Method for real-time analysis of low-temperature plasma cracking thermodynamic process

Abstract

The invention provides a method for real-time analysis of a low-temperature plasma cracking thermodynamic process, which comprises the following steps of: acquiring emission spectrum information in a cracking cavity in the low-temperature plasma cracking process of a carbon fiber precursor by utilizing hyperspectral imaging equipment to obtain a cracking cavity space image containing the emission spectrum information; the spectral characteristics are used for representing the spectral characteristics in the carbon fiber molecular chain breakage process; the method comprises the following steps: monitoring a gray level image of a local region of vortex circulation, calculating curvature characteristics of a vortex form, identifying a geometrical shape and a change rule of a vortex region, determining a region of which the curvature characteristic similarity is higher than a similarity threshold as a vortex region with a self-organization characteristic, and representing the rearrangement and graphitization degree of a carbon fiber molecular chain by the vortex region.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for real-time analysis of a low-temperature plasma cracking thermodynamic process. Background Art

[0002] During the low-temperature plasma cracking process of carbon fiber precursor method, self-organized structures such as vortices and ripples will appear in specific areas of the plasma boundary layer and the cracking cavity. These features are closely related to the fracture, rearrangement and graphitization process of carbon fiber molecular chains. The formation mechanism of these self-organized structures is still unclear, and their influence on the carbon fiber cracking process and the structural evolution of carbon fiber molecular chains needs to be further studied. The current technical contradictions are: on the one hand, the emergence of self-organized structures is closely related to parameters such as plasma temperature, pressure, and electric field strength. These parameters not only affect the plasma state, but also directly act on the carbon fiber surface active groups, but these parameters are dynamically changing during the cracking process. How to monitor and control these parameters in real time to control the formation of self-organized structures, so as to solve the problem of how to accurately identify and locate unstable vortex circulation in the cavity; on the other hand, the frequency of spectral feature data will directly affect the accuracy of dominant frequency judgment. How to accurately extract the dominant frequency of vortex circulation, so as to use the vortex dominant frequency combined with the center position information to obtain the coefficient of variation and grayscale image that can show the distribution law of its structural evolution, and then characterize the degree of fracture and orientation state of carbon fiber molecular chains, is a technical problem that needs to be solved urgently. When the grayscale image presents a vortex morphology with self-organizing characteristics, how to accurately calculate its curvature characteristics to quantify the complexity of the morphological structure, and accurately collect its ripple wavelength and circulation velocity data, combined with the charge distribution characteristics of the active groups on the carbon fiber surface, by studying the formation mechanism of these self-organizing structures and their influence on the carbon fiber cracking process, the real-time analysis of the thermodynamic process of carbon fiber low-temperature plasma cracking can be achieved. Summary of the invention

[0003] The present invention provides a method for real-time analysis of low-temperature plasma pyrolysis thermodynamic process, which mainly comprises:

[0004] Using hyperspectral imaging equipment, the emission spectrum information in the cracking cavity during the low-temperature plasma cracking of carbon fiber precursors is collected to obtain a spatial image of the cracking cavity containing the emission spectrum information, which is used to characterize the spectral characteristics of the carbon fiber molecular chain breakage process;

[0005] The spectral intensity value of the spectral characteristic band of the emission spectral information of the cracking cavity spatial image is obtained, including the characteristic spectral line generated when the carbon fiber molecular chain is broken. If the spectral intensity value of the spectral characteristic band exceeds a preset threshold, it is confirmed that an unstable vortex circulation occurs inside the cracking cavity and the vortex circulation is located, and the central position information of the vortex circulation is recorded and integrated into the high-intensity spectral data set;

[0006] The local area image is subjected to Fourier transform to obtain a spectral image. The spatial positions are grouped according to the frequency peak and the spectrum characteristics are obtained. The local area includes the vortex circulation center and the fracture area of ​​the active group on the surface of the carbon fiber precursor. When the amplitude of a certain spatial frequency exceeds the amplitude threshold, it is determined as the dominant frequency of the unstable vortex circulation.

[0007] Based on the dominant frequency as the time sampling benchmark, the center position of the vortex circulation is mapped to the spatial coordinate system, and the coefficient of variation is used as the weight to perform weighted average on the grayscale values ​​of the pixels in the high-intensity spectral data set to obtain the grayscale image and spatial position image of the local area of ​​the vortex circulation, which are used to characterize the fracture degree and orientation state of the carbon fiber molecular chain.

[0008] By monitoring the grayscale image of the local area of ​​the vortex circulation, the curvature characteristics of the vortex morphology are calculated, the geometric shape and change law of the vortex area are identified, and the area with curvature feature similarity higher than the similarity threshold is determined as a vortex area with self-organization characteristics. The vortex area represents the rearrangement and graphitization degree of the carbon fiber molecular chain.

[0009] The wall charge accumulation data of the vortex area and the charge distribution characteristics of the carbon fiber surface active groups are obtained, and a machine learning model is constructed in combination with the high-intensity spectral data set to predict the correlation between the charge distribution and the self-organizing structure. When the rate of change of the wall charge distribution exceeds the rate threshold, it is determined to be a nearby area that is prone to forming a self-organizing structure, and the relevant data are integrated into an information chain of the vortex area.

[0010] The arc discharge frequency and vortex center magnetic field intensity data of the cracking chamber are obtained and associated with the vortex area information chain. By measuring the changes in vortex behavior in arc discharge, the correlation between vortex structure evolution, wall charge distribution and cracking thermodynamic parameters is established, and the breakage, rearrangement and graphitization processes of the carbon fiber molecular chain are analyzed to form a complete analysis link.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0012] The present invention discloses a method for real-time analysis of a low-temperature plasma cracking thermodynamic process, which uses a hyperspectral imaging device to collect emission spectrum information in a cracking chamber, and identifies and locates unstable vortex circulation by analyzing the intensity value of the spectral characteristic band. A Fourier transform is performed on a local area of ​​the vortex circulation to determine the dominant frequency, and the vortex center position is mapped based on this. By monitoring the curvature characteristics and wall charge distribution of the vortex area, the area with self-organization characteristics is identified, and the correlation between the charge distribution and the self-organization structure is predicted. Combined with the arc discharge frequency and magnetic field intensity data, the correlation between the vortex structure evolution, the wall charge distribution and the cracking thermodynamic parameters is established, and a comprehensive analysis of the carbon fiber molecular chain breakage, rearrangement and graphitization process is achieved, providing a basis for optimizing the cracking process. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 The present invention is a flow chart of a method for real-time analysis of a low-temperature plasma cracking thermodynamic process.

[0014] Figure 2 It is a schematic diagram of a method for real-time analysis of a low-temperature plasma cracking thermodynamic process of the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0016] like Figure 1-2 In this embodiment, a method for real-time analysis of a low-temperature plasma cracking thermodynamic process may specifically include:

[0017] Step S101, using a hyperspectral imaging device, collecting emission spectrum information in a cracking cavity during low-temperature plasma cracking of carbon fiber precursors, obtaining a cracking cavity spatial image containing emission spectrum information, which is used to characterize the spectral characteristics of the carbon fiber molecular chain breakage process.

[0018] A hyperspectral imager is used to obtain spectral data in a carbon fiber precursor cracking cavity, and the spectral data is decomposed in the frequency domain by Fourier transform to obtain a first spectral characteristic frequency distribution diagram; wavelet transform is performed to remove background noise based on the first spectral characteristic frequency distribution diagram to obtain a second spectral characteristic frequency distribution diagram, and spectral peak position and intensity data are extracted from the second spectral characteristic frequency distribution diagram; the second spectral characteristic frequency distribution diagram is measured by a spectral absorption method to obtain gas component concentration data in the cavity, and a second gas state data set is generated by data processing; a deep neural network model is established for the second gas state data set, and the spectral peak position and intensity data are input into the deep neural network model to obtain a predicted value of the gas component concentration, and the carbon fiber molecular chain break position is judged according to a pre-established standard carbon fiber molecular fragment spectral database.

[0019] Exemplarily, a three-dimensional spatial spectral database is established according to the emission spectrum intensity distribution mode in the carbon fiber precursor cracking cavity, a hyperspectral imager is used to obtain spectral data in the cavity from multiple angles, and the spectral data is decomposed in the frequency domain by Fourier transform to obtain a first spectral characteristic frequency distribution diagram. The spectral signal is preprocessed for the first spectral characteristic frequency distribution diagram, and the background noise is removed by wavelet transform to obtain a second spectral characteristic frequency distribution diagram, and the spectral peak position and intensity data are extracted from the second spectral characteristic frequency distribution diagram. The two-photon absorption spectroscopy method is used to monitor the relationship between the gas density and the plasma state in the cracking cavity in real time, and the ratio of the number of excited state particles to the number of ground state particles is recorded to establish a first gas state data set. A gas flow numerical model is established based on the first gas state data set, the cavity boundary temperature field distribution is set as the boundary condition, and the spectral absorption method is used to measure the gas component concentration in the cavity to obtain a second gas state data set. Data fusion is performed on the second gas state data set and the second spectral characteristic frequency distribution diagram, and a gas component concentration and spectral feature correspondence model is established through a deep neural network. The model input is the spectral peak position and intensity data, and the output is the predicted value of the gas component concentration. The predicted values ​​of gas component concentrations were compared and verified using a pre-established standard carbon fiber molecular fragment spectral database, and the similarity score was calculated using a spectral matching algorithm to determine the position of the carbon fiber molecular chain break. In the monitoring of the carbon fiber precursor cracking process, the hyperspectral imaging device acquires the emission spectral data in the cavity through multi-angle sampling. The sampling angles include five positions: 0 degrees, 45 degrees, 90 degrees, 135 degrees, and 180 degrees. The spectral data of 100 wavelength points are acquired at each angle to form a 500×100 spectral data matrix. After Fourier transformation, the frequency domain characteristic distribution is obtained. In the spectral data matrix, the wavelength range is 200-800nm, the wavelength interval is 6nm, and the spectral intensity value range is 0-4095 digital quantization units. The acquired spectral data was subjected to wavelet transform denoising, and the db4 wavelet basis function was selected to perform 5-layer decomposition on the original spectral data. The soft threshold was set to 15, and the denoised spectral data was obtained after reconstruction. The main characteristic peaks were located at 308nm, 336nm, and 391nm, corresponding to the characteristic emission lines of OH free radicals, NH free radicals, and N2+ ions, respectively. Two-photon absorption spectroscopy was used to monitor the gas state in the cracking chamber. The laser wavelength was 532nm and the pulse energy density was 2.5J / cm 2 The absorption signal intensity was recorded by a photomultiplier tube, and the excited state particle density was calculated to be 3.2×10 16 / cm 3 , the ground state particle density is 8.5×10 18 / cm 3, the ratio is 0.0038. In the gas flow numerical model, the cavity boundary temperature is set to 1200K, the pressure is standard atmospheric pressure, and the NS equations are used to describe the gas flow, where the gas viscosity coefficient μ is 1.8×10 -5 Pa·s, thermal conductivity λ is 0.025W / (m·K), and the temperature field and velocity field distribution in the cavity are obtained by solving the finite volume method. The deep neural network model adopts a 5-layer fully connected network structure. The input layer contains 100 spectral peak position and intensity data nodes, the hidden layers are 64, 32, and 16 nodes respectively, and the output layer is 10 nodes corresponding to different gas component concentrations. The activation function uses the ReLU function, the loss function uses the mean square error function, and the learning rate is set to 0.001. After 10,000 iterations of training, the model converges. In the standard carbon fiber molecular fragment spectrum database, the characteristic spectral data of 20 typical molecular fragments are included, including fragments generated by the breakage of CC bonds, CH bonds, and CN bonds. The spectrum matching algorithm uses the cosine similarity calculation method, and the similarity threshold is set to 0.85. When the similarity between the measured spectrum and the standard spectrum exceeds the threshold, it is determined to be this type of molecular fragment. The break position and degree of the carbon fiber molecular chain can be judged by the type and concentration distribution of molecular fragments.

[0020] Step S102, obtaining the spectral intensity value of the spectral characteristic band of the emission spectral information of the cracking cavity spatial image, including the characteristic spectral line generated when the carbon fiber molecular chain breaks. If the spectral intensity value of the spectral characteristic band exceeds a preset threshold, it is confirmed that an unstable vortex circulation appears inside the cracking cavity and the vortex circulation is located, and the center position information of the vortex circulation is recorded and integrated into the high-intensity spectral data set.

[0021] A multi-angle spectral acquisition device is used to obtain emission spectral data of carbon fiber molecular chain fracture, and the spectral data includes characteristic spectral line positions and spectral intensity values; it is judged whether the spectral intensity value exceeds a preset threshold value, and if the spectral intensity value exceeds the preset threshold value, a vortex field distribution map is obtained by particle image velocimetry; a regional growing algorithm is used to extract the vortex circulation core area for the vortex field distribution map, and a vortex characteristic vector is calculated by a centroid positioning algorithm, and the vortex characteristic vector includes vortex center coordinates, vortex peak value, and vortex core area; according to the vortex characteristic vector and the spectral data, a time series correlation is performed, and a vortex center motion trajectory curve is fitted by the least squares method, and the vortex center motion trajectory curve records the vortex circulation evolution process.

[0022] Exemplarily, according to the emission spectrum information generated by the breakage of the carbon fiber molecular chain in the cracking cavity, the spectral characteristic band data is collected at 0 degrees, 45 degrees, 90 degrees, and 135 degrees, and the characteristic spectral line position and spectral intensity value are extracted by the peak and valley recognition method to obtain the first spectral data set. A threshold is judged for the spectral intensity value in the first spectral data set, and a threshold is set according to 80% of the standard intensity of the characteristic spectral line. If the spectral intensity value exceeds the threshold, the particle image velocimetry method is used to obtain the velocity distribution diagram of the flow field in the cracking cavity, and the first vortex field distribution diagram is calculated by the curl operator. The first vortex field distribution diagram is subjected to Gaussian filtering and noise reduction processing to obtain a second vortex field distribution diagram, and the region with a vortex value greater than 2 times the average vortex is extracted as the vortex circulation core area by the regional growth algorithm. The center of mass positioning algorithm is used to calculate the vortex center coordinates for the vortex circulation core area, and a vortex characteristic vector is constructed, including the vortex center coordinates, the vortex peak value, and the vortex core area area. According to the vortex characteristic vector and the first spectral data set, the light intensity correction algorithm is used to compensate the vortex area spectral intensity to generate a high-intensity spectral data set. According to the vortex circulation center position information in the high-intensity spectral data set, the least squares method is used to fit the vortex center motion trajectory curve to record the vortex circulation evolution process. When collecting spectra in the carbon fiber cracking cavity, four optical fiber probes at different angles are used to synchronously collect the emission spectrum. The acquisition aperture of each probe is 2mm, the sampling frequency is 100Hz, the spectral range covers 200-800nm, and the spectral resolution is 0.1nm. The characteristic spectral lines are extracted by the peak and valley recognition algorithm, mainly including three characteristic spectral lines of CH (431.4nm), CN (388.3nm), and C2 (516.5nm). When the characteristic spectral line intensity exceeds the 80% threshold of the standard spectral line intensity, the particle image velocimetry system is started, and alumina tracer particles with a particle size of 0.5μm, a pulsed laser wavelength of 532nm, a pulse interval of 100μs, a camera resolution of 2048×2048 pixels, and a spatial resolution of 0.05mm / pixel are used to measure the flow field velocity field. The spatial gradient of the velocity field is calculated by the central difference method, and then the curl field distribution is solved. The curl calculation formula is: Where u and v are the velocity components in the x and y directions, respectively. The calculated curl field is subjected to denoising by a Gaussian filter, with a Gaussian kernel size of 5×5 and a standard deviation of σ=1.5. In the filtered curl field, the vorticity value is greater than twice the average vorticity, about 150s -1 The area is marked as the vortex core area. The vortex core area is usually elliptical in shape, with a major axis of about 4 mm, a minor axis of about 2 mm, and an area of ​​about 6.28 mm. 2. The vortex characteristic vector includes parameters such as the vortex center coordinates (x, y), the vortex peak value ωmax, and the core area S. The vortex center coordinates are calculated using the regional centroid formula xc=∑(xi·ωi) / ∑ωi, yc=∑(yi·ωi) / ∑ωi, and ωi is the vortex value of the i-th point. The influence of the vortex region on spectral acquisition is mainly manifested as light intensity attenuation. The attenuation coefficient is proportional to the vortex intensity, and the typical value is between 0.85-0.95. When performing least squares fitting on the vortex center trajectory, the second-order polynomial model y=ax 2 +bx+c, where the typical coefficient value is a=-0.05mm -1 , b = 0.8, c = 2.5mm. The root mean square error of the fitting curve is usually less than 0.2mm, indicating that the vortex motion has good regularity. The movement range of the vortex center in the x direction is 10-30mm, the movement range in the y direction is 5-15mm, and the movement period is about 0.1s. The evolution process of the vortex circulation is characterized by an increase in intensity first and then a decrease in intensity. The maximum vorticity value can reach 300s-1 and the duration is about 0.2s.

[0023] Step S103, perform Fourier transform on the local area image to obtain a spectral image, group the spatial positions according to the frequency peak and obtain the spectrum characteristics, the local area includes the vortex circulation center and the carbon fiber precursor surface active group fracture area, when the amplitude of a certain spatial frequency exceeds the amplitude threshold, it is determined as the dominant frequency of the unstable vortex circulation.

[0024] According to the central position of the vortex circulation, the local area image is divided into blocks using a sliding window, and the first frequency domain image is obtained through a two-dimensional fast Fourier transform; a Gaussian filter is used to suppress noise in the first frequency domain image, and the regions with adjacent frequency differences less than a difference threshold are clustered using a region growing clustering algorithm to obtain a second frequency domain image; based on the second frequency domain image, a random forest algorithm is used to classify the frequency domain features at different spatial positions, and feature vectors are extracted from the frequency domain features to obtain a frequency domain feature data set; a dual threshold detection is performed on the frequency amplitude in the frequency domain feature data set, and the maximum entropy algorithm is used to sort the frequency domain features to obtain the dominant frequency components of the vortex circulation.

[0025] Exemplarily, according to the coordinates of the vortex circulation center position and the fracture area of ​​the active group on the carbon fiber precursor surface, a 32×32 pixel sliding window is used to divide the local area image into blocks, and the sliding step is set to 16 pixels. Each image block is transformed in the frequency domain by a two-dimensional fast Fourier transform to obtain a first frequency domain image. A 5×5 pixel Gaussian filter is used for noise suppression for the first frequency domain image, and the standard deviation is set to 1.2. The local area spatial frequency peak is extracted by the peak detection algorithm, and the region with a frequency difference of less than 10% between adjacent positions is clustered by regional growth to generate a second frequency domain image. According to the statistical frequency amplitude distribution of the second frequency domain image, the random forest algorithm is used to classify the frequency domain features at different spatial positions, and feature vectors are extracted from the frequency domain features, including frequency peak, amplitude, bandwidth, symmetry, and skewness parameters, to obtain a frequency domain feature data set. For the frequency amplitude in the frequency domain feature data set, a double threshold detection is performed with 3 times the background amplitude mean as the low threshold and 5 times as the high threshold to obtain a third frequency domain image. According to the third frequency domain image, the maximum entropy algorithm is used to sort the frequency domain features. The sorting is based on the entropy value. The entropy value is calculated using the Shannon entropy formula to determine the dominant frequency component of the vortex circulation. Time-frequency analysis is performed on the dominant frequency component, and the frequency change characteristics over time are calculated using continuous wavelet transform to record the evolution of the dominant frequency. In the analysis of vortex circulation in the cracking process of carbon fiber precursor, the frequency domain feature extraction of local area images is the key. When using a 32×32 pixel sliding window for image segmentation, the selection of a sliding step of 16 pixels ensures a 50% overlap rate between adjacent blocks, which is conducive to capturing continuously changing frequency domain features. The frequency spectrum obtained after each image block is subjected to a two-dimensional fast Fourier transform, and the frequency resolution is the sampling frequency divided by the window size. For a sampling frequency of 10kHz, the frequency resolution is 312.5Hz. In the noise suppression process, the standard deviation of the 5×5 Gaussian filter is 1.2, which is an optimized parameter value. This parameter can effectively suppress high-frequency noise while maintaining the main characteristics of the spectrum. For Gaussian white noise with an amplitude of 100, the noise amplitude drops below 15 after filtering. The frequency peak detection adopts the local maximum method, sets the minimum peak-to-valley ratio to 3:1, and the interval between adjacent peaks is not less than 5 frequency points to avoid the detection of false peaks. When performing regional growth clustering, the frequency difference of 10% is used as the threshold, which reflects the spatial continuity of the frequency domain characteristics of the vortex circulation. The typical vortex circulation area is about 1000 pixels and contains 3-5 main frequency components. In the frequency domain feature vector, the frequency peak reflects the vortex rotation frequency, usually in the range of 200-500Hz, the bandwidth characterizes the volatility of the vortex intensity, and the symmetry and skewness describe the statistical characteristics of the spectrum shape. In the dual threshold detection, the low threshold is 3 times the background amplitude mean, about 45 units, and the high threshold is 5 times, about 75 units, based on the optimal segmentation ratio obtained from experimental statistics.When the maximum entropy algorithm sorts the frequency domain features, the entropy value is calculated using the formula H = -Σ(pi×log2pi), where pi is the normalized frequency amplitude. The entropy value of the dominant frequency is usually between 4.5 and 5.5, which is significantly higher than the secondary frequency component. The continuous wavelet transform uses Morlet wavelet, with a scale parameter range of 1-64, corresponding to a frequency range of 50-1000Hz. The time-frequency analysis results show that the dominant frequency of the vortex circulation shows an upward trend in the early stage of formation, increasing from 200Hz to 350Hz, with a duration of about 50ms, and then tends to stabilize, and the frequency fluctuation range is controlled within ±20Hz. The slope of the frequency evolution curve reflects the formation speed of the vortex circulation, and the fluctuation amplitude of the stable section indicates the stability of the vortex structure.

[0026] Step S104, based on the dominant frequency as the time sampling benchmark, the center position of the vortex circulation is mapped to the spatial coordinate system, and the coefficient of variation is used as a weight to perform weighted averaging on the grayscale values ​​of the pixels in the high-intensity spectral data set to obtain a grayscale image and a spatial position image of the local area of ​​the vortex circulation, which are used to characterize the degree of breakage and orientation state of the carbon fiber molecular chain.

[0027] The cubic spline interpolation method is used to resample the time series according to the center position of the vortex circulation, and the first spatial position image is obtained by mapping the rotation and translation matrix to the spatial rectangular coordinate system; for the high-intensity spectral data set, a sliding window is used to calculate the statistical parameters of the pixel grayscale values ​​in the local area, and the first weighted image is obtained by the ratio of the standard deviation to the mean; according to the first weighted image, the high-intensity spectral data set is subjected to pixel-level weighted averaging operation, and the weighted grayscale value distribution is clustered by the random forest algorithm to obtain the second grayscale image; the second grayscale image is subjected to feature-level fusion with the first spatial position image, and the spatial feature vector containing position coordinates, grayscale value, and gradient direction is extracted by a deep neural network.

[0028] Exemplarily, the sampling time interval is set according to the dominant frequency of the vortex circulation, the reciprocal of the period is used as the reference time unit, the time series of the center position of the vortex circulation is resampled by the cubic spline interpolation method, and the rotation and translation matrix is ​​constructed to map the center position to the spatial rectangular coordinate system to obtain the first spatial position image. For the pixel points in the high-intensity spectral data set, a 16×16 pixel sliding window is used to calculate the statistical parameters of the pixel grayscale value in the local area, and the window sliding step is 8 pixels. The coefficient of variation is calculated by the ratio of the standard deviation to the mean to generate the first weighted image. According to the first weighted image, the high-intensity spectral data set is subjected to pixel-level weighted averaging operation, the weighting coefficient is normalized to the range of 0 to 1, and the random forest algorithm is used to perform cluster analysis on the weighted grayscale value distribution to obtain the second grayscale image. The gradient distribution of the grayscale value is calculated for the second grayscale image, the edge features are extracted by the sobel operator, and the gradient direction histogram is established to obtain the third grayscale image. According to the feature level fusion of the third grayscale image and the first spatial position image, a deep neural network is used to extract the spatial feature vector, which includes three components: position coordinates, grayscale value, and gradient direction. A quadratic curve equation group is constructed for the spatial eigenvector, and the spatial orientation parameters of the carbon fiber molecular chain are fitted by the least squares method. The curvature value and the inflection point position are calculated to determine the degree of molecular chain breakage. In the process of carbon fiber cracking, the dominant frequency of the vortex circulation is usually in the range of 300-500Hz. Taking the frequency of 400Hz as an example, the corresponding benchmark sampling time interval is 2.5ms. When the vortex center position is resampled by cubic spline interpolation, 4 control points are selected to construct a cubic polynomial, and the interpolation node interval is 1 / 4 of the benchmark time interval, that is, 0.625ms, so as to obtain a smooth center trajectory curve. The rotation and translation matrix uses a 4×4 homogeneous coordinate transformation matrix, which contains the rotation angle θ and the translation vector (tx, ty, tz). In the processing of high-intensity spectral data sets, the sliding window of 16×16 pixels corresponds to an actual spatial size of approximately 0.8mm×0.8mm, and each window contains 256 pixels. The coefficient of variation is calculated by the ratio of the standard deviation σ divided by the mean μ, that is, CV=σ / μ. For images with grayscale values ​​ranging from 0 to 255, the typical coefficient of variation is distributed between 0.1 and 0.5. A larger coefficient of variation indicates that the grayscale value in this area fluctuates significantly. When the grayscale value is weighted averaged, the weight coefficient w is normalized by the sigmoid function, w = 1 / (1+e^(-CV)), so that the weight value is limited to the range of 0-1. The formula for calculating the weighted grayscale value g' is g' = Σ(wi×gi) / Σwi, where gi is the original grayscale value. The sobel operator is used for edge detection based on the weighted result. The horizontal operator is [-1,-2,-1; 0,0,0; 1,2,1], and the vertical operator is [-1,0,1; -2,0,2; -1,0,1], and the gradient amplitude and direction are calculated.When the gradient direction histogram is counted, the 360-degree direction is divided into 18 intervals, each with 20 degrees, and the number of gradient directions falling into each interval is counted and weighted. The feature vector fusion uses a three-layer deep neural network, with 5 input layer nodes, including x, y coordinates, grayscale values, gradient amplitude and direction, 16 and 8 hidden layer nodes, and 3 output layer nodes. The spatial orientation curve uses the quadratic function y=ax. 2 +bx+c fitting, where coefficient a reflects the curvature of the curve, b represents the overall tilt trend, c is the longitudinal offset, and the position where the curvature value is greater than 0.5mm-1 is determined as a potential break point. In practical applications, the typical curvature value of carbon fiber precursor is in the range of 0.2-0.8mm-1, the curvature value of the fracture area often exceeds 0.6mm-1, and the curvature value at the fracture point shows a mutation characteristic.

[0029] Step S105, by monitoring the grayscale image of the local area of ​​the vortex circulation, calculating the curvature characteristics of the vortex morphology, identifying the geometric shape and change law of the vortex area, and determining the area where the curvature feature similarity is higher than the similarity threshold as a vortex area with self-organizing characteristics. The vortex area represents the rearrangement and graphitization degree of the carbon fiber molecular chain.

[0030] The Sobel operator is used to extract the vortex edge contour point sequence from the vortex circulation grayscale image, and the first curvature characteristic image is obtained by calculating the vortex edge contour point sequence through the cubic spline curve; the random forest algorithm is used to extract the shape characteristics of the vortex area based on the first curvature characteristic image, and the shape characteristics include roundness, eccentricity, and area ratio, to obtain the second curvature characteristic image; the vortex area feature similarity is calculated by the cosine distance based on the second curvature characteristic image, if the feature similarity is higher than the preset threshold, the regional growing method is used to obtain the third curvature characteristic image; the vortex area morphological parameters are calculated based on the third curvature characteristic image, and the morphological parameters include regional area, perimeter, and equivalent diameter, and the interlayer spacing parameters are obtained through a deep neural network to judge the graphitization conversion level.

[0031] Exemplarily, based on the grayscale image of the local area of ​​vortex circulation, the sobel operator is used to extract the vortex edge contour, the edge contour point sequence is fitted by a cubic spline curve, the tangent vectors and normal vectors of adjacent points on the curve are calculated, and the discrete curvature formula k = (x'y"-y'x") / ((x'^2+y'^2)^(3 / 2)) is used to calculate the curvature distribution of the vortex morphology to obtain a first curvature characteristic image. Morphological processing is performed on the first curvature characteristic image, and a circular structure element with a radius of 3 pixels is used for opening operations. The edge contour gaps are filled by closing operations. The random forest algorithm is used to extract shape features such as roundness, eccentricity, and area ratio of the vortex area to generate a second curvature characteristic image. A feature vector is constructed based on the second curvature characteristic image, including the mean, standard deviation, maximum, and minimum values ​​of the curvature. The feature similarity of adjacent vortex areas is calculated by cosine distance to obtain a first similarity matrix. The first similarity matrix is ​​normalized. If the similarity value is higher than the preset threshold value of 0.85, the vortex region is merged by the regional growing method. The region with gray value difference less than the standard deviation is set as the growth condition to obtain the third curvature characteristic image. The morphological parameters of the self-organized vortex region are calculated according to the third curvature characteristic image, including the regional area, perimeter, and equivalent diameter. The spatial distribution characteristics of the regional gray value are extracted using a deep neural network to generate the second characteristic matrix. The interlayer spacing parameter d = λ / (2sinθ) of the carbon fiber molecular chain is calculated for the second characteristic matrix, where λ is the wavelength of the incident light and θ is the diffraction angle. The graphitization conversion level is judged by the deviation between the interlayer spacing and the standard graphite lattice spacing. During the carbon fiber cracking process, the morphological characteristic analysis of the vortex circulation begins with the extraction of the edge contour. When using the Sobel operator for edge detection, the differential operators in the horizontal and vertical directions are [-1,-2,-1; 0,0,0; 1,2,1] and [-1,0,1; -2,0,2; -1,0,1] respectively. For a 256-grayscale image, the typical edge response intensity threshold is set to 50. The extracted edge point sequence is fitted by a cubic spline curve, with each control point separated by 4 pixels to ensure the smoothness of the curve. The curvature calculation is expressed parametrically. For a point P(x(t), y(t)) on the curve, its curvature k = (x'y”-y'x”) / ((x'^2+y'^2)^(3 / 2)). Taking the vortex core area as an example, the curvature value usually fluctuates between 0.2-0.8, and the curvature peak appears at the sharp turn of the vortex circulation. In morphological processing, a circular structural element with a radius of 3 pixels corresponds to an actual size of about 150 microns, and the area of ​​the noise spots eliminated by the opening operation is less than 28 square pixels. Shape features include roundness, 4π×area / circumference 2 The roundness of the ideal circle is 1, and the roundness of the actual vortex area is between 0.7-0.9; the eccentricity reflects the degree of vortex ellipse, and the calculation formula is e=√(1-b 2 / a 2), where a and b are the lengths of the major and minor axes, respectively, with typical values ​​between 0.3-0.6; the area ratio is defined as the ratio of the area of ​​the vortex region to the area of ​​its convex hull, which is usually greater than 0.85. The similarity is calculated using the cosine distance. For the eigenvectors vi and vj, the similarity s = vi·vj / (|vi|×|vj|). The eigenvector contains curvature statistics with a mean range of 0.4-0.6, a standard deviation of 0.1-0.2, a kurtosis of 2.5-4.0, and a skewness of ±0.5. The threshold of 0.85 is set to ensure that the merged vortex regions have similar morphological features. During the region growing process, the seed point selects the local curvature maximum point, and the growth conditions include the grayscale value difference being less than the current region standard deviation, 15-25 grayscale levels, and the curvature change rate being less than 30%. The input feature dimension of the deep neural network is 128, including a normalized grayscale histogram of 64 dimensions and a morphological feature of 64 dimensions. In the calculation of the interlayer spacing of carbon fiber molecular chains, the incident light wavelength λ = 0.154nm, copper target X-ray, diffraction angle θ changes from 22° to 26°, and the corresponding interlayer spacing d decreases from 0.395nm to 0.335nm. The interlayer spacing of standard graphite is 0.3354nm. The deviation of the interlayer spacing of the actual sample from the standard value reflects the degree of graphitization. A deviation of less than 0.002nm indicates that the structure is highly graphitized.

[0032] Step S106, obtain the wall charge accumulation data of the vortex area and the charge distribution characteristics of the carbon fiber surface active groups, build a machine learning model based on the high-intensity spectral data set, predict the correlation between the charge distribution and the self-organizing structure, and when the rate of change of the wall charge distribution exceeds the rate threshold, it is determined to be an adjacent area that is prone to forming a self-organizing structure, and the relevant data are integrated into an information chain of the vortex area.

[0033] The Laplace operator is used to calculate the charge spatial distribution gradient obtained by the wall charge detector in the vortex region, and the first charge distribution characteristic map is obtained by the Poisson equation solver; based on the electric field strength data in the first charge distribution characteristic map, a Gaussian smoothing filter is used to obtain the second charge distribution characteristic map; data alignment is performed on the second charge distribution characteristic map and the high-intensity spectral data set, and a mapping relationship between charge distribution and spectral characteristics is established using the random forest algorithm to obtain the third charge distribution characteristic map; based on the charge density, accumulation rate, and spectral feature data of the third charge distribution characteristic map, a deep neural network is used to construct a prediction model, and if the accumulation rate exceeds the rate threshold, it is marked as a self-organized adjacent area.

[0034] Exemplarily, based on the charge density data collected by the wall charge detector in the vortex region, the five-point Laplace operator is used to calculate the charge spatial distribution gradient, and the spatial distribution of the wall charge density and the electric field strength is calculated by the Poisson equation solver to obtain the first charge distribution characteristic diagram. For the electric field strength data in the first charge distribution characteristic diagram, Gaussian smoothing filtering is used to remove noise, and the kernel size is set to 5×5 to obtain the second charge distribution characteristic diagram. According to the data alignment of the second charge distribution characteristic diagram and the high-intensity spectral data set, the random forest algorithm is used to establish the mapping relationship between the charge distribution and the spectral characteristics. The feature vector contains the electric field strength, charge density, spectral peak position, and peak intensity to generate the third charge distribution characteristic diagram. For the third charge distribution characteristic diagram, the wall charge change rate dρ / dt is calculated, and a sliding window of 32 time points is set, with a window step of 8 time points. The charge accumulation rate is calculated by the central difference method to obtain the first rate distribution diagram. According to the first rate distribution diagram, the charge accumulation characteristics are extracted, and a prediction model is constructed using a deep neural network. The input layer contains three sets of data: charge density, accumulation rate, and spectral characteristics. The number of hidden layer nodes is 64-32-16. When the accumulation rate exceeds the preset threshold, it is marked as a self-organized adjacent area. A feature data table is constructed for the self-organized adjacent area, which contains four data blocks: spatial coordinates, charge distribution, spectral characteristics, and accumulation rate. The information chain structure in a unified format is generated through a data merger. During the carbon fiber cracking process, the wall charge distribution is calculated using a five-point Laplace operator. For discrete points with a grid spacing of h = 0.1 mm, the typical potential value φ is distributed between -500 V and 500 V, and the calculated charge density ρ ranges from -2×10 -6 Up to 2×10 -6 C / m 3 When solving the Poisson equation, the dielectric constant ε0 is taken as 8.85×10 -12 F / m, and the boundary condition is set to wall potential φ = 0 V. Gaussian smoothing filter uses a 5×5 kernel matrix and a standard deviation σ = 1.2 to perform noise reduction on the electric field intensity data. The electric field intensity fluctuation range before processing is 4×10 5 ±2×10 4 V / m, after filtering, the fluctuation is reduced to ±5×10 3 V / m, maintaining the main characteristics of the electric field distribution. The random forest algorithm constructs 100 decision trees, and the feature vector contains electric field strength, charge density, spectral peak position, CH is 431.4nm, CN is 388.3nm, C2 is 516.5nm and their corresponding peak intensities. The charge accumulation rate is calculated using a 32-point sliding window, corresponding to an actual time span of 80ms, a window step of 8 points, and 20ms. The central difference calculation formula is dρ / dt=(ρi+1-ρi-1) / (2Δt), where Δt is the sampling time interval of 2.5ms. The typical charge accumulation rate is within ±5×10-4 C / (m 3 ·s) range, when the accumulation rate exceeds 1×10 -3 C / (m 3 ·s), it is marked as a self-organizing neighboring area. The input layer of the deep neural network model contains 72 nodes, of which the charge density and accumulation rate each occupy 16 nodes, and the spectral features occupy 40 nodes, including the intensity values ​​of 20 wavelength points. The hidden layer uses the ReLU activation function, the dropout rate is set to 0.3, and the initial value of the learning rate is 0.001. The training of the prediction model uses the mean square error loss function, and the accuracy of the validation set reaches 92% after 10,000 iterations. In the structural design of the feature data table, the spatial coordinates record the vortex center position (x, y, z) with an accuracy of 0.1mm; the charge distribution data includes charge density ρ, electric field intensity E, and accumulation rate dρ / dt; the spectral features record the wavelength position λ, peak intensity I, and half-peak width Δλ of the characteristic peak. The information chain is stored in binary format. The header of each data block contains a 4-byte identifier and 4-byte length information to ensure the integrity and traceability of the data. The typical size of the information chain structure is 4KB, which contains the key parameters required to characterize the self-organizing process.

[0035] Step S107, obtain the arc discharge frequency and vortex center magnetic field strength data of the cracking chamber, associate them with the vortex area information chain, and establish the correlation between the vortex structure evolution, wall charge distribution and cracking thermodynamic parameters by measuring the changes in vortex behavior in arc discharge, analyze the breakage, rearrangement and graphitization process of the carbon fiber molecular chain, and form a complete analysis link.

[0036] The discharge frequency data obtained by the arc detector and the magnetic field intensity distribution data measured by the Hall sensor array are received, and a first frequency spectrum characteristic diagram is obtained by fast Fourier transform; a Butterworth low-pass filter is used to process the frequency domain components in the first frequency spectrum characteristic diagram to obtain a second frequency spectrum characteristic diagram; the second frequency spectrum characteristic diagram is time-series correlated with the magnetic field intensity distribution data, and the vortex motion characteristics are extracted by the random forest algorithm to obtain a first evolution characteristic diagram; a characteristic matrix is ​​constructed according to the first evolution characteristic diagram and the wall charge distribution data, a deep neural network is used to establish a thermodynamic parameter mapping relationship, and the fracture position is determined by the bond angle change rate.

[0037] Exemplarily, the arc detectors arranged in a circular array in the cracking chamber are used to obtain the discharge frequency data, and the magnetic field intensity distribution in the vortex center area is measured by a 4×4 Hall sensor array with a spacing of 5 mm. The frequency domain components of the arc discharge are calculated by a 1024-point fast Fourier transform, and the first spectrum feature map is generated in combination with the vortex area information chain. For the frequency domain components in the first spectrum feature map, a Butterworth low-pass filter is used to eliminate noise, and the cutoff frequency is set to 1 / 4 of the sampling frequency to obtain a second spectrum feature map. According to the time series association between the second spectrum feature map and the magnetic field intensity data, the random forest algorithm is used to extract the vortex motion characteristics, including the vortex center trajectory, angular velocity, and magnetic field gradient, and the first evolution feature map is generated. For the first evolution feature map, the vortex structure parameters are extracted, and the principal component analysis method is used to reduce the dimension of the parameters. The eigenvectors with a cumulative contribution rate of more than 85% are selected as the principal components to obtain the second evolution feature map. A feature matrix is ​​constructed based on the second evolution feature map and the wall charge distribution data, and a deep neural network is used to establish a thermodynamic parameter mapping relationship. The input layer includes three sets of data: temperature field, pressure field, and electric field intensity. The molecular chain stress distribution is calculated based on the thermodynamic parameter mapping relationship, the fracture position is determined by the bond angle change rate, and the degree of graphitization is quantified by the lattice interlayer spacing calculation method. A complete link table is established based on the molecular chain structure evolution data, which contains four data blocks: vortex characteristics, field strength distribution, thermodynamic parameters, and structural characteristics. A hierarchical storage structure is used to record the corresponding relationship of each link. In the monitoring of carbon fiber cracking process, arc discharge frequency and magnetic field intensity are key parameters. The 4×4 Hall sensor array covers a measurement area of ​​20×20mm, and the sensitivity of each sensor is 50mV / mT, with a measurement range of 0-100mT. The arc discharge frequency sampling rate is set to 10kHz, and a 1024-point FFT is used to obtain a 0-5kHz spectrum distribution. The main discharge frequency components are concentrated in the range of 500-2000Hz. The Butterworth low-pass filter adopts a 4th-order design with a cutoff frequency of 2.5kHz. The high-frequency noise is suppressed by the -3dB amplitude-frequency characteristic at the cutoff frequency. The signal-to-noise ratio of the filtered signal is increased from the original 15dB to 28dB, maintaining the main characteristics of the discharge frequency. In the vortex motion feature extraction, the center trajectory is located by the maximum point of magnetic field intensity, and the angular velocity is calculated by the position change at adjacent moments. The typical value is in the range of 200-600rad / s, and the maximum value of magnetic field gradient reaches 5mT / mm. When principal component analysis is used to reduce the dimension of vortex structure parameters, the original features contain 15 parameters, including position coordinates (x, y, z), velocity components (vx, vy, vz), acceleration components (ax, ay, az), magnetic field intensity (Bx, By, Bz) and its gradient The cumulative contribution rate of the first four principal components is 87%, of which the first principal component, about 45%, mainly reflects the spatial position characteristics of the vortex, the second principal component, about 23%, represents the motion state, and the third and fourth principal components, about 12% and 7%, respectively, correspond to the magnetic field distribution characteristics. The thermodynamic parameter mapping uses a five-layer neural network, the input layer includes temperature field 1000-2500K, pressure field 0.1-1MPa, electric field strength 1-5×10 5 V / m three groups of data, each with 20 sampling points. The number of hidden layer nodes is 128, 64, 32, and 16, respectively, and the ReLU activation function is used. The molecular chain breakage criterion is based on the CC bond angle change rate. When the bond angle changes by more than 15° / ps, it is determined to be the break position. The degree of graphitization is characterized by the interlayer spacing of the crystal plane. The standard value of complete graphitization is 0.3354nm, and the actual measured value varies in the range of 0.3354-0.344nm. The complete link table adopts a hierarchical storage structure. The first layer stores 512 bytes of vortex characteristics, including position, velocity, and acceleration data; the second layer records 1024 bytes of field strength distribution, including electromagnetic field strength and gradient; the third layer stores 2048 bytes of thermodynamic parameters, including temperature, pressure, and energy data; the fourth layer stores 1024 bytes of structural characteristics, including bond angle, bond length, and interlayer spacing information.

[0038] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for real-time analysis of low-temperature plasma pyrolysis thermodynamic process, characterized in that: The method comprises: Using hyperspectral imaging equipment, the emission spectrum information in the cracking cavity during the low-temperature plasma cracking of carbon fiber precursors is collected to obtain a spatial image of the cracking cavity containing the emission spectrum information, which is used to characterize the spectral characteristics of the carbon fiber molecular chain breakage process; The spectral intensity value of the spectral characteristic band of the emission spectral information of the cracking cavity spatial image is obtained, including the characteristic spectral line generated when the carbon fiber molecular chain is broken. If the spectral intensity value of the spectral characteristic band exceeds a preset threshold, it is confirmed that an unstable vortex circulation occurs inside the cracking cavity and the vortex circulation is located, and the central position information of the vortex circulation is recorded and integrated into the high-intensity spectral data set; The local area image is subjected to Fourier transform to obtain a spectral image. The spatial positions are grouped according to the frequency peak and the spectrum characteristics are obtained. The local area includes the vortex circulation center and the fracture area of ​​the active group on the surface of the carbon fiber precursor. When the amplitude of a certain spatial frequency exceeds the amplitude threshold, it is determined as the dominant frequency of the unstable vortex circulation. Based on the dominant frequency as the time sampling benchmark, the center position of the vortex circulation is mapped to the spatial coordinate system, and the coefficient of variation is used as the weight to perform weighted average on the grayscale values ​​of the pixels in the high-intensity spectral data set to obtain the grayscale image and spatial position image of the local area of ​​the vortex circulation, which are used to characterize the fracture degree and orientation state of the carbon fiber molecular chain. By monitoring the grayscale image of the local area of ​​the vortex circulation, the curvature characteristics of the vortex morphology are calculated, the geometric shape and change law of the vortex area are identified, and the area with curvature feature similarity higher than the similarity threshold is determined as a vortex area with self-organization characteristics. The vortex area represents the rearrangement and graphitization degree of the carbon fiber molecular chain. The wall charge accumulation data of the vortex area and the charge distribution characteristics of the carbon fiber surface active groups are obtained, and a machine learning model is constructed in combination with the high-intensity spectral data set to predict the correlation between the charge distribution and the self-organizing structure. When the rate of change of the wall charge distribution exceeds the rate threshold, it is determined to be a nearby area that is prone to forming a self-organizing structure, and the relevant data are integrated into an information chain of the vortex area. The arc discharge frequency and vortex center magnetic field intensity data of the cracking chamber are obtained and associated with the vortex area information chain. By measuring the changes in vortex behavior in arc discharge, the correlation between vortex structure evolution, wall charge distribution and cracking thermodynamic parameters is established, and the breakage, rearrangement and graphitization processes of the carbon fiber molecular chain are analyzed to form a complete analysis link.

2. The method according to claim 1, characterized in that The method uses a hyperspectral imaging device to collect emission spectrum information in the cracking cavity during the low-temperature plasma cracking of carbon fiber precursors, and obtains a cracking cavity spatial image containing the emission spectrum information, which is used to characterize the spectral characteristics of the carbon fiber molecular chain breakage process, including: A hyperspectral imager is used to obtain spectral data in a carbon fiber precursor cracking cavity, and the spectral data is decomposed in the frequency domain by Fourier transform to obtain a first spectral characteristic frequency distribution diagram; Performing wavelet transform on the first spectral characteristic frequency distribution graph to remove background noise, obtaining a second spectral characteristic frequency distribution graph, and extracting spectral peak position and intensity data from the second spectral characteristic frequency distribution graph; The second spectral characteristic frequency distribution diagram is measured by using a spectral absorption method to obtain gas component concentration data in the cavity, and a second gas state data set is generated by data processing; A deep neural network model is established for the second gas state data set. The spectral peak position and intensity data are input into the deep neural network model to obtain the predicted value of the gas component concentration, and the carbon fiber molecular chain breakage position is determined according to the pre-established standard carbon fiber molecular fragment spectral database.

3. The method according to claim 1, characterized in that The spectral intensity value of the spectral characteristic band of the emission spectrum information of the cracking cavity space image is obtained, including the characteristic spectral line generated when the carbon fiber molecular chain is broken. If the spectral intensity value of the spectral characteristic band exceeds a preset threshold, it is confirmed that an unstable vortex circulation occurs inside the cracking cavity and the vortex circulation is located, and the vortex circulation center position information is recorded and integrated into a high-intensity spectral data set, including: A multi-angle spectrum acquisition device is used to obtain emission spectrum data of carbon fiber molecular chain fracture, wherein the spectrum data includes characteristic spectrum line positions and spectrum intensity values; Determine whether the spectral intensity value exceeds a preset threshold value, and if the spectral intensity value exceeds the preset threshold value, obtain a vortex field distribution map by particle image velocimetry; A regional growing algorithm is used to extract the core area of ​​the vortex circulation according to the vortex field distribution diagram, and a vortex characteristic vector is calculated by a centroid positioning algorithm, wherein the vortex characteristic vector includes the coordinates of the vortex center, the vortex peak value and the area of ​​the vortex core area; The vortex characteristic vector is temporally correlated with the spectral data, and a vortex center motion trajectory curve is fitted using the least squares method. The vortex center motion trajectory curve records the vortex circulation evolution process.

4. The method according to claim 1, characterized in that The local area image is subjected to Fourier transform to obtain a spectral image, and the spatial positions are grouped according to the frequency peak and the spectrum characteristics are obtained. The local area includes the vortex circulation center and the fracture area of ​​the active group on the surface of the carbon fiber precursor. When the amplitude of a certain spatial frequency exceeds the amplitude threshold, it is determined as the dominant frequency of the unstable vortex circulation, including: According to the center position of the vortex circulation, the local area image is divided into blocks using a sliding window, and the first frequency domain image is obtained by two-dimensional fast Fourier transform; A Gaussian filter is used to suppress noise in the first frequency domain image, and regions whose frequency differences between adjacent positions are less than a difference threshold are clustered using a region growing clustering algorithm to obtain a second frequency domain image; classifying frequency domain features at different spatial positions using a random forest algorithm according to the second frequency domain image, extracting feature vectors from the frequency domain features, and obtaining a frequency domain feature data set; A double threshold detection is performed on the frequency amplitude in the frequency domain feature data set, and the frequency domain features are sorted using a maximum entropy algorithm to obtain the dominant frequency components of the vortex circulation.

5. The method according to claim 1, characterized in that The method uses the dominant frequency as the time sampling benchmark, maps the center position of the vortex circulation to the spatial coordinate system, and uses the coefficient of variation as the weight to perform weighted averaging on the grayscale values ​​of the pixels in the high-intensity spectral data set to obtain a grayscale image and a spatial position image of the local area of ​​the vortex circulation, which are used to characterize the fracture degree and orientation state of the carbon fiber molecular chain, including: The time series is resampled using the cubic spline interpolation method according to the center position of the vortex circulation, and the first spatial position image is obtained by mapping the rotation and translation matrix to the spatial rectangular coordinate system; For high-intensity spectral data sets, a sliding window is used to calculate the statistical parameters of pixel grayscale values ​​in the local area, and the first weighted image is obtained by the ratio of the standard deviation to the mean; Performing a pixel-level weighted average operation on the high-intensity spectral data set according to the first weight image, and performing a cluster analysis on the weighted grayscale value distribution by a random forest algorithm to obtain a second grayscale image; The second grayscale image is fused with the first spatial position image at a feature level, and a spatial feature vector including position coordinates, grayscale values ​​and gradient directions is extracted through a deep neural network.

6. The method according to claim 1, characterized in that The method comprises: monitoring the grayscale image of the local area of ​​the vortex circulation, calculating the curvature characteristics of the vortex morphology, identifying the geometric shape and change law of the vortex area, and determining the area with the curvature characteristic similarity higher than the similarity threshold as the vortex area with self-organization characteristics, and the vortex area represents the rearrangement and graphitization degree of the carbon fiber molecular chain, including: A sobel operator is used to extract a vortex edge contour point sequence from the vortex circulation grayscale image, and a first curvature characteristic image is obtained by calculating the vortex edge contour point sequence through a cubic spline curve; Extracting shape features of the vortex region using a random forest algorithm according to the first curvature feature image, wherein the shape features include roundness, eccentricity and area ratio, to obtain a second curvature feature image; Calculating the vortex region feature similarity according to the second curvature feature image by using the cosine distance, and if the feature similarity is higher than a preset threshold, obtaining a third curvature feature image by using a region growing method; The vortex region morphological parameters are calculated according to the third curvature characteristic image, and the morphological parameters include region area, perimeter and equivalent diameter. The interlayer spacing parameters are obtained through a deep neural network to determine the graphitization conversion level.

7. The method according to claim 1, characterized in that The wall charge accumulation data of the vortex region and the charge distribution characteristics of the carbon fiber surface active groups are obtained, and a machine learning model is constructed in combination with a high-intensity spectral data set to predict the correlation between the charge distribution and the self-organizing structure. When the rate of change of the wall charge distribution exceeds a rate threshold, it is determined to be a neighboring area that is prone to forming a self-organizing structure, and the relevant data are integrated into an information chain of the vortex region, including: The Laplace operator is used to calculate the charge spatial distribution gradient obtained by the wall charge detector in the vortex region, and the first charge distribution characteristic diagram is obtained by the Poisson equation solver; According to the electric field intensity data in the first charge distribution characteristic diagram, a second charge distribution characteristic diagram is obtained by using a Gaussian smoothing filter; Performing data alignment on the second charge distribution characteristic map and the high-intensity spectral data set, and using a random forest algorithm to establish a mapping relationship between charge distribution and spectral features to obtain a third charge distribution characteristic map; A prediction model is constructed using a deep neural network based on the charge density, accumulation rate and spectral feature data of the third charge distribution characteristic diagram. If the accumulation rate exceeds a rate threshold, it is marked as a self-organizing adjacent area.

8. The method according to claim 1, characterized in that: The arc discharge frequency and vortex center magnetic field intensity data of the cracking cavity are obtained and associated with the vortex area information chain. By measuring the vortex behavior changes in the arc discharge, the correlation relationship between the vortex structure evolution, the wall charge distribution and the cracking thermodynamic parameters is established, and the fracture, rearrangement and graphitization process of the carbon fiber molecular chain is analyzed to form a complete analysis link, including: Receiving the discharge frequency data acquired by the arc detector and the magnetic field intensity distribution data measured by the Hall sensor array, and obtaining a first frequency spectrum characteristic diagram through fast Fourier transform; According to the frequency domain components in the first frequency spectrum characteristic graph, a Butterworth low-pass filter is used to process the frequency domain components to obtain a second frequency spectrum characteristic graph; Performing time series correlation on the second frequency spectrum characteristic graph and the magnetic field intensity distribution data, extracting vortex motion characteristics using a random forest algorithm, and obtaining a first evolution characteristic graph; A characteristic matrix is ​​constructed based on the first evolution characteristic graph and the wall charge distribution data, a deep neural network is used to establish a thermodynamic parameter mapping relationship, and the fracture position is determined by the bond angle change rate.

Images