Control method of pressure control device for uniform distribution of multi-layer composite copper foil

Abstract

The application provides a control method of a pressure control device for uniform distribution of a multi-layer composite copper foil, comprising: obtaining material attribute parameters of each layer of the multi-layer composite copper foil, and establishing a composite copper foil material attribute database; through a finite element analysis method, combining the multi-layer structure of the composite copper foil and the material attribute data, analyzing stress distribution and deformation of each layer of the copper foil under different pressure conditions, and obtaining a quantitative evaluation index of pressure distribution uniformity; according to the quantitative evaluation index of pressure distribution uniformity, analyzing the mapping relationship between pressure distribution uniformity and process parameters and equipment mechanical characteristics, and obtaining optimal process parameter combinations and equipment control strategies for realizing uniform pressure distribution; when process requirements change, extracting a historical optimization case similar to the current process requirements from a knowledge base as an initial optimization scheme, and then optimizing through a reinforcement learning algorithm to respond to the change in process requirements.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a control method for a pressure control device with uniformly distributed multilayer composite copper foil. Background Technology

[0002] In the manufacturing process of multilayer composite copper foil, the multilayer structure and material properties of the copper foil easily lead to uneven pressure distribution when pressure is applied. Due to the differences in material properties between the layers of the copper foil, such as thickness, hardness, and surface roughness, the deformation and stress distribution between different layers are not consistent under the same pressure. This uneven pressure distribution results in differences in the bonding strength between the layers of the copper foil, thus affecting the overall performance and quality of the composite copper foil. Furthermore, in actual production, due to variations in process parameters and the mechanical characteristics of the equipment, it is difficult for the pressure application device to achieve precise control and dynamic adjustment of the pressure. This necessitates a control method that can achieve uniform pressure distribution and dynamic adjustment based on the multilayer structure and material properties of the copper foil. Simultaneously, this control method must also consider changes in process requirements and the mechanical characteristics of the equipment during production to ensure precise pressure control and rapid response under different operating conditions, thereby guaranteeing the production quality and efficiency of the composite copper foil. Summary of the Invention

[0003] This invention provides a control method for a pressure control device with uniformly distributed multilayer composite copper foil, mainly including:

[0004] Obtain the material property parameters of each layer of multilayer composite copper foil and establish a composite copper foil material property database;

[0005] By using the finite element method and combining the multilayer structure and material property data of composite copper foil, the stress distribution and deformation of each layer of copper foil under different pressure conditions are analyzed, and a quantitative evaluation index of the pressure distribution uniformity is obtained.

[0006] Based on the quantitative evaluation index of pressure distribution uniformity, the mapping relationship between pressure distribution uniformity and process parameters and equipment mechanical characteristics is analyzed to obtain the optimal combination of process parameters and equipment control strategy to achieve pressure uniformity.

[0007] The optimal combination of process parameters and equipment control strategies to achieve uniform pressure distribution are applied to the copper foil production line. During the production of composite copper foil, process parameters and equipment operating status data are collected in real time to dynamically predict the pressure distribution under the current operating conditions. If the prediction results meet the requirements of the quantitative evaluation index of pressure distribution uniformity, the current process parameters and equipment control strategies are maintained unchanged.

[0008] If the predicted pressure distribution does not meet the uniformity requirement, the process parameters and equipment control strategies are dynamically adjusted through reinforcement learning algorithms to make the pressure distribution gradually become more uniform, and the optimized combination of process parameters and control strategies are updated in the knowledge base.

[0009] When process requirements change, historical optimization cases similar to the current process requirements are extracted from the knowledge base as initial optimization schemes, and then optimized through reinforcement learning algorithms to respond to changes in process requirements.

[0010] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0011] This invention discloses a control method for a pressure control device that achieves uniform distribution of pressure in multilayer composite copper foil. The method establishes a composite copper foil material property database, builds a pressure distribution model through finite element analysis, and employs machine learning algorithms to establish a mapping relationship model between pressure distribution uniformity and process parameters. During production, online thickness gauges and multi-directional pressure detection devices are used to collect copper foil thickness and pressure data in real time, and multiphysics coupling simulation is combined to simulate transient pressure distribution under high-speed motion. Adaptive control algorithms are used to adjust pressure roller parameters in real time to ensure the stability of copper foil thickness under dynamic conditions. Reinforcement learning algorithms are used to dynamically optimize process parameters and equipment control strategies, gradually making the pressure distribution more uniform. When process requirements change, similar historical cases are extracted from the knowledge base and fine-tuned to quickly respond to process changes. This invention achieves precise control and dynamic optimization of pressure distribution during composite copper foil production, effectively improving copper foil thickness uniformity and product quality stability, and providing technical support for the large-scale production of high-performance composite copper foil. Attached Figure Description

[0012] Figure 1 This is a flowchart of the control method for the pressure control device with uniformly distributed multilayer composite copper foil according to the present invention. Detailed Implementation

[0013] 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 with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0014] like Figure 1 The control method of the pressure control device with uniformly distributed multilayer composite copper foil in this embodiment may specifically include:

[0015] S101. Obtain the material property parameters of each layer of the multilayer composite copper foil and establish a composite copper foil material property database.

[0016] Scanning electron microscope (SEM) images of each layer of the composite copper foil structure were obtained; surface topography point cloud data were obtained based on the measurement data from the SEM images; the point cloud data were processed using a Gaussian filter to obtain surface profile values; cyclic loads were applied to the surface of the composite copper foil based on the surface profile values, and the Young's modulus and hardness values ​​of each layer were calculated through load-displacement curves to establish a composite copper foil material property database.

[0017] For example, three-dimensional digital point cloud data of the copper foil surface morphology is obtained based on high-magnification scanning electron microscope (SEM) images of each layer in the composite copper foil structure. The root mean square (RMS) value of the surface roughness and surface morphology characteristic values ​​of each layer are calculated from the point cloud data. A Gaussian filter with a cutoff wavelength of 0.8 micrometers is used to process the point cloud data to obtain the surface profile values ​​of each layer. A nanoindenter is used to apply cyclic loads to the surface of each layer of the composite copper foil. The elastoplastic deformation behavior of each layer is quantitatively analyzed through load-displacement curves. The Young's modulus and hardness values ​​of each layer are calculated according to the elastic modulus equation in the Oliver-Fal model, establishing a composite copper foil material property database. Based on the surface profile values ​​and load-strain relationship data, a digital model of the cross-sectional profile of each layer of the composite copper foil is established, and the interlayer stress distribution is calculated using contact mechanics equations. For the interlayer stress distribution data, an acoustic emission detector is used to record the interface stress wave signal under pressure load. The signal is decomposed and reconstructed using the db6 wavelet basis function, and a fuzzy clustering algorithm is used to classify and identify the stress wave features to determine the interface stress concentration areas. Based on the surface profile values ​​of each layer, load-strain relationship, and data on interface stress concentration areas, a multilayer mechanical model of the composite copper foil structure was constructed using finite element analysis software. The deformation of each layer was calculated using nonlinear stress-strain relationships. Based on the deformation data of each layer, the strain energy density criterion J=(1 / 2E)(σ^2+τ^2) was used to determine the location of interface stress concentration, thus obtaining the stress distribution law of each layer of the composite copper foil under pressure. In the surface morphology analysis of the composite copper foil structure, high-magnification scanning electron microscopy (SEM) was used to acquire surface micromorphology images at 5000x magnification. Surface morphology contour data points were extracted using image processing software to form a three-dimensional point cloud data matrix. For surface roughness measurement, a 10×10 micrometer scanning area was selected with a sampling interval of 0.1 micrometers, obtaining height data for 100×100 measurement points. A Gaussian filter was used to filter the raw data, with a cutoff wavelength of 0.8 micrometers to effectively remove high-frequency noise. After filtering, the root mean square value Ra of the surface roughness was calculated to be 0.35 micrometers. In the material mechanical property testing process, a nanoindenter was used with a triangular pyramidal diamond indenter. The maximum load was set to 500 mN, and the loading rate was 10 mN / s. A 5×5 dot matrix was selected on the sample surface for indentation testing. The load-displacement curve was analyzed using the Oliver-Fal model, where the elastic modulus was calculated as E = 0.5(dP / dh)(π / A)^0.5, where P is the load, h is the indentation depth, and A is the contact area. The calculated elastic modulus of the surface copper foil was 110 GPa, and the hardness was 1.2 GPa. In the interface stress analysis, acoustic emission detection was performed using a piezoelectric sensor array with a frequency response range of 100-1000 kHz. The acquired acoustic emission signals were processed using wavelet transform, and a 5-level decomposition was performed using the db6 wavelet basis function to extract energy characteristic parameters.Acoustic emission signal characteristics were classified using a fuzzy clustering algorithm, with 3 cluster centers corresponding to three states: intact interface, slightly damaged interface, and interface delamination. In the finite element analysis, a multi-layer structural mechanical model was established based on measured material parameters, using 8-node hexahedral elements with a mesh size of 0.1 micrometers. Under pressure loading, stress concentration locations were determined using the strain energy density criterion J = (1 / 2E)(σ^2 + τ^2), where σ is the normal stress and τ is the shear stress. A stress concentration region was identified when the strain energy density J exceeded the critical value of 0.5 joules per cubic millimeter. Calculation results show that stress concentration easily occurs at the interface due to abrupt changes in material properties, with the maximum strain energy density reaching 0.8 joules per cubic millimeter.

[0018] S102. Using the finite element analysis method, combined with the multilayer structure and material property data of the composite copper foil, the stress distribution and deformation of each layer of the copper foil under different pressure conditions are analyzed to obtain a quantitative evaluation index of the pressure distribution uniformity.

[0019] A three-dimensional finite element geometry is established based on the number of layers and material density of the composite copper foil. A tetrahedral mesh is generated using a mesh generator, and the surface morphology features of the copper foil are obtained from the scanned data. The contact area is calculated using the surface morphology features of the copper foil, and the interlayer frictional stress is obtained using Coulomb's law of friction. The nodal displacement field is calculated using the principle of virtual work based on the interlayer frictional stress. If the local pressure deviation of the nodal displacement field exceeds a preset threshold, the corrected nodal displacement field is obtained using the residual iteration method. The strain energy of each element is calculated using the strain energy density function based on the corrected nodal displacement field, and the integral value of the strain energy of each element is obtained by three-point Gaussian integration.

[0020] For example, a three-dimensional finite element geometric structure is established based on the number of composite copper foil layers, material density, interlayer contact state, and interface bonding strength data. A tetrahedral mesh is generated using a mesh generator, and the mesh node displacements are assigned using a four-node bilinear interpolation function. The surface morphology and roughness characteristics of the copper foil are obtained from the scanning data. The contact area is calculated using the copper foil surface morphology data, and the interlayer frictional stress is calculated using Coulomb's friction law τ = μσ, where τ is the frictional stress, μ is the friction coefficient, and σ is the normal stress. The stress distribution of each layer of the composite copper foil structure is solved using a piecewise cubic Hermitian interpolation function. Based on the stress distribution data of each layer, the node displacement field of each layer is calculated using the principle of virtual work, and the strain compatibility equation is applied. Determine the deformation variables for each layer, where ε is strain, u is displacement in the x-direction, and v is displacement in the y-direction. If the local pressure deviation exceeds a preset threshold, the displacement field is corrected using the residual iteration method, based on the stress balance equation. Solving for nodal force equilibrium, σx represents the normal stress in the x-direction, and τxy represents the shear stress. Based on the corrected displacement field data, the strain energy density function U = 1 / 2(σε + τγ) is used to calculate the element strain energy, where U is the strain energy density, ε is the normal strain, and γ is the shear strain. The integral value of the strain energy of each layer of elements is obtained through the three-point Gaussian integration. The pressure distribution uniformity evaluation index σ = √[Σ(Ui - Uavg)^2 / n] is used to calculate the integral value of the strain energy, where Ui is the element strain energy, Uavg is the average strain energy, and n is the total number of elements. In the pressure distribution modeling of composite copper foil, the copper foil surface is first meshed. For the analysis area of ​​0.1 square millimeters, tetrahedral elements are used for meshing, with the element size set to 5 micrometers, generating a total of 20,000 mesh elements. Based on the scanning data showing a root mean square value of 0.35 micrometers for surface roughness, the mesh nodes are assigned displacement values ​​through a four-node bilinear interpolation function, which uses a shape function expression. In the interlaminar contact analysis, surface morphology data showed that the actual contact area was approximately 75% of the nominal contact area. The friction coefficient μ was set to 0.3. When the normal pressure was 100 MPa, the interlaminar frictional stress τ was calculated to be 30 MPa according to Coulomb's friction law. A piecewise cubic Hermitian interpolation function was used to interpolate the stress field, with the interpolation interval [-1, 1] divided into 50 sub-intervals. During deformation calculations, the nodal displacement field was calculated using the virtual work principle δW = Σ(σδε + τδγ), where σ is the normal stress, τ is the shear stress, ε is the normal strain, and γ is the shear strain. The calculation results showed that the maximum nodal displacement was 2.5 μm, occurring in the boundary load region. The strain compatibility equation was used to verify the continuity of deformation, and the maximum strain was calculated to be 0.025. When the local pressure deviation exceeded the set threshold of 10%, the displacement field was corrected using the residual iteration method, with the iteration convergence criterion set to 0.001. The nodal force balance was solved using the stress balance equation, and the calculation showed that the maximum stress concentration factor was reduced to 2.1 after correction. Subsequently, the strain energy of the unit was calculated using the strain energy density function U = 1 / 2(σε + τγ). The coordinates of the Gaussian integration point were selected as ±0.774597, and the weighting coefficient was 0.555556 for numerical integration. Finally, the pressure distribution uniformity evaluation index σ = √[Σ(Ui - Uavg)^2 / n] was obtained through strain energy calculation, where the unit average strain energy Uavg is 0.45 joules per cubic millimeter, and the standard deviation σ is 0.08 joules per cubic millimeter, indicating that the pressure distribution non-uniformity is 17.8%.

[0021] S103. Based on the quantitative evaluation index of pressure distribution uniformity, analyze the mapping relationship between pressure distribution uniformity and process parameters and equipment mechanical characteristics, and obtain the optimal combination of process parameters and equipment control strategy to achieve pressure uniformity.

[0022] The pressure distribution data collected by the pressure sensor is denoised using a median filter. The denoised data is then decomposed to obtain a pressure distribution uniformity feature vector. Based on this feature vector, a correlation matrix between the pressure distribution uniformity feature vector and the process parameters is established using the Pearson correlation coefficient method. A parameter coupling coefficient matrix is ​​obtained by nonlinearly mapping the correlation matrix using a radial basis function kernel support vector machine. For the parameter coupling coefficient matrix, an orthogonal design method is used to generate process parameter test schemes. The test schemes are evaluated and calculated using the analytic hierarchy process (AHP) to obtain a parameter optimization dataset. The parameter optimization dataset is trained using a neural network based on backpropagation. If the weight matrix after training satisfies the convergence condition, a genetic algorithm using single-point crossover and Gaussian mutation is used to combine and optimize the process parameters to obtain a combination of pressure uniformity control parameters.

[0023] For example, based on real-time pressure data collected by pressure sensors, a median filter is used to denoise the pressure distribution data. The denoised data is then decomposed into five levels using the db4 wavelet basis function, and the pressure distribution uniformity feature vector is obtained from the reconstructed coefficient matrix. The Pearson correlation coefficient method is used to establish the correlation matrix R(i,j) between the pressure distribution uniformity feature vector and the process parameters and mechanical characteristic parameters. A radial basis function kernel support vector machine is used to perform nonlinear mapping on the correlation matrix data to determine the coupling coefficient values ​​between parameters. Based on the parameter coupling coefficient matrix, an L16(4^5) process parameter test scheme is generated using orthogonal design. The analytic hierarchy process (AHP) is used to evaluate and calculate each combination scheme, and a parameter optimization dataset is established based on the calculation results. A neural network based on error backpropagation is used to train the parameter optimization dataset. Hidden layer neurons are used to fit the correspondence between process parameters and equipment parameters, and the parameter mapping relationship is obtained from the trained weight matrix. Based on the parameter mapping relationship matrix, a genetic algorithm using single-point crossover and Gaussian mutation is used to optimize the combination of process parameters. The optimal solution is calculated using the minimum variance fitness function, obtaining the combination of control parameters under uniform pressure distribution conditions. In pressure distribution data processing, a 5×5 window median filter was used to denoise the original pressure data. The standard deviation of the data before filtering was 0.85 MPa, which was reduced to 0.32 MPa after filtering. Subsequently, a 5-level decomposition was performed using the db4 wavelet basis function. The high-frequency coefficients obtained after decomposition reflected the pressure fluctuation characteristics, while the low-frequency coefficients reflected the pressure distribution trend. The eigenvalues ​​of the reconstructed coefficient matrix were 2.45, 1.87, 1.32, 0.95, and 0.68, respectively. In the correlation analysis between the pressure distribution uniformity feature vector and process parameters, the Pearson correlation coefficient method was used to construct the correlation matrix. The calculation results showed that the correlation coefficient between pressure and temperature was 0.82, with velocity 0.75, and with equipment vibration 0.68. The radial basis function kernel was selected as the kernel function for the support vector machine, with σ set to 1.5. After cross-validation, the classification accuracy reached 92.5%. In the parameter combination optimization stage, an L16 (4^5) orthogonal array was used to design the experimental scheme, including four levels of temperature (120°C, 140°C, 160°C, 180°C), four levels of pressure (8 MPa, 10 MPa, 12 MPa, 14 MPa), and four levels of speed (5 m / min, 8 m / min, 11 m / min, 14 m / min). The weights of each scheme were calculated using the analytic hierarchy process (AHP), and the consistency ratio (CR) was 0.047, meeting the consistency requirement. The neural network training used a backpropagation (BP) network structure with two hidden layers: 15 neurons in the first hidden layer and 8 neurons in the second hidden layer. The learning rate was set to 0.05, and the momentum factor was 0.8. After 2000 iterations on 1000 training samples, the training error converged to 0.0025, and the prediction accuracy on the test set reached 95.3%.Finally, in the genetic algorithm optimization, the population size was set to 100, the crossover probability to 0.85, and the mutation probability to 0.05. The fitness function f = 1 / sqrt(Σ(xi-μ)^2 / n) was used for fitness evaluation, where xi is the single-point pressure value, μ is the average pressure, and n is the number of measurement points. After 200 generations of evolution, the fitness value of the optimal solution increased from the initial 0.62 to 0.93. The corresponding optimal combination of process parameters was: temperature 155 degrees Celsius, pressure 11.5 MPa, speed 9.5 m / min, equipment vibration frequency 35 Hz, and roller gap 0.8 mm.

[0024] The equipment control strategy includes: collecting copper foil thickness data in real time using an online thickness gauge on the multilayer composite copper foil production line; processing the collected data using data analysis methods to obtain a spatial distribution map of the copper foil thickness; judging the uniformity of the thickness distribution; and adjusting the pressure of the pressure rollers in the corresponding area when an abnormal thickness is found in a local area.

[0025] A sliding mean filter is used to process the data collected by the thickness gauge, and the copper foil thickness distribution value is obtained by fitting using the least squares method. Based on the copper foil thickness distribution value, a bicubic interpolation algorithm is used for spatial transformation and reconstruction, and the reconstructed data is gridded using the Kriging interpolation method to obtain a thickness distribution map. For the thickness distribution map, a support vector regression algorithm is used to establish a mapping relationship between pressure and thickness, and the uniformity of copper foil thickness is judged by the variance calculation formula. If the variance of a local area exceeds a preset standard deviation threshold, the grid coordinates of that area are obtained. Based on the grid coordinates, the pressure compensation value is updated using a recursive least squares formula, and the pressure roller pressure is dynamically adjusted using a PI controller to obtain the pressure adjustment amount.

[0026] For example, based on the original data collected by the multi-point scanning thickness gauge, a sliding mean filter with a window length of 128 points is used to preprocess the thickness measurement data. A third-order polynomial curve is fitted using the least squares method, and the copper foil thickness distribution values ​​are obtained from the fitted curve. A 16×16-point bicubic interpolation algorithm is used to spatially transform and reconstruct the thickness distribution values. The reconstructed data is then gridded using the Kriging interpolation method, and a spatial distribution map of the copper foil thickness is obtained from the gridded data. Based on the spatial distribution map of the thickness, a Gaussian kernel function support vector regression algorithm is used to establish the mapping relationship between pressure and thickness, where the kernel function parameter σ is determined through cross-validation. The variance is calculated using the formula... Calculate the thickness uniformity index, where xi is the thickness at the i-th measurement point. The average thickness is given by n, where n is the number of measurement points. If the variance of a local area exceeds the preset standard deviation threshold, the grid coordinates of that area are recorded. For the grid coordinates of the area exceeding the standard deviation threshold, a recursive least squares formula is used to update the pressure compensation value. Based on the pressure compensation value, a PI controller with a proportional coefficient Kp = 0.8 and an integral time Ti = 0.5 is used to perform closed-loop adjustment of the pressure roller pressure, obtaining the dynamic pressure adjustment amount from the feedback data. During the copper foil thickness measurement process, the scanning thickness gauge continuously scans the copper foil at a sampling frequency of 100 Hz, and the raw data exhibits random fluctuations of ±0.15 micrometers. A 128-point moving average filter is used to preprocess the data, reducing the measurement noise to [value missing].

[0027] ±0.05 μm. A third-order polynomial was fitted to the filtered data, resulting in the equation y = 2.35 × 10⁻⁶ x³ - 4.12 × 10⁻⁴ x² + 0.0183 x + 12.53, where x represents the measurement location coordinates and y represents the copper foil thickness. The goodness-of-fit R² reached 0.985. In spatial data reconstruction, a 16 × 16 point bicubic interpolation algorithm was used to resample the thickness data. The interpolation function weighting coefficient matrix was 4 × 4, and the sampling interval was 2 mm. Kriging interpolation employed a spherical semi-variogram model γ(h) = C₀ + C[1.5(h / a) - 0.5(h / a)³], where C₀ is the nugget value (0.01), C is the sill value (1.2), and a is the range (50 mm). After interpolation, the grid resolution was improved to 0.5 mm. The support vector regression algorithm uses a Gaussian kernel function, and the kernel parameter σ = 0.8 and the penalty factor C = 10 are determined through five-fold cross-validation. The training data contains 5000 pressure-thickness sample pairs, and the root mean square error of the prediction after training is 0.08 μm. Thickness uniformity is evaluated using variance calculation. For a 100×100 mm area, it is divided into 400 grid cells, and the calculated standard deviation threshold is 0.12 μm. In pressure compensation control, the forgetting factor λ of the recursive least squares algorithm is set to 0.95, the initial covariance matrix P0 = 100I, and the initial estimated parameter value θ0 = [0.5, 0.3]^T. When the standard deviation of the thickness in a certain area is detected to exceed 0.12 μm, pressure compensation calculation is triggered, and the parameter estimate θ(k) is updated. The pressure closed-loop regulation uses a PI controller, with a proportional coefficient Kp = 0.8 for fast response and an integral time Ti = 0.5 seconds to eliminate steady-state error. Actual measurement data shows that, with a pressure fluctuation amplitude of ±0.5 MPa, the thickness control accuracy remains within ±0.1 μm. Data analysis reveals that the copper foil thickness distribution exhibits a slightly thicker central region and a slightly thinner edge region, with a maximum thickness difference of approximately 0.25 μm. Comparison data before and after pressure compensation indicates that dynamic pressure adjustment significantly improves thickness uniformity, reducing the standard deviation from 0.15 μm to 0.08 μm, meeting the requirements for high-precision copper foil production.

[0028] The equipment control strategy includes: collecting pressure data of copper foil during production in real time through pressure sensors, establishing a pressure-thickness mapping relationship through machine learning algorithms, predicting the deformation of copper foil under different pressure conditions, and guiding the dynamic adjustment of pressure rollers to suppress thickness differences caused by anisotropy.

[0029] A Butterworth low-pass filter is used to filter the pressure signal collected by the multi-directional pressure sensor array. The filtered pressure signal is then decomposed, and a pressure distribution feature matrix is ​​obtained from the reconstruction coefficients. Based on the pressure distribution feature matrix, a convolutional neural network is used for feature extraction, and the mapping relationship between pressure and deformation is obtained from the last layer of the convolutional feature map. For the pressure-deformation mapping relationship, a principal component analysis model is constructed using the covariance matrix, and the principal components of deformation in each direction are obtained from the eigenvalues. Based on the principal components of deformation in each direction, the deformation feature weight coefficients are calculated through singular value decomposition. An adaptive PID controller is used to compensate and adjust the pressure roller, where the proportional coefficient is determined by the deformation feature weight coefficients.

[0030] For example, based on the pressure signals collected by a multi-directional pressure sensor array, a Butterworth low-pass filter is used to preprocess the pressure data in different directions. The filtered data is then decomposed into three layers using the db4 wavelet basis function, and the pressure distribution feature matrix is ​​obtained from the reconstruction coefficients. A four-layer convolutional neural network is used to extract features from the pressure distribution feature matrix, where the first layer has a 5×5 kernel size, the second layer has a 3×3 kernel size, the third layer has a 3×3 kernel size, and the fourth layer has a 2×2 kernel size. The mapping relationship between pressure and deformation is obtained from the feature map of the last convolutional layer. Based on the pressure-deformation mapping data, a principal component analysis model U = XV is constructed using the covariance matrix, where X is the standardized deformation data matrix and V is the eigenvector matrix. The principal components of deformation in each direction are obtained from the eigenvalues ​​λi. For the deformation principal component data, the deformation eigenvalues ​​are calculated using singular value decomposition W = USV^T. The deformation sensitivity in each direction is determined from the singular value sequence Si, and the deformation feature weight coefficients are obtained. Based on the deformation characteristic weighting coefficient, an adaptive PID controller is used to compensate and adjust the pressure roller. The proportional coefficient Kp is determined by the deformation sensitivity, the integral time Ti by the response characteristics, and the derivative time Td by the dynamic characteristics. The PID parameters are updated in real time using a recursive least squares method, and the compensation effect is evaluated using a prediction error criterion. The dynamic pressure compensation parameters are obtained from the evaluation results. In the multi-directional pressure detection of copper foil, the pressure sensor array is arranged along four directions: 0°, 45°, 90°, and 135°, with eight sensors in each direction and a sampling frequency of 200 Hz. The original pressure signal exhibits random fluctuations of ±0.2 MPa. A Butterworth fourth-order low-pass filter with a cutoff frequency of 50 Hz is used for preprocessing, improving the signal-to-noise ratio to 35 dB. The filtered data is then decomposed into three levels using the db4 wavelet basis function to obtain high-frequency coefficients d1, d2, and d3 and low-frequency coefficient a3, reflecting pressure changes at different scales. In the four-layer convolutional neural network architecture, the first layer uses 16 5×5 convolutional kernels to extract local features, the second layer uses 32 3×3 convolutional kernels to extract medium-scale features, the third layer uses 64 3×3 convolutional kernels to extract large-scale features, and the fourth layer uses 128 2×2 convolutional kernels to synthesize features across all scales. Each convolutional layer is followed by a ReLU activation function, and 2×2 max pooling is used. The network was trained using 5000 stress-deformation data pairs, and the training error converged to 0.015 after 2000 iterations. In principal component analysis, the deformed data was first standardized, and the eigenvalues ​​of the covariance matrix were calculated.

[0031] The values ​​are [2.85, 1.62, 0.38, 0.15], with corresponding cumulative contribution rates of [56.8%, 89.2%, 96.8%, 100%]. The first two principal components are selected as the main deformation features, and their feature vectors are respectively...

[0032] [0.82, 0.45, 0.21, 0.12] and [0.15, 0.76, 0.52, 0.23]. Singular value decomposition results show that the singular values ​​in the four directions are 6.82, 4.35, 1.85, and 0.92, respectively, indicating higher deformation sensitivity in the 0-degree and 45-degree directions. Based on this, the adaptive PID controller parameters are set as follows: proportional gain Kp = 1.2, integral time Ti = 0.8 seconds, and derivative time Td = 0.2 seconds in the 0-degree direction; the control parameters in the 45-degree direction are 1.0, 0.6 seconds, and 0.15 seconds, respectively. The recursive least squares method uses a forgetting factor λ = 0.95 for parameter updates, and the prediction error criterion is the root mean square error (RMSE). When the RMSE exceeds 0.1 MPa, adaptive parameter adjustment is triggered. Actual test data show that after adopting dynamic pressure compensation, the thickness variation coefficient in the 0-degree direction decreased from 4.2% to 2.1%, in the 45-degree direction from 3.8% to 1.9%, in the 90-degree direction from 2.5% to 1.4%, and in the 135-degree direction from 2.3% to 1.3%, significantly improving the overall uniformity of pressure distribution.

[0033] The equipment control strategy also includes: establishing a mathematical model of the dynamic contact process between the pressure roller and the copper foil based on the production line speed and the deformation characteristics of the copper foil material; using a multi-physics coupling simulation method to simulate the transient pressure distribution under high-speed motion; obtaining the variation law of copper foil thickness with time and position; and adjusting the operating parameters of the pressure roller in real time to ensure the stability of copper foil thickness under dynamic working conditions.

[0034] The dynamic equations are solved using a second-order Newmark time integral algorithm to obtain displacement field distribution data during the contact process between the pressure roller and the copper foil. Based on this displacement field distribution data, the stress field distribution is calculated using the stress-strain constitutive equation of a viscoelastic model, and dynamic contact pressure data is obtained from this stress field distribution. For this dynamic contact pressure data, an autoencoder neural network with three hidden layers is used for feature extraction, and the pressure-deformation mapping relationship data is obtained from the feature extraction results. Based on this pressure-deformation mapping relationship data, the fourth-order Runge-Kutta method is used to solve the differential equations for the copper foil thickness, and the spatiotemporal distribution characteristic data of the thickness is obtained from the numerical solution.

[0035] For example, based on the production line operation data collected by the speed sensor, the second-order Newmark time integration algorithm is used to apply the dynamic equation M(d 2The displacement field distribution during the contact process between the pressure roller and the copper foil is obtained by solving the equation u / dt² + C(du / dt) + Ku = F(t). The stress field distribution during the deformation process of the copper foil is calculated using the stress-strain constitutive equation σ = Dε + ηdε / dt. The material deformation characteristics are described using a viscoelastic model, and the dynamic contact pressure distribution is obtained from the stress field data. σ is the strain tensor, D is the elastic modulus matrix, ε is the strain tensor, and η is the viscosity coefficient. A self-encoding neural network with three hidden layers is used to extract features from the dynamic pressure distribution data. The number of neurons in each hidden layer is 64, 32, and 16, respectively. The mapping relationship between pressure and deformation is obtained through the encoder. Based on the pressure-deformation mapping relationship, the fourth-order Runge-Kutta method is used to solve the differential equation system of copper foil thickness dx / dt = f(x,t), where x is the thickness, and the integration step size t is taken as one-quarter of the control period. The spatiotemporal distribution characteristics of the thickness are obtained from the numerical solution. The temperature field distribution during the rolling process is solved using a thermo-coupling equation set, and the temperature stress is calculated through thermoelastic constitutive relations. The deformation correction amount of the copper foil is obtained from the coupled field data. Based on the deformation correction amount, a Kalman filter state equation is constructed, and the pressure roll parameters are estimated and predicted using the state observation equation. An adaptive pole placement controller is used to adjust the pressure roll parameters in real time. The controller transfer function adopts the second-order form G(s)=ωn2 / (s 2+2ξωns+ωn2), where s is the complex frequency variable in the Laplace transform, and the pressure compensation parameters are obtained from the closed-loop response. During the dynamic rolling of copper foil, a second-order Newmark time-integration algorithm is used to numerically solve the dynamic equations. The mass matrix M reflects the density distribution of the copper foil material, the damping matrix C characterizes the internal resistance of the material, the stiffness matrix K describes the elastic properties of the material, the external force vector F(t) represents the force exerted by the pressure roller, and u is the displacement vector. When the production line speed is 20 m / min and the integration time step is 0.001 s, the calculation results show that the maximum displacement in the pressure application area reaches 0.15 mm. During material deformation, the viscoelastic constitutive equation is used to describe the stress-strain relationship, where the elastic modulus D is 110 GPa and the viscosity coefficient η is 0.8. The calculated pressure distribution in the dynamic contact area exhibits typical asymmetric characteristics, with a maximum pressure of 180 MPa in the inlet area and 150 MPa in the outlet area. Feature extraction was performed using an autoencoder neural network. The input layer had 64 neurons corresponding to pressure sensor data, while the three hidden layers had 64, 32, and 16 neurons respectively. After 2000 iterations of training, the reconstruction error was reduced to 0.025. The fourth-order Runge-Kutta method was used to solve the differential equation for copper foil thickness, with an integration step size of 0.00025 seconds. Calculations showed that with a rolling speed fluctuation of ±5%, the thickness variation exhibited a clear periodicity, with a fluctuation period of approximately 0.1 seconds. Simultaneously, thermo-mechanical coupling calculations showed that a temperature increase of approximately 25 degrees Celsius in the rolling zone resulted in a thermal expansion deformation of 0.03 mm. This thermal effect led to a 15% increase in the actual thickness deviation compared to the purely mechanical calculation result. In the state equation of the Kalman filter, the state transition matrix A was obtained through system identification, the control input matrix B was determined by the actuator characteristics, and the covariance matrices of the process noise w(k) and measurement noise v(k) were Q = diag[0.01, 0.01] and R = 0.04, respectively. Filtering results show that the root mean square error of pressure prediction is 0.8 MPa, and the prediction lead is 0.05 seconds. With the characteristic frequency ωn of the adaptive pole placement controller set to 20 Hz and the damping ratio ξ to 0.7, the pressure fluctuation amplitude is reduced by 60% after closed-loop control, and the thickness control accuracy is improved to ±0.5 μm. This multiphysics coupling method comprehensively considers mechanical deformation, thermal effects, and dynamic response characteristics, achieving precise control of copper foil thickness during high-speed rolling. Measured data shows that at a production line speed of 20 m / min, the standard deviation of the copper foil thickness uniformity index decreases from 2.8 μm to 1.2 μm, significantly improving dynamic rolling accuracy.

[0036] S104. The optimal combination of process parameters and equipment control strategy to achieve uniform pressure distribution is applied to the copper foil production line. During the production of composite copper foil, process parameters and equipment operating status data are collected in real time to dynamically predict the pressure distribution under the current working conditions. If the prediction results meet the requirements of the quantitative evaluation index of pressure distribution uniformity, the current process parameters and equipment control strategy are maintained unchanged.

[0037] Based on the real-time data stream acquired by the pressure sensor, a Kalman filter using the state equation is employed for state estimation, and a preprocessed data sequence is obtained from the recursive formula. For the preprocessed data sequence, the evolution of parameters is extracted using memory units and hidden states in a Long Short-Term Memory (LSTM) network, and pressure distribution prediction data is obtained from the output layer. Based on the pressure distribution prediction data, a Gaussian kernel function is used to construct a pressure distribution probability density curve, and the kernel function bandwidth parameter is determined using cross-validation. The pressure distribution variance is calculated from the probability density function. For the pressure distribution prediction data and process parameters, singular value decomposition is performed using matrix factorization, and a parameter importance sequence is obtained by eigenvalue sorting. Parameter sensitivity coefficients are extracted from the eigenvector matrix.

[0038] For example, based on the real-time data stream collected by pressure sensors and equipment condition monitoring instruments, a Kalman filter with a state equation is used for state estimation. The data sequence is smoothed using the recursive formula θ(k) = θ(k-1) + K(k)[y(k) - θ(k-1)], where θ(k) is the state estimate at the current moment, θ(k-1) is the state estimate at the previous moment, K(k) is a weighting factor of 1 / k+1, and y(k) is the measured value at the current moment. The time-series characteristics of process parameters are obtained from the preprocessed data. A long short-term memory network with two hidden layers is used to dynamically predict the time-series characteristics of process parameters. The number of hidden layer units is 128 and 64, respectively. The parameter evolution law is extracted through the memory unit ct and the hidden state ht, and the pressure distribution prediction data is obtained from the output layer. Based on the pressure distribution prediction data, a pressure distribution probability density curve is constructed using a Gaussian kernel function. The kernel function bandwidth parameter h is determined by cross-validation, and the pressure distribution variance is calculated from the probability density function. A matrix factorization method is used to perform singular value decomposition (U = WSV^T) on the process parameters and pressure distribution data. Parameter importance is ranked by eigenvalue sorting, and parameter sensitivity coefficients are extracted from the eigenvector matrix. If the pressure distribution variance is less than a variance threshold, an autoregressive moving average method is used to test the stationarity of the current process parameters. Unit root tests are used to determine parameter stability, and parameter maintenance strategies are determined from the test results. A principal component regression method is used to establish the process parameter optimization equation system Y = XB + E, where Y is the pressure uniformity index and X is the process parameter matrix. Parameter compensation weights are obtained from the regression coefficients B. In the real-time monitoring of the composite copper foil production line, pressure sensors collect pressure data at a sampling frequency of 100 Hz, and equipment condition monitoring instruments collect parameters such as speed, temperature, and vibration. Kalman filters were used for state estimation. The state vector contained three components: pressure, velocity, and temperature. The state transition matrix A was obtained through system identification. The process noise covariance Q = diag[0.01, 0.02, 0.015] and the measurement noise covariance R = diag[0.02, 0.03, 0.025]. The forgetting factor λ for recursive least squares was set to 0.95, and the data smoothing window length was 50 sampling points. The Long Short-Term Memory (LSTM) network adopted a two-layer hidden layer structure: the first layer had 128 memory units, and the second layer had 64 memory units. The input layer contained 10 process parameter nodes. The network was trained using 5000 sets of historical data, each set containing a sequence of 100 time steps. After training, the root mean square error of prediction on the validation set was 0.085, the prediction lead was 0.5 seconds, and the correlation coefficient reached 0.92. The pressure distribution probability density estimation used a Gaussian kernel function, and the kernel function bandwidth was determined to be 0.8 through five-fold cross-validation. The calculation results show that when the pressure distribution is uniform, the probability density curve exhibits a unimodal symmetric characteristic with a variance of less than 0.1; when the pressure is uneven, the curve exhibits a skewed or bimodal characteristic with a variance that increases to more than 0.25.Singular value decomposition is performed on the process parameter matrix to obtain the eigenvalue sequence.

[0039] The values ​​[8.52, 4.31, 2.15, 1.08, 0.54, 0.27] indicate that the cumulative contribution rate of the first three principal components reaches 85%. Eigenvector analysis shows that rolling speed, temperature, and pressure are the most significant parameters affecting uniformity, with sensitivity coefficients of 0.82, 0.65, and 0.58, respectively. In the parameter stability test, the augmented Dickey-Fuller test was used, and the test statistic ADF was -3.85, which is less than the critical value of -3.43 at the 1% significance level, indicating that the parameter series has good stationarity. Principal component regression analysis results show that there is a significant nonlinear relationship between the pressure uniformity index and the process parameters, with a multivariate correlation coefficient R0. 2 The coefficient reached 0.89. The regression coefficient matrix indicates that the compensation weight for pressure is 0.45, for velocity is 0.32, and for temperature is 0.23, providing a quantitative basis for precise control of process parameters. In actual production, when the pressure distribution variance is less than 0.1 and the parameter sequence passes the stationarity test, the current process parameters are maintained unchanged; when an anomaly is detected, dynamic adjustments are made according to the parameter compensation weights, significantly improving the uniformity and stability of the pressure distribution.

[0040] S105. If the predicted pressure distribution does not meet the uniformity requirement, the process parameters and equipment control strategy are dynamically adjusted through reinforcement learning algorithm to make the pressure distribution gradually become uniform, and the optimized process parameter combination and control strategy are updated to the knowledge base.

[0041] A working condition feature space is constructed based on the state vector, and the working condition state representation is obtained through principal component analysis. A parameter optimization function is constructed using the policy gradient method for the working condition state representation, and the policy gradient value is obtained through Monte Carlo sampling. An adaptive learning rate optimizer is used to iteratively calculate the policy gradient value, and the parameter adjustment amount is obtained through Taylor expansion. A neural network model is constructed using the deep deterministic policy gradient method for the parameter adjustment amount, and the optimal combination of process parameters is obtained from the policy network.

[0042] For example, based on the deviation between the predicted pressure distribution and the uniformity threshold, a state vector is used.

[0043] The feature space of the operating conditions is constructed using the x1, x2, ..., xn region. Principal component analysis is used to reduce the dimensionality of the state data, and eigenvectors corresponding to eigenvalues ​​with a cumulative contribution rate greater than 95% are selected to construct the operating condition state representation. Based on the operating condition state representation, the policy gradient method is used to construct the parameter optimization function. θ represents the policy parameter, and Rτ represents the total reward of trajectory τ. The policy gradient is calculated through 500 Monte Carlo sampling iterations, and the parameter adjustment direction is obtained from the historical database. An adaptive learning rate optimizer is used to iteratively update the process parameters. The parameter adjustment amount Δθ = f'(x) / 1! + f""x) / 2! + f"'(x) / 3! is calculated using a third-order Taylor expansion. The reward value r(s,a) is calculated based on the degree of improvement in pressure uniformity. A temporal difference objective function is constructed based on the reward value, and the compensation parameters are reinforced using a deep deterministic policy gradient method. The commentator network adopts a three-layer structure [256,128,64], and the actor network adopts a symmetric structure [64,128,256]. An experience replay buffer is used to store the optimized trajectory, and a priority sampling method is used to select high-value experiences for learning. The optimal combination of process parameters is obtained from the trained policy network. Based on the optimal combination of process parameters, a process knowledge base is constructed using a knowledge graph structured approach. Optimization experience is described through semantic triples (operating condition, parameter, effect), and process parameter control specifications are derived from rule-based reasoning. In the pressure optimization control of composite copper foil, the operating condition feature space contains 10 state variables, including pressure non-uniformity, velocity fluctuation rate, temperature distribution, and vibration amplitude. Principal component analysis results show that the eigenvalue sequence is...

[0044] [4.82, 2.56, 1.33, 0.65, 0.34, 0.15, 0.08, 0.04, 0.02, 0.01], the first four principal components account for 95.2% of the total contribution. These principal components are selected to construct a low-dimensional working condition representation. During parameter optimization, the policy gradient method uses a Gaussian policy function πθ(a|s) = N(μθ(s), σ²), where the mean function μθ(s) is fitted by a three-layer neural network. The policy gradient is calculated through 500 Monte Carlo samplings. The sampling results show that the gradient for the pressure parameter is 0.85, the gradient for the velocity parameter is 0.62, and the gradient for the temperature parameter is 0.43, indicating that pressure regulation is most sensitive to uniformity improvement. Adaptive learning rate optimization uses a third-order Taylor expansion to approximate the parameter adjustment amount. The derivatives of each order of the expansion are f'(x) = 2.5, f""x" = -0.8, and f""'(x) = 0.15, respectively. The calculated adjustment amounts for pressure parameters were 0.42 MPa, velocity parameters were 0.25 m / min, and temperature parameters were 0.18 degrees Celsius. The reward function was designed as r(s,a)=-w1σ2-w2|Δa|, where σ2 is the pressure variance, Δa is the parameter adjustment amount, and the weight coefficients w1=0.7 and w2=0.3. The deep deterministic policy gradient network adopted a symmetric structure, with the commentator network having three hidden layers with 256, 128, and 64 neurons respectively, and the actor network adopting a mirror structure. The experience replay buffer capacity was set to 10000, and the priority sampling probability was proportional to the temporal difference error. After training for 5000 rounds, the network converged to a stable policy, and the accuracy of pressure uniformity prediction reached 92.5%. The knowledge graph used semantic triples to store optimization experience, with the triple structure being (work condition type, parameter combination, optimization effect). For example, when the operating condition is "high speed, light load," the optimal parameter combination is a pressure of 10.5 MPa, a speed of 25 m / min, and a temperature of 155 degrees Celsius, with a corresponding pressure non-uniformity of less than 5%. Parameter configuration specifications for typical operating conditions such as "high speed-light load-low pressure" are extracted through rule-based reasoning, forming a complete process knowledge base. Practical applications show that the parameter optimization method based on reinforcement learning and knowledge graphs significantly improves the pressure uniformity of composite copper foil, reducing the non-uniformity from 12% to 4.5%, and also significantly improving process stability.

[0045] S106. When process requirements change, extract historical optimization cases similar to the current process requirements from the knowledge base as the initial optimization scheme, and then optimize them through reinforcement learning algorithms to respond to changes in process requirements.

[0046] Feature vectors are constructed based on process parameters, equipment parameters, and pressure distribution. The similarity between these feature vectors is calculated using cosine similarity, and the most similar process scheme is obtained using Euclidean distance hierarchical clustering. For the most similar process scheme, a dual-network structure comprising a critic network and an actor network is constructed. The critic network calculates parameter gradient values ​​using a cross-entropy loss function. A Gaussian kernel function is used to spatially map the parameter gradient values, and the kernel function parameters are determined through grid search. Pressure distribution prediction values ​​are obtained from the regression equation. A reward function containing pressure bias and variance is constructed based on the pressure distribution prediction values. A deep substitution strategy optimization algorithm is used to iteratively update the parameters, and the optimal parameter combination is obtained through random exploration.

[0047] For example, based on the data showing changes in process requirements, a cosine similarity calculation formula is constructed using feature vectors [process parameters, equipment parameters, pressure distribution].

[0048] s im(A,B)=Σ(Ai×Bi) / sqrt(Σ(Ai 2 )×Σ(Bi 2 Let Ai be the i-th feature of vector A and Bi be the i-th feature of vector B. Similar cases are grouped using Euclidean distance hierarchical clustering, and the closest process solution is obtained from the cluster centers. Based on the closest process solution, a dual-network structure is constructed using the soft actor-critic algorithm. The critic network contains three layers [256, 128, 1], and the actor network contains three layers.

[0049] [128,64,32], parameter gradients are calculated using the cross-entropy loss function. A Gaussian kernel function is used to map the parameter space, and the kernel parameter σ is determined through grid search. The predicted pressure distribution value is obtained from the regression equation. Based on the predicted pressure distribution value, a reward function r(s,a)=-k1×D(p)+k2×var(p) is constructed, where D(p) represents the pressure deviation and var(p) represents the variance. The parameter adjustment amount is calculated using the gradient ascent method. A deep substitution strategy optimization algorithm is used to iteratively update the parameters. An ε-greedy strategy is used to randomly explore the parameter space and obtain the optimal parameter combination from the exploration data. Based on the optimal parameter combination, a knowledge distillation method is used to integrate new optimization experience into the knowledge base. The fusion and update of new and old knowledge is achieved through soft-label cross-entropy, forming a process parameter optimization specification from the updated knowledge base. In the knowledge base case retrieval, the feature vector contains 12 process parameters, such as speed, temperature, and pressure, 8 equipment parameters, such as equipment vibration and roll gap, and 16 pressure distribution feature values. When the process requirements change from "high speed and light load" to "medium speed and heavy load", the calculated cosine similarity distribution with historical cases is [0.92, 0.85, 0.78, 0.65]. Clustering is performed using a threshold of 0.8, and cases with a similarity of 0.92 are selected as the initial optimization values. In the dual-network structure of the soft actor / critic algorithm, the critic network adopts a three-layer structure [256, 128, 1], uses the ReLU activation function, and outputs Q-value estimates; the actor network adopts a three-layer structure [128, 64, 32], and outputs the action probability distribution. The cross-entropy loss function calculation results show that the gradient of the pressure parameter is 0.45, the gradient of the speed parameter is 0.32, and the gradient of the temperature parameter is 0.28, indicating that pressure regulation is the most critical. The parameter space mapping uses a Gaussian kernel function, and the kernel parameter σ = 0.8 is determined through grid search. Kernel principal component regression prediction shows that the uniformity of pressure distribution has a non-linear relationship with process parameters, and the correlation coefficient R 2The coefficient reached 0.88. In the reward function, the weighting coefficients k1 = 0.6 and k2 = 0.4, yielding the maximum parameter adjustments: pressure ±0.5 MPa, speed ±2 m / min, and temperature ±5 degrees Celsius. The deep substitution strategy optimization employed an ε-greedy strategy for parameter exploration, with the exploration probability ε linearly decreasing from an initial 0.3 to 0.05. A total of 500 random samples were performed in the parameter space, obtaining the optimal parameter combination of 12.5 MPa pressure, 18 m / min speed, and 165 degrees Celsius temperature. This parameter set improved the pressure distribution uniformity index by 45%. During knowledge distillation, soft labels with a temperature parameter T = 3 were used for knowledge transfer, reducing the cross-entropy loss from an initial 1.25 to 0.15. The fusion of old and new knowledge used an exponential moving average to update the knowledge base parameters, with an update coefficient α = 0.9. The updated knowledge base contains 350 sets of optimized parameter combinations for typical operating conditions, covering 95% of process variation scenarios. In this way, the response time for process parameter optimization was reduced from 15 minutes to 2 minutes, significantly improving optimization efficiency. Optimization results show that, despite changes in process requirements, the pressure distribution uniformity index remained within ±5%, meeting the requirements for high-precision copper foil production.

[0050] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A control method of a pressure control apparatus for uniformly distributing a plurality of composite copper foils, characterized by, The method comprises: Obtaining material attribute parameters of each layer of the multi-layer composite copper foil, and establishing a composite copper foil material attribute database; Through a finite element analysis method, in combination with the multi-layer structure and material attribute data of the composite copper foil, stress distribution and deformation of each layer of the copper foil under different pressure conditions are analyzed, and a quantitative evaluation index of pressure distribution uniformity is obtained; According to the quantitative evaluation index of pressure distribution uniformity, a mapping relationship between pressure distribution uniformity and process parameters and equipment mechanical characteristics is analyzed, and an optimal process parameter combination and equipment control strategy for realizing uniform pressure distribution are obtained; The optimal process parameter combination and equipment control strategy for realizing uniform pressure distribution are applied to the copper foil production line, process parameters and equipment operating state data are collected in real time during the production process of the composite copper foil, the pressure distribution under the current working condition is dynamically predicted, and if the prediction result meets the requirements of the quantitative evaluation index of pressure distribution uniformity, the current process parameters and equipment control strategy are maintained unchanged; If the predicted pressure distribution does not meet the uniformity requirement, the process parameters and equipment control strategy are dynamically adjusted through a reinforcement learning algorithm, so that the pressure distribution gradually tends to be uniform, and the optimized process parameter combination and control strategy are updated to the knowledge base; wherein the reinforcement learning algorithm adopts a deep deterministic policy gradient method; When the process requirement changes, a historical optimization case similar to the current process requirement is extracted from the knowledge base as an initial optimization scheme, and then the reinforcement learning algorithm is used for optimization; The method comprises: Obtaining scanning electron microscope photos of each layer of the composite copper foil structure; Obtaining surface topography point cloud data from the scanning electron microscope photo measurement data; Using a Gaussian filter to process the point cloud data to obtain surface profile degree values; Applying cyclic load on the surface of the composite copper foil according to the surface profile degree values, calculating the Young's modulus and hardness values of each layer through the load displacement curve to establish a composite copper foil material attribute database; wherein a nanoindenter is used to apply cyclic load, and the Young's modulus and hardness values are calculated based on the Oliver-Pharr model; The method comprises: According to the number of layers and the material density of the composite copper foil, a three-dimensional finite element geometric structure is established, and a tetrahedral mesh is generated through a meshing device to obtain copper foil surface topography characteristic values from the scanned data; The copper foil surface topography characteristic values are used to calculate the contact area, and the interlayer friction stress value is obtained through the Coulomb friction law; According to the interlayer friction stress value, the node displacement field is calculated through the virtual work principle, and if the local pressure deviation of the node displacement field exceeds a preset threshold, a residual iteration method is used to obtain a corrected node displacement field; Adopt the revised node displacement field to calculate the unit strain energy through the strain energy density function unit, and obtain the integral value of each layer unit strain energy through three-point Gaussian integration, and calculate the quantitative evaluation index of pressure distribution uniformity by using the integral value of strain energy; wherein the stress concentration position is judged by using the strain energy density criterion.

2. The method of claim 1, wherein, According to the quantitative evaluation index of pressure distribution uniformity, the mapping relationship between pressure distribution uniformity and process parameters and equipment mechanical characteristics is analyzed, and the optimal process parameter combination for realizing uniform pressure distribution is obtained, including: Adopt the median filter to denoise the pressure distribution data collected by the pressure sensor, and decompose the data after denoising to obtain the pressure distribution uniformity feature vector; According to the pressure distribution uniformity feature vector, a correlation matrix between the process parameters is established by using the Pearson correlation coefficient method, and a parameter coupling coefficient matrix is obtained by nonlinear mapping of the correlation matrix through a radial basis kernel function support vector machine; For the parameter coupling coefficient matrix, an orthogonal design method is used to generate a process parameter test scheme, and the test scheme is evaluated and calculated by an analytic hierarchy process to obtain a parameter optimization data set; The parameter optimization data set is trained by a neural network based on error back propagation, and if the weight matrix after training meets the convergence condition, the optimal process parameter combination for realizing uniform pressure distribution is obtained by combining and optimizing the process parameters through a genetic algorithm with single-point crossover and Gaussian mutation.

3. The method of claim 1, wherein, The device control strategy includes: collecting copper foil thickness data in real time through an online thickness gauge on a multi-layer composite copper foil production line, processing the collected data through a data analysis method to obtain a spatial distribution map of the copper foil thickness, and judging the uniformity of the thickness distribution; when local area thickness abnormalities are found, adjusting the pressure roller pressure of the corresponding area.

4. The method of claim 1, wherein, The device control strategy includes: collecting pressure data in real time through a pressure sensor, establishing a pressure and thickness mapping relationship through a machine learning algorithm, predicting the deformation amount of the copper foil under different pressure conditions, and guiding the dynamic adjustment of the pressure roller to suppress the thickness difference caused by anisotropy.

5. The method of claim 1, wherein, The device control strategy includes: establishing a mathematical model of the dynamic contact process between the pressure roller and the copper foil according to the production line speed and the deformation characteristics of the copper foil material, simulating the transient pressure distribution under high-speed motion by using a multi-physical field coupling simulation method, obtaining the change law of the copper foil thickness with time and position, and adjusting the operating parameters of the pressure roller in real time to ensure the stability of the copper foil thickness under dynamic working conditions.

6. The method of claim 1, wherein, The optimal process parameter combination and device control strategy for realizing uniform pressure distribution are applied to the copper foil production line, real-time collection of process parameters and equipment operating state data in the composite copper foil production process is performed, the pressure distribution under the current working condition is dynamically predicted, and if the prediction result meets the requirements in the quantitative evaluation index of pressure distribution uniformity, the current process parameters and device control strategy are maintained unchanged, including: According to the real-time data stream collected by the pressure sensor, a Kalman filter of a state equation is used for state estimation, and a preprocessed data sequence is obtained from a recursive formula; For the pre-processed data sequence, the memory unit in the long short-term memory network and the parameter evolution law of the hidden state are used to obtain pressure distribution prediction data from the output layer; According to the pressure distribution prediction data, a Gaussian kernel function is used to construct a pressure distribution probability density curve, a cross-validation method is used to determine the kernel function bandwidth parameter, and a pressure distribution variance value is calculated from the probability density function; For the pressure distribution prediction data and process parameters, a matrix decomposition method is used for singular value decomposition, an eigenvalue sequence is obtained by sorting the eigenvalues, and a parameter sensitivity coefficient is extracted from the feature vector matrix.

7. The method of claim 1, wherein, If the predicted pressure distribution does not meet the uniformity requirement, the process parameters and equipment control strategy are dynamically adjusted through the reinforcement learning algorithm, the pressure distribution is gradually made uniform, and the optimized process parameter combination and control strategy are updated to the knowledge base, including: According to the state vector, a working condition feature space is constructed, and a working condition state representation is obtained by principal component analysis; For the working condition state representation, a policy gradient method is used to construct a parameter optimization function, and a policy gradient value is obtained by Monte Carlo sampling; According to the policy gradient value, an adaptive learning rate optimizer is used for iterative calculation, and a parameter adjustment amount is obtained by Taylor expansion; For the parameter adjustment amount, a deep deterministic policy gradient method is used to construct a neural network model, and an optimized process parameter combination is obtained from the policy network.

8. The method of claim 1, wherein, When the process requirements change, similar historical optimization cases to the current process requirements are extracted from the knowledge base as the initial optimization scheme, and then optimized through the reinforcement learning algorithm, including: According to the process parameters, equipment parameters and pressure distribution, a feature vector is constructed, the similarity value between the feature vectors is calculated by cosine similarity, and the most similar process scheme is obtained by Euclidean distance hierarchical clustering method; For the most similar process scheme, a double network structure containing a critic network and an actor network is constructed, and the critic network calculates the parameter gradient value through the cross-entropy loss function; The parameter gradient value is spatially mapped using a Gaussian kernel function, the kernel function parameters are determined by grid search, and the pressure distribution prediction value is obtained from the regression equation; According to the pressure distribution prediction value, a reward function containing pressure deviation and variance is constructed, a deep substitute policy optimization algorithm is used to update the parameters iteratively, and the optimal parameter combination is obtained by random exploration.

Images