High-robustness flame stability measurement and evaluation method and system for complex working conditions

Abstract

The invention relates to a high-robustness flame stability measurement and evaluation method and system for complex working conditions, and the method comprises the steps: obtaining an image sequence of flame of a combustor, and carrying out the preprocessing of the image sequence; carrying out flame characteristic parameter extraction on the preprocessed image sequence, carrying out adaptive characteristic fusion on the flame characteristic parameters, and calculating a flame stability index; and performing decision analysis on the flame stability index through double-threshold hysteresis comparison, and outputting a flame stability measurement and evaluation result and early warning information. According to the method, invalid images can be effectively identified, eliminated and compensated under complex working conditions, the feature extraction method and the fusion weight are dynamically adjusted according to working condition changes, robustness and predictability are achieved at the same time in the state output link, and therefore accurate and reliable quantitative measurement and evaluation results are provided in a changeable industrial operation environment.

Description

Technical Field

[0001] This invention relates to the field of combustion process measurement and control technology, and in particular to a highly robust method and system for measuring and evaluating flame stability under complex operating conditions. Background Technology

[0002] In industrial production processes, the quantitative measurement and assessment of burner flames is a crucial technology for achieving safe, efficient, and clean operation of industrial kilns, boilers, and engines. However, in real-world industrial scenarios with complex operating conditions, existing flame monitoring technologies still have significant shortcomings in providing operators with online and reliable quantitative assessment results. Complex operating conditions may include, but are not limited to, various unstable and non-ideal operating states such as fuel composition fluctuations, drastic load changes, combustion air distribution adjustments, high-dust environments, and contamination of the observation window. Under these conditions, existing monitoring technologies mainly suffer from the following three problems:

[0003] First, the data acquisition process suffers from insufficient anti-interference capabilities. Flame images acquired in industrial settings commonly exhibit high-intensity noise and transient interference, while existing technologies lack effective mechanisms for identifying and filtering contaminant data during the data preprocessing stage. For instance, under conditions of dust interference, moisture obscuring the image, or momentary vibration of the optical probe, distorted or invalid image frames may directly enter the analysis model, reducing the accuracy of subsequent flame stability determination.

[0004] Secondly, the model's adaptability to different operating conditions is limited. Existing flame monitoring methods based on visible light imaging mostly rely on single flame features, such as root area, brightness, profile, or flicker frequency. These methods are typically only applicable to specific operating conditions and are easily affected by the equipment's installation location. Even with multi-feature fusion models, their parameters and weights are usually only effective under the designed operating conditions. When actual operating conditions deviate from the preset conditions, the intrinsic relationship between feature parameters and flame stability changes. If the model cannot adaptively adjust, the accuracy of stability determination will significantly decrease.

[0005] Finally, the reliability and predictability of the status output are insufficient. Current research on flame stability measurement and assessment in industrial settings is limited, and existing assessment methods mostly employ a single threshold. When stability indicators approach critical values, even minor fluctuations can lead to frequent state switching, causing repeated alarm signal activation and deactivation. Furthermore, the single threshold method lacks time-series-based trend analysis capabilities, making it difficult to effectively filter out transient interference and unable to provide early warnings of instability trends. This significantly impacts the robustness and predictability of the assessment results under complex operating conditions.

[0006] Therefore, there is an urgent need for a robust flame stability measurement and evaluation method and system for complex operating conditions, which can effectively identify, eliminate and compensate invalid images under complex operating conditions, and dynamically adjust the feature extraction method and fusion weight according to changes in operating conditions. It should have both robustness and predictability in the state output stage, thereby providing accurate and reliable quantitative measurement and evaluation results in a variable industrial operating environment. Summary of the Invention

[0007] The purpose of this invention is to provide a robust flame stability measurement and evaluation method and system for complex operating conditions, covering all key steps from image acquisition, data preprocessing, feature extraction and fusion calculation, stability analysis to result output. It can effectively measure and evaluate the flame stability of various burners and is suitable for a variety of industrial application scenarios.

[0008] To achieve the above objectives, the present invention provides the following solution:

[0009] A robust method for measuring and evaluating flame stability under complex working conditions includes:

[0010] Acquire an image sequence of the burner flame, and preprocess the image sequence;

[0011] Flame feature parameters are extracted from the preprocessed image sequence, and adaptive feature fusion is performed on the flame feature parameters to calculate the flame stability index.

[0012] The flame stability index is analyzed by comparing two threshold hysteresis, and the flame stability measurement and evaluation results and early warning information are output.

[0013] Optionally, preprocessing the image sequence includes:

[0014] The main and secondary ROI collaborative method is used to determine the main flame region and the flame root region of the image sequence;

[0015] Within the flame body area and the flame root area, the validity of an image frame is determined based on the red channel pixel ratio, average brightness and contrast, edge intensity and noise level. Image frames that do not meet the quality requirements are removed. Among them, image frames that do not meet the quality requirements include distorted frames, contaminated frames and invalid frames.

[0016] When an image frame is removed, interpolation compensation is performed based on the feature information of the adjacent valid frames to obtain a complete image sequence.

[0017] The image sequence is obtained by dynamically adjusting the image capture frame rate based on the flame stability index change rate and the flame dominant frequency, combined with Nyquist sampling.

[0018] Optionally, the primary and secondary ROI collaborative method is used to determine the flame body region and root region, including:

[0019] The color flame images in the image sequence are converted into grayscale images. Local adaptive threshold segmentation is performed on the grayscale images to initially identify the flame region. The flame edges are extracted based on gradient operators. Morphological processing is performed on the edge detection results to remove isolated noise and fill small holes, thereby obtaining the processed flame region.

[0020] The minimum bounding rectangle is calculated based on the processed flame region outline, and the rectangle is used as the main ROI region to obtain the main flame region.

[0021] Based on the visibility of the burner outlet and ignition point in the image sequence, three cases are identified for obtaining the secondary ROI region, i.e., obtaining the flame root region. The three cases are: burner outlet visible, burner outlet not visible but ignition point visible, and burner outlet not visible and ignition point not visible.

[0022] Optionally, dynamically adjusting the image capture frame rate of the image sequence based on the flame stability index change rate and the flame dominant frequency, combined with Nyquist sampling, includes:

[0023] The highest frame rate is used when the rate of change of the flame stability index is higher than the upper limit threshold, the lowest frame rate is used when it is lower than the lower limit threshold, and the intermediate frame rate is used when it is between the two thresholds. The highest frame rate is set based on twice the flame main frequency, according to the Nyquist sampling theorem.

[0024] Optionally, the flame characteristic parameters include: geometric parameters, brightness parameters, and thermodynamic parameters, wherein the geometric parameters include ignition point, root area, flaming angle, flame length, and flame area; the brightness parameters include brightness and non-uniformity determined by the contrast characteristics of gray intensity within the flame area; and the thermodynamic parameters include maximum temperature, minimum temperature, average temperature, and flicker frequency.

[0025] Optionally, adaptive feature fusion of the flame feature parameters includes:

[0026] The correlation threshold is determined based on the combustion conditions and observation location to obtain effective feature parameters;

[0027] The flame stability index is calculated based on the normal and abnormal rates of change of the effective feature parameters, combined with adaptive weights. The normal rate of change is the rate of change of feature parameters within a preset confidence interval in the flame feature parameter sequence, while the abnormal rate of change is the rate of change of feature parameters outside the preset confidence interval in the flame feature parameter sequence.

[0028] Optionally, the decision analysis of the flame stability index includes:

[0029] Several intervals are defined based on the average and standard deviation of the flame stability index within a preset time window and the dual-threshold hysteresis comparison factor. Within these intervals, flame stability figures are obtained based on the volatility assessment factor. The volatility assessment factor is determined based on the standard deviation of the flame stability index within the preset time window and the current standard deviation. The average and standard deviation of the flame stability index within the preset time window are updated based on the average value of the flame stability index sequence.

[0030] Based on the flame stability numbers, the flame stability level can be classified.

[0031] Optionally, the output of flame stability measurement and evaluation results and early warning information includes: outputting flame stability index curves, current status levels and trend prediction results, and the output format includes at least one of a visual interface, an audible and visual alarm signal, and a data communication interface.

[0032] This invention also provides a highly robust flame stability measurement and evaluation system for complex working conditions, comprising:

[0033] The image acquisition unit is used to acquire image sequences of the burner flame;

[0034] A data preprocessing unit is used to preprocess the image sequence;

[0035] The core computing unit is used to extract flame feature parameters from the preprocessed image sequence, perform adaptive feature fusion on the flame feature parameters, and calculate the flame stability index.

[0036] The decision analysis unit is used to perform decision analysis on the flame stability index through dual-threshold hysteresis comparison.

[0037] The output unit is used to output flame stability measurement and evaluation results and early warning information.

[0038] The beneficial effects of this invention are as follows:

[0039] (1) By combining ROI region determination with frame quality assessment, distorted, contaminated, and invalid image frames can be effectively identified and removed, improving the effectiveness and stability of input data from the source and providing a reliable foundation for subsequent feature analysis; (2) An interpolation-based image frame compensation strategy is adopted to maintain the continuity of the time series after removing invalid frames, thereby ensuring the accuracy and completeness of feature extraction and trend analysis; (3) Multi-feature adaptive fusion calculation is used, and the fusion weights are dynamically adjusted according to changes in working conditions, so that the evaluation model can still maintain high accuracy and robustness under complex working conditions, avoiding performance degradation of the fixed-weight model under non-design working conditions; (4) Introduction The dual threshold hysteresis comparison method, especially when the stability index is close to the critical value, effectively reduces false alarms and false alarms caused by small fluctuations and improves the stability of the judgment result; (5) Combined with the trend extrapolation prediction algorithm, it can issue an early warning before the flame stability drops significantly, giving operators or control systems time to react, thereby improving the safety of combustion operation; (6) Adaptive state judgment method is adopted to divide the flame state into five levels: stable, relatively stable, unstable, extremely unstable and flameout, and dynamically adjust the judgment threshold according to the working conditions, providing operators with clear state level information to support the rapid adjustment of combustion control strategy.

[0040] In summary, the method and system of this invention can effectively identify and filter input flame images, eliminating invalid or distorted data. Based on this, multiple flame feature parameters are adaptively extracted and fused according to the data preprocessing results, and the adaptability of the model under different working conditions is improved by dynamically adjusting the fusion weights. The obtained flame stability index sequence is subjected to trend extrapolation prediction and dual threshold hysteresis comparison processing, thereby achieving high-precision determination and early warning of flame stability under complex working conditions, and improving the robustness and accuracy of the evaluation results. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart of a highly robust flame stability measurement and evaluation method for complex working conditions, according to an embodiment of the present invention.

[0043] Figure 2 This is a schematic diagram of the flame root region according to an embodiment of the present invention;

[0044] Figure 3 This is a schematic diagram of the flame root region according to an embodiment of the present invention;

[0045] Figure 4 This is a flowchart of the flame stability index measurement method according to an embodiment of the present invention;

[0046] Figure 5 This is a framework diagram of a highly robust flame stability measurement and evaluation system for complex working conditions, according to an embodiment of the present invention.

[0047] Among them, 1-flame area, 2-main ROI area, 3-visible secondary ROI area at burner outlet, 4-burner outlet, 5-ignition point, 6-flame root area, 7-flame geometric centroid, 8-flame edge, 9-invisible secondary ROI area at burner outlet, 10-image edge. Detailed Implementation

[0048] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] like Figure 1 As shown in the figure. This embodiment proposes a highly robust flame stability measurement and evaluation method for complex working conditions, including:

[0051] Acquire an image sequence of the burner flame and preprocess the image sequence;

[0052] Flame feature parameters are extracted from the preprocessed image sequence, and adaptive feature fusion is performed on the flame feature parameters to calculate the flame stability index.

[0053] The decision analysis of the flame stability index is performed by comparing the dual threshold hysteresis, and the flame stability measurement and evaluation results and early warning information are output.

[0054] Further preprocessing of the image sequence includes:

[0055] The main and secondary ROI collaborative method is used to determine the main flame region and the root region of the flame in the image sequence;

[0056] Within the main flame area and the flame root area, the validity of an image frame is determined based on the red channel pixel ratio, average brightness and contrast, edge intensity and noise level. Image frames that do not meet the quality requirements are removed. Among them, image frames that do not meet the quality requirements include distorted frames, contaminated frames and invalid frames.

[0057] When an image frame is removed, interpolation compensation is performed based on the feature information of the adjacent valid frames to obtain a complete image sequence.

[0058] Based on the rate of change of the flame stability index and the flame dominant frequency, the image capture frame rate is dynamically adjusted in conjunction with Nyquist sampling to obtain a preprocessed image sequence.

[0059] Specifically, the image sequence undergoes preprocessing to identify and filter distorted, contaminated, or invalid image frames before subsequent analysis, while maintaining the integrity of the time series. The image preprocessing includes: ① ROI region determination: A primary and secondary ROI collaborative method is used to determine the flame region and root region, reducing background interference and limiting the scope of subsequent analysis; ② Frame quality assessment: Within the determined ROI region, the validity of image frames is determined based on quantitative indicators such as the proportion of red channel pixels, average brightness and contrast, edge intensity, and noise level, discarding image frames that do not meet the quality requirements. Among them, distorted image frames are blurry or deformed images caused by optical probe jitter, focus shift, etc.; contaminated image frames are images with reduced contrast or missing flame information due to smoke, water vapor or window dirt obscuring the image; invalid image frames are images that fail to correctly contain the flame area or whose flame area is severely obscured; ③ Image frame compensation: when an image frame is removed, interpolation compensation is performed based on the feature information of the adjacent valid frames to maintain the integrity of the time series; ④ Dynamic frame rate control: the image shooting frame rate is dynamically adjusted according to the flame stability index change rate and the flame main frequency, combined with the Nyquist sampling principle, to balance the flame change capture accuracy and storage resource utilization.

[0060] Furthermore, the primary and secondary ROI collaborative method is used to determine the main flame region and root region, including:

[0061] The color image of the flame in the image sequence is converted into a grayscale image. Local adaptive threshold segmentation is performed on the grayscale image to initially identify the flame region. The flame edge is extracted based on the gradient operator. Morphological processing is performed on the edge detection results to remove isolated noise and fill small area holes to obtain the processed flame region.

[0062] The minimum bounding rectangle is calculated based on the processed flame region outline, and the rectangle is used as the main ROI region, that is, the main flame region is obtained.

[0063] Based on the visibility of the burner outlet and ignition point in the image sequence, three cases are identified for obtaining the secondary ROI region, i.e., the flame root region. The three cases are: burner outlet visible, burner outlet invisible but ignition point visible, and burner outlet invisible and ignition point invisible.

[0064] Furthermore, the frame rate of image capture is dynamically adjusted based on the rate of change of the flame stability index and the flame dominant frequency, combined with Nyquist sampling, including:

[0065] The highest frame rate is used when the rate of change of the flame stability index is higher than the upper limit threshold, the lowest frame rate is used when it is lower than the lower limit threshold, and the intermediate frame rate is used when it is between the two thresholds. The highest frame rate is set based on twice the flame frequency according to the Nyquist sampling theorem.

[0066] Furthermore, the flame characteristic parameters include: geometric parameters, brightness parameters, and thermodynamic parameters. Among them, the geometric parameters include ignition point, root area, flaming angle, flame length, and flame area; the brightness parameters include brightness and non-uniformity determined by the contrast characteristics of gray intensity within the flame region; and the thermodynamic parameters include maximum temperature, minimum temperature, average temperature, and flicker frequency.

[0067] Furthermore, adaptive feature fusion of flame feature parameters includes:

[0068] The correlation threshold is determined based on the combustion conditions and observation location to obtain effective feature parameters;

[0069] The flame stability index is calculated based on the normal and abnormal rates of change of the effective feature parameters, combined with adaptive weights. The normal rate of change is the rate of change of the feature parameters in the flame feature parameter sequence that are within the preset confidence interval, and the abnormal rate of change is the rate of change of the feature parameters in the flame feature parameter sequence that are outside the preset confidence interval.

[0070] Specifically, feature extraction and fusion calculation includes: extracting multiple flame feature parameters from the effective image sequence obtained through image preprocessing, including at least one of the following parameters: ignition point, root area, flame angle, flame length, flame area, brightness, non-uniformity, highest temperature, lowest temperature, average temperature, and flicker frequency; inputting the feature parameters into an adaptive feature fusion formula, dynamically adjusting the fusion weights according to changes in operating conditions, and calculating the flame stability index.

[0071] Further decision analysis of the flame stability index includes:

[0072] Several intervals are defined based on the average and standard deviation of the flame stability index within a preset time window and the dual-threshold hysteresis comparison factor. Within these intervals, flame stability figures are obtained based on the volatility assessment factor. The volatility assessment factor is determined based on the standard deviation of the flame stability index within the preset time window and the current standard deviation. The average and standard deviation of the flame stability index within the preset time window are updated based on the average value of the flame stability index sequence.

[0073] Based on flame stability data, the flame stability level can be classified.

[0074] Specifically, stability analysis and state determination include: analyzing the original flame stability index sequence, including: ① dual threshold hysteresis comparison: reducing frequent fluctuations of the stability index near the critical value; ② trend extrapolation prediction: predicting the trend of flame stability changes, and triggering an early warning when the prediction result is lower than the set threshold within the warning time window; ③ adaptive flame stability state determination: determining the flame state in real time based on the flame stability index and trend information, and classifying the state into five levels: stable, relatively stable, unstable, extremely unstable, and flameout.

[0075] Furthermore, the output of flame stability measurement and evaluation results and early warning information includes: outputting flame stability index curves, current status levels and trend prediction results, and the output format includes at least one of the following: a visual interface, audible and visual alarm signals, and a data communication interface.

[0076] like Figure 5 As shown, this embodiment also provides a highly robust flame stability measurement and evaluation system for complex working conditions, comprising the following components connected in sequence:

[0077] The image acquisition unit is used to acquire image sequences of the burner flame;

[0078] Specifically, the image acquisition unit is used to acquire image sequences of the burner flame in real time according to preset acquisition parameters (including shooting frame rate, image resolution, exposure time, etc.), and transmit the image sequences to the data preprocessing unit; the image acquisition unit can be composed of an industrial camera, an image acquisition card and its driver, or implemented by an embedded device with image acquisition function.

[0079] The data preprocessing unit is used to preprocess the image sequence;

[0080] Specifically, the data preprocessing unit preprocesses the raw flame images transmitted by the image acquisition unit. It identifies and filters distorted, contaminated, or invalid image frames through ROI region determination and frame quality assessment. After removing invalid frames, it maintains the integrity of the time series through image frame compensation, thereby obtaining a valid image sequence. Distorted image frames are those that are blurred or deformed due to optical probe jitter, focus shift, etc., and cannot accurately reflect the characteristics of the flame. Contaminated image frames are those whose image contrast is reduced or whose flame area information is missing due to smoke, moisture, or window grime obscuring the image. Invalid image frames are those that fail to correctly contain the flame area or whose flame area is severely obscured.

[0081] The data preprocessing unit includes: ① ROI determination subunit: used to determine the flame region and root region using a primary and secondary ROI collaborative method, reducing background interference and limiting the scope of subsequent analysis; ② Frame quality evaluation subunit: used to determine the validity of image frames within the determined ROI region based on quantitative indicators such as the proportion of red channel pixels, average brightness and contrast, edge intensity, and noise level; ③ Image frame compensation subunit: used to perform interpolation compensation based on the valid frames before and after the image frames are removed, in order to maintain the integrity of the time series; ④ Dynamic frame rate control subunit: used to adaptively adjust the image capture frame rate based on the change rate of the flame stability index and the flame main frequency, combined with the Nyquist sampling principle, in order to balance the accuracy of flame change capture and the utilization rate of storage resources.

[0082] The core computing unit is used to extract flame feature parameters from the preprocessed image sequence, perform adaptive feature fusion on the flame feature parameters, and calculate the flame stability index.

[0083] Specifically, the core computing unit includes: a multi-feature extraction subunit for extracting multiple flame feature parameters from the effective image sequence output by the data preprocessing unit, including at least one of the following parameters: ignition point location, root area, flame angle, flame length, flame area, brightness, non-uniformity, highest temperature, lowest temperature, average temperature, and flicker frequency; and a feature fusion calculation subunit for inputting the flame feature parameters into an adaptive feature fusion formula, dynamically adjusting the fusion weights according to changes in operating conditions, and calculating the flame stability index.

[0084] The decision analysis unit is used to perform decision analysis on the flame stability index through dual-threshold hysteresis comparison.

[0085] Specifically, the decision analysis unit is used to analyze and process the flame stability index output by the core calculation unit, including: ① Decision logic subunit: using a dual threshold hysteresis comparison method to reduce frequent fluctuations of the evaluation results near the critical value; ② Trend prediction subunit: using a trend extrapolation prediction method to predict the trend of flame stability changes, and triggering an early warning when the prediction result is below the set threshold within the warning time window; ③ Adaptive state determination subunit: adaptively determining the flame stability state based on the flame stability index, and classifying the state into five levels: stable, relatively stable, unstable, extremely unstable, and flameout.

[0086] The output unit is used to output flame stability measurement and evaluation results and early warning information;

[0087] Specifically, the result output unit is used to output the flame stability measurement and evaluation results and early warning information generated by the decision analysis unit; the output includes the flame stability index curve, the current state level and trend prediction results, and the output form includes at least one of the following: a visual interface, an audible and visual alarm signal, and a data communication interface, which is used to assist operators or the upper control system in adjusting the combustion.

[0088] Example 1:

[0089] like Figure 1 As shown, a robust flame stability measurement and evaluation method for complex working conditions includes:

[0090] Step 1: First, the image acquisition unit continuously acquires a sequence of images of the burner flame according to preset acquisition parameters, including frame rate, image resolution, and exposure time. To ensure stable and clear image data under complex operating conditions such as fuel composition fluctuations, load changes, and high dust levels, the image acquisition unit can be implemented using an industrial camera or an embedded device with image acquisition capabilities. The acquisition parameters can be automatically or manually adjusted according to the operating environment to adapt to different operating conditions.

[0091] Step 2: The acquired image sequence is transmitted to the data preprocessing unit for processing. The data preprocessing includes ROI region determination, frame quality assessment, image frame compensation, and dynamic frame rate control.

[0092] The purpose of ROI (Region of Interest) determination is to accurately identify the flame region and segment its edges within the entire image, providing spatial localization support for subsequent feature parameter extraction. ROI determination employs a primary and secondary ROI collaborative localization method, comprising a primary ROI region and a secondary ROI region.

[0093] The main ROI region is used to cover the main body of the flame. Its determination method is as follows: the color image of the flame is converted to a grayscale image; local adaptive thresholding is performed on the grayscale image to initially identify flame region 1; then, flame edges are extracted based on gradient operators; morphological processing (such as erosion, dilation, and closing operations) is performed on the edge detection results to remove isolated noise and fill small holes. The minimum bounding rectangle is calculated based on the processed flame region contour, and this rectangle serves as the main ROI region 2.

[0094] The secondary ROI region is mainly used for calculating the area of ​​the flame root region 6. Its location and extent are divided into three cases based on the visibility of the burner outlet and ignition point in the image:

[0095] Case 1 (burner outlet visible): such as Figure 2As shown, if the ignition point 5 is visible, use the center of the ignition point area as the starting point; if the ignition point is not visible, use the center of the burner outlet 4 as the starting point. Along the flame propagation direction, measure the distance from the starting point to the geometric centroid 7 of the flame, and draw a reference line parallel to the burner outlet at the 1 / 2 of this distance. The straight line segment between the two intersection points of this reference line and the flame edge 8 is used as the long side of the secondary ROI rectangular frame; draw perpendicular lines from the two endpoints of this straight line segment in the opposite direction of the flame propagation direction to the burner outlet plane, and the length of the perpendicular line segment is used as the short side of the secondary ROI rectangular frame. The rectangular area enclosed by the above long side and short side is the visible secondary ROI area 3 of the burner outlet.

[0096] Case 2 (the burner outlet is not visible but the ignition point is visible): As Figure 3 shown, use the center of the ignition point 5 area as the starting point, measure the distance to the geometric centroid 7 of the flame along the flame propagation direction, and draw a reference line at the 1 / 3 of this distance. The straight line segment between the two intersection points of this reference line and the flame edge 8 is used as the long side of the secondary ROI rectangular frame; draw perpendicular lines from the two endpoints of this straight line segment in the opposite direction of the flame propagation direction to the image edge 10 on the side corresponding to the burner outlet direction in the image, and the length of the perpendicular line segment is used as the short side of the secondary ROI rectangular frame. The rectangular area enclosed by the above long side and short side is the invisible secondary ROI area 9 of the burner outlet.

[0097] Case 3 (the burner outlet is not visible and the ignition point is not visible): The secondary ROI is not automatically extracted and can be determined or marked manually in the image.

[0098] Frame quality assessment is used to determine whether there are distortions, contaminations or invalid situations in the image frame, so as to ensure that the image quality entering the subsequent feature extraction step meets the analysis requirements. A distorted frame refers to an image frame that is blurred or deformed due to the jitter of the optical probe or the focus shift; a contaminated frame refers to an image frame whose contrast is reduced or the flame information is missing due to the occlusion of soot, water vapor or window dirt; an invalid frame refers to an image frame that lacks a flame area or the flame area is severely occluded. The evaluation process adopts a multi-index comprehensive scoring method, specifically including the following quantitative indexes:

[0099] 1) Proportion of the red channel value range of the image: Statistically analyze the red channel values of the flame area in the image and calculate the pixel proportion that satisfies . When is higher than the proportion threshold of 95%, it is determined that this index is qualified. Within this red channel range, there is a good linear relationship between the ratio of the red and blue channel pixels of the pixels and the flame temperature, which is conducive to obtaining the flame temperature.

[0100] 2) Mean brightness and contrast: Convert the image into a grayscale value map I and calculate the average grayscale value within the main ROI area with standard deviation ,

[0101] ;

[0102] ;

[0103] Where A is the number of pixels in the rectangle and the average brightness. Contrast needs to be set within the empirical range of 60–200. The value needs to be greater than the threshold of 20 to ensure that the image brightness is not overexposed and that the layers are distinct, making it easier to extract the flame edges.

[0104] 3) Edge intensity: Based on the proportion of edge pixels in the flame region. ,like A value exceeding the proportion threshold of 5% indicates that sufficient flame edge information can be identified from the flame image, and it can be judged as qualified.

[0105] 4) Noise level: The image noise level is calculated using the local variance method. ,like If the noise level is less than the upper noise threshold, the image noise is considered to meet the requirements. This threshold is set based on the noise baseline of the acquisition device under standard operating conditions plus a tolerance, and can be adjusted according to the operating conditions.

[0106] In this embodiment, the weights of each indicator are... The sum is 1. Calculate the overall score:

[0107] ;

[0108] in The scores are obtained from the normalized values ​​of the red channel ratio, brightness and contrast, edge intensity, and noise level, respectively. Overall Score A frame is considered valid if its score exceeds the total threshold of 90; otherwise, it is discarded. It should be noted that due to the complexity of industrial environments, optical units may be difficult to install in the optimal flame monitoring position, leading to some indicators failing to meet standards. In such cases, the thresholds and weights of the four indicators can be adjusted based on the actual image quality and flame characteristics.

[0109] For images deemed invalid frames, compensation is needed to maintain the continuity of the flame image time series and prevent gaps in the time series from interfering with subsequent feature extraction, trend analysis, and stability determination. The compensation method employs interpolation, calculating the result for each pixel in the missing frame using linear interpolation between the corresponding pixel values ​​of the previous and next frames.

[0110] ;

[0111] in and These are images of the adjacent valid frames before and after the missing frame.

[0112] Dynamic frame rate control is used to adaptively adjust the image acquisition and processing frame rate to balance equipment cost and data storage cost. It utilizes the rate of change of the flame stability index per unit time obtained through feature parameter fusion. Determine whether the shooting frame rate increases or decreases.

[0113] ;

[0114] in, Let be the flame stability index at time t. The time interval used to calculate the rate of change is set by default to the acquisition interval between two adjacent frames of images. It can be manually set according to actual working conditions and data smoothing requirements.

[0115] when Above the upper limit threshold At that time, the highest shooting frame rate is used to capture more details; when Below the lower threshold When using the lowest possible frame rate, save on storage and computational costs; when When the frame rate fluctuates between the upper and lower thresholds, an intermediate frame rate is used to balance capture speed and resource consumption. Considering that the flame spectrum distribution in an industrial furnace already contains most of the combustion information at 50 Hz, according to the Nyquist sampling theorem, a maximum frame rate of 100 Hz is sufficient to meet the engineering requirements for capturing most of the combustion information. This dynamic frame rate adjustment method ensures capture accuracy when the flame state changes rapidly, reduces data volume and processing pressure under stable operating conditions, thereby improving system data storage efficiency and extending the service life of the image acquisition unit.

[0116] Step 3: The effective flame image sequence obtained after data preprocessing is transmitted to the core computing unit, the flame feature parameter sequence is extracted, and the flame feature parameters are fused to obtain the flame stability index. The following section, in conjunction with the appendix... Figure 4 This will be explained. Flame characteristic parameters are divided into three categories: geometric parameters, brightness parameters, and thermodynamic parameters.

[0117] The geometric parameters include ignition point, root area, angle, flame length, and flame area. The ignition point is the absolute distance from the burner outlet to the nearest bright point along the central axis. The root area is the ratio of the number of bright pixels within the sub-ROI region to the total pixels of the bounding box. The angle is the angle between the fitted lines of the bright points on both sides of the combustion direction. The flame length is the absolute distance between the burner outlet and the farthest bright point. The flame area is the ratio of the number of bright pixels in the image to the total number of pixels in the image. The brightness parameters include brightness and non-uniformity. Brightness is determined by the average grayscale value of the flame region in the flame grayscale image, and non-uniformity is the contrast feature of grayscale intensity within the flame region. The thermodynamic parameters include maximum temperature, minimum temperature, average temperature, and flicker frequency. The maximum temperature, minimum temperature, and average temperature are determined by the flame temperature distribution, and the flicker frequency is determined by the weighted average of the frequency components of the average grayscale sequence of the flame determined by the flame grayscale image sequence across the entire frequency range. When some of the feature parameters cannot be extracted from the flame image, only the extractable feature parameters are analyzed and fused.

[0118] Correlation analysis of the rate of change of flame characteristic parameters and flame flicker frequency is used to determine effective parameters that contain sufficient flame stability information. The calculation of the flame stability index requires integrating various effective characteristic parameters over a certain time series length; therefore, the rate of change of a flame characteristic parameter is defined as the ratio of its standard deviation to its mean over that time series length. A high correlation coefficient between the rate of change of a flame characteristic parameter and the flame flicker frequency indicates that the parameter contains sufficient flame stability information under the current operating conditions. It should be clarified that a higher correlation coefficient does not necessarily mean that the characteristic parameter contains more flame stability information. Based on measurement experience, the initial correlation threshold for determining the effectiveness of characteristic parameters is set at 0.6. This threshold can be manually adjusted according to actual operating conditions to accommodate different fuel types, load conditions, and observation angles.

[0119] By analyzing the statistical characteristics of the effective characteristic parameter, the normal and abnormal rates of change of that parameter can be obtained. Normal rate of change ( The percentage of change of characteristic parameters within the 95% confidence interval of the flame characteristic parameter sequence is denoted as . Abnormal percentage of change ( ). ) represents the rate of change of characteristic parameters outside the 95% confidence interval in the flame characteristic parameter sequence, i.e.,

[0120]

[0121] Among them, the unstable factor The definition of

[0122]

[0123]

[0124] In the above formula, As an instability factor, the fluctuation characteristics of the characteristic parameter x outside the 95% confidence interval reflect the sensitivity of this parameter to flame stability; and These are the average values ​​of the larger and smaller portions outside the 95% confidence interval of the feature parameter x sequence, respectively. Let x be the standard deviation of parameter x.

[0125] By integrating the effective characteristic parameters and the normal and abnormal rates of change of flame flicker frequency, the flame stability index (δ) is obtained, i.e.

[0126]

[0127] in , The weights of the feature parameter x(i) are determined by the corresponding instability factor. The corrected value is used to quantify the information contribution of this feature parameter in flame stability measurement. A higher weight value indicates that the feature parameter contains more flame stability information. The number of effective feature parameters, and These represent the weights of the flame flicker frequency and the normal rate of change, respectively.

[0128] Step Four: In actual combustion, even if the flame reaches the optimal stability achievable by the combustion system, the flame stability index calculated through characteristic parameter fusion will always be greater than 0 due to the flame's pulsating characteristics. Furthermore, the flame stability index does not have a fixed theoretical upper limit; its value may increase significantly in a near-extinguished flame state. Based on these characteristics, industrial operators cannot directly judge the flame stability state solely based on the stability index value. Moreover, slight changes in the stability index are usually not significant for combustion control operations, as minor fluctuations in flame stability within industrial furnaces are normal. In view of these problems, this invention introduces a simpler and more intuitive judgment indicator than the stability index, enabling operators to quickly and clearly obtain the flame stability level within the furnace and perform corresponding operations.

[0129] In industrial combustion systems, when the stability of the flame within the furnace meets production requirements and the stability index does not show a significant trend within a certain time window, the average value of the stability index within that window can be used. and standard deviation This serves as the optimal reference value for flame stability under current combustion conditions. The flame stability index may increase or decrease when combustion conditions change. An increase in the flame stability index indicates worsening flame stability. When the flame stability index decreases, a larger standard deviation indicates potentially worse flame stability; a smaller standard deviation indicates improved flame stability, and the reference value should be redefined. Based on this correspondence between the flame stability index and flame stability, the "flame stability number" is further defined as a simplified judgment indicator for industrial applications:

[0130] ;

[0131] in, To avoid repeated state switching caused by frequent fluctuations in the stability index near the critical value, a dual-threshold hysteresis comparison factor is introduced. When the value is between the two thresholds, the original judgment state remains unchanged. The mean value of the flame stability index series is obtained by using a sliding window to further eliminate the influence of transient disturbances. A volatility assessment factor is introduced. Based on the current standard deviation Compared to Divide into grades according to proportions:

[0132] ;

[0133] in, denoted as the standard deviation of the flame stability index. Based on the fluctuations in the flame stability index, the stability of the flame can be further assessed. A flame stability index with larger fluctuations may indicate that the flame is less stable.

[0134] When the flame stability index meets the following conditions, the flame is considered to have reached a more stable state, and the optimal flame stability reference value is redefined:

[0135] 1) Mean of stability index ;

[0136] 2) Volatility Standard Deviation (That is, there is no significant fluctuation).

[0137] The formula for updating the reference value is:

[0138]

[0139] in, The length of the flame stability index sequence required to redetermine the optimal flame stability.

[0140] Converting the flame stability index into flame stability numbers allows for the scientific classification of flame stability levels.

[0141] 1) Stable flame, ∈ [8, 10];

[0142] 2) The flame is relatively stable. ∈ [6, 8);

[0143] 3) The flame is unstable. ∈ [3, 6);

[0144] 4) The flame is extremely unstable. ∈ (0, 3);

[0145] 5) The flame went out. = 0.

[0146] A smaller stability number indicates that the flame is closer to an unstable or extinguished state; a larger stability number indicates better flame stability. This method only requires manual initial parameter setting upon first use; subsequent judgment thresholds can be adaptively adjusted based on real-time flame status data.

[0147] To achieve early warning, this invention introduces a trend extrapolation prediction function into the decision analysis. This function performs trend fitting based on the flame stability index sequence, for example, using a linear regression model to obtain the slope of the fitted curve. When the slope is consistently positive and its absolute value exceeds a preset threshold, it is determined that the flame stability is showing a continuous downward trend, triggering an early warning signal in advance.

[0148] Step 5: After determining the flame stability, the system will provide the evaluation results and early warning information through the result output module, so that operators or the upper control system can promptly obtain flame operating status information and take corresponding measures. In this embodiment, the result output includes one or more of the following forms:

[0149] The monitoring terminal displays the flame stability index curve, current status level (stable, relatively stable, unstable, extremely unstable, flameout) and trend prediction results in real time. Different status levels can be distinguished using color coding; for example, green indicates stable, blue indicates relatively stable, yellow indicates unstable, red indicates extremely unstable, and gray indicates flameout. Historical data playback is provided to facilitate operators' analysis of flame status changes and their causes.

[0150] Audible and visual alarm: When the status is determined to be "unstable", "extremely unstable", or "fire off", the audible and visual alarm device will be triggered to issue a prompt or warning signal. The alarm mode can be adjusted according to the status level. For example, only a yellow indicator light will be emitted in a relatively stable state, an intermittent buzzer will be emitted in an unstable state, and a continuous buzzer accompanied by a flashing red light will be emitted in an extremely unstable state or when the fire is off.

[0151] Data communication output: Evaluation results, flame stability index values, trend prediction information, etc., are transmitted in real time to a host computer, DCS (Distributed Control System), or remote monitoring center via industrial communication interfaces (such as Modbus, Profibus, Ethernet, etc.). Data uploads are supported at fixed time intervals or triggered by status change events.

[0152] Data storage and archiving: The evaluation results and related images and feature parameters are stored in a local database or cloud server for long-term historical analysis, algorithm optimization, and operation and maintenance records.

[0153] See Figure 5 This embodiment provides a highly robust flame stability measurement and evaluation system for complex working conditions, comprising an image acquisition unit, a data preprocessing unit, a core computing unit, a decision analysis unit, and a result output unit connected in sequence. The functions of each unit are as follows:

[0154] Image acquisition unit: Used to perform the image acquisition operation as described in step one of the method embodiments, including acquiring a sequence of flame images according to preset parameters (shooting frame rate, image resolution, exposure time, etc.) and transmitting the image data to the data preprocessing unit. This unit may consist of an industrial camera, an image acquisition card and its driver, or be implemented by an embedded device with image acquisition capabilities.

[0155] Data preprocessing unit: Used to perform data preprocessing operations as described in step two of the method embodiment, including functions such as ROI region determination, frame quality assessment, image frame compensation, and dynamic frame rate adjustment. ROI region determination can adopt a primary and secondary ROI collaborative localization method to clearly identify the flame region and root region in the image. Frame quality assessment can be based on quantitative indicators such as red channel ratio, brightness and contrast, edge intensity, and noise level. Image frame compensation can adopt an interpolation compensation method to restore the continuity of the time series after removing invalid frames. Dynamic frame rate adjustment can adaptively adjust the frame rate by combining the flame flicker frequency and the Nyquist sampling principle to balance capture accuracy and storage resource utilization.

[0156] The core computing unit is used to perform the flame feature extraction and feature parameter fusion calculation operations as described in step three of the method embodiment. Geometric parameters, brightness parameters, and thermodynamic parameters are extracted from the effective image sequence. Based on the quantitative relationship and statistical analysis characteristics between each feature parameter and flame stability, effective parameters containing sufficient flame stability information are fused to obtain the flame stability index, generating a flame stability index sequence.

[0157] Decision Analysis Unit: Used to perform the decision analysis operations as described in step four of the method embodiment, including dual-threshold hysteresis comparison, trend extrapolation prediction, and flame stability level determination functions. The flame state is divided into five levels: stable, relatively stable, unstable, extremely unstable, and flameout, and adaptive adjustment is supported.

[0158] Result output unit: used to perform the result output operation as described in step five of the method embodiment, including visual interface display, audible and visual alarm, data communication output and historical data storage, etc., to provide real-time evaluation results and historical analysis data to operators and the upper system.

[0159] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A robust method for measuring and evaluating flame stability under complex working conditions, characterized in that, include: Acquire an image sequence of the burner flame, and preprocess the image sequence; Flame feature parameters are extracted from the preprocessed image sequence, and adaptive feature fusion is performed on the flame feature parameters to calculate the flame stability index. The flame stability index is analyzed by comparing two threshold hysteresis, and the flame stability measurement and evaluation results and early warning information are output.

2. The method for measuring and evaluating the robust flame stability under complex working conditions according to claim 1, characterized in that, Preprocessing the image sequence includes: The main and secondary ROI collaborative method is used to determine the main flame region and the flame root region of the image sequence; Within the flame body area and the flame root area, the validity of an image frame is determined based on the red channel pixel ratio, average brightness and contrast, edge intensity and noise level. Image frames that do not meet the quality requirements are removed. Among them, image frames that do not meet the quality requirements include distorted frames, contaminated frames and invalid frames. When an image frame is removed, interpolation compensation is performed based on the feature information of the adjacent valid frames to obtain a complete image sequence. The image sequence is obtained by dynamically adjusting the image capture frame rate based on the flame stability index change rate and the flame dominant frequency, combined with Nyquist sampling.

3. The method for measuring and evaluating high-robust flame stability under complex working conditions according to claim 2, characterized in that, The primary and secondary ROI collaborative method was used to determine the flame body region and root region, including: The color flame images in the image sequence are converted into grayscale images. Local adaptive threshold segmentation is performed on the grayscale images to initially identify the flame region. The flame edges are extracted based on gradient operators. Morphological processing is performed on the edge detection results to remove isolated noise and fill small holes, thereby obtaining the processed flame region. The minimum bounding rectangle is calculated based on the processed flame region outline, and the rectangle is used as the main ROI region to obtain the main flame region. Based on the visibility of the burner outlet and ignition point in the image sequence, three cases are identified for obtaining the secondary ROI region, i.e., obtaining the flame root region. The three cases are: burner outlet visible, burner outlet not visible but ignition point visible, and burner outlet not visible and ignition point not visible.

4. The method for measuring and evaluating the robust flame stability under complex working conditions according to claim 2, characterized in that, The image sequence is dynamically adjusted based on the rate of change of the flame stability index and the flame dominant frequency, combined with Nyquist sampling, including: The highest frame rate is used when the rate of change of the flame stability index is higher than the upper limit threshold, the lowest frame rate is used when it is lower than the lower limit threshold, and the intermediate frame rate is used when it is between the two thresholds. The highest frame rate is set based on twice the flame main frequency, according to the Nyquist sampling theorem.

5. The method for measuring and evaluating the robust flame stability under complex working conditions according to claim 1, characterized in that, The flame characteristic parameters include: geometric parameters, brightness parameters, and thermodynamic parameters. The geometric parameters include ignition point, root area, flare angle, flame length, and flame area. The brightness parameters include brightness and non-uniformity determined by the contrast characteristics of gray intensity within the flame area. The thermodynamic parameters include maximum temperature, minimum temperature, average temperature, and flicker frequency.

6. The method for measuring and evaluating the robust flame stability under complex working conditions according to claim 1, characterized in that, Adaptive feature fusion of the flame feature parameters includes: The correlation threshold is determined based on the combustion conditions and observation location to obtain effective feature parameters; The flame stability index is calculated based on the normal and abnormal rates of change of the effective feature parameters, combined with adaptive weights. The normal rate of change is the rate of change of feature parameters within a preset confidence interval in the flame feature parameter sequence, while the abnormal rate of change is the rate of change of feature parameters outside the preset confidence interval in the flame feature parameter sequence.

7. The method for measuring and evaluating the robust flame stability under complex working conditions according to claim 1, characterized in that, Decision analysis of the flame stability index includes: Several intervals are defined based on the average and standard deviation of the flame stability index within a preset time window and the dual-threshold hysteresis comparison factor. Within these intervals, flame stability figures are obtained based on the volatility assessment factor. The volatility assessment factor is determined based on the standard deviation of the flame stability index within the preset time window and the current standard deviation. The average and standard deviation of the flame stability index within the preset time window are updated based on the average value of the flame stability index sequence. Based on the flame stability numbers, the flame stability level can be classified.

8. The method for measuring and evaluating the robust flame stability under complex working conditions according to claim 1, characterized in that, The output of flame stability measurement and evaluation results and early warning information includes: output flame stability index curve, current status level and trend prediction results, and the output format includes at least one of the following: visual interface, audible and visual alarm signal, and data communication interface.

9. A highly robust flame stability measurement and evaluation system for complex working conditions, used to implement the method described in any one of claims 1-8, characterized in that, include: The image acquisition unit is used to acquire image sequences of the burner flame; A data preprocessing unit is used to preprocess the image sequence; The core computing unit is used to extract flame feature parameters from the preprocessed image sequence, perform adaptive feature fusion on the flame feature parameters, and calculate the flame stability index. The decision analysis unit is used to perform decision analysis on the flame stability index through dual-threshold hysteresis comparison. The output unit is used to output flame stability measurement and evaluation results and early warning information.

Images