Digital twinning combined cracking equipment data processing method

Abstract

The invention provides a cracking equipment data processing method combined with digital twinning, which comprises the following steps: acquiring initial state data and environmental data of a carbon fiber composite material in a cracking furnace, and performing time synchronization processing to obtain a time consistency data set; extracting pyrolysis characteristic parameters of interaction between the carbon fiber bundle and the matrix from the spatial registration data set, and calculating a reaction rate and pyrolysis behavior change of a microscopic scale of a fiber matrix interface to obtain a pyrolysis characteristic parameter set; if a deviation exists between the prediction result and a preset target carbon powder quality index, determining a heating power distribution adjustment amount and a change trend of an air inlet valve opening degree based on a deviation degree, and obtaining a control parameter adjustment set of the cracking furnace; and carrying out constraint and smoothing processing on the control parameter adjustment set to obtain a control set value, formatting the control set value into a closed-loop feedback instruction, and updating the closed-loop feedback instruction to a cracking furnace control system to obtain adjusted pyrolysis operation state data.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular, to a data processing method for pyrolysis equipment combined with digital twin. Background Art

[0002] Recycling carbon fiber through the pyrolysis process and generating high-value carbon powder can not only reduce resource waste but also significantly improve the economic benefits of material recycling. However, there are still many deficiencies in the existing pyrolysis methods in achieving refined control. Traditional technologies mostly rely on adjusting empirical parameters, lacking a comprehensive description of the dynamic evolution of the pyrolysis behavior of materials. Especially in complex environments, the real-time monitoring and feedback control capabilities are weak, resulting in large fluctuations in carbon powder quality and poor process stability. These limitations stem from the lack of understanding of the multi-scale mechanisms of the pyrolysis process and the insufficiency of data-driven regulation means. Specifically, the current methods are difficult to accurately capture the interaction law between carbon fiber bundles and the matrix at the mesoscopic scale, nor can they finely describe the dynamic characteristics of the interfacial reaction kinetics at the microscopic scale. Since the dynamic changes of key state variables during the pyrolysis process are difficult to be accurately captured; at the same time, the lack of efficient integration of sensor data leads to an increase in the mapping deviation between the digital model and the physical process; more importantly, the existing feedback mechanism has a lag response and is difficult to adjust process parameters in a timely manner according to the degradation behavior of materials. These unsolved problems directly restrict the intelligence and efficiency of the pyrolysis process. Therefore, how to construct a digital twin system that can real-time characterize the multi-scale pyrolysis behavior of carbon fiber composites, integrate high-precision sensor data, and achieve closed-loop feedback control has become a key issue in improving the refined control ability of the pyrolysis process. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides a data processing method for pyrolysis equipment combined with digital twin, mainly including:

[0004] Obtain the initial state data and environmental data of the carbon fiber composite material in the pyrolysis furnace, and perform time synchronization processing to obtain a time-consistent data set;

[0005] Obtain the geometric structure, material properties, and pyrolysis process principle of the pyrolysis furnace, establish a digital twin model, and perform spatial registration operations according to the time-consistent data set. Map the physical positions of the sensor arrays to the virtual space in the digital twin model through a preset spatial coordinate mapping algorithm to obtain a spatially registered data set;

[0006] Extract the pyrolysis characteristic parameters of the interaction between carbon fiber bundles and the matrix from the spatially registered data set, calculate the reaction rate and pyrolysis behavior changes at the microscopic scale of the fiber matrix interface to obtain a set of pyrolysis characteristic parameters;

[0007] The pyrolysis characteristic parameter set is used to drive the digital twin model for high-frequency iterative calculations to obtain the dynamic update state of the digital twin body, simulate the temperature field, pressure field and material degradation state in the pyrolysis furnace, predict the carbon powder purity distribution, particle size distribution prediction, and residual matrix content, and obtain the pyrolysis furnace state prediction result;

[0008] If there is a deviation between the prediction result and the preset target carbon powder quality index, the adjustment amount of the heating power distribution and the change trend of the intake valve opening are determined based on the deviation degree to obtain the control parameter adjustment set of the pyrolysis furnace;

[0009] The control parameter adjustment set is constrained and smoothed to obtain the control set value, which is formatted into a closed-loop feedback instruction and updated to the pyrolysis furnace control system to obtain the adjusted pyrolysis operation state data;

[0010] The degradation rate of the carbon fiber bundle and the matrix decomposition ratio are extracted from the adjusted pyrolysis operation state data, and the mass distribution of the pyrolysis products is verified by gas analysis to obtain the quality evaluation data. Compared with the target product quality index, the parameter weights of the digital twin model are updated using an optimization algorithm to obtain the optimized pyrolysis control scheme.

[0011] Further, the initial state data and environmental data of the carbon fiber composite material in the cracking furnace are obtained, and time synchronization processing is performed to obtain a time consistency data set, including: dividing grid points on the surface of the carbon fiber composite material according to the sensor, arranging a thermocouple sensor array from the inner wall of the cracking furnace to the center of the composite material according to the grid point spacing, detecting the gas component concentration through a high-precision gas chromatograph, using a pressure sensor to measure the pressure value of each point, and using a distributed data collector to collect all sensor signals to obtain a first group of raw data. The data validity is verified for the first group of raw data, the temperature data range is set between minus 50 degrees and 2000 degrees, the pressure data range is set between 0 and 1000 kilopascals, and the gas component concentration data range is set between 0 and 100%, and the abnormal values are eliminated to obtain a second group of valid data. The second group of valid data is sorted according to the sensor acquisition timestamp, and the temperature, pressure, and gas component data are denoised using a Kalman filter. The state vector of the filter includes temperature values, pressure values, and gas component concentration values. The observation matrix is determined by the sensor layout position to obtain a third group of filtered data. The time series correlation coefficient matrix between each point of the sensor array is calculated for the third group of filtered data, and the data is processed in time synchronization according to the correlation coefficient. The missing data is supplemented by the linear interpolation method to obtain the fourth group of time synchronization data. The three-dimensional space grid model is established using the fourth group of time synchronization data, and the Kriging interpolation function is constructed. The range parameter of the interpolation function is obtained by the variance analysis of the experimental data. The spherical semivariogram model is used to calculate the spatial correlation, and the temperature field, pressure field and gas component distribution at the grid node are interpolated to obtain the fifth group of spatial distribution data. A time series database table is established for the fifth group of spatial distribution data. The table structure contains timestamp, coordinate position, temperature value, pressure value, and gas component concentration fields. The database transaction mechanism is used to ensure data integrity, and the carbon fiber composite material state monitoring data set at the specified time is extracted from the database.

[0012] Furthermore, the geometric structure, material properties and cracking process principle of the cracking furnace are obtained, a digital twin model is established, and a spatial registration operation is performed according to the time consistency data set. The physical position of the sensor array is matched with the virtual space in the digital twin model through a preset spatial coordinate mapping algorithm to obtain a spatial registration data set, including: importing the cracking furnace point cloud data into the three-dimensional modeling software, generating the first layer of mesh on the wall of the cracking furnace cavity from the point cloud, importing the thermal parameter library of carbon fiber composite materials to obtain the thermal conductivity, specific heat capacity, and density values, and using a tetrahedral mesher to mesh the cracking furnace cavity and composite materials to obtain the second layer of mesh. The mesh quality of the second layer of mesh is tested, the aspect ratio, distortion, and orthogonality index of the mesh unit are calculated, unqualified mesh units are removed for local encryption reconstruction, and the mesh quality is optimized by a mesh smoothing algorithm to obtain the third layer of mesh. According to the physical space coordinates of the sensor array, the Harris corner detection algorithm is used to identify the characteristic points of the sensor layout position, the local coordinate system origin and basis vectors are calculated based on the characteristic point coordinates, and the Euler angle rotation matrix is used to calculate the transformation of the sensor position from the local coordinate system to the global coordinate system to obtain the first set of unified coordinate data. Using the first set of unified coordinate data and the third-layer grid, the iterative closest point registration algorithm is used to calculate the Euclidean distance matrix from the sensor node to the grid node. The spatial transformation matrix is solved according to the least squares criterion, and the grid nodes are transformed into the fourth-layer grid by cubic spline. The registration error evaluation index is established for the fourth-layer grid, the average deviation between the sensor physical position and the grid node position is calculated, the cross-validation method is used to evaluate the registration accuracy, and the area where the deviation exceeds the threshold is locally optimized to obtain the fifth-layer grid. The spatial position of the corresponding node of the sensor is extracted from the fifth-layer grid, and a radial basis function interpolator is constructed to establish the mapping function from the physical space to the virtual space. The corresponding node state parameters of the actual sensor temperature field, pressure field, and gas composition field in the virtual space are calculated, and a physical quantity and state parameter mapping database is established.

[0013] Furthermore, pyrolysis characteristic parameters of the interaction between carbon fiber bundles and the matrix are extracted from the spatially registered dataset, and the reaction rate and pyrolysis behavior changes at the microscale of the fiber-matrix interface are calculated to obtain a set of pyrolysis characteristic parameters, including: according to the temperature field data of the carbon fiber bundle-matrix interface in the spatially registered dataset, the central difference method is used to calculate the interface temperature gradient distribution, and Fourier's law of heat conduction is used to calculate the interface heat flux density. The first batch of heat conduction data is obtained by calculating the temperature gradient measurement error from the deviation of the temperature values at adjacent measurement points. For the first batch of heat conduction data, according to the component concentration information collected by the gas composition on-line monitor, Fick's law of diffusion is used to calculate the gas concentration distribution at the interface, and the amount of matrix decomposition per unit time is calculated from the change in gas concentration to obtain the second batch of gas diffusion data. Based on the second batch of gas diffusion data, the law of conservation of mass is used to calculate the gas generation rate during the reaction process, and an Arrhenius reaction equation is established from the corresponding relationship between gas concentration and temperature. The pre-exponential factor and activation energy in the equation are obtained by least-squares fitting to obtain the third batch of reaction kinetic data. For the third batch of reaction kinetic data, a mapping function of temperature and reaction rate is constructed, the interface movement displacement is calculated using the reaction rate, and the interface movement speed is calculated from the displacement change to obtain the fourth batch of interface evolution data. According to the fourth batch of interface evolution data, a reaction interface parameter data table is established. The data table contains four fields: temperature gradient, gas concentration, reaction rate, and interface displacement. All parameters are stored using a timestamp index, and a pyrolysis process parameter database is established.

[0014] Furthermore, the pyrolysis characteristic parameter set is used to drive the digital twin model for high-frequency iterative calculations to obtain the dynamic update state of the digital twin body, simulate the temperature field, pressure field, and material degradation state in the cracking furnace, predict the carbon powder purity distribution, particle size distribution prediction, and residual matrix content, and obtain the cracking furnace state prediction results, including: dividing the computational domain grid according to the pyrolysis characteristic parameter set, where the grid node spacing is determined by the temperature gradient and pressure gradient, constructing the temperature field and pressure field control equations using the finite element method, and calculating the temperature field and pressure field distributions within the time step using the fourth-order Runge-Kutta integration algorithm to obtain the first batch of physical field data. For the first batch of physical field data, calculate the material degradation rate based on the temperature field and pressure field gradient distributions, calculate the stress distribution within the fiber bundle using the stress tensor, and calculate the material fracture probability from the stress distribution to obtain the second batch of stress-strain data. Based on the second batch of stress-strain data, calculate the stress distribution and displacement field at the gas-solid two-phase interface, calculate the diffusion flux of the gas-phase components using the law of conservation of mass, and calculate the migration rate of carbon elements from the diffusion flux to obtain the third batch of mass transfer data. According to the third batch of mass transfer data, use the Gaussian mixture clustering algorithm to calculate the probability density distribution of the carbon powder particle size, calculate the carbon powder content in different particle size intervals from the probability density function, and calculate the carbon powder purity using the carbon element balance relationship to obtain the fourth batch of product data. For the fourth batch of product data, establish a calculation function for the residual matrix content, use support vector regression to predict the matrix decomposition rate, and calculate the remaining matrix spatial distribution based on the decomposition rate to obtain the fifth batch of degradation data. Use the least squares method to compare the error between the calculation results and the measured data, adjust the boundary conditions of the control equation, and repeat the calculation steps until the error converges, and obtain the cracking furnace state prediction results from the convergent solution.

[0015] Further, if there is a deviation between the prediction result and the preset target toner quality index, the adjustment amount of the heating power distribution and the change trend of the intake valve opening are determined based on the deviation degree, and a control parameter adjustment set for the pyrolysis furnace is obtained, including: calculating the toner purity deviation value, particle size distribution deviation value, and residue content deviation value according to the prediction result, establishing a weighted sum of squared errors function for the three deviation values, and calculating the total deviation degree of the quality index from the error function to obtain the first batch of quality deviation data. For the first batch of quality deviation data, the gradient descent method is used to calculate the adjustment direction of the heating power. The heating power adjustment area is divided according to the temperature field uniformity index, and the power adjustment ratio of each area is determined using the temperature gradient distribution to obtain the second batch of power adjustment data. The change trend of the pyrolysis gas components is calculated according to the second batch of power adjustment data, the gas concentration distribution in the furnace is calculated using the gas diffusion equation, and the adjustment amount of the intake valve opening is determined based on the concentration uniformity index to obtain the third batch of flow field adjustment data. For the third batch of flow field adjustment data, a state transition equation for the pyrolysis process is established, the state response relationship before and after the process parameter adjustment is calculated, and the sequential execution order of the parameter adjustment is obtained from the state response relationship to obtain the fourth batch of control sequence data. The parameter optimization function is constructed using the fourth batch of control sequence data, the change trend of the quality index after the parameter adjustment is predicted based on the random forest regression, and the reliability of the adjustment scheme is evaluated using the cross-validation method to obtain the fifth batch of verification data. An adjustment parameter database is established according to the fifth batch of verification data. The database structure includes four data tables: power adjustment amount, valve opening change amount, adjustment execution order, and expected deviation correction rate. The control parameter adjustment scheme for the pyrolysis furnace is obtained from the data tables.

[0016] Further, the control parameter adjustment set is constrained and smoothed to obtain a control setpoint, which is formatted into a closed-loop feedback instruction and updated to the pyrolysis furnace control system to obtain the adjusted pyrolysis operation state data, including: extracting the heating power adjustment amount and the valve opening change amount according to the control parameter adjustment set, limiting the heating power adjustment range within 10% of the rated power, limiting the valve opening change range within 5% of the current opening, and using cubic spline interpolation to smooth the adjustment curve to obtain the first set of smoothing parameters. For the first set of smoothing parameters, the control upper limit is set using the rated power of the pyrolysis furnace heater and the maximum valve opening, and the control lower limit is set using the minimum startup power of the heater and the minimum valve opening. The safety interval for parameter adjustment is calculated from the control limits to obtain the second set of constraint parameters. A closed-loop controller is constructed according to the second set of constraint parameters, the proportional coefficient and the integral time constant are set, the control deviation is calculated using the sensor sampling data, and the control output amount is calculated from the deviation value to obtain the third set of control quantities. A control instruction structure is established for the third set of control quantities, the power control field and the valve control field of the instruction frame are set, the instruction is encoded using the communication protocol, and the control instruction sequence is generated from the encoded data to obtain the fourth set of instruction data. Control operations are performed according to the fourth set of instruction data, the temperature field, pressure field, and gas component data are collected in real time using a feedback sensor, and the collected data is denoised using Kalman filtering to obtain the fifth set of state data. The validity of the fifth set of state data is verified, the temperature field uniformity index and the pressure field stability index are calculated, the execution effect of the control instruction is judged, and the pyrolysis operation state database is updated from the verification result.

[0017] Further, extract the degradation rate of the carbon fiber bundle and the matrix decomposition ratio from the adjusted pyrolysis operating state data, verify the mass distribution of the pyrolysis products through gas analysis to obtain the mass evaluation data, compare it with the quality indicators of the target product, and use an optimization algorithm to update the parameter weights of the digital twin model to obtain an optimized pyrolysis control scheme, including: calculating the change in the mass of the carbon fiber bundle over time based on the pyrolysis operating state data, using an electronic balance measuring device to record the mass loss value per unit time, solving the corresponding relationship between temperature and degradation rate based on the Arrhenius reaction kinetics equation, and extracting the degradation rate of the carbon fiber bundle from the mass change curve to obtain the first set of reaction data. For the first set of reaction data, use a gas chromatograph to detect the content distribution of hydrocarbons in the pyrolysis gas, use a mass analyzer to measure the content of carbon atoms in the gaseous products, and calculate the decomposition amount of the matrix material from the carbon atom balance relationship to obtain the second set of conversion data. According to the second set of conversion data, use an X-ray diffractometer to determine the crystallinity of the carbon powder, use a scanning electron microscope to observe the morphological characteristics of the carbon powder, and calculate the product quality indicators from the carbon powder purity test results to obtain the third set of mass data. For the third set of mass data, establish a deviation function between the product quality indicators and the target indicators, use an adaptive particle swarm optimization algorithm to calculate the correction direction of the mass deviation, and construct an update rule for the digital twin parameters from the optimization results to obtain the fourth set of optimization data. Based on the fourth set of optimization data, use a genetic algorithm to search for the heating power distribution scheme, calculate the power setting values for each heating zone, and determine the intake valve opening sequence from the temperature field distribution to obtain the fifth set of control data. Establish a closed-loop control scheme according to the fifth set of control data, set the control period for the heating power and valve opening, generate standard control instructions from the control parameter sequence, and construct a pyrolysis process control database.

[0018] The technical solution provided by the embodiment of the present invention may include the following beneficial effects:

[0019] The present invention discloses a data processing method for a pyrolysis device combined with digital twin. By deploying a sensor array to obtain the initial state and environmental data, establishing a digital twin model and performing spatial registration, extracting pyrolysis characteristic parameters to drive the iterative calculation of the model, and predicting the state of the pyrolysis furnace and the quality indicators of the carbon powder. According to the deviation between the prediction result and the target indicator, adjust the control parameters and update them to the pyrolysis furnace control system. The present invention realizes the precise control of the pyrolysis process and the optimization of product quality through real-time monitoring and digital twin technology, improving the efficiency and quality of carbon fiber recycling. This method combines traditional pyrolysis technology with intelligent control, providing a new technical path for the efficient recycling and utilization of carbon fiber composites. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flowchart of a data processing method for a pyrolysis device combined with digital twin according to the present invention.

[0021] Figure 2 This is a schematic diagram of a cracking equipment data processing method combined with digital twins according to the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] like Figure 1-2 In this embodiment, a cracking equipment data processing method combined with digital twins may specifically include:

[0024] S101, placing a sensor array in the cracking furnace, collecting initial state data and environmental data of the carbon fiber composite material and performing time synchronization processing to generate a time-consistent data set.

[0025] At the beginning of the operation of the cracking furnace, it is necessary to obtain the real-time status of the carbon fiber composite material. To this end, a three-dimensional grid is divided according to the distance from the inner wall of the cracking furnace to the center of the composite material, and the grid spacing can be set according to the material size and monitoring accuracy requirements. For example, for a composite material plate with a size of 1000 mm × 1000 mm, a 50×50 grid can be divided on the surface, with a spacing of 20 mm between each grid, and measuring points are set every 5 mm in the thickness direction to form a three-dimensional sensing network. The sensor array uses K-type thermocouples for temperature measurement, with a range covering -200 to 1370 degrees Celsius and an accuracy of ±0.75%. At the same time, a strain gauge pressure sensor is configured with a range of 0 to 500 kPa and an accuracy of ±0.1%, and a six-channel gas chromatograph is used to detect the concentrations of components such as hydrogen and carbon monoxide.

[0026] S1011. Collect sensor signals through distributed data collectors to generate the first set of raw data, verify the validity and preliminarily process the data, and remove outliers to improve data reliability.

[0027] During the collection process, the distributed data collector is used to synchronously record the signals of each sensor to form the first set of raw data. To ensure the quality of the data, the validity verification range is set: the temperature data is limited to -50 to 2000 degrees Celsius, the pressure data is 0 to 1000 kPa, and the gas component concentration is 0 to 100%. The removal of abnormal values is based on the cracking process rules, such as the temperature change rate in the heating stage does not exceed 50 degrees Celsius per minute, the temperature difference between adjacent measuring points does not exceed 100 degrees Celsius, the pressure difference does not exceed 50 kPa, and the sum of gas components remains 100%. After verification, the second set of valid data is obtained and sorted by timestamp, laying the foundation for subsequent processing.

[0028] S1012. The second set of valid data is denoised using a Kalman filter and the time series correlation is calculated. Missing data are supplemented using linear interpolation to generate a time-consistent data set. A three-dimensional grid model is constructed to support spatial analysis.

[0029] In the data processing stage, the Kalman filter is applied to denoise the second set of valid data. The filter state vector includes temperature, pressure, and gas component concentration. The observation matrix is determined by the sensor positions, and the noise covariance is obtained through calibration experiments. After filtering, the temperature noise is reduced to ±2 degrees Celsius, and the pressure noise is reduced to ±1 kPa, significantly improving the data smoothness. Then, the Pearson correlation coefficient between each measurement point is calculated with a 15-minute time window, and the data groups with a correlation coefficient greater than 0.8 are synchronized. For missing data, quadratic polynomial interpolation is used for the temperature field, and linear interpolation is used for the pressure field and gas components with an interpolation interval of 1 second to generate the third set of time-synchronized data. Based on this, a three-dimensional spatial grid model is established using Kriging interpolation. The range parameter is determined through experimental variance analysis. For example, the range of the temperature field is 80 mm, and the interpolation error is controlled within 5 degrees Celsius.

[0030] The layout of the sensor array and data processing fully consider the dynamics of the pyrolysis process. Through high-precision acquisition and time synchronization technology, the time-consistent dataset can accurately reflect the state changes of the carbon fiber composite material, providing reliable data support for subsequent digital twin modeling. The obtained dataset is stored in a time-series database, sliced by hour, with each slice containing 36,000 data points. The data integrity is ensured through batch writing and transaction mechanisms, and the query response time is less than 100 milliseconds, laying a foundation for real-time monitoring.

[0031] S102. Obtain the geometric structure, material properties, and process principle of the pyrolysis furnace to establish a digital twin model, and perform a spatial registration operation using the time-consistent dataset. Map the physical positions of the sensor array to the virtual space through a preset spatial coordinate mapping algorithm to generate a spatial registration dataset.

[0032] To achieve an accurate simulation of the pyrolysis process, first, a digital twin model is constructed based on the physical characteristics of the pyrolysis furnace. A laser scanner is used to collect point cloud data from multiple angles with a scanning resolution of 0.5 mm and a total of 5 million points. The complete geometric structure of the pyrolysis furnace is generated through multiple scans and stitching. Subsequently, the point cloud data is imported into 3D modeling software, and the first-layer surface mesh of the inner cavity wall is generated using the Marching Cubes algorithm with an element size set to 2 mm. The initial mesh contains approximately 1 million triangular elements. To reflect the thermal behavior of the material, the thermal conductivity, specific heat capacity, and density data are extracted from the carbon fiber composite material parameter library. The thermal conductivity increases from 10 W / (m·K) to 25 W / (m·K) with temperature, the specific heat capacity increases from 800 J / (kg·K) to 1500 J / (kg·K), and the density remains stable at 1600 kg / m³.

[0033] S1021. Use a tetrahedral mesh generator to divide the inner cavity of the cracking furnace and the composite material into grids to generate the second layer of grids. Adopt a mesh quality detection and optimization algorithm to improve the applicability of the grids. At the same time, extract feature points according to the physical positions of the sensors and calculate unified coordinate data to lay a foundation for subsequent registration.

[0034] In the grid division stage, import the first layer of grids into the tetrahedral mesh generator, perform three-dimensional division on the internal space of the cracking furnace and the carbon fiber composite material to generate the second layer of volume grids. To ensure the simulation accuracy, conduct a quality assessment on the second layer of grids, calculate the aspect ratio, distortion, and orthogonality indicators, and set the aspect ratio threshold to 5, the distortion threshold to 0.7, and the orthogonality threshold to 0.3. The unqualified grid elements are mainly concentrated in areas with drastic curvature changes such as the feed inlet and the discharge outlet, accounting for about 15%. Locally encrypt these areas, divide the unqualified elements into 4 sub-elements, halve the side length, and adjust the node positions through the Laplace smoothing algorithm. After optimization, the third layer of grids is obtained. At the same time, according to the 8×8×4 layout of the sensor array, with a horizontal spacing of 50 mm and a vertical spacing of 100 mm, use the Harris corner detection algorithm to identify the feature points of the sensor positions. The algorithm response threshold is set to 0.1, the non-maximum suppression window is 3×3 pixels, and the feature point positioning accuracy is better than 0.5 mm. Taking the first sensor in the array as the origin, determine the basis vectors along the arrangement direction, and calculate the transformation from the local coordinate system to the global coordinate system through the Euler angle rotation matrix to generate the first set of unified coordinate data.

[0035] S1022. Use the first set of unified coordinate data and the third layer of grids to perform iterative closest point registration. Solve the spatial transformation matrix by the least squares method and perform deformation optimization to generate the fourth layer of grids. At the same time, construct a radial basis function interpolator to map physical quantities to the virtual space and establish a database of physical quantities and state parameters.

[0036] Based on the first set of unified coordinate data and the third - layer grid, the Iterative Closest Point (ICP) registration algorithm is used to calculate the Euclidean distance matrix between sensor nodes and grid nodes. The maximum number of iterations of the algorithm is set to 50, and the convergence threshold is 0.01 mm. In each iteration, the closest point pairs are first matched, and the rigid - body transformation matrix is solved through singular - value decomposition. Then, the spatial transformation is optimized based on the least - squares criterion to obtain the preliminary registration result. To further improve the accuracy, cubic spline deformation is performed based on the thin - plate spline function. The control points are selected from the grid nodes within a radius of 20 mm around the sensor to generate the fourth - layer grid. For the fourth - layer grid, a registration error evaluation index is defined, and the average deviation between the sensor position and the grid nodes is calculated. Through leave - one - out cross - validation, 80% of the sensor points are randomly selected for registration, and the remaining 20% are used for verification. The average error is controlled within 1 mm, and the maximum error in 95% of the area is less than 2 mm. The control - point density is increased in the areas where the deviation exceeds the standard to optimize and obtain the fifth - layer grid. Subsequently, the corresponding node positions of the sensor are extracted from the fifth - layer grid to construct a radial - basis - function interpolator. The influence radius of the Gaussian kernel function is set to twice the sensor spacing, and interpolation is performed on the temperature field, pressure field, and gas - component field respectively. To avoid dimensional differences, the input data are pre - normalized, and the state parameters of the corresponding nodes in the virtual space are calculated. Finally, a mapping database in the form of key - value pairs is established, where the key is the sensor coordinate, and the value is the temperature, pressure, and gas - component data.

[0037] The spatial registration operation is realized through multi - level grid optimization and precise algorithms, ensuring a high degree of consistency between the digital twin model and the physical cracking furnace state. The improvement of grid quality and the control of registration accuracy make the mapping of physical quantities such as temperature field and pressure field in the virtual space more realistic, providing a solid foundation for subsequent dynamic simulations. This step does not overly limit the specific implementation details of the algorithm, and technicians can adjust the optimization strategy according to actual needs.

[0038] S103: Extract pyrolysis characteristic parameters from the spatial registration dataset and calculate the reaction rate and pyrolysis behavior changes at the micro - scale to generate a set of pyrolysis characteristic parameters.

[0039] Obtain the temperature - field data of the interface between carbon - fiber bundles and the matrix from the spatial registration dataset, and use the central - difference method to calculate the interface temperature - gradient distribution. To ensure the calculation accuracy, the adjacent measurement - point spacing is selected as 0.5 mm, and the temperature measurement range is set from 200 to 1000 degrees Celsius. The central - difference method calculates the gradient through the ratio of the temperature difference between adjacent measurement points to the distance. The interface temperature gradient can reach 200 degrees Celsius per millimeter at the initial stage of pyrolysis and gradually flattens to 50 degrees Celsius per millimeter in the later stage. To evaluate the error, analyze the temperature deviation between adjacent measurement points, which is usually controlled within ±5 degrees Celsius, and the converted gradient error is about 10 degrees Celsius per millimeter. Subsequently, based on Fourier's law of heat conduction, calculate the interface heat - flux density, combine the temperature gradient and the thermal conductivity of the material to generate the first batch of heat - conduction data reflecting the heat - transfer intensity for subsequent gas - diffusion analysis.

[0040] S1031. Calculate the interfacial gas concentration distribution based on the first batch of heat conduction data combined with the gas component monitoring results, deduce the gas generation rate through the principle of mass conservation, and construct an Arrhenius reaction equation to characterize the reaction kinetic characteristics.

[0041] Use a four-channel on-line gas monitor to collect the concentrations of components such as hydrogen, carbon monoxide, and methane at the interface, with a monitoring frequency of 1 Hz. Extract the heat flux density from the first batch of heat conduction data, and calculate the gas concentration distribution using Fick's diffusion law. Fick's law determines the gas migration amount per unit time based on the product of the concentration gradient and the diffusion coefficient. The calculation results show that within 10 mm from the interface, the hydrogen concentration decreases from 15% to 5%, and the carbon monoxide concentration decreases from 25% to 8%. According to the change in concentration, estimate the decomposition amount of the substrate per unit time. For example, the decomposition rate reaches 0.2 g / (cm²·min) at 800 °C. Then, based on the principle of mass conservation, calculate the gas generation rate by the difference between the initial mass of the substrate and the real-time gas generation amount. To reveal the relationship between temperature and reaction rate, establish an Arrhenius reaction equation, and obtain the pre-exponential factor and activation energy in the equation through least squares fitting. Divide the temperature into a low-temperature region of 200 to 400 °C, a medium-temperature region of 400 to 700 °C, and a high-temperature region of 700 to 1000 °C. The fitting results show that the activation energies are 80 kJ / mol, 120 kJ / mol, and 150 kJ / mol respectively, generating the second batch of reaction kinetic data.

[0042] S1032. Based on the second batch of reaction kinetic data, construct a mapping function between temperature and reaction rate, calculate the interfacial movement displacement and velocity, and generate a set of pyrolysis characteristic parameters.

[0043] Extract the reaction rate from the second batch of reaction kinetics data, construct the mapping function between temperature and reaction rate, and calculate the interface movement displacement through numerical integration. To improve the calculation accuracy, the interface region is divided into a 100×100 grid, and the spatial discretization method is used to monitor the displacement change. The calculation shows that the interface movement speed in the initial stage of pyrolysis is about 0.01 mm per minute, and it increases to 0.1 mm per minute when the temperature rises to 800 °C. By further deriving the interface movement speed from the relationship between the displacement change and time, it is found that for every 0.1 g / (cm²·min) increase in the reaction rate, the movement speed increases by about 0.02 mm per minute. Based on this, a set of pyrolysis characteristic parameters including temperature gradient, gas concentration, reaction rate, and interface displacement is generated. For easy management, the parameter set is stored in a database with a timestamp index, and the hierarchical storage structure ensures data accuracy: the temperature gradient is accurate to 0.1 °C / mm, the gas concentration is accurate to 0.1%, the reaction rate is accurate to 0.01 g / (cm²·min), and the displacement is accurate to 0.001 mm. The database update frequency is 10 Hz, the data is processed by a lossless compression algorithm with a compression ratio of 5:1, and the retrieval response time is less than 100 ms.

[0044] Through multi-physics field coupling and kinetic analysis, the extraction of pyrolysis characteristic parameters is more comprehensive, reflecting the microscopic behavior of the interaction between carbon fiber bundles and the matrix. The introduction of the Arrhenius equation makes the correlation between the reaction rate and temperature clearer, providing a theoretical basis for the calculation of the interface movement speed. The efficient storage and fast access capabilities of the parameter database support the real-time monitoring and optimization adjustment of the pyrolysis process. This step does not limit the specific numerical range too much and can be adjusted by technicians according to the actual pyrolysis conditions.

[0045] S104. Use the pyrolysis characteristic parameter set to drive the digital twin model for high-frequency iterative calculations, simulate the temperature field, pressure field, and material degradation behavior in the pyrolysis furnace, predict the carbon powder purity, particle size distribution, and residual matrix content, generate the pyrolysis furnace state prediction results, and ensure the prediction accuracy through multi-field coupling calculations and error optimization.

[0046] Based on the pyrolysis characteristic parameter set, the calculation domain grid is divided, and the grid node spacing is determined by the temperature and pressure gradients. In regions with large gradients, such as where the temperature change exceeds 100 °C / mm or the pressure change exceeds 10 kPa / mm, a refined grid is used, and the minimum size is set to 0.1 mm. The control equations for the temperature field and pressure field are constructed using the finite element method, considering heat conduction, convection, and radiation heat transfer mechanisms, and the fourth-order Runge-Kutta integration method is used to calculate the distribution state within the time step. The initial time step is set to 0.01 s and is adaptively adjusted according to the temperature change rate, with a minimum step size of up to 0.001 s. The calculation results show that heat conduction is the main mechanism in the initial stage of pyrolysis, and the proportion of radiation heat transfer increases significantly after the temperature rises to 600 °C, generating the first batch of physical field data, which provides a basis for subsequent degradation analysis.

[0047] S1041. Calculate the material degradation rate and fiber bundle stress distribution for the first batch of physical field data. Through gas-solid interface stress analysis and mass transfer calculation, obtain the carbon element migration rate and carbon powder particle size distribution, and gradually deduce the carbon powder quality index.

[0048] Extract the temperature field and pressure field distributions from the first batch of physical field data, and calculate the material degradation rate by combining the coupling effect of temperature and pressure. Under the conditions of 800 degrees Celsius and 200 kPa, the degradation rate can reach 0.5 grams per square centimeter per minute. Calculate the fiber bundle stress distribution using the anisotropic constitutive relationship, with the Young's modulus in the fiber direction set to 240 GPa and the transverse modulus to 20 GPa. Through stress tensor analysis, the stress distribution conforms to the Weibull distribution, with a characteristic stress of 2 GPa and a Weibull modulus of 10, and then calculate the material fracture probability. Based on the fracture probability, use the slip boundary condition to calculate the gas-solid interface stress, and the tangential stress is usually less than 0.1 MPa. Combining the law of conservation of mass, calculate the diffusion flux of gas-phase components by multiplying the concentration gradient and the diffusion coefficient, and the diffusion coefficient increases to 0.1 square centimeters per second when the temperature rises to 1000 degrees Celsius. Deduce the carbon element migration rate from the flux data and find that it has an exponential relationship with temperature, with an activation energy of about 150 kJ / mol. Then, use the Gaussian mixture clustering algorithm to analyze the carbon powder particle size probability density distribution, set 3 Gaussian components, and optimize the mixture weights by the expectation maximization algorithm. The particle size distribution shows a bimodal characteristic, with the main peak at 5 microns and the secondary peak at 20 microns, divided into 5 particle size intervals with a width of 10 microns, and generate the second batch of product data.

[0049] Through the second batch of product data, calculate the carbon powder content using the carbon element balance relationship, and consider the influence of the residual matrix and impurities to estimate the carbon powder purity, which ranges from 95% to 99.5%. To further analyze the residual matrix, construct a support vector regression model to predict the matrix decomposition rate, select the radial basis function as the kernel function, and optimize the parameters through grid search. The prediction results show that the matrix content decays exponentially with pyrolysis time, with a time constant of about 30 minutes, and gradually decreases from the center to the edge in terms of spatial distribution, generating the third batch of degradation data.

[0050] S1042. Iteratively optimize the digital twin model based on multi-batch data, adjust the boundary conditions of the control equation to reduce the prediction error, and generate high-precision pyrolysis furnace state prediction results.

[0051] Compare the first batch of physical field data, the second batch of product data, and the third batch of degradation data with the measured values, and calculate the error using the least squares method. The initial error may be manifested as a temperature field deviation of ±10 °C, a pressure field deviation of ±5 kPa, and a toner purity deviation of ±0.5%. By adjusting the boundary conditions of the control equations, such as heat flux boundaries or pressure constraints, iterative calculations are performed repeatedly, with an interval of 0.1 s for each iteration. Under typical operating conditions, 50 to 100 iterations are required, and the total duration covers a 4-hour pyrolysis process. The convergence criterion is set as a relative temperature field error of less than 0.1%, a relative pressure field error of less than 1%, and a relative toner purity error of less than 0.5%. After optimization, the model can accurately reflect the dynamic state inside the pyrolysis furnace, providing a reliable basis for process control.

[0052] High-frequency iterative calculations and multi-field coupling analysis enable the digital twin model to truly simulate the pyrolysis process, and the accuracy of the prediction results is significantly improved through error optimization. The quantitative analysis of toner particle size and purity provides data support for product quality control, while the prediction of matrix residue content helps to evaluate resource utilization efficiency. The specific parameters of this step can be flexibly adjusted by technicians according to the actual operating conditions.

[0053] S105. If there is a deviation between the predicted result of the pyrolysis furnace state and the preset target toner quality index, determine the adjustment directions of the heating power distribution and the intake valve opening by analyzing the deviation degree, generate an adjustment set of control parameters for the pyrolysis furnace, and at the same time use multi-objective optimization and prediction algorithms to ensure the effectiveness and reliability of the adjustment plan.

[0054] When the prediction result shows that the toner quality index deviates from the target value, first calculate the specific value of the deviation. The target toner quality index includes a purity range set above 98%, a particle size range of 20 to 50 microns, and a maximum residue content not exceeding 2%. By comparing with the prediction result, calculate the purity deviation value, the particle size distribution deviation value, and the residue content deviation value respectively. Establish a weighted sum of squares error function using the multi-objective weighting method, where the purity weight is 0.5, the particle size distribution weight is 0.3, and the residue content weight is 0.2. If the total deviation degree exceeds 0.1, trigger the adjustment process. The deviation calculation is based on a percentage form. For example, when the purity is lower than 98%, it is calculated according to the difference ratio; when the particle size distribution deviates from the target interval, it is calculated according to the distribution ratio outside the interval; when the residue content exceeds the standard, it is calculated according to the absolute value of the excess part, generating the first batch of quality deviation data to provide a basis for optimization.

[0055] S1051. Use the gradient descent method to determine the heating power adjustment direction for the first batch of quality deviation data and divide the adjustment area. At the same time, calculate the flow field adjustment amount based on the gas diffusion law, and gradually deduce the change trend of the control parameters.

[0056] Based on the first batch of quality deviation data, the adjustment direction of the heating power is calculated using the gradient descent method. The cracking furnace is longitudinally divided into 4 independent heating zones, and each zone is equipped with a separate heating element. The temperature field uniformity is evaluated by the standard deviation of the measured point temperatures. The area where the standard deviation exceeds 20 degrees Celsius is determined as the non-uniform area and needs to be adjusted for power preferentially. The adjustment ratio is positively correlated with the temperature gradient. For example, when the gradient increases by 100 degrees Celsius per meter, the power increases by 10%, generating the second batch of power adjustment data. Then, according to the change in the cracking gas components caused by the power adjustment, the concentration distribution in the furnace is calculated using the gas diffusion equation. The diffusion process comprehensively considers convection and molecular diffusion, and the diffusion coefficient is approximately 0.1 square centimeters per second at 800 degrees Celsius. The concentration uniformity is judged by the coefficient of variation. If the coefficient is greater than 0.2, it is considered that the distribution is uneven and the intake valve opening needs to be adjusted. The change in the opening is inversely proportional to the concentration gradient. For example, when the concentration gradient increases by 5%, the opening decreases by 2%, generating the third batch of flow field adjustment data. This process aims to quickly correct the quality deviation through the coordinated optimization of the temperature and the flow field.

[0057] S1052. Establish the state transition equation of the pyrolysis process using the third batch of flow field adjustment data and generate the control sequence. At the same time, predict the adjustment effect through random forest regression and construct the parameter database to provide support for closed-loop control.

[0058] Starting from the third batch of flow field adjustment data, construct the state transition equation of the pyrolysis process to describe the dynamic responses of the temperature field, pressure field, and gas components before and after parameter adjustment. The response time varies due to differences in physical processes. The pressure field adjustment has the fastest response, about 1 minute, the gas component adjustment takes 5 minutes, and the temperature field adjustment takes 10 minutes. Determine the timing of parameter adjustment based on the response speed. First, adjust the parameters with fast response such as the pressure field, and then adjust the temperature field to ensure a smooth transition of the process, generating the fourth batch of control sequence data. Then, construct the parameter optimization function based on the fourth batch of control sequence data and use random forest regression to predict the change in the quality index after adjustment. The model is trained with historical operation data. The input features include the current state variables and adjustment parameters, and the output is the adjusted index value. The accuracy is evaluated through 5-fold cross-validation. The purity prediction error is controlled within ±0.5%, the particle size distribution error is within ±2 microns, and the residue content error is within ±0.2%, generating the fifth batch of verification data. To efficiently manage the adjustment plan, establish a parameter database, which includes four data tables: the power adjustment amount, the change in valve opening, the execution order, and the expected deviation correction rate. The update period is 1 hour, and the last 100 adjustment records are retained to optimize the strategy.

[0059] Through deviation analysis and multi-level optimization, the control parameter adjustment set can quickly respond to quality deviations and improve the stability of toner quality. The combination of gradient descent method and random forest regression ensures the scientific nature of the adjustment direction and the predictability of the effect. The real-time update of the parameter database provides data support for the continuous improvement of the process, and the specific adjustment range can be flexibly set by technicians according to actual needs.

[0060] S106. Constrain and smooth the control parameter adjustment set to generate a closed-loop feedback instruction and update the control system, and at the same time optimize the operating state of the pyrolysis process to ensure that the pyrolysis effect after adjustment is stable and controllable.

[0061] Extract the heating power adjustment amount and the valve opening change amount based on the control parameter adjustment set. To avoid sudden changes in equipment operation, limit the power adjustment range within 10% of the rated power, and limit the valve opening change range within 5% of the current opening. Use cubic spline interpolation to smooth the adjustment curve, set the control point interval to 10 seconds, ensure smooth transition of parameter changes, avoid drastic fluctuations in temperature or flow rate, and generate the first set of smoothed parameters. This process aims to balance the adjustment speed and system stability and provide a reliable basis for safety constraints.

[0062] For the first set of smoothed parameters, determine the safe operating range in combination with the physical characteristics of the pyrolysis furnace heater and the valve. Taking a heater with a rated power of 100 kW as an example, the minimum starting power is set at 20 kW, the maximum power is limited to 95 kW, the minimum valve opening is 10% to maintain basic gas flow, and the maximum is 90% to reserve adjustment margin. The parameter change slope is controlled within 2% per second to match the equipment response time. Calculate the safe range of parameter adjustment through upper and lower limit constraints to generate the second set of constraint parameters, ensuring that the adjustment will not exceed the equipment's tolerance or cause operation risks.

[0063] S1061. Build a closed-loop controller based on the second set of constraint parameters and calculate the control output. At the same time, use feedback sensors to collect data and perform noise reduction processing to generate state data reflecting the adjustment effect to optimize the operating state.

[0064] A closed-loop controller is designed using a second set of constraint parameters. The proportional-integral control strategy is adopted, with the proportional coefficient set to 0.8 and the integral time constant set to 60 seconds to achieve a balance between fast response and error elimination. The controller calculates the control deviation by sampling data from sensors. The sampling period is set to 1 second, and the deviation is calculated based on the difference between the measured value and the set value of the temperature field. The output signal range is from 0 to 100%, corresponding to the actual adjustment range of the heater and the valve. To compensate for the sensor response delay, feedforward control is introduced, and the delay time is set to 5 seconds according to the device characteristics. The calculated control output generates a third set of control quantities. Subsequently, the third set of control quantities is formatted into control instructions. The instruction frame contains a power control field and a valve control field. The power resolution is 0.1 kW, and the valve opening resolution is 0.1%. It is encoded through a standard communication protocol with a baud rate of 9600, and 16-bit cyclic redundancy check is used to ensure the transmission reliability, generating a fourth set of instruction data. After the instructions are executed, the temperature field, pressure field, and gas component data are collected in real-time using a distributed sensor network. The accuracy of the temperature sensor is 0.1 °C, the accuracy of the pressure sensor is 0.1 kPa, and the sampling period of the gas chromatograph is 10 seconds. The collected data is denoised by a Kalman filter. The state noise covariance is set to 1% of the measured value, and the observation noise covariance is 0.1%, effectively reducing the noise interference and generating a fifth set of state data.

[0065] The validity of the fifth set of state data is verified, and the temperature field uniformity and pressure field stability indicators are calculated. The temperature field uniformity requires that the temperature difference between adjacent measurement points is less than 20 °C, and the pressure field stability requires that the fluctuation amplitude is less than 5 kPa. If an abnormality occurs, such as an excessive temperature difference or excessive pressure fluctuation, an alarm is triggered and the state information at the abnormal moment is recorded. The qualified verified data is updated to the pyrolysis operation status database with an update period of 1 minute, retaining the records of the most recent 24 hours, supporting fast retrieval by timestamp or parameter type, providing data support for subsequent process optimization.

[0066] The constraint and smoothing processing of the control parameters ensure the safety and stability of the adjustment process. The closed-loop feedback mechanism optimizes the operation status through real-time data, improving the control accuracy and response ability of the pyrolysis furnace. The application of the Kalman filter further improves the reliability of the state data, laying a foundation for achieving high-quality pyrolysis products. The specific parameter settings can be adjusted and optimized by technicians according to the actual situation of the device.

[0067] S107: Extract key reaction parameters from the adjusted pyrolysis operation status data and evaluate the quality of the pyrolysis products. Update the digital twin model through multi-dimensional analysis and optimization algorithms to generate an optimized pyrolysis control scheme to improve the process effect.

[0068] Obtain the mass change information of the carbon fiber bundle from the adjusted pyrolysis operating state data, use an electronic balance to record the mass loss value per unit time in real time, and calculate the relationship between temperature and degradation rate in combination with the Arrhenius reaction kinetic equation. The measurement by the electronic balance shows that the peak degradation rate is 0.5 grams per minute at 800 degrees Celsius, and then gradually slows down with the increase in temperature. The Arrhenius equation determines the activation energy to be 150 kJ / mol and the frequency factor to be 10 to the power of 12 per second through fitting. Extract the degradation rate of the carbon fiber bundle from the mass loss curve to generate the first set of reaction data. This process not only quantifies the pyrolysis progress but also lays a foundation for subsequent matrix decomposition analysis.

[0069] S1071. Analyze the cracked gas components of the first set of reaction data and calculate the matrix decomposition ratio. At the same time, evaluate the product quality through the microscopic characteristics of the carbon powder to generate multi-dimensional quality data to verify the pyrolysis effect.

[0070] Use a gas chromatograph to detect the cracked gas components in the first set of reaction data and find that methane accounts for about 30%, ethane accounts for 15%, and benzene compounds account for 10%. The gas components show regularity with temperature changes, with small molecule hydrocarbons mainly in the low-temperature stage and benzene ring compounds increasing in the high-temperature stage. Measure the carbon atom content in the gaseous products using a mass analyzer, with a mass fraction of 75%. Calculate the matrix decomposition ratio through the carbon atom balance relationship, and the decomposition degree can reach 95% in the later stage of pyrolysis to generate the second set of conversion data. Based on the second set of conversion data, use an X-ray diffractometer to determine the crystallinity of the carbon powder. The characteristic peak appears at 2θ angle of 26 degrees, and the full width at half maximum is about 0.8 degrees. The calculated crystallinity is 85%. Observe the morphology of the carbon powder through a scanning electron microscope. The particles are in the shape of short fibers, with a length distribution of 10 to 50 microns and a diameter of 5 to 10 microns. The purity test shows that the carbon content reaches 98.5%, and the main impurities are hydrogen and oxygen to generate the third set of quality data. These data comprehensively reflect the physical and chemical properties of the cracked products and provide a basis for quality optimization.

[0071] Compare the third set of quality data with the target indicators and establish a deviation function, including purity, particle size, and residual matrix, with a weight ratio set to 6:3:1. Use the adaptive particle swarm optimization algorithm to calculate the deviation correction direction, set the population size to 50, the number of iterations to 200, and the inertia weight to decrease from 0.9 to 0.4. The optimization results show that when the heating power distribution is parabolic and the power density in the central region is 30% higher than that at the edge, the quality indicators are closest to the target values. Then, based on the optimization direction, use the genetic algorithm to search for the best heating power distribution scheme. The chromosome uses real number coding, with a length twice the number of heating zones, a crossover probability of 0.8, a mutation probability of 0.1, and a population size of 100. After 500 generations of evolution, the power setting values of each heating zone fluctuate between 40 and 90 kW, and the intake valve opening is adjusted between 30% and 70% to generate the fourth set of control data.

[0072] S1072. Construct a closed-loop control scheme based on the fourth set of control data and update the database to improve the stability of the cracking process and the consistency of product quality through the optimized parameters.

[0073] Design a closed-loop control scheme according to the fourth set of control data. The control period is set to 10 seconds, and the adjustment of heating power and valve opening is completed within each period. The control instructions follow the standard industrial communication protocol, and the data frame contains a 16-bit power control word and an 8-bit valve control word to ensure the efficiency and accuracy of instruction transmission. The optimized control parameter sequence is stored in the cracking process control database, which adopts a time series structure to support retrieval by timestamp and parameter type. The data compression ratio reaches 10:1, improving the storage efficiency and access speed. Through repeated verification and adjustment, ensure that the quality of the cracking products is stably achieved the target indicators, while enhancing the adaptability and controllability of the process.

[0074] The combination of multi-dimensional quality assessment and optimization algorithms enables the digital twin model to continuously improve according to real-time data. The generated control scheme effectively improves the carbon powder purity and the uniformity of particle size distribution. The optimized heating power distribution and valve opening sequence further reduce the residual matrix content, providing technical support for the efficient recycling of carbon fiber composites. The specific optimization parameters can be flexibly adjusted according to the actual working conditions to meet different requirements.

[0075] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor limit the present invention to the specific implementation manners. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification to better explain the principle and practical application of the present invention, so that those skilled in the art in the relevant technical field can well understand and utilize the present invention. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A data processing method for a cracking device combined with digital twins, characterized in that The method includes: Obtaining the initial state data and environmental data of the carbon fiber composite material in the pyrolysis furnace, and performing time synchronization processing to obtain a time-consistent data set; Obtaining the geometric structure, material properties, and pyrolysis process principle of the pyrolysis furnace, establishing a digital twin model, and performing a spatial registration operation according to the time-consistent data set. By using a preset spatial coordinate mapping algorithm, the physical positions of the sensor arrays are corresponded to the virtual space in the digital twin model to obtain a spatial registration data set; Extracting the pyrolysis characteristic parameters of the interaction between carbon fiber bundles and the matrix from the spatial registration data set, calculating the reaction rate and pyrolysis behavior changes at the microscale of the fiber-matrix interface to obtain a pyrolysis characteristic parameter set; Using the pyrolysis characteristic parameter set to drive the digital twin model for high-frequency iterative calculations to obtain the dynamic update state of the digital twin, simulating the temperature field, pressure field, and material degradation state in the pyrolysis furnace, and predicting the carbon powder purity distribution, particle size distribution prediction, and residual matrix content to obtain the pyrolysis furnace state prediction result; If there is a deviation between the prediction result and the preset target carbon powder quality index, then based on the deviation degree, determine the adjustment amount of the heating power distribution and the change trend of the intake valve opening to obtain a control parameter adjustment set for the pyrolysis furnace; Perform constraint and smoothing processing on the control parameter adjustment set to obtain a control set value, format it into a closed-loop feedback instruction, and update it to the pyrolysis furnace control system to obtain the adjusted pyrolysis operation state data; Extract the degradation rate of the carbon fiber bundle and the matrix decomposition ratio from the adjusted pyrolysis operation state data, verify the mass distribution of the pyrolysis products through gas analysis to obtain mass evaluation data, compare it with the target product quality index, and use an optimization algorithm to update the parameter weights of the digital twin model to obtain an optimized pyrolysis control scheme.

2. The method according to claim 1, characterized in that, The obtaining of the initial state data and environmental data of the carbon fiber composite material in the pyrolysis furnace, and performing time synchronization processing to obtain a time-consistent data set includes: Arranging a thermocouple sensor array according to the distance from the inner wall of the pyrolysis furnace to the center grid points of the composite material, and using a distributed data collector to collect sensor signals to obtain a first set of original data; Performing denoising processing on the first set of original data using a Kalman filter, where the state vector of the Kalman filter includes temperature values, pressure values, and gas component concentration values, to obtain a second set of filtered data; Calculating the temporal correlation coefficient matrix between points of the sensor array according to the second set of filtered data, and using linear interpolation to supplement the missing data in the correlation coefficient matrix to obtain a third set of time-synchronized data; Establishing a three-dimensional spatial grid model for the third set of time-synchronized data, calculating the spatial correlation between grid nodes using a spherical semivariogram model, and performing interpolation calculations on the temperature field, pressure field, and gas component distribution at the grid nodes to obtain a time-consistent data set.

3. The method according to claim 1, wherein The obtaining of the geometric structure, material properties, and pyrolysis process principle of the pyrolysis furnace, establishing a digital twin model, and performing a spatial registration operation according to the time-consistent data set. By using a preset spatial coordinate mapping algorithm, the physical positions of the sensor arrays are corresponded to the virtual space in the digital twin model to obtain a spatial registration data set includes: Generate the first-layer mesh of the inner cavity wall of the cracking furnace using point cloud data, and obtain the second-layer mesh through a tetrahedral mesh divider based on the first-layer mesh; Identify the characteristic points of the sensor layout position for the second-layer mesh using the Harris corner detection algorithm, and calculate the first set of unified coordinate data through the Euler angle rotation matrix based on the characteristic point coordinates; Calculate the Euclidean distance matrix from the sensor nodes to the mesh nodes using the iterative closest point registration algorithm based on the first set of unified coordinate data, and obtain the fourth-layer mesh by solving the spatial transformation matrix through the least squares criterion; Construct a radial basis function interpolator for the spatial positions of the corresponding sensor nodes in the fourth-layer mesh, and calculate the state parameters of the corresponding nodes in the virtual space of the actual sensor temperature field, pressure field, and gas component field through the interpolator to obtain the spatial registration data set.

4. The method according to claim 1, wherein Extract the pyrolysis characteristic parameters of the interaction between the carbon fiber bundle and the matrix from the spatial registration data set, calculate the reaction rate and the change in pyrolysis behavior at the microscale of the fiber matrix interface, and obtain the pyrolysis characteristic parameter set, including: Calculate the interface temperature gradient distribution using the central difference method based on the temperature field data of the carbon fiber bundle and the matrix interface in the spatial registration data set, and obtain the interface heat flux density value from the temperature gradient distribution; Calculate the gas concentration distribution at the interface using Fick's diffusion law for the interface heat flux density value, and obtain the amount of matrix decomposition per unit time from the gas concentration distribution; Calculate the gas generation rate using the law of conservation of mass based on the amount of matrix decomposition, and establish the Arrhenius reaction equation from the corresponding relationship between the gas generation rate and temperature; Construct a mapping function between temperature and reaction rate for the Arrhenius reaction equation, calculate the interface movement displacement from the reaction rate, and form the pyrolysis characteristic parameter set.

5. The method according to claim 1, characterized in that Drive the digital twin model to perform high-frequency iterative calculations using the pyrolysis characteristic parameter set, obtain the dynamic update state of the digital twin body, simulate the temperature field, pressure field, and material degradation state in the cracking furnace, predict the carbon powder purity distribution, particle size distribution prediction, and residual matrix content, and obtain the cracking furnace state prediction result, including: Divide the mesh nodes of the computational domain according to the pyrolysis characteristic parameter set, construct the control equations for the temperature field and pressure field using the finite element method, and obtain the temperature field and pressure field distribution data from the control equations; Calculate the material degradation rate for the temperature field and pressure field distribution data, calculate the stress distribution of the fiber bundle using the stress tensor, and obtain the material fracture probability data from the stress distribution; Calculate the stress distribution at the gas-solid two-phase interface based on the material fracture probability data, calculate the gas-phase component diffusion flux using the law of conservation of mass, and obtain the carbon element migration rate data from the diffusion flux; Calculate the probability density distribution of the carbon powder particle size using the Gaussian mixture clustering algorithm for the carbon element migration rate data, calculate the carbon powder content using the carbon element balance relationship, and obtain the carbon powder purity data from the carbon powder content; Use the least squares method to compare the error between the calculation result and the measured data, and obtain the cracking furnace state prediction result.

6. The method according to claim 1, wherein If there is a deviation between the predicted result and the preset target toner quality index, the adjustment amount of the heating power distribution and the change trend of the intake valve opening are determined based on the deviation degree, and a control parameter adjustment set for the pyrolysis furnace is obtained, including: Obtain the toner purity deviation value, particle size distribution deviation value, and residue content deviation value for the toner quality index, and establish a weighted sum of squared errors function based on the three deviation values to obtain the quality deviation data; Use the gradient descent method to calculate the heating power adjustment direction according to the quality deviation data, and divide the heating power adjustment area through the temperature field uniformity index to obtain the power adjustment data; Calculate the change trend of the pyrolysis gas components using the power adjustment data, and calculate the gas concentration distribution in the furnace according to the gas diffusion equation to obtain the flow field adjustment data; Establish a state transition equation for the pyrolysis process for the flow field adjustment data, and obtain the control sequence data through the state response relationship before and after the adjustment of the process parameters; Construct a parameter optimization function according to the control sequence data, and use random forest regression to predict the change trend of the quality index after parameter adjustment to obtain the control parameter adjustment set for the pyrolysis furnace.

7. The method according to claim 1, wherein Perform constraint and smoothing processing on the control parameter adjustment set to obtain the control setpoint, format it into a closed-loop feedback command, and update it to the pyrolysis furnace control system to obtain the adjusted pyrolysis operation state data, including: Obtain the heating power adjustment amount and valve opening change amount according to the control parameter adjustment set, and perform smoothing processing on the adjustment curve using cubic spline interpolation to obtain the first set of smoothing parameters; For the first set of smoothing parameters, calculate the parameter adjustment safety interval from the rated power of the pyrolysis furnace heater and the upper and lower limits of the valve opening to obtain the second set of constraint parameters; Construct a closed-loop controller according to the second set of constraint parameters, calculate the control deviation value using the sensor sampling data, and obtain the third set of control quantities from the deviation value; For the third set of control quantities, collect the temperature field and pressure field data through the feedback sensor, perform noise reduction processing on the collected data using the Kalman filter and then perform validity verification, calculate the temperature field uniformity index and the pressure field stability index, judge the execution effect of the control command, and obtain the adjusted pyrolysis operation state data from the verification result.

8. The method according to claim 1, wherein Extract the degradation rate of the carbon fiber bundle and the matrix decomposition ratio from the adjusted pyrolysis operation state data, verify the mass distribution of the pyrolysis products through gas analysis to obtain the quality evaluation data, compare it with the target product quality index, and use the optimization algorithm to update the parameter weights of the digital twin model to obtain the optimized pyrolysis control scheme, including: Obtain the mass loss value signal recorded by the electronic balance measurement device, and solve the corresponding relationship between temperature and degradation rate according to the mass loss value signal and the Arrhenius reaction kinetics equation to obtain the first set of reaction data; Detect the hydrocarbon content of the pyrolysis gas in the first set of reaction data using a gas chromatograph, and measure the carbon atom content in the pyrolysis gas through a mass analyzer to obtain the second set of conversion data; According to the second set of conversion data, determine the carbon powder crystallinity parameter using an X-ray diffractometer, and obtain the third set of quality data through a scanning electron microscope to obtain the morphological characteristics of the carbon powder; For the third group of quality data, an adaptive particle swarm optimization algorithm is used to calculate the quality deviation correction direction, a genetic algorithm is used to search for the heating power distribution scheme, the power setting values of each heating zone are calculated, and the intake valve opening sequence is determined from the temperature field distribution to obtain an optimized cracking control scheme.

Images