Real-time detection method and system for combustion emissions based on methanol fuel

Abstract

The present invention relates to the technical field of component detection, and discloses a real-time detection method and system for combustion emissions based on methanol fuel, which are used to improve the accuracy of real-time detection of combustion emissions based on methanol fuel. The method comprises: collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas, obtaining filtered gas, and extracting particle diameter data of the particulate matter; performing spectral analysis on the filtered gas to obtain a corresponding spectrum; performing gas component analysis on the filtered gas according to the spectrum to obtain gas component data; performing fuel purity analysis on the methanol fuel according to the gas component data and the particle diameter of the particulate matter to obtain fuel purity data; performing fuel component ratio analysis on the methanol fuel according to the fuel purity data to obtain target fuel component ratio data.

Description

Technical Field

[0001] The present invention relates to the technical field of component detection, and in particular to a real-time detection method and system for combustion emissions based on methanol fuel. Background Art

[0002] In the prior art, the detection of combustion emissions mainly relies on traditional laboratory analysis methods. By collecting exhaust gas samples after combustion and analyzing them in the laboratory, the composition and concentration of the emissions can be obtained. These methods include gas chromatography, mass spectrometry, and infrared spectroscopy, which can provide relatively accurate detection results. However, traditional methods usually have problems such as long detection cycle, complex operation, and high cost, making it difficult to achieve real-time online monitoring.

[0003] The shortcomings of existing technologies in real-time monitoring are mainly reflected in the following aspects. First, the traditional laboratory analysis method collects and transports gas samples, which cannot reflect the emission changes during the combustion process in real time. Secondly, due to the limitations of the detection equipment, it is difficult to analyze multiple gas components and the physical properties of particulate matter at the same time, resulting in incomplete detection data. In addition, the existing technology lacks a comprehensive analysis of the diameter and composition of particulate matter, and cannot accurately evaluate the purity and combustion efficiency of the fuel. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a method and system for real-time detection of combustion emissions based on methanol fuel, which are used to improve the accuracy of real-time detection of combustion emissions based on methanol fuel.

[0005] The present invention provides a real-time detection method for combustion emissions based on methanol fuel, comprising: collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas to obtain filtered gas, and extracting particle diameter data of the particulate matter; performing spectral analysis on the filtered gas to obtain a corresponding spectrum graph; performing gas composition analysis on the filtered gas according to the spectrum graph to obtain gas composition data; performing fuel purity analysis on the methanol fuel according to the gas composition data and the particle diameter of the particulate matter to obtain fuel purity data; performing fuel composition ratio analysis on the methanol fuel according to the fuel purity data to obtain target fuel composition ratio data.

[0006] In the present invention, the steps of collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas, obtaining filtered gas, and extracting particle diameter data of the particulate matter include: collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas, obtaining filtered gas; collecting particle images of the particulate matter, performing grayscale conversion on the particle image, obtaining a grayscale particle image; performing image denoising on the grayscale particle image by a Gaussian filtering algorithm, obtaining a denoised particle image; performing particle diameter analysis on the denoised particle image, obtaining particle diameter data of the particulate matter.

[0007] In the present invention, the step of performing particle diameter analysis on the denoised particle image to obtain particle diameter data of the particles includes: performing grayscale gradient analysis on the denoised particle image to obtain grayscale gradient data corresponding to the denoised particle image; performing edge detection on the grayscale particle image through the grayscale gradient data to obtain particle contour data corresponding to the grayscale particle image; performing contour smoothing on the particle contour data corresponding to the grayscale particle image to obtain smoothed contour data corresponding to the grayscale particle image; performing minimum circumscribed circle analysis on the smoothed contour data to obtain multiple minimum circumscribed circles; extracting the diameter of each minimum circumscribed circle, and using the diameter of each minimum circumscribed circle as the particle diameter data of the particles.

[0008] In the present invention, the step of performing spectral analysis on the filtered gas to obtain a corresponding spectrum graph includes: irradiating the filtered gas through a preset infrared light source, and collecting a first absorption spectrum signal in real time; extracting absorption peak characteristics of the first absorption spectrum signal to obtain a first absorption peak characteristic corresponding to the first absorption spectrum signal; generating a first spectrum graph according to the first absorption peak characteristic, and irradiating the filtered gas through a preset ultraviolet light source, and collecting a second absorption spectrum signal in real time; extracting absorption peak characteristics of the second absorption spectrum signal to obtain a second absorption peak characteristic corresponding to the second absorption spectrum signal; generating a second spectrum graph according to the second absorption peak characteristic, and performing spectrum fusion on the first spectrum graph and the second spectrum graph to obtain the spectrum graph.

[0009] In the present invention, the step of performing gas composition analysis on the filtered gas according to the spectrogram to obtain gas composition data includes: performing image segmentation on the spectrogram to obtain multiple sub-spectrograms; inputting the multiple sub-spectrograms into the convolution layer of a convolutional neural network to extract local features to obtain a local feature set of each sub-spectrogram, and the convolutional neural network includes: a convolution layer, a pooling layer, and a fully connected layer; inputting the local feature set of each sub-spectrogram into the pooling layer for dimensionality reduction to obtain a reduced dimensionality feature set of each sub-spectrogram; inputting the reduced dimensionality feature set of each sub-spectrogram into the fully connected layer for feature fusion to obtain a fused vector set; Perform gas component mapping on each fusion vector in the fusion vector set to obtain the gas component corresponding to each fusion vector; perform absorption peak intensity analysis on the first absorption peak feature to obtain the first absorption peak intensity; perform absorption peak intensity analysis on the second absorption peak feature to obtain the second absorption peak intensity; perform component concentration analysis on the gas component corresponding to each fusion vector according to the first absorption peak intensity and the second absorption peak intensity to obtain component concentration data of the gas component corresponding to each fusion vector; merge the gas component corresponding to each fusion vector and the component concentration data of the gas component corresponding to each fusion vector into the gas composition data.

[0010] In the present invention, the step of performing fuel purity analysis on the methanol fuel according to the gas composition data and the particle diameter of the particulate matter to obtain fuel purity data includes: performing data traversal on the gas composition data to obtain each gas component and component concentration data of each gas component; generating initial proportion data of each gas component according to each gas component and component concentration data of each gas component; performing proportion analysis on the particulate matter according to the particle diameter of the particulate matter to obtain particulate matter proportion data; correcting the initial proportion data according to the particulate matter proportion data to obtain target proportion data, and performing fuel purity analysis on the methanol fuel according to the target proportion data to obtain the fuel purity data.

[0011] In the present invention, the step of performing a fuel composition ratio analysis on the methanol fuel according to the fuel purity data to obtain target fuel composition ratio data includes: performing a harmful substance analysis on the target ratio data to obtain harmful substance proportion data; performing a purity correction value analysis on the methanol fuel according to the harmful substance proportion data to obtain a target purity correction value; correcting the fuel purity data according to the purity correction value to obtain a target fuel purity; performing a fuel composition ratio analysis on the methanol fuel according to the target fuel purity to obtain the target fuel composition ratio data.

[0012] The present invention also provides a real-time detection system for combustion emissions based on methanol fuel, comprising:

[0013] A gas filtering module, used for collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas to obtain filtered gas, and extracting particle diameter data of the particulate matter;

[0014] A spectrum analysis module, used for performing spectrum analysis on the filtered gas to obtain a corresponding spectrum graph;

[0015] A component analysis module, used for performing gas component analysis on the filtered gas according to the spectrum to obtain gas component data;

[0016] A purity analysis module, used for performing fuel purity analysis on the methanol fuel according to the gas composition data and the particle diameter of the particulate matter to obtain fuel purity data;

[0017] The ratio analysis module is used to perform fuel composition ratio analysis on the methanol fuel according to the fuel purity data to obtain target fuel composition ratio data.

[0018] In the technical solution provided by the present invention, by collecting the exhaust gas of methanol fuel and filtering the particulate matter, the particulate impurities in the gas can be effectively removed, and the particle diameter data of the particulate matter can be extracted. The particle image is denoised using the Gaussian filtering algorithm to ensure the accuracy and reliability of the particle diameter data. In addition, the filtered gas is analyzed by infrared spectrum and ultraviolet spectrum, and the concentrations of water and carbon dioxide, as well as methanol and formaldehyde are detected respectively to generate corresponding spectrograms. The convolutional neural network is used to perform image segmentation and feature extraction on the spectrogram, and the convolution layer, pooling layer and fully connected layer are used to perform local feature extraction, dimension reduction and feature fusion on multiple sub-spectral graphs, and finally the concentration data of each gas component is obtained. Not only the detection accuracy and efficiency are improved, but also the ability of simultaneous detection of multiple components is realized. Based on the combination of spectrogram and image processing technology, the components and particle characteristics in the exhaust gas can be comprehensively analyzed. By performing intensity analysis on the first absorption peak feature and the second absorption peak feature in the spectrogram, the fuel purity analysis is performed according to the gas composition data and the particle diameter data of the particulate matter to obtain the fuel purity data. By traversing the gas composition data, the initial proportion data of each gas component is obtained; then the proportion analysis is performed according to the particle diameter of the particulate matter to obtain the particulate matter proportion data; then the initial proportion data is corrected according to the particulate matter proportion data to obtain the target proportion data, and the fuel purity analysis of methanol fuel is performed accordingly to obtain the fuel purity data. It can comprehensively consider the gas composition and particulate matter characteristics to provide a more accurate fuel purity assessment. The target proportion data is analyzed for harmful substances, and the proportion data of harmful substances is calculated; the purity correction value of methanol fuel is analyzed according to the harmful substance proportion data to obtain the target purity correction value; then the fuel purity data is corrected according to the purity correction value to obtain the target fuel purity; finally, the fuel composition ratio of methanol fuel is analyzed according to the target fuel purity to obtain the target fuel composition ratio data. It can effectively identify and correct the harmful substance components in the fuel. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0020] Figure 1 This is a flow chart of a method for real-time detection of combustion emissions based on methanol fuel in an embodiment of the present invention;

[0021] Figure 2 Schematic diagram of a grayscale particle image according to an embodiment of the present invention.

[0022] Figure 3 Schematic diagram of a denoised particle image according to an embodiment of the present invention.

[0023] Figure 4 Schematic diagram of a first spectrum diagram according to an embodiment of the present invention.

[0024] Figure 5 is a schematic diagram of a second spectrum diagram according to an embodiment of the present invention.

[0025] Figure 6 Schematic diagram of a real-time detection system for combustion emissions based on methanol fuel in an embodiment of the present invention. DETAILED DESCRIPTION

[0026] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0028] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0029] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 , Figure 1 is a flow chart of a method for real-time detection of combustion emissions based on methanol fuel according to an embodiment of the present invention, such as Figure 1 As shown, the following steps are included:

[0030] S101, collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas to obtain filtered gas, and extracting particle diameter data of the particulate matter;

[0031] S102, performing spectral analysis on the filtered gas to obtain a corresponding spectrum graph;

[0032] S103, performing gas composition analysis on the filtered gas according to the spectrum to obtain gas composition data;

[0033] S104, performing fuel purity analysis on the methanol fuel according to the gas composition data and the particle diameter of the particulate matter to obtain fuel purity data;

[0034] S105. Perform fuel composition ratio analysis on the methanol fuel according to the fuel purity data to obtain target fuel composition ratio data.

[0035] It should be noted that the exhaust gas generated after the combustion of methanol fuel is initially filtered to remove large particles. The accuracy of gas composition detection is improved. The filtered gas is subjected to spectral analysis. At the same time, the particles are collected and their images are collected through a microscope or a high-resolution camera. They are processed by an image processing algorithm. First, grayscale conversion is performed to simplify the processing steps. Then, the Gaussian filter algorithm is used to perform image denoising to remove noise in the image and obtain a clear particle image. Next, the processed particle image is subjected to grayscale gradient analysis to obtain grayscale gradient data. The edge of the particle image is detected by the grayscale gradient data to obtain the contour data of the particle. In order to ensure the smoothness and accuracy of the contour, the contour data is further smoothed to obtain smooth contour data. Finally, the minimum circumscribed circle analysis is performed to obtain multiple minimum circumscribed circles, and the diameter of each minimum circumscribed circle is extracted as the particle diameter data of the particle.

[0036] After the filtered gas is obtained, the gas is spectrally analyzed by an infrared light source and an ultraviolet light source respectively. The specific steps are as follows: the filtered gas is irradiated by a preset infrared light source, and the first absorption spectrum signal is collected in real time; the absorption peak feature of the first absorption spectrum signal is extracted to obtain the first absorption peak feature corresponding to the first absorption spectrum signal, and a first spectrum graph is generated. This spectrum graph is mainly used to detect the concentration of water and carbon dioxide. Similarly, the filtered gas is irradiated by a preset ultraviolet light source, and the second absorption spectrum signal is collected in real time; the absorption peak feature of the second absorption spectrum signal is extracted to obtain the second absorption peak feature corresponding to the second absorption spectrum signal, and a second spectrum graph is generated, which is used to detect the concentration of methanol and formaldehyde. Next, the first spectrum graph and the second spectrum graph are spectrally fused to obtain a comprehensive spectrum graph that fully reflects the composition information of the filtered gas.

[0037] The convolutional neural network (CNN) is used to perform image segmentation on the spectrum and divide the spectrum into multiple sub-spectra. Each sub-spectra is input into the convolution layer of the convolutional neural network for local feature extraction to obtain the local feature set of each sub-spectra. The convolutional neural network includes a convolutional layer, a pooling layer, and a fully connected layer. After the convolution layer extracts the local features, the pooling layer performs dimensionality reduction processing on the local feature set to obtain a reduced dimensionality feature set, which is further input into the fully connected layer for feature fusion to obtain a fusion vector set. Next, gas component mapping is performed on each fusion vector to obtain the gas component corresponding to each vector. The absorption peak intensity of the first absorption peak feature is analyzed to obtain the first absorption peak intensity, and the absorption peak intensity of the second absorption peak feature is analyzed to obtain the second absorption peak intensity. According to these absorption peak intensities, the component concentration of the gas component corresponding to each fusion vector is analyzed respectively, and finally the component concentration data of the gas component corresponding to each fusion vector is obtained, and these data are merged into gas composition data.

[0038] On this basis, the fuel purity analysis of methanol fuel is carried out according to the gas composition data and the particle diameter of the particulate matter. The specific steps include traversing the gas composition data to obtain each gas component and its concentration data; generating the initial proportion data of each gas component based on these data. At the same time, the proportion analysis is carried out according to the particle diameter data of the particulate matter to obtain the particulate matter proportion data. The initial proportion data is corrected using the particulate matter proportion data to obtain the target proportion data. Based on this, the fuel purity analysis is carried out to obtain the fuel purity data.

[0039] For example, suppose that the gas composition after methanol combustion is detected to include 20% water, 30% carbon dioxide, 25% methanol and 25% formaldehyde. Through the analysis of particulate matter, the average particle diameter is 2 microns, accounting for 10%. Based on these data, the initial ratio is corrected, the corrected ratio is calculated, and finally the fuel purity data is obtained. Finally, the fuel composition ratio analysis is performed based on the fuel purity data to obtain the target fuel composition ratio data, including the analysis of harmful substances on the target ratio data, the calculation of the proportion of harmful substances, and the purity correction of methanol fuel based on these data to obtain the target purity correction value, and then the fuel purity data is corrected according to the correction value to finally obtain the target fuel purity. Then, the fuel composition ratio analysis of methanol fuel is performed according to the target fuel purity to determine the optimal fuel composition ratio to achieve the goal of optimizing combustion and reducing harmful emissions.

[0040] By executing the above steps, the particulate impurities in the gas can be effectively removed by collecting the exhaust gas of methanol fuel and filtering the particulate matter, and the particle diameter data of the particulate matter can be extracted. The particle image is denoised by using the Gaussian filter algorithm to ensure the accuracy and reliability of the particle diameter data. In addition, the filtered gas is analyzed by infrared spectrum and ultraviolet spectrum through spectral analysis technology to detect the concentrations of water and carbon dioxide, as well as methanol and formaldehyde, respectively, and generate corresponding spectra. The convolutional neural network is used to perform image segmentation and feature extraction on the spectra, and the convolution layer, pooling layer and fully connected layer are used to perform local feature extraction, dimension reduction and feature fusion on multiple sub-spectra, and finally the concentration data of each gas component is obtained. Not only the detection accuracy and efficiency are improved, but also the ability of simultaneous detection of multiple components is realized. Based on the combination of spectra and image processing technology, the components and particle characteristics in the exhaust gas can be comprehensively analyzed. By performing intensity analysis on the first absorption peak feature and the second absorption peak feature in the spectra, the fuel purity analysis is performed according to the gas composition data and the particle diameter data of the particulate matter to obtain the fuel purity data. By traversing the gas composition data, the initial proportion data of each gas component is obtained; then the proportion analysis is performed according to the particle diameter of the particulate matter to obtain the particulate matter proportion data; then the initial proportion data is corrected according to the particulate matter proportion data to obtain the target proportion data, and the fuel purity analysis of methanol fuel is performed accordingly to obtain the fuel purity data. It can comprehensively consider the gas composition and particulate matter characteristics to provide a more accurate fuel purity assessment. The target proportion data is analyzed for harmful substances, and the proportion data of harmful substances is calculated; the purity correction value of methanol fuel is analyzed according to the harmful substance proportion data to obtain the target purity correction value; then the fuel purity data is corrected according to the purity correction value to obtain the target fuel purity; finally, the fuel composition ratio of methanol fuel is analyzed according to the target fuel purity to obtain the target fuel composition ratio data. It can effectively identify and correct the harmful substance components in the fuel.

[0041] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0042] (1) Collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas, and obtaining filtered gas;

[0043] (2) Collecting particle images of the particulate matter and performing grayscale conversion on the particle images to obtain grayscale particle images;

[0044] (3) Perform image denoising on the grayscale particle image using a Gaussian filtering algorithm to obtain a denoised particle image;

[0045] (4) Perform particle diameter analysis on the denoised particle image to obtain particle diameter data of the particles.

[0046] Specifically, in the real-time detection process of methanol fuel combustion emissions, the exhaust gas after the combustion of methanol fuel is first collected, and the particulate matter in the exhaust gas is filtered through a high-efficiency filter to obtain filtered gas. This step uses continuous sampling technology to collect gas samples from the exhaust port of the combustion device, and then removes the particulate matter in the gas through a multi-stage filtration system. The filtration system includes a primary filter and a precision filter. The primary filter is used to remove larger particles, and the precision filter further removes tiny particles to ensure that a high-purity filtered gas sample is obtained. Collect particle images of the filtered particles. This step uses a high-resolution microscope or a digital camera to perform microscopic imaging of the collected particles. In order to improve the efficiency of image processing, the collected particle image is gray-converted to convert the color image into a gray-scale image. In the gray-scale conversion process, the weighted average method is usually used to weight the values ​​of the three RGB channels according to the sensitivity of the human eye to different colors. The conversion formula is: Gray = 0.299R + 0.587G + 0.114B, where R, G, and B represent the pixel values ​​of the red, green, and blue channels respectively, and Gray is the grayscale value after conversion. This conversion method retains the main structure and brightness information of the image while reducing the amount of data, which is beneficial to subsequent image processing.

[0047] like Figure 2 The figure shows a grayscale particle image. In order to improve the image quality, the grayscale particle image is denoised. Here, the Gaussian filter algorithm is used. The algorithm uses a two-dimensional Gaussian function as the convolution kernel to smooth the image. The kernel function of the Gaussian filter is:

[0048] ;

[0049] Where (x, y) represents the pixel coordinates and σ is the standard deviation of the Gaussian distribution. By adjusting the σ value, the intensity of the filter can be controlled. Gaussian filtering has a good inhibitory effect on high-frequency noise in the image, while retaining the edge information of the image and avoiding excessive blurring of the image. After applying Gaussian filtering, a denoised particle image is obtained, such as Figure 3 The figure shows a denoised particle image, which has a higher signal-to-noise ratio and is beneficial for subsequent particle diameter analysis.

[0050] The particle diameter analysis is performed on the denoised particle image to obtain the particle diameter data of the particles. This step first performs edge detection on the image and uses the Canny edge detection algorithm to identify the contours of the particles. The Canny algorithm includes steps such as Gaussian filtering, calculating gradient amplitude and direction, non-maximum suppression, dual threshold detection, and edge tracking. After edge detection, the detected contour is subjected to shape analysis to calculate the equivalent circular diameter of each particle. The calculation formula for the equivalent circular diameter is:

[0051] ;

[0052] Where D is the equivalent circle diameter and A is the projected area of ​​the particle. In this way, the diameter data of each particle can be obtained, which reflects the size distribution of the particles.

[0053] For example, in a methanol fuel combustion emission test, a gas sample was collected continuously from the combustion device exhaust port for 10 minutes, and a pure gas sample was obtained after filtering through a multi-stage filtration system. At the same time, about 500 particles were collected, and these particles were imaged using a high-resolution microscope to obtain a color image with an original resolution of 2048x1536 pixels. After converting the color image to a grayscale image, the image data volume was reduced by about 66%. The Gaussian filter algorithm was applied for denoising, and σ=1.5 was selected, and the convolution kernel size was 5x5 pixels. After denoising, the signal-to-noise ratio of the image was increased from the original 20dB to 28dB. Finally, 486 effective particles were identified through Canny edge detection and shape analysis, and the equivalent circular diameters of these particles were calculated to be distributed between 0.5μm and 5μm, with an average diameter of 2.3μm.

[0054] In a specific embodiment, the process of performing the particle diameter analysis step on the denoised particle image may specifically include the following steps:

[0055] (1) Perform grayscale gradient analysis on the denoised particle image to obtain grayscale gradient data corresponding to the denoised particle image;

[0056] (2) Perform edge detection on the grayscale particle image through grayscale gradient data to obtain particle contour data corresponding to the grayscale particle image;

[0057] (3) performing contour smoothing on the particle contour data corresponding to the grayscale particle image to obtain smoothed contour data corresponding to the grayscale particle image;

[0058] (4) Performing minimum circumscribed circle analysis on the smooth contour data to obtain multiple minimum circumscribed circles;

[0059] (5) Extract the diameter of each minimum circumscribed circle and use the diameter of each minimum circumscribed circle as the particle diameter data of the particulate matter.

[0060] Specifically, in the real-time detection of methanol fuel combustion emissions, in-depth analysis of denoised particle images is a crucial step. First, grayscale gradient analysis is performed to calculate the grayscale change rate of each pixel in the image. The grayscale gradient reflects the degree of brightness change in the image and plays an important role in identifying the edges of particles. Grayscale gradient analysis uses the Sobel operator to calculate the grayscale changes in the horizontal and vertical directions respectively. For each pixel (x, y) in the image, the horizontal gradient Gx and the vertical gradient Gy are calculated as follows:

[0061] ;

[0062] Where I(x,y) represents the grayscale value of the pixel (x,y). The gradient amplitude G and direction θ are:

[0063] ;

[0064] ;

[0065] The calculated grayscale gradient data is used to perform edge detection and identify the contour of particles. The Canny algorithm is used for edge detection, which includes Gaussian filtering, calculation of gradient amplitude and direction, non-maximum suppression, dual threshold detection and edge tracking. The Canny algorithm can effectively detect the real edge while suppressing noise, and is particularly suitable for the extraction of particle contours.

[0066] The particle contour data obtained after edge detection may have some discontinuous or jagged edges, so contour smoothing is required. Contour smoothing uses a combination of Gaussian filtering and morphological operations. First, apply Gaussian filtering to the contour data to smooth small fluctuations on the contour. The Gaussian filter kernel function is:

[0067] ;

[0068] Where σ is the standard deviation of the Gaussian distribution, which controls the degree of smoothness. Then, the contour is further smoothed by the morphological closing operation, which includes two steps: dilation and erosion, which can fill small holes and gaps in the contour to make the contour smoother and more continuous.

[0069] The minimum circumscribed circle analysis is performed on the smoothed contour data. This step aims to approximate irregular particles into circles to facilitate the calculation of diameters. The minimum circumscribed circle analysis uses a randomized incremental algorithm. The basic idea of ​​the algorithm is to randomly select points on the contour and gradually construct and update the minimum circle that surrounds all points. The time complexity of the algorithm is O(n), where n is the number of points on the contour.

[0070] Finally, the diameter of each minimum circumscribed circle is extracted as the particle diameter data of the particles. The diameter calculation formula is:

[0071] ;

[0072] Where D is the diameter and A is the area of ​​the smallest circumscribed circle.

[0073] For example, in a methanol fuel combustion emission test, the denoised particle image was collected with a resolution of 1024x768 pixels. The image was subjected to grayscale gradient analysis, and the gradient of each pixel was calculated using a 3x3 Sobel operator. The calculation results showed that the gradient values ​​in the image ranged from 0 to 255, with an average gradient value of 45.6. Canny edge detection was then performed, with a low threshold of 30, a high threshold of 90, and a sigma value of 1.4. The edge detection results identified 374 particle contours. These contours were smoothed using a Gaussian filter with a standard deviation of σ=1.2, and then a 3x3 structural element was applied for morphological closing. After smoothing, the average perimeter of the contours was reduced by 8.3%, indicating that some subtle jagged edges were successfully removed. In the minimum circumscribed circle analysis, the 374 smoothed contours were processed one by one, with an average time of 0.5 milliseconds per contour. Finally, 374 minimum circumscribed circles were obtained, with diameters ranging from 1μm to 10μm and an average diameter of 4.2μm.

[0074] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0075] (1) Irradiating the filtered gas with a preset infrared light source and collecting a first absorption spectrum signal in real time;

[0076] (2) extracting absorption peak features of the first absorption spectrum signal to obtain first absorption peak features corresponding to the first absorption spectrum signal;

[0077] (3) generating a first spectrum according to the first absorption peak characteristics, irradiating the filtered gas with a preset ultraviolet light source, and collecting a second absorption spectrum signal in real time;

[0078] (4) extracting absorption peak characteristics of the second absorption spectrum signal to obtain second absorption peak characteristics corresponding to the second absorption spectrum signal;

[0079] (5) Generate a second spectrum according to the characteristics of the second absorption peak, and perform spectrum fusion on the first spectrum and the second spectrum to obtain a spectrum.

[0080] Specifically, the filtered gas is irradiated by a preset infrared light source, and the first absorption spectrum signal is collected in real time. The infrared light source is used to irradiate the gas sample so that the gas molecules absorb infrared light within a specific wavelength range, thereby generating an absorption spectrum. Specifically, when the gas molecules absorb infrared light of a specific wavelength, their internal vibration or rotation energy levels undergo transitions, and the absorbed light energy corresponds to the energy difference of these transitions. By detecting these absorption peaks, the presence and concentration of different components in the gas can be determined. The first absorption spectrum signal collected in real time contains all the absorption information of the filtered gas under the irradiation of the infrared light source, which reflects the characteristic absorption peaks of the gas components. In order to extract useful features from it, the absorption peak feature extraction is performed on the first absorption spectrum signal. In this step, Fourier transform infrared spectroscopy (FTIR) technology can be used to convert the time domain signal into a frequency domain signal through Fourier transform, thereby obtaining a spectrum. In the spectrum, each absorption peak corresponds to the characteristic absorption wavelength and absorption intensity of a specific gas component. By analyzing the characteristics of these absorption peaks, the type and concentration of the gas component can be determined.

[0081] Next, a first spectrum is generated based on the extracted absorption peak features. Figure 4 As shown in the figure, this spectrum shows the absorption characteristics of the filtered gas in the infrared spectrum, especially the absorption peaks of water and carbon dioxide. In order to further analyze the gas composition, it is necessary to irradiate the filtered gas with a preset ultraviolet light source and collect the second absorption spectrum signal in real time. The ultraviolet light source is used to irradiate the gas sample, so that the gas molecules absorb light energy in the ultraviolet spectrum, which will also cause the transition of the internal electronic energy level. By detecting the absorption peaks of these transitions, more gas composition information can be obtained, especially the concentration of methanol and formaldehyde. The second absorption spectrum signal collected in real time also needs to be extracted for absorption peak characteristics. Using a method similar to infrared spectroscopy, the ultraviolet spectrum signal is processed to extract characteristic absorption peaks. These characteristic absorption peaks reflect the absorption characteristics of the gas components in the ultraviolet spectrum. By analyzing these characteristic absorption peaks, the concentration information of methanol and formaldehyde can be determined.

[0082] After extracting the characteristic absorption peaks of the UV spectrum, a second spectrum is generated based on these characteristics, such as Figure 5As shown. This spectrum shows the absorption characteristics of the filtered gas in the ultraviolet spectrum range, supplementing some gas composition information that is not covered by the infrared spectrum. In order to obtain a more comprehensive gas composition analysis result, it is necessary to fuse the first spectrum and the second spectrum to generate a comprehensive spectrum. The information of the infrared spectrum and the ultraviolet spectrum is integrated to provide a more complete gas composition map. In the fusion process, a variety of techniques can be used, such as weighted averaging, interpolation, and principal component analysis (PCA). The weighted averaging method assigns weights according to the signal-to-noise ratio and importance of the spectrum, and obtains a comprehensive spectrum by weighted averaging. The interpolation rule interpolates between different spectra to fill the uncovered wavelength range and obtain a smoother comprehensive spectrum. Principal component analysis (PCA) is a data dimensionality reduction technique that extracts the main components in the spectrum, removes redundant information, and generates a high-quality comprehensive spectrum.

[0083] For example, in an actual methanol fuel combustion emission test, the filtered gas is irradiated by an infrared light source, and the first absorption spectrum signal is collected in real time. After Fourier transform and absorption peak feature extraction, the first spectrum is generated. The first spectrum shows the characteristic absorption peaks of water and carbon dioxide, which are located at 2.7 microns and 4.26 microns, respectively. The intensity of these absorption peaks reflects the concentration of water and carbon dioxide. Then, the same filtered gas sample is irradiated by an ultraviolet light source, and the second absorption spectrum signal is collected in real time. The absorption peak feature extraction of the ultraviolet spectrum signal is performed in the same way to generate the second spectrum. The second spectrum shows the characteristic absorption peaks of methanol and formaldehyde, which are located at 205 nanometers and 284 nanometers, respectively. The intensity of these absorption peaks reflects the concentration of methanol and formaldehyde. After the two spectra are generated, the principal component analysis (PCA) technology is used to fuse the spectra. PCA extracts the main components of the two spectra, removes noise and redundant information, and generates a spectrum.

[0084] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0085] (1) Perform image segmentation on the spectrum to obtain multiple sub-spectrum graphs;

[0086] (2) Inputting multiple sub-spectral images into the convolutional layer of the convolutional neural network for local feature extraction to obtain a local feature set of each sub-spectral image. The convolutional neural network includes: a convolutional layer, a pooling layer, and a fully connected layer;

[0087] (3) Input the local feature set of each sub-spectral image into the pooling layer for dimensionality reduction to obtain the reduced dimensionality feature set of each sub-spectral image;

[0088] (4) Input the dimension-reduced feature set of each sub-spectral image into the fully connected layer for feature fusion to obtain a fused vector set;

[0089] (5) Mapping the gas composition of each fusion vector in the fusion vector set to obtain the gas composition corresponding to each fusion vector;

[0090] (6) performing absorption peak intensity analysis on the first absorption peak characteristics to obtain the first absorption peak intensity;

[0091] (7) performing an absorption peak intensity analysis on the second absorption peak characteristics to obtain the second absorption peak intensity;

[0092] (8) performing component concentration analysis on the gas component corresponding to each fusion vector according to the first absorption peak intensity and the second absorption peak intensity, and obtaining component concentration data of the gas component corresponding to each fusion vector;

[0093] (9) The gas component corresponding to each fusion vector and the component concentration data of the gas component corresponding to each fusion vector are combined into gas composition data.

[0094] Specifically, the comprehensive spectrum is segmented to obtain multiple sub-spectral graphs. The purpose of image segmentation is to decompose the large spectrum into smaller parts so that each part focuses on specific spectral features. The sliding window method or the fixed area segmentation method can be used to divide the comprehensive spectrum into several sub-graphs, each covering a different wavelength range. These sub-spectral graphs are input into the convolution layer of the convolutional neural network for local feature extraction. The convolution layer performs a convolution operation on the input image through a set of convolution kernels (filters) to extract local features in the image. Each convolution kernel can capture specific image features, such as edges, corners, etc. Through multi-layer convolution operations, the convolutional neural network can extract deep features in the spectrum and generate a local feature set for each sub-spectral graph. The activation function after the convolution layer (such as ReLU) introduces nonlinearity, allowing the model to express more complex features.

[0095] After obtaining the local feature set, these features are input into the pooling layer for dimensionality reduction. The pooling layer reduces the dimension of the feature map by downsampling while retaining the main feature information. Commonly used pooling operations include maximum pooling and average pooling. Maximum pooling selects the maximum value in the pooling window, which can retain the most significant features; average pooling takes the average of all values ​​in the window, which can smooth the feature map. Through the pooling operation, the local feature set of each sub-spectral map is converted into a reduced-dimensional feature set, which significantly reduces the amount of data and improves computational efficiency.

[0096] Then, the reduced dimension feature set of each sub-spectrum is input into the fully connected layer for feature fusion. The fully connected layer expands the reduced dimension feature set into a one-dimensional vector, combines features and adjusts weights, and finally outputs a fused vector set. The fully connected layer can integrate different features through a combination of linear transformation and activation functions to generate a fused vector with higher representation capabilities. Each fused vector contains all the key information of the sub-spectrum, providing a basis for subsequent gas composition analysis.

[0097] For each fusion vector in the fusion vector set, gas composition mapping is required to obtain the gas composition corresponding to each fusion vector. In this step, the high-dimensional feature vector is converted into a specific gas composition using a trained gas composition mapping model. The mapping model can use linear regression, support vector machine (SVM) or other regression models to learn the relationship between spectral features and gas component concentrations through training data. Through gas composition mapping, the specific gas component and its preliminary concentration corresponding to each fusion vector can be obtained.

[0098] In order to improve the accuracy of gas component concentration analysis, the absorption peak intensity analysis is also performed on the first absorption peak feature and the second absorption peak feature. By analyzing the intensity of the absorption peak, the concentrations of water, carbon dioxide, methanol and formaldehyde can be accurately calculated. In the specific operation, the corresponding absorption peak positions are found in the fusion vector set, and the intensity values ​​of these positions are calculated. The absorption peak intensity reflects the light absorption ability of gas molecules and can be directly used for concentration calculation.

[0099] According to the first absorption peak intensity and the second absorption peak intensity, the gas component corresponding to each fusion vector is analyzed for component concentration, and the absorption peak intensity is converted into a concentration value using the Beer-Lambert Law. The concentration value of the gas component can be calculated using the known absorbance and molar absorptivity. For each fusion vector, the above process is repeated to obtain the concentration data of the gas component corresponding to each fusion vector. Finally, the gas component and its concentration data corresponding to each fusion vector are merged into gas component data. These data contain all detected gas components and their corresponding concentrations, providing reliable data support for the comprehensive analysis of methanol fuel combustion emissions.

[0100] For example, firstly, the comprehensive spectrum is segmented to obtain several sub-spectra. These sub-spectra are input into the pre-trained convolutional neural network for local feature extraction to obtain the local feature set of each sub-spectra. Through the maximum pooling operation, the size of the reduced dimension feature set is significantly reduced, but the main feature information is still retained. Then, the reduced dimension feature set is input into the fully connected layer for feature fusion to generate a fusion vector set. Each fusion vector represents the high-dimensional features of a sub-spectra, which are converted into specific gas components through the mapping model.

[0101] During the analysis, it was found that the fusion vector of a sub-spectrum showed obvious absorption peaks of methanol and formaldehyde, located at 205 nanometers and 284 nanometers respectively. The concentrations of methanol and formaldehyde were calculated by analyzing the absorption peak intensity. Combined with the Beer-Lambert law, the methanol concentration was 0.25 mol / L and the formaldehyde concentration was 0.15 mol / L. Similarly, by analyzing the fusion vectors of other sub-spectrums, the concentrations of water and carbon dioxide were calculated.

[0102] Finally, the gas components and their concentration data corresponding to all fusion vectors are integrated to obtain comprehensive gas composition data. These data include the concentrations of water, carbon dioxide, methanol and formaldehyde, providing accurate information support for the purity analysis and composition ratio of methanol fuel combustion emissions. Through this comprehensive method, real-time and comprehensive detection of methanol fuel combustion emissions is achieved, ensuring the optimization of fuel purity and combustion efficiency.

[0103] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0104] (1) Traverse the gas composition data to obtain each gas component and the component concentration data of each gas component;

[0105] (2) generating initial ratio data of each gas component according to each gas component and component concentration data of each gas component;

[0106] (3) Analyze the particle proportion according to the particle diameter of the particles to obtain the particle proportion data;

[0107] (4) The initial ratio data is corrected according to the particulate matter ratio data to obtain the target ratio data, and the fuel purity analysis of the methanol fuel is performed according to the target ratio data to obtain the fuel purity data.

[0108] Specifically, the gas component data is traversed to obtain each gas component and its corresponding component concentration data. In the previous steps, the concentration data of each gas component in the filtered gas has been obtained through spectral analysis and convolutional neural network processing. At this time, it is necessary to traverse these data and extract the name of each gas component and its concentration value one by one. Suppose that in a certain detection, the gas component data obtained include the concentrations of methanol, carbon dioxide, water and formaldehyde, which are 0.25mol / L, 0.15mol / L, 0.20mol / L and 0.10mol / L respectively. After obtaining each gas component and its concentration data, the next step is to generate the initial proportion data of each gas component based on these data. The initial proportion data reflects the relative proportion of each gas component in the total gas. In specific operations, the initial proportion of each gas component can be calculated by the following formula: initial proportion = gas component concentration / total gas concentration. Among them, the total gas concentration is the sum of the concentrations of all gas components, as shown below;

[0109] The total gas concentration is 0.25mol / L+0.15mol / L+0.20mol / L+0.10mol / L=0.70mol / L. Therefore, the initial proportion of methanol is 0.25mol / L / 0.70mol / L≈0.357, and the initial proportion of carbon dioxide is 0.15mol / L / 0.70mol / L≈0.214.

[0110] The initial ratio of water is 0.20mol / L / 0.70mol / L≈0.286, and the initial ratio of formaldehyde is 0.10mol / L / 0.70mol / L≈0.143. Next, the proportion of particles is analyzed according to their particle diameters, with the aim of evaluating the proportion of particles in the exhaust gas. Through the previous image processing steps, the particle diameter data of the particles has been obtained. Assume that in a certain test, the average diameter of the particles is measured to be 2 microns, and the total number of particles is 100. In order to perform a proportion analysis, the proportion of particles in the total exhaust gas can be calculated. In specific operations, the proportion of particles can be calculated by the following formula: Particle proportion = total particle volume / total exhaust gas volume. Assume that the volume of each particle is V=(4 / 3)πr³, where r is the radius of the particle. For particles with an average diameter of 2 microns, their radius is 1 micron, and the volume V≈4.19 microns³. The total particle volume is 4.19 microns³*100≈419 microns³. Assuming the total exhaust gas volume is 1 cubic meter, the particulate matter accounts for ≈419 microns³ / Micrometer³.

[0111] Then, the initial ratio data is corrected according to the particulate matter ratio data to obtain more accurate target ratio data. In specific operations, the correction can be made by the following formula: target ratio = initial ratio × (1-particulate matter ratio). In the previous example, since the particulate matter ratio is very small, its effect on the initial ratio can be ignored. Therefore, the target ratio is approximately equal to the initial ratio: the target ratio of methanol is ≈0.357, the target ratio of carbon dioxide is ≈0.214, the target ratio of water is ≈0.286, and the target ratio of formaldehyde is ≈0.143.

[0112] Finally, the fuel purity analysis of methanol fuel is performed according to the target ratio data to obtain fuel purity data. Fuel purity reflects the purity of methanol in methanol fuel and the ratio of its mixture components. In specific operations, fuel purity analysis can be performed through the following steps: First, convert the target ratio data into mass fraction. Assuming that the molar mass of each gas component is M1, M2, M3 and M4, the mass fraction of each gas component is: mass fraction 1 = target ratio 1 × M1 / ​​Σ (target ratio × molar mass).

[0113] In the previous example, assuming that the molar mass of methanol is 32 g / mol, carbon dioxide is 44 g / mol, water is 18 g / mol, and formaldehyde is 30 g / mol, the mass fraction of methanol = 0.357×32 / (0.357×32+0.214×44+0.286×18+0.143×30). The calculated mass fraction is used to determine the purity of methanol. Assuming that the final calculation result shows that the mass fraction of methanol in the mixture is 50%, the purity of methanol is 50%. Through this series of steps, accurate fuel purity data can be obtained to ensure the quality and combustion efficiency of methanol fuel.

[0114] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0115] (1) Analyze the target ratio data for harmful substances and obtain the harmful substance percentage data;

[0116] (2) Analyze the purity correction value of methanol fuel based on the harmful substance ratio data to obtain the target purity correction value;

[0117] (3) Correcting the fuel purity data according to the purity correction value to obtain the target fuel purity;

[0118] (4) Perform fuel composition analysis on methanol fuel according to the target fuel purity to obtain target fuel composition ratio data.

[0119] Specifically, in the real-time detection of methanol fuel combustion emissions, the analysis of harmful substances in the target ratio data is a key step to ensure fuel quality and safety. First, the fuel sample is comprehensively analyzed by gas chromatography-mass spectrometry (GC-MS) to identify and quantify the harmful substances therein. GC-MS technology combines the efficient separation capability of gas chromatography with the precise qualitative and quantitative capabilities of mass spectrometry, and can accurately detect trace harmful substances. During the analysis process, the fuel sample is gasified and separated by a chromatographic column. Different components are eluted at different times according to their physical and chemical properties, and then enter the mass spectrometer for ionization and mass analysis. By comparing with the standard spectrum library, the chemical structure of each component is determined, and its content is calculated based on the peak area. After obtaining the harmful substance ratio data, the purity correction value analysis is performed. The purity correction value reflects the actual purity of the fuel after removing the harmful substances, and the calculation formula is: purity correction value = 1 - Σ(harmful substance ratio). Among them, the harmful substance ratio refers to the mass fraction of each harmful substance in the fuel. This step takes into account the cumulative impact of all detected harmful substances and provides a basis for subsequent fuel purity corrections.

[0120] The fuel purity data is corrected according to the purity correction value to obtain the target fuel purity. The correction formula is: target fuel purity = original fuel purity × purity correction value. Original fuel purity refers to the fuel purity without considering the influence of harmful substances, which is usually obtained by measuring the content of main components (such as methanol). This correction process ensures the accuracy of fuel purity assessment and provides a reliable basis for fuel quality control.

[0121] According to the target fuel purity, the fuel composition ratio of methanol fuel is analyzed to obtain the target fuel composition ratio data. This step uses a multivariable optimization algorithm to calculate the optimal fuel composition ratio by considering multiple factors such as fuel purity, combustion efficiency, and emission standards. The optimization objective function includes maximizing the combustion heat value, minimizing harmful emissions, and meeting the target fuel purity requirements. The algorithm uses a hybrid method of genetic algorithm (GA) combined with particle swarm optimization (PSO), which can quickly find the global optimal solution in a complex multidimensional parameter space.

[0122] For example, in a methanol fuel quality assessment, GC-MS analysis revealed trace amounts of harmful substances such as formaldehyde (0.05%), formic acid (0.03%), and ethanol (0.12%) in the fuel sample. The total percentage of harmful substances was 0.2%. Based on this, the calculated purity correction value was 0.998 (1 - 0.002). The original fuel purity was determined by gas chromatography to be 99.5%, and the corrected target fuel purity was 99.3% (99.5% × 0.998).

[0123] In the process of optimizing the fuel composition ratio, the optimization goal is set to maximize the combustion heat value, while ensuring that NOx emissions are less than 30ppm and particulate matter emissions are less than 5mg / m³. Genetic algorithms are used for optimization, with a population size of 100 and 500 iterations. Each individual encodes information about the proportion of components such as methanol, water, and additives. Through genetic operations such as crossover and mutation, the fitness of individuals is optimized generation by generation. The optimal ratio finally obtained is: methanol 98.5%, water 1.2%, and additives 0.3%. This ratio is expected to control NOx emissions below 25ppm and particulate matter emissions below 3mg / m³ while ensuring 99.3% purity.

[0124] The embodiment of the present invention also provides a real-time detection system for combustion emissions based on methanol fuel, such as Figure 6 As shown, the real-time detection system for combustion emissions based on methanol fuel specifically includes:

[0125] The gas filtering module 201 is used to collect exhaust gas after the combustion of methanol fuel, filter particulate matter in the exhaust gas to obtain filtered gas, and extract particle diameter data of the particulate matter;

[0126] A spectrum analysis module 202, used for performing spectrum analysis on the filtered gas to obtain a corresponding spectrum graph;

[0127] A component analysis module 203, used for performing gas component analysis on the filtered gas according to the spectrum to obtain gas component data;

[0128] Purity analysis module 204, used to perform fuel purity analysis on the methanol fuel according to the gas composition data and the particle diameter of the particulate matter to obtain fuel purity data;

[0129] The ratio analysis module 205 is used to perform fuel composition ratio analysis on the methanol fuel according to the fuel purity data to obtain target fuel composition ratio data.

[0130] Through the collaborative work of the above modules, the particulate impurities in the gas can be effectively removed by collecting the exhaust gas of methanol fuel and filtering the particulate matter, and the particle diameter data of the particulate matter can be extracted. The particle image is denoised by using the Gaussian filter algorithm to ensure the accuracy and reliability of the particle diameter data. In addition, the infrared spectrum and ultraviolet spectrum analysis of the filtered gas are performed by spectral analysis technology to detect the concentrations of water and carbon dioxide, as well as methanol and formaldehyde, respectively, and generate corresponding spectra. The convolutional neural network is used to perform image segmentation and feature extraction on the spectra, and the convolution layer, pooling layer and fully connected layer are used to perform local feature extraction, dimension reduction and feature fusion on multiple sub-spectra, and finally the concentration data of each gas component is obtained. Not only the detection accuracy and efficiency are improved, but also the ability of simultaneous detection of multiple components is realized. Based on the combination of spectra and image processing technology, the components and particle characteristics in the exhaust gas can be comprehensively analyzed. By performing intensity analysis on the first absorption peak feature and the second absorption peak feature in the spectra, the fuel purity analysis is performed according to the gas composition data and the particle diameter data of the particulate matter to obtain the fuel purity data. By traversing the gas composition data, the initial proportion data of each gas component is obtained; then the proportion analysis is performed according to the particle diameter of the particulate matter to obtain the particulate matter proportion data; then the initial proportion data is corrected according to the particulate matter proportion data to obtain the target proportion data, and the fuel purity analysis of methanol fuel is performed accordingly to obtain the fuel purity data. It can comprehensively consider the gas composition and particulate matter characteristics to provide a more accurate fuel purity assessment. The target proportion data is analyzed for harmful substances, and the proportion data of harmful substances is calculated; the purity correction value of methanol fuel is analyzed according to the harmful substance proportion data to obtain the target purity correction value; then the fuel purity data is corrected according to the purity correction value to obtain the target fuel purity; finally, the fuel composition ratio of methanol fuel is analyzed according to the target fuel purity to obtain the target fuel composition ratio data. It can effectively identify and correct the harmful substance components in the fuel.

[0131] The above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit the same. Although the present invention has been described in detail with reference to the embodiments, a person skilled in the art should understand that the specific implementation modes of the present invention can still be modified or replaced by equivalents, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be included in the scope of the claims of the present invention.

Claims

1. A real-time detection method for combustion emissions based on methanol fuel, characterized in that: include: Collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas to obtain filtered gas, and extracting particle diameter data of the particulate matter; Performing spectral analysis on the filtered gas to obtain a corresponding spectrum graph, including: irradiating the filtered gas through a preset infrared light source and collecting a first absorption spectrum signal in real time; extracting absorption peak characteristics of the first absorption spectrum signal to obtain a first absorption peak characteristic corresponding to the first absorption spectrum signal; generating a first spectrum graph according to the first absorption peak characteristic, wherein characteristic absorption peaks of water and carbon dioxide are displayed in the first spectrum graph, and irradiating the filtered gas through a preset ultraviolet light source and collecting a second absorption spectrum signal in real time; extracting absorption peak characteristics of the second absorption spectrum signal to obtain a second absorption peak characteristic corresponding to the second absorption spectrum signal; generating a second spectrum graph according to the second absorption peak characteristic, wherein characteristic absorption peaks of formaldehyde and methanol are displayed in the second spectrum graph, and performing spectrum fusion on the first spectrum graph and the second spectrum graph to obtain the spectrum graph; The gas composition data of the filtered gas is obtained by performing gas composition analysis on the filtered gas according to the spectrum, including: performing image segmentation on the spectrum to obtain a plurality of sub-spectrums; inputting the plurality of sub-spectrums into a convolutional layer of a convolutional neural network to extract local features to obtain a local feature set of each sub-spectrum, wherein the convolutional neural network includes a convolutional layer, a pooling layer, and a fully connected layer; inputting the local feature set of each sub-spectrum into the pooling layer to reduce dimension to obtain a reduced dimension feature set of each sub-spectrum; inputting the reduced dimension feature set of each sub-spectrum into the fully connected layer to perform feature fusion to obtain a fused vector set; and merging the fused vector set. Perform gas component mapping on each fusion vector to obtain the gas component corresponding to each fusion vector; perform absorption peak intensity analysis on the first absorption peak feature to obtain the first absorption peak intensity; perform absorption peak intensity analysis on the second absorption peak feature to obtain the second absorption peak intensity; perform component concentration analysis on the gas component corresponding to each fusion vector according to the first absorption peak intensity and the second absorption peak intensity to obtain component concentration data of the gas component corresponding to each fusion vector; merge the gas component corresponding to each fusion vector and the component concentration data of the gas component corresponding to each fusion vector into the gas composition data; The fuel purity analysis of the methanol fuel is performed according to the gas composition data and the particle diameter of the particulate matter to obtain the fuel purity data, including: performing data traversal on the gas composition data to obtain each gas component and the component concentration data of each gas component; generating initial proportion data of each gas component according to each gas component and the component concentration data of each gas component; performing proportion analysis on the particulate matter according to the particle diameter of the particulate matter to obtain the particulate matter proportion data; correcting the initial proportion data according to the particulate matter proportion data to obtain the target proportion data, and performing fuel purity analysis on the methanol fuel according to the target proportion data to obtain the fuel purity data.

2. The method for real-time detection of combustion emissions based on methanol fuel according to claim 1, characterized in that: The steps of collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas to obtain filtered gas, and extracting particle diameter data of the particulate matter include: Collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas to obtain filtered gas; Collecting a particle image of the particulate matter, and performing grayscale conversion on the particle image to obtain a grayscale particle image; Performing image denoising on the grayscale particle image by using a Gaussian filtering algorithm to obtain a denoised particle image; The particle diameter analysis is performed on the denoised particle image to obtain particle diameter data of the particles.

3. The method for real-time detection of combustion emissions based on methanol fuel according to claim 2, characterized in that: The step of performing particle diameter analysis on the denoised particle image to obtain particle diameter data of the particles comprises: Performing grayscale gradient analysis on the denoised particle image to obtain grayscale gradient data corresponding to the denoised particle image; Performing edge detection on the grayscale particle image using the grayscale gradient data to obtain particle contour data corresponding to the grayscale particle image; Performing contour smoothing on the particle contour data corresponding to the grayscale particle image to obtain smoothed contour data corresponding to the grayscale particle image; Performing minimum circumscribed circle analysis on the smoothed contour data to obtain multiple minimum circumscribed circles; The diameter of each minimum circumscribed circle is extracted, and the diameter of each minimum circumscribed circle is used as the particle diameter data of the particle.

4. A real-time detection system for combustion emissions based on methanol fuel, used to execute the real-time detection method for combustion emissions based on methanol fuel as claimed in any one of claims 1 to 3, characterized in that: include: A gas filtering module, used for collecting exhaust gas after combustion of methanol fuel, filtering particulate matter in the exhaust gas to obtain filtered gas, and extracting particle diameter data of the particulate matter; A spectral analysis module, for performing spectral analysis on the filtered gas to obtain a corresponding spectrum graph, including: irradiating the filtered gas through a preset infrared light source and collecting a first absorption spectrum signal in real time; extracting absorption peak characteristics of the first absorption spectrum signal to obtain a first absorption peak characteristic corresponding to the first absorption spectrum signal; generating a first spectrum graph according to the first absorption peak characteristic, wherein the first spectrum graph displays characteristic absorption peaks of water and carbon dioxide, and irradiating the filtered gas through a preset ultraviolet light source and collecting a second absorption spectrum signal in real time; extracting absorption peak characteristics of the second absorption spectrum signal to obtain a second absorption peak characteristic corresponding to the second absorption spectrum signal; generating a second spectrum graph according to the second absorption peak characteristic, wherein the second spectrum graph displays characteristic absorption peaks of formaldehyde and methanol, and performing spectrum fusion on the first spectrum graph and the second spectrum graph to obtain the spectrum graph; The component analysis module is used to perform gas component analysis on the filtered gas according to the spectrum to obtain gas component data, including: performing image segmentation on the spectrum to obtain multiple sub-spectrums; inputting the multiple sub-spectrums into the convolution layer of the convolutional neural network to extract local features to obtain a local feature set of each sub-spectrum, and the convolutional neural network includes: a convolution layer, a pooling layer and a fully connected layer; inputting the local feature set of each sub-spectrum into the pooling layer for dimensionality reduction to obtain a reduced dimensionality feature set of each sub-spectrum; inputting the reduced dimensionality feature set of each sub-spectrum into the fully connected layer for feature fusion to obtain a fusion vector set; Performing gas component mapping on each fusion vector in the fusion vector set to obtain the gas component corresponding to each fusion vector; performing absorption peak intensity analysis on the first absorption peak feature to obtain the first absorption peak intensity; performing absorption peak intensity analysis on the second absorption peak feature to obtain the second absorption peak intensity; performing component concentration analysis on the gas component corresponding to each fusion vector according to the first absorption peak intensity and the second absorption peak intensity to obtain component concentration data of the gas component corresponding to each fusion vector; merging the gas component corresponding to each fusion vector and the component concentration data of the gas component corresponding to each fusion vector into the gas composition data; The purity analysis module is used to perform fuel purity analysis on the methanol fuel according to the gas composition data and the particle diameter of the particulate matter to obtain fuel purity data, including: performing data traversal on the gas composition data to obtain each gas component and component concentration data of each gas component; generating initial proportion data of each gas component according to each gas component and component concentration data of each gas component; performing proportion analysis on the particulate matter according to the particle diameter of the particulate matter to obtain particulate matter proportion data; correcting the initial proportion data according to the particulate matter proportion data to obtain target proportion data, and performing fuel purity analysis on the methanol fuel according to the target proportion data to obtain the fuel purity data.

Images