different ammonia doping ratios;

[0008] According to the change curves and propagation characteristics, establish a discriminant criterion for the key points of flame stability, set a threshold for flame stability, and judge whether the flame is in a stable state at the current ammonia doping ratio by comparing the deviation between the actual flame parameters and the threshold, and determine the occurrence conditions of the key points of flame stability;

[0009] For the occurrence conditions of the key points of flame stability, simulate the flow field distribution under different burner structures, obtain the fuel injection velocity field, measure the temperature field and species concentration field in the combustion area, and judge the influence of the burner structure on flame stability;

[0010] According to the analysis results of the influence of the burner structure on flame stability, use the genetic algorithm to optimize the geometric parameters of the burner. By adjusting the fuel injection angle and the shape of the combustion chamber, use the numerical simulation method to calculate the flame morphology evolution process under the optimized burner structure, and judge whether the optimized burner can effectively suppress the occurrence of the key points of flame stability, and obtain the optimal ammonia doping ratio.

[0011] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0012] The present invention discloses a method for detecting the stability of a torch flame. First, the present invention precisely adjusts the mixing ratio of hydrogen and ammonia through a mass flow controller, and uses a gas analyzer to real-time monitor the component ratio of the mixed fuel. In this way, it realizes the precise control and acquisition of mixed fuel samples with different hydrogen ammonia doping ratios, and solves the problem of accurate regulation of the mixed fuel ratio;

[0013] Then, through the flame stability test on the obtained mixed fuel samples, comprehensive flame stability parameter data are obtained, providing a basis for subsequent model training; Secondly, through model analysis, the relationship between the hydrogen ammonia doping ratio and flame stability is determined, and a relationship model between the hydrogen ammonia doping ratio and flame characteristics is established, realizing the prediction of flame stability; Finally, through the analysis of the influence on flame stability, the discriminant criterion and threshold of the key points of flame stability are determined, which can accurately judge the stable state of the flame, and provide a direction for the optimized design of the burner, obtain the optimal ammonia doping ratio and burner structure, and improve the flame stability. Generally speaking, the present invention effectively improves the stability of the flame at different hydrogen ammonia doping ratios, providing a practical scientific basis and technical support for fuel mixing and burner design. Brief Description of the Drawings

[0014] Figure 1It is a flowchart of a method for detecting the stability of a torch flame according to the present invention.

[0015] Figure 2 It is a schematic diagram of a method for detecting the stability of a torch flame according to the present invention.

[0016] Figure 3 It is another schematic diagram of a method for detecting the stability of a torch flame according to the present invention. Detailed implementation manners

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and detailedly described with reference to the accompanying drawings in the embodiments of the present invention. The described embodiments are only a part of the embodiments of the present invention.

[0018] As Figures 1-3 , a method for detecting the stability of a torch flame in this embodiment may specifically include:

[0019] S101. Obtain mixed fuel samples with different ratios from the storage tank according to a preset hydrogen-ammonia mixing ratio range, adjust the mixing ratio of hydrogen and ammonia by using a mass flow controller, and monitor the concentrations of hydrogen and ammonia in the mixed fuel in real time through a gas analyzer to determine whether the components of the mixed fuel meet the predetermined ratio requirements.

[0020] Obtain a mixed fuel sample within a preset hydrogen-ammonia mixing ratio range, where the mixed fuel sample is supplied by a storage tank; according to the preset hydrogen-ammonia mixing ratio range, adjust the mixing ratio of hydrogen and ammonia in the mixed fuel sample by using a mass flow controller; use a gas analyzer to monitor the mixed fuel sample in real time to obtain concentration data of hydrogen and ammonia; calculate the deviation value between the actual components of the mixed fuel sample and the preset ratio requirements by using the direct comparison method. If the deviation value exceeds the preset threshold, trigger the automatic correction program of the mass flow controller; perform concentration fluctuation analysis based on the deviation value to quantitatively evaluate the concentration stability of the mixed fuel sample; use the Kalman filter algorithm to smooth the concentration data to obtain the smoothed concentration data; combine the concentration fluctuation analysis results and the smoothed concentration data to determine whether the components of the mixed fuel sample meet the predetermined ratio requirements; if the measurement results for a continuous preset number of times are within the allowable deviation range, determine that the components of the mixed fuel sample meet the predetermined ratio requirements.

[0021] Exemplarily, a mixed fuel sample with a preset hydrogen-ammonia blending ratio range of 30%-70% is obtained from a storage tank. The mixing ratio of hydrogen and ammonia is precisely adjusted through a mass flow controller. According to the preset fuel sample volume parameter, the adjustment accuracy of the mass flow controller is controlled to 0.1%. A gas analyzer is used to monitor the mixed fuel sample in real time to obtain the concentration data of hydrogen and ammonia. The monitored data is processed by the standard gas calibration method to obtain the calibrated concentration data of the hydrogen-ammonia mixed fuel. For the calibrated concentration data, the direct comparison method is used to calculate the deviation value between the actual mixed fuel components and the preset ratio requirements. The allowable deviation range is set to ±1%. Based on the calculated deviation value, a concentration fluctuation analysis is carried out to quantitatively evaluate the concentration stability of the mixed fuel sample. If the deviation value exceeds the preset threshold, the automatic correction program of the mass flow controller is triggered to re-adjust the mixing ratio to meet the predetermined ratio requirements. The Kalman filter algorithm is used to smooth the continuously monitored concentration data to reduce the influence of random fluctuations on the judgment result. Combining the concentration fluctuation analysis results and the data after smoothing, it is judged whether the components of the mixed fuel meet the predetermined ratio requirements. If the results of 10 consecutive measurements are within the allowable deviation range, it is determined that the components of the mixed fuel meet the predetermined ratio requirements. When obtaining the mixed fuel sample with a preset hydrogen-ammonia blending ratio range of 30%-70% from the storage tank, the mass flow controller adopts the principle of a thermal mass flowmeter to precisely adjust the mixing ratio of hydrogen and ammonia. According to the preset fuel sample volume parameter of 500 mL, the adjustment accuracy of the controller is set to 0.1%, that is, the hydrogen flow rate is 150-350 mL / min and the ammonia flow rate is 150-350 mL / min. The gas analyzer uses the infrared absorption spectroscopy method to monitor the mixed fuel sample in real time at a sampling frequency of 1 Hz to obtain the concentration data of hydrogen and ammonia. Through the standard gas calibration method, calibration is carried out using hydrogen and ammonia standard gases with a purity of 99.999% to obtain the calibrated concentration data of the hydrogen-ammonia mixed fuel. The direct comparison method is used to calculate the deviation value between the actual mixed fuel components and the preset ratio requirements. The allowable deviation range is set to ±1%. For example, when the preset hydrogen ratio is 60%, the actual measurement is 62%, and the deviation value is 2%, exceeding the allowable range. A concentration fluctuation analysis is carried out, and the standard deviation method is used to calculate the standard deviation of 100 consecutive measurement data to evaluate the concentration stability of the mixed fuel sample. If the deviation value exceeds the preset threshold, the PID automatic correction program of the mass flow controller is triggered to adjust the P, I, and D parameters and re-adjust the mixing ratio to meet the predetermined ratio requirements. Using the Kalman filter algorithm, the initial state estimate value and the error covariance matrix are set to smooth the continuously monitored concentration data to reduce the influence of random fluctuations on the judgment result. Combining the concentration fluctuation analysis results and the data after smoothing, it is judged whether the components of the mixed fuel meet the predetermined ratio requirements. If the results of 10 consecutive measurements are within the allowable deviation range, it is determined that the components of the mixed fuel meet the predetermined ratio requirements.

[0022] S102. Conduct a flame stability test on the obtained mixed fuel sample. Use a high-speed imaging system to capture the flame morphology, analyze the change in flame length, measure the flame color characteristics, and determine the flame temperature distribution to obtain a dataset of flame stability parameters at different ammonia doping ratios.

[0023] Obtain the mixed fuel sample, test the mixed fuel with different ammonia doping ratios, capture the flame morphology through a high-speed imaging system to obtain a continuous sequence of flame images. Use the Canny edge detection algorithm to extract the flame contour from the flame image sequence and calculate the characteristic parameters of the flame area and height. According to the flame image sequence, analyze the oscillation frequency and amplitude of the flame length through fast Fourier transform to obtain the flame length oscillation characteristic data. Use a multispectral imaging device to collect images of the flame in multiple bands, calculate the color temperature and radiation intensity distribution of the flame through two-color method temperature measurement, and obtain the flame color characteristic and temperature distribution data. Calculate the flame stability index. If the ratio of the standard deviation of the flame length to the average flame length is less than a preset threshold, it is determined that the flame is stable. Construct a dataset of flame stability parameters based on the flame oscillation frequency, propagation speed, brightness uniformity, and stability index.

[0024] Exemplarily, for the obtained mixed fuel sample, the mixed fuels with ammonia doping ratios of 30%, 50%, and 70% are successively tested. The mixed fuel is ignited by a burner, and a high-speed imaging system with 10,000 frames per second is used to capture the flame morphology, obtaining a continuous 1000-frame flame image sequence. The Canny edge detection algorithm is used to extract the flame contour from the flame images, and characteristic parameters such as the flame area and height are calculated. According to the flame image sequence, the flame length change curve is calculated, and the oscillation frequency and amplitude of the flame length are analyzed through fast Fourier transform to obtain the flame length oscillation characteristic data. A multispectral imaging device is used to collect images of the flame in three bands of 340 nm, 455 nm, and 655 nm. The color temperature and radiation intensity distribution of the flame are calculated through two-color method temperature measurement, obtaining the flame color characteristics and temperature distribution data. The flame propagation speed is calculated from the change in the position of the flame front between consecutive frames. Combining the flame morphology, length change, color characteristics, and temperature distribution data, a flame stability parameter dataset is constructed. The flame stability index is calculated, defined as 1 minus the ratio of the standard deviation of the flame length to the average flame length. For each ammonia doping ratio, the flame oscillation frequency, propagation speed, brightness uniformity, and stability index are recorded to form a complete flame stability parameter dataset. For the obtained mixed fuel sample, the automated test system successively prepares the mixed fuels with ammonia doping ratios of 30%, 50%, and 70%, and precisely controls the flow rate to be 10 L / min. After the burner is ignited, the high-speed imaging system with 10,000 frames per second captures the flame morphology, continuously collecting a 1000-frame flame image sequence with a resolution of 1280x1024. The image processing unit applies the Canny edge detection algorithm, sets the low threshold to 50 and the high threshold to 150, and extracts the flame contour. The number of pixels of the flame area and the maximum height in each frame of the image are calculated and converted into actual dimensions, such as an area of 300 cm 2, with a height of 40 cm. Perform a fast Fourier transform on the flame length data of 1000 frames of images, with a sampling frequency of 10 kHz, to obtain a spectrogram, and identify the main oscillation frequencies such as 100 Hz and an amplitude of 0.5 cm. The multispectral imaging device synchronously acquires 1000 frames of flame images in three bands of 340 nm, 455 nm, and 655 nm. Apply the two-color method for temperature measurement, select the 455 nm and 655 nm bands, calculate the flame temperature distribution, and obtain a maximum temperature of 2000 K and an average temperature of 1800 K. By comparing the changes in the position of the flame front between adjacent frames, calculate the flame propagation speed, such as 0.5 m / s. Combine all the data to construct a dataset of flame stability parameters. Calculate the flame stability index. Assuming an average flame length of 40 cm and a standard deviation of 2 cm, the stability index is 1 - (2 / 40) = 0.95. For each ammonia doping ratio, record the flame oscillation frequency such as 100 Hz, the propagation speed such as 0.5 m / s, the standard deviation of brightness uniformity / average brightness = 0.1, and the stability index such as 0.95 to form a complete dataset of flame stability parameters for subsequent analysis and optimization of combustion conditions.

[0025] S103. According to the dataset of flame stability parameters, determine the relationship function between the hydrogen ammonia doping ratio and the flame length, color, and temperature. Train a flame stability prediction model through the support vector machine algorithm to judge the stability state of the flame under different doping ratios.

[0026] According to the hydrogen ammonia doping ratio and the flame parameter dataset, use the polynomial regression method to fit the relationship function between the hydrogen ammonia doping ratio and the flame length, color, and temperature to obtain three relationship function equations. Perform feature engineering processing on the dataset of flame stability parameters, select the doping ratio, flame length, color, and temperature as input features, and use the flame stability index as the label. Train a flame stability prediction model through the support vector machine algorithm, and optimize the kernel function parameters and regularization parameters through the grid search method. If the flame parameter data with different ammonia doping ratios are input, then use the support vector machine model for prediction to obtain the corresponding flame stability prediction values. Judge the stability state of the flame according to the pre-established stability threshold, set multiple stability levels, and determine different prediction value ranges.

[0027] Exemplarily, according to the flame stability parameter dataset, the relationship functions between the hydrogen-ammonia blending ratio and the flame length, color, and temperature are fitted using the polynomial regression method. The optimal order is selected through the cross-validation method, and the polynomial coefficients are determined using the least squares method to obtain three relationship function equations. Feature engineering is performed on the flame stability parameter dataset. The ammonia blending ratio, flame length, color, and temperature are selected as input features, and the flame stability index is used as the label. The optimal feature subset is selected through the recursive feature elimination method, and at the same time, the calculation results of the relationship functions are added as new features to obtain the processed training dataset. Using the processed training dataset, a support vector machine algorithm is used to train the flame stability prediction model. The kernel function parameters and regularization parameters are optimized through the grid search method, and the L2 regularization term is introduced to handle the overfitting problem to obtain the optimal support vector machine model. Input the flame parameter data with different ammonia blending ratios, and use the trained support vector machine model for prediction to obtain the corresponding flame stability prediction values. The stability state of the flame is judged according to the stability threshold determined by the statistical analysis of historical data. Multiple stability levels are set, corresponding to different prediction value ranges, to achieve an accurate judgment of the flame stability state under different ammonia blending ratios. Based on the flame stability parameter dataset, which contains 1000 sets of flame parameter records at different ammonia blending ratios from 30% to 70%, the relationship functions are fitted using the polynomial regression method. Through 5-fold cross-validation, the mean squared errors of polynomials of orders 1 to 5 are compared, and the 3rd-order polynomial with the smallest mean squared error is selected as the optimal model. The polynomial coefficients are determined using the least squares method to obtain three relationship function equations between the hydrogen-ammonia blending ratio and the flame length, color, and temperature. Feature engineering is performed on the dataset. The initial features include the ammonia blending ratio, flame length, RGB values of the color, and temperature, a total of 6 features. The recursive feature elimination method is applied, with a step size of 1 and 5 cross-validation times, to select 4 optimal features. The calculation results of the relationship functions are added as new features to form a 7-dimensional feature vector. A support vector machine algorithm is used to train the prediction model, using the RBF kernel function. The parameters are optimized through the grid search method within the ranges of C = [0.1, 1, 10, 100] and gamma = [0.01, 0.1, 1, 10]. The L2 regularization term is introduced, and the regularization coefficient is set to 0.01 to prevent overfitting. 80% of the data is used as the training set, and 20% is used as the test set. The optimal model is obtained after 1000 iterations of training, and the accuracy of the test set reaches 95%. Input the flame parameter data of a new ammonia blending ratio, such as 55%, and the model outputs a stability prediction value of 0.82. According to the statistics of historical data, the stability threshold is set to [0.7, 0.8, 0.9], corresponding to three levels: unstable, relatively stable, and stable. The judgment result shows that the flame is in a relatively stable state at a 55% ammonia blending ratio, providing a basis for subsequent combustion optimization.

[0028] S104. Conduct kinetic analysis on the output results of the flame stability prediction model, simulate the combustion processes of hydrogen and ammonia, obtain the curves of the concentrations of key intermediate products changing with time, and judge the propagation characteristics of the reaction chain under different ammonia doping ratios.

[0029] Modify the GRI-Mech3.0 mechanism according to the output results of the preset flame stability prediction model, and construct the combustion kinetic mechanism of the hydrogen and ammonia mixed fuel. The combustion kinetic mechanism includes elementary reaction equations and reaction rate parameters; solve the combustion kinetic equations to obtain the curves of the concentrations of key intermediate products OH, H, and NH2 changing with time; use the local sensitivity analysis method to calculate the contribution degree of each reaction process to the overall reaction rate under different ammonia doping ratios, and obtain the reaction chain length and branching ratio; establish an association model between the ammonia doping ratio and the flame propagation speed and ignition delay time according to the reaction chain length and branching ratio, and predict the flame propagation characteristics at any ammonia doping ratio through cubic spline interpolation; if the reaction chain length is greater than the preset reaction chain length threshold and the branching ratio is greater than the preset branching ratio threshold, it is judged that the flame is in a stable state; if the reaction chain length is less than the preset reaction chain length threshold or the branching ratio is less than the preset branching ratio threshold, it is judged that the flame is in an unstable state.

[0030] Exemplarily, based on the output results of the flame stability prediction model, the GRI-Mech3.0 mechanism is modified to construct the combustion kinetic mechanism of hydrogen and ammonia blended fuels, including elementary reaction equations and reaction rate parameters, and the reaction mechanism and initial conditions are imported through CHEMKIN software. The fourth-order Runge-Kutta method is used to solve the combustion kinetic equations to obtain the concentration-time curves of key intermediate products such as OH, H, NH2, etc., and calculate the peak concentrations and appearance times of each intermediate product. Using the local sensitivity analysis method, calculate the contribution of each reaction step to the overall reaction rate at different ammonia blending ratios, identify the main reaction paths and rate-controlling steps, and obtain quantitative descriptions of reaction chain propagation characteristics such as reaction chain length and branching ratio. Based on the reaction chain propagation characteristic data, establish an association model between the ammonia blending ratio and the flame propagation speed and ignition delay time, and predict the flame propagation characteristics at any ammonia blending ratio through cubic spline interpolation. Set the reaction chain length threshold and branching ratio threshold, and by comparing the relationship between the reaction chain length and branching ratio at different ammonia blending ratios and the preset thresholds, combined with the output results of the flame stability prediction model, comprehensively judge the flame stability state. Based on the stability indices at ammonia blending ratios of 30%, 50%, and 70% output by the flame stability prediction model, modify the GRI-Mech3.0 mechanism using CHEMKIN software to construct a combustion kinetic mechanism of hydrogen-ammonia blended fuels containing 253 elementary reactions. Set the initial conditions as temperature 1000K, pressure 1atm, and equivalence ratio 1.0. Using the fourth-order Runge-Kutta method with a time step of 1e-6 seconds, solve the combustion kinetic equations to obtain the concentration-time curves of 10 key intermediate products such as OH, H, NH2, etc. Calculate the peak concentrations and appearance times of the intermediate products, such as the OH peak concentration of 1.5e-3 mol / L and the appearance time of 0.5 ms. Using the local sensitivity analysis method, calculate the contribution of each reaction step to the overall reaction rate, and identify 3 main reaction paths and 2 rate-controlling steps. Quantitatively describe the reaction chain propagation characteristics, such as the reaction chain length of 15 and the branching ratio of 0.8 at 30% ammonia blending. Establish an association model between the ammonia blending ratio and the flame propagation speed and ignition delay time, and predict the flame propagation characteristics at ammonia blending ratios of 40% and 60% through cubic spline interpolation. Set the reaction chain length threshold to 20 and the branching ratio threshold to 0.7. Compare the relationship between the reaction chain length and branching ratio at different ammonia blending ratios and the preset thresholds, and combined with the output results of the flame stability prediction model such as the stability index of 0.85 at 40% ammonia blending, comprehensively judge that the flame is in a stable state at 40% ammonia blending and the flame stability decreases at 60% ammonia blending.

[0031] S105. According to the change curve and propagation characteristics, establish a criterion for judging the key points of flame stability, set a threshold for flame stability, and judge whether the flame is in a stable state under the current ammonia blending ratio by comparing the deviation between the actual flame parameters and the threshold, and determine the occurrence conditions of the key points of flame stability.

[0032] Obtain the flame characteristic data, extract the key features of flame stability by using the principal component analysis method, classify the key features by using the K-means clustering algorithm, and obtain a feature vector including flame length, temperature and oscillation frequency. According to the feature vector, establish a flame stability discrimination model by using the support vector machine algorithm. The discrimination model takes the key features as input variables and the flame stability state as the output variable. For the discrimination model, set a threshold for flame stability and construct a multi-dimensional threshold space, and the boundary of the multi-dimensional threshold space is determined by the output result of the support vector machine model. Receive the actual flame parameters under the current ammonia blending ratio, calculate the Mahalanobis distance between the actual flame parameters and the multi-dimensional threshold space, and the Mahalanobis distance is obtained by inverting the covariance matrix and calculating the sum of squares of the differences of the multi-dimensional feature vectors. If the Mahalanobis distance is less than the preset threshold, it is determined that the flame is in a stable state; when 10 consecutive data points are all within the stable region, it is determined as the key point of flame stability, and record the occurrence conditions of the key point of flame stability.

[0033] Exemplarily, according to the change curve and propagation characteristic data, the principal component analysis method is used to extract the key features of flame stability. The features are classified by the K-means clustering algorithm to obtain the feature vectors of the key points of flame stability, including parameters such as flame length, temperature, and oscillation frequency. The support vector machine algorithm is used to establish a flame stability discrimination model, with the key features as input variables and the flame stability state as the output variable. The optimal model parameters are determined through 5-fold cross-validation to obtain the discrimination criterion for the key points of flame stability. Based on historical data and expert knowledge, the Delphi method is used to synthesize expert opinions, set the flame stability threshold, construct a multi-dimensional threshold space, and use the output result of the support vector machine model to assist in determining the boundary of the multi-dimensional threshold space. Input the actual flame parameters under the current ammonia blending ratio, and calculate the Mahalanobis distance from the threshold space, specifically obtained by inverting the covariance matrix and calculating the sum of the squares of the differences of the multi-dimensional feature vectors. If the distance is less than the preset threshold, determined by the statistical distribution of historical data, it is determined that the flame is in a stable state. When 10 consecutive data points are all within the stable region, it is determined as the key point of flame stability, and the current conditions are recorded as the occurrence conditions of the key point of flame stability. Based on 1000 sets of flame stability experimental data, the principal component analysis method is used to extract the key features, and the principal components with a cumulative contribution rate of 95% are retained to obtain 5 main features: flame length, temperature, oscillation frequency, propagation speed, and OH radical concentration. The K-means clustering algorithm is used, with the number of clusters k = 3, and after 50 iterations, the feature space is divided into three categories: stable, metastable, and unstable. The support vector machine algorithm is used to construct a discrimination model, with the input being a 5-dimensional feature vector and the output being the stability state, 0 - unstable, 1 - metastable, 2 - stable. The RBF kernel function is used, and through 5-fold cross-validation, the parameters are optimized to C = 10 and gamma = 0.1, and the model accuracy reaches 92%. The Delphi method is applied, inviting 7 experts for 3 rounds of evaluation, and combined with historical data statistics, a multi-dimensional threshold space is determined: flame length 30 - 50 cm, temperature 1800 - 2200 K, oscillation frequency 100 - 150 Hz, propagation speed 0.5 - 1.5 m / s, OH concentration (1 - 5)×10^-3 mol / L. Calculate the Mahalanobis distance between the actual parameters and the threshold space, use the inverse matrix of the covariance matrix, and set the distance threshold to 2.5 based on the 95% confidence interval of historical data. At an ammonia blending ratio of 30%, input the measured parameters: flame length 40 cm, temperature 2000 K, oscillation frequency 120 Hz, propagation speed 1.0 m / s, OH concentration 3×10^-3 mol / L. The calculated Mahalanobis distance is 1.8, which is less than the threshold 2.5, so it is determined to be in a stable state. Continuously monitor 100 data points. When 10 consecutive points are all within the stable region, the condition of 30% ammonia blending and the current parameter combination are recorded as the key point of flame stability.

[0034] S106. For the occurrence conditions of the key points of flame stability, simulate the flow field distribution under different burner structures, obtain the fuel injection velocity field, measure the temperature field and species concentration field in the combustion region, and judge the influence of the burner structure on flame stability.

[0035] Establish a three-dimensional model of the burner structure according to the fuel mixing ratio, combustion temperature and pressure parameters, including the nozzle diameter, combustion chamber length and cross-sectional area; use ANSYS Fluent software to perform mesh generation and boundary condition setting for the burner structure model; solve the Navier-Stokes equation and k-ε turbulence model of fluid motion to obtain the fuel injection velocity field of the burner structure model; according to the fuel injection velocity field, obtain the velocity vector distribution and turbulence intensity distribution in the combustion region of the burner structure model; use the simulated combustion model and chemical reaction kinetic equation to calculate the temperature field and species concentration field in the combustion region; obtain the temperature distribution cloud map of the temperature field and the concentration distribution of key species in the species concentration field; calculate the flame oscillation frequency and amplitude according to the temperature distribution cloud map and the concentration distribution of the key species; use the random forest regression model to analyze the relationship between the burner structure parameters and the flame oscillation frequency and amplitude; if the flame oscillation frequency is lower than the preset threshold and the flame oscillation amplitude is lower than the preset threshold, it is judged that the burner structure corresponding to the burner structure parameters has good flame stability.

[0036] Exemplarily, according to the occurrence conditions of the key points of flame stability, parameters such as fuel mixture ratio, combustion temperature, and pressure are converted into burner structure parameters, and three-dimensional models of different burner structures are established, including nozzle diameter, combustion chamber length, and cross-sectional area, etc. The ANSYS Fluent software is used for mesh generation and boundary condition setting to simulate the flow field distribution under different structures. By solving the basic equations of fluid motion such as the Navier-Stokes equation and the k-ε turbulence model equation, the fuel injection velocity field is calculated, and the velocity vector distribution and turbulence intensity distribution in the combustion region are obtained to judge the flow field characteristics under different structures. Using the simulated combustion model Eddy Dissipation Concept combustion model and chemical reaction kinetics equations, the temperature field and species concentration field in the combustion region are calculated, and the temperature distribution contour map and the concentration distributions of key species such as OH and NO are obtained. Based on the simulation results, the flame oscillation frequency and amplitude are calculated to evaluate the flame stability. Using the random forest regression model, the relationship between the burner structure parameters and the flame stability index is analyzed to judge the influence degree of different structures on the flame stability. Through ANOVA (Analysis of Variance), the differences in flame stability under different burner structures are compared to quantitatively evaluate the influence of structural changes on stability. Based on the key points of flame stability conditions: fuel mixture ratio 40%, combustion temperature 2000K, pressure 2 atm, three burner structures are designed. Type A: nozzle diameter 5 mm, combustion chamber length 200 mm, cross-sectional area 100 cm 2 , Type B: nozzle diameter 8 mm, combustion chamber length 250 mm, cross-sectional area 150 cm 2 , Type C: nozzle diameter 6 mm, combustion chamber length 180 mm, cross-sectional area 80 cm 2Use ANSYS Fluent software to establish a three-dimensional model. Tetrahedral elements are used for mesh generation, and the number of meshes is approximately 500,000. For boundary condition settings, the inlet velocity is 20 m / s, the outlet pressure is 1 atm, and the wall is adiabatic. Solve the Navier-Stokes equations and the k-ε turbulence model equations, with 1000 iteration times and a residual less than 1e-6. The calculation results show that the maximum turbulence intensity of type A is 15%, type B is 12%, and type C is 18%. Apply the EDC combustion model, and the reaction mechanism includes 30 species and 200 elementary reactions. For the calculation results of the temperature field, the highest temperature of type A is 2150 K, type B is 2080 K, and type C is 2210 K. The peak OH concentration is 2.5e-3 mol / L for type A, 2.1e-3 mol / L for type B, and 2.8e-3 mol / L for type C. Analyze the temperature fluctuations through fast Fourier transform to obtain the flame oscillation frequency, which is 120 Hz for type A, 95 Hz for type B, and 135 Hz for type C; the amplitude is 50 K for type A, 40 K for type B, and 65 K for type C. Use a random forest regression model with 100 trees and a maximum depth of 10 to analyze the relationship between structural parameters and stability indicators. The feature importance ranking is that the nozzle diameter is 0.4, the combustion chamber length is 0.3, and the cross-sectional area is 0.3. The ANOVA analysis results show that the flame stability differences among the three structures are significant with a p-value < 0.01, among which type B has the highest stability and type C has the lowest. The comprehensive evaluation shows that the structure of type B burner has the most positive impact on flame stability.

[0037] S107. According to the analysis results of the influence of the burner structure on flame stability, use the genetic algorithm to optimize the geometric parameters of the burner. By adjusting the fuel injection angle and the combustion chamber shape, use the numerical simulation method to calculate the flame morphology evolution process under the optimized burner structure, and judge whether the optimized burner can effectively suppress the occurrence of key points of flame stability to obtain the best ammonia blending ratio.

[0038] Obtain the burner geometric parameters, fuel injection angle, and combustion chamber shape, and construct the genetic algorithm optimization model; wherein, the genetic algorithm optimization model sets the population size, crossover rate, and mutation rate. Based on the structural parameters obtained from the optimization model, establish a three-dimensional computational fluid dynamics model; the three-dimensional computational fluid dynamics model adopts a turbulence model and the GRI-Mech3.0 reaction mechanism. Obtain the spatio-temporal distribution data of the flame morphology evolution through the three-dimensional computational fluid dynamics model; calculate the flame oscillation frequency, flame propagation speed, and flame temperature distribution according to the spatio-temporal distribution data; if the flame oscillation frequency is less than the preset frequency threshold and the amplitude is less than the preset amplitude threshold, determine the ammonia blending ratio range for stable combustion. Use the NSGA-II algorithm for multi-objective optimization, and set the flame stability, combustion efficiency, and NOx emission level as the optimization objectives; obtain the Pareto optimal solution set based on the multi-objective optimization. Obtain the performance data of the optimized burner through a small test bench; compare the performance data with the numerical simulation prediction results; adjust the optimization parameters according to the comparison results.

[0039] Exemplarily, according to the analysis results of the influence of the burner structure on the flame stability, construct a genetic algorithm optimization model, set the population size to 100, the crossover rate to 0.8, and the mutation rate to 0.1. Set the burner geometric parameters such as nozzle diameter, combustion chamber length, cross-sectional area, fuel injection angle, and combustion chamber shape as optimization variables, and use the weighted sum of the flame oscillation frequency and amplitude as the fitness function to perform the optimization calculation of the burner structure parameters. Using the optimized burner structure parameters, establish a three-dimensional computational fluid dynamics model, adopt the k-ε turbulence model and the GRI-Mech3.0 reaction mechanism, simulate the combustion process under different ammonia blending ratios, and obtain the spatio-temporal distribution data of the flame morphology evolution. By analyzing the flame morphology evolution data, calculate characteristic parameters such as the flame oscillation frequency, flame propagation speed, and flame temperature distribution. When the flame oscillation frequency is less than 100 Hz and the amplitude is less than 50 K, it is determined to be stable, and the ammonia blending ratio range for stable combustion is determined. Based on the ammonia blending ratio range for stable combustion, combined with the combustion efficiency and emission characteristic data, use the NSGA-II algorithm for multi-objective optimization, set the flame stability, combustion efficiency, and NOx emission level as the optimization objectives, calculate the Pareto optimal solution set, and select the best ammonia blending ratio from it. Verify the performance of the optimized burner through a small test bench, compare the experimental results with the numerical simulation prediction, and further adjust the optimization parameters to ensure the accuracy and reliability of the optimization results. Based on the previous analysis results, construct a genetic algorithm optimization model, set the population size to 100, the crossover rate to 0.8, the mutation rate to 0.1, and the number of iterations to 500. The optimization variables include a nozzle diameter of 5 - 10 mm, a combustion chamber length of 150 - 300 mm, and a cross-sectional area of 50 - 200 cm 2, the fuel injection angle is 15 - 45°. The fitness function is 0.7×(1 / flame oscillation frequency) + 0.3×(1 / flame amplitude). After 500 generations of evolution, the optimal structure is obtained, with a nozzle diameter of 7.5 mm, a combustion chamber length of 220 mm, and a cross-sectional area of 120 cm 2 , and the injection angle is 30°. An optimized burner model is established using ANSYS Fluent software. The k-ε turbulence model and the GRI-Mech 3.0 reaction mechanism with 53 species and 325 reactions are adopted to simulate the combustion process at ammonia doping ratios of 30%, 40%, 50%, and 60%. The time step is 0.1 ms, and the total simulation time is 100 ms. Analyze the data of the flame morphology evolution, and calculate that the flame oscillation frequency is 85 Hz and the amplitude is 45 K when the ammonia doping ratio is 40%, which is determined to be stable; the frequency is 120 Hz and the amplitude is 70 K when the ammonia doping ratio is 60%, which is determined to be unstable. The stable combustion ammonia doping ratio range is determined to be 35% - 55%. Within this range, the NSGA-II algorithm is used for multi-objective optimization, with a population size of 50 and 200 iterations. The optimization objectives are to maximize the flame stability index 1 - oscillation frequency / 100 Hz - amplitude

[0040] / 100 K, maximize the combustion efficiency, and minimize the NOx emissions. The Pareto optimal solution set is obtained, and the solution with a flame stability index of 0.85, a combustion efficiency of 98%, and NOx emissions of 30 ppm is selected, corresponding to the optimal ammonia doping ratio of 45%. It is verified on a small test bench that the measured flame stability index is 0.82, the combustion efficiency is 97.5%, and the NOx emissions are 32 ppm, with an error less than 5% compared with the simulation results, verifying the reliability of the optimization results.

[0041] As described above, it is only the preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent replacements or changes, and all should be covered within the protection scope of the present invention.

Claims

1. A method for detecting the stability of a torch flame, characterized in that: The method comprises: According to the preset hydrogen-ammonia ratio range, mixed fuel samples with different ratios are obtained from the storage tank, the mixing ratio of hydrogen and ammonia is adjusted by a mass flow controller, and the concentration of hydrogen and ammonia in the mixed fuel is monitored in real time by a gas analyzer to determine whether the components of the mixed fuel meet the preset ratio requirements; The flame stability test was carried out on the obtained mixed fuel samples. A high-speed camera system was used to capture the flame morphology, analyze the change of flame length, measure the flame color characteristics, determine the flame temperature distribution, and obtain the flame stability parameter data set under different ammonia mixing ratios; According to the flame stability parameter data set, the relationship function between the hydrogen-ammonia ratio and the flame length, color and temperature is determined. The flame stability prediction model is trained by the support vector machine algorithm to determine the flame stability state under different ammonia ratios. Conduct kinetic analysis on the output results of the flame stability prediction model, simulate the combustion process of hydrogen and ammonia, obtain the time-varying curve of the concentration of key intermediate products, and determine the propagation characteristics of the reaction chain under different ammonia mixing ratios; According to the change curve and propagation characteristics, the flame stability key point discrimination criteria are established, the flame stability threshold is set, and by comparing the deviation between the actual flame parameters and the threshold, it is judged whether the flame is in a stable state under the current ammonia mixing ratio, and the conditions for the occurrence of the flame stability key point are determined; According to the conditions for the occurrence of key points of flame stability, the flow field distribution under different burner structures is simulated, the fuel injection velocity field is obtained, the temperature field and species concentration field of the combustion area are measured, and the influence of the burner structure on the flame stability is determined; According to the analysis results of the influence of burner structure on flame stability, genetic algorithm is used to optimize the geometric parameters of burner. By adjusting the fuel injection angle and the shape of the combustion chamber, the flame morphology evolution process under the optimized burner structure is calculated using numerical simulation method to determine whether the optimized burner can effectively suppress the occurrence of key points of flame stability and obtain the optimal ammonia blending ratio.

2. The method according to claim 1, wherein: The method comprises obtaining mixed fuel samples with different proportions from a storage tank according to a preset hydrogen-ammonia ratio range, adjusting the mixing ratio of hydrogen and ammonia by a mass flow controller, monitoring the concentration of hydrogen and ammonia in the mixed fuel in real time by a gas analyzer, and judging whether the components of the mixed fuel meet the preset ratio requirements, including: Obtaining a mixed fuel sample within a preset hydrogen-ammonia ratio range, wherein the mixed fuel sample is supplied from a storage tank; According to the preset hydrogen-ammonia ratio range, adjusting the mixing ratio of hydrogen and ammonia in the mixed fuel sample by a mass flow controller; Using a gas analyzer to monitor the mixed fuel sample in real time to obtain concentration data of hydrogen and ammonia; Calculating the deviation between the actual components of the mixed fuel sample and the preset ratio requirement by using a direct comparison method, and triggering the automatic correction procedure of the mass flow controller if the deviation exceeds a preset threshold; Performing concentration fluctuation analysis according to the deviation value to quantitatively evaluate the concentration stability of the mixed fuel sample; Using a Kalman filter algorithm to smooth the concentration data to obtain smoothed concentration data; Combining the concentration fluctuation analysis result and the smoothed concentration data, determining whether the components of the mixed fuel sample meet the predetermined ratio requirements; If the measurement results for a preset number of consecutive times are all within the allowable deviation range, it is determined that the components of the mixed fuel sample meet the predetermined ratio requirements.

3. The method according to claim 1, wherein: The flame stability test is performed on the obtained mixed fuel sample, and a high-speed camera system is used to capture the flame morphology, analyze the change of flame length, measure the flame color characteristics, and determine the flame temperature distribution to obtain a flame stability parameter data set under different ammonia mixing ratios, including: Obtaining the mixed fuel sample, testing the mixed fuel with different ammonia blending ratios, capturing the flame morphology through a high-speed camera system, and obtaining a continuous flame image sequence; The Canny edge detection algorithm is used to extract the flame contour from the flame image sequence, and the characteristic parameters of the flame area and height are calculated; According to the flame image sequence, analyzing the oscillation frequency and amplitude of the flame length by fast Fourier transform to obtain flame length oscillation characteristic data; A multi-spectral imaging device is used to collect images of the flame in multiple bands, and the color temperature and radiation intensity distribution of the flame are calculated by two-color temperature measurement to obtain flame color characteristics and temperature distribution data; Calculating the flame stability index, and if the ratio of the flame length standard deviation to the average flame length is less than a preset threshold, determining that the flame is stable; A flame stability parameter data set is constructed based on the flame oscillation frequency, propagation speed, brightness uniformity and stability index.

4. The method according to claim 1, wherein: The method comprises: determining the relationship function between the hydrogen-ammonia ratio and the flame length, color and temperature according to the flame stability parameter data set, training the flame stability prediction model by the support vector machine algorithm, and judging the flame stability state under different ammonia ratios, including: According to the hydrogen-ammonia ratio and flame parameter data set, a polynomial regression method is used to fit the relationship function between the hydrogen-ammonia ratio and the flame length, color and temperature to obtain three relationship function equations; Performing feature engineering processing on the flame stability parameter data set, selecting the ammonia mixing ratio, flame length, color and temperature as input features, and taking the flame stability index as a label; The flame stability prediction model was trained using support vector machine algorithm, and the kernel function parameters and regularization parameters were optimized using grid search method. If flame parameter data with different ammonia blending ratios are input, the support vector machine model is used to perform predictions to obtain corresponding flame stability prediction values; The stability state of the flame is judged according to a pre-established stability threshold, a plurality of stability levels are set, and different prediction value ranges are determined.

5. The method according to claim 1, wherein: The output results of the flame stability prediction model are subjected to dynamic analysis, the combustion process of hydrogen and ammonia is simulated, the concentration curve of key intermediate products is obtained over time, and the propagation characteristics of the reaction chain under different ammonia mixing ratios are determined, including: Modify the GRI-Mech3.0 mechanism according to the output results of the preset flame stability prediction model to construct the combustion kinetics mechanism of the hydrogen and ammonia mixed fuel, wherein the combustion kinetics mechanism includes elementary reaction equations and reaction rate parameters; Solve the combustion kinetic equations to obtain the concentration curves of the key intermediate products OH, H and NH2 over time; The contribution of each reaction process to the overall reaction rate under different ammonia doping ratios was calculated using the local sensitivity analysis method, and the reaction chain length and branching rate were obtained. A correlation model between the ammonia ratio, the flame propagation speed and the ignition delay time is established according to the reaction chain length and the branching rate, and the flame propagation characteristics under any ammonia ratio are predicted by the cubic spline interpolation method; If the reaction chain length is greater than a preset reaction chain length threshold and the branching rate is greater than a preset branching rate threshold, it is determined that the flame is in a stable state; If the reaction chain length is less than a preset reaction chain length threshold or the branching rate is less than a preset branching rate threshold, it is determined that the flame is in an unstable state.

6. The method according to claim 1, wherein: According to the change curve and propagation characteristics, a flame stability key point discrimination criterion is established, a flame stability threshold is set, and by comparing the deviation between the actual flame parameters and the threshold, it is judged whether the flame is in a stable state under the current ammonia mixing ratio, and the occurrence conditions of the flame stability key point are determined, including: Obtaining flame characteristic data, extracting key features of flame stability using principal component analysis, classifying the key features using a K-means clustering algorithm, and obtaining feature vectors including flame length, temperature, and oscillation frequency; According to the feature vector, a flame stability discrimination model is established using a support vector machine algorithm, wherein the discrimination model uses key features as input variables and flame stability status as output variables; For the discrimination model, a flame stability threshold is set to construct a multidimensional threshold space, wherein the boundary of the multidimensional threshold space is determined by the output result of the support vector machine model; receiving an actual flame parameter under a current ammonia blending ratio, and calculating a Mahalanobis distance between the actual flame parameter and a multidimensional threshold space, wherein the Mahalanobis distance is obtained by inverting a covariance matrix and calculating a sum of squares of multidimensional eigenvector differences; If the Mahalanobis distance is less than a preset threshold, it is determined that the flame is in a stable state; When 10 consecutive data points are all located in the stable region, they are determined to be key points of flame stability, and the conditions for the occurrence of the key points of flame stability are recorded.

7. The method according to claim 1, wherein: The method aims at the occurrence conditions of the key points of flame stability, simulates the flow field distribution under different burner structures, obtains the fuel injection velocity field, measures the temperature field and species concentration field of the combustion area, and determines the influence of the burner structure on the flame stability, including: According to the fuel mixture ratio, combustion temperature and pressure parameters, a three-dimensional model of the burner structure is established, including the nozzle diameter, combustion chamber length and cross-sectional area; ANSYS Fluent software is used to perform meshing and boundary condition setting of the burner structure model; Solving the Navier-Stokes equations of fluid motion and the k-ε turbulence model to obtain the fuel injection velocity field of the burner structure model; According to the fuel injection velocity field, obtaining the velocity vector distribution and turbulence intensity distribution of the combustion area in the burner structure model; Calculating the temperature field and species concentration field of the combustion area by using a simulated combustion model and a chemical reaction kinetic equation; Obtaining a temperature distribution cloud map of the temperature field and a concentration distribution of key species in the species concentration field; Calculating the flame oscillation frequency and amplitude according to the temperature distribution cloud map and the concentration distribution of the key species; Analyzing the relationship between the burner structural parameters and the flame oscillation frequency and amplitude using a random forest regression model; If the flame oscillation frequency is lower than a preset threshold and the flame oscillation amplitude is lower than a preset threshold, it is determined that the burner structure corresponding to the burner structural parameters has good flame stability.

8. The method according to claim 1, wherein: The method uses a genetic algorithm to optimize the geometric parameters of the burner based on the analysis results of the influence of the burner structure on the flame stability, adjusts the fuel injection angle and the shape of the combustion chamber, and uses a numerical simulation method to calculate the evolution of the flame morphology under the optimized burner structure, and determines whether the optimized burner can effectively suppress the occurrence of key points of flame stability, and obtains the optimal ammonia blending ratio, including: Obtaining burner geometric parameters, fuel injection angle and combustion chamber shape, and constructing the genetic algorithm optimization model; Wherein, the genetic algorithm optimization model sets the population size, crossover rate and mutation rate; A three-dimensional computational fluid dynamics model is established based on the structural parameters obtained from the optimization model; The three-dimensional computational fluid dynamics model adopts a turbulence model and a GRI-Mech3.0 reaction mechanism; Acquiring the spatiotemporal distribution data of flame morphology evolution through the three-dimensional computational fluid dynamics model; Calculating flame oscillation frequency, flame propagation speed and flame temperature distribution according to the spatiotemporal distribution data; If the flame oscillation frequency is less than a preset frequency threshold and the amplitude is less than a preset amplitude threshold, then the ammonia blending ratio range for stable combustion is determined; The NSGA-II algorithm is used for multi-objective optimization, and flame stability, combustion efficiency and NOx emission level are set as optimization targets; Obtaining a Pareto optimal solution set based on the multi-objective optimization; Obtain performance data for optimized afterburners through small test benches; comparing the performance data with numerical simulation prediction results; Adjust the optimization parameters according to the comparison results.

Images