Intelligent control system and method for production process of dimming film

Abstract

The present application relates to the field of intelligent control of production process, and specifically discloses an intelligent control system and method for production process of light-adjustable film, which comprises: obtaining rheological response data and solvent concentration distribution data of wet coating film in film thickness and film surface direction online, identifying rheological anisotropy critical transition characteristic vector and constructing residual solvent anisotropic space-time distribution matrix; inputting the above characteristic into a prediction model, outputting bicontinuous phase structure anisotropy index and thickness direction droplet orientation degree anisotropy index, generating time-varying shear stress waveform parameter set and transmembrane thickness direction negative pressure gradient sequence; applying in-plane shear stress field and transmembrane negative pressure suction force according to the above parameters within the critical relaxation window period, forming a gel network structure with a preset anisotropy degree. The present application realizes accurate pre-programming of gel network anisotropy by actively regulating the anisotropy of solvent evaporation, significantly improving the three-dimensional optical uniformity and low side-view haze performance of light-adjustable film.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology for production processes, specifically to an intelligent control system and method for the production process of dimming films. Background Technology

[0002] Dimming films, such as polymer-dispersed liquid crystal films and electrochromic films, are widely used in automotive window films, architectural glass functional films, and special functional films. Their core principle is to control the alignment and orientation of liquid crystal droplets or functional molecules within the film using an external electric field, thereby adjusting light transmittance. In recent years, with the rapid development of high-end applications such as new energy vehicle roof glass and intelligent building curtain walls, the market has placed higher demands on the optical uniformity, low side-view haze, and large-format consistency of dimming films.

[0003] Currently, the main technical challenges faced in the production of dimming films are as follows: Firstly, during the transition of a coated wet film from a viscous flow state to a gel state, there is a significant difference in the solvent evaporation rate between the film thickness direction and the film surface direction. This difference leads to the formation of an anisotropic gel network structure in three-dimensional space within the polymer matrix. This structure serves as a template for the growth of liquid crystal microdroplets during subsequent curing, directly determining the optical anisotropic properties of the final product. However, existing technologies only focus on the overall solvent content or average evaporation rate, neglecting the essential difference in solvent evaporation rates between the film thickness direction and the film surface direction and its decisive influence on the anisotropy of the gel network. Consequently, they cannot effectively characterize and control the anisotropic mass transfer behavior in three-dimensional space.

[0004] Secondly, existing control methods typically manage temperature, voltage, and tension as independent loops, failing to differentiate the varying anisotropy requirements of the gel network along the film surface and thickness directions. During wet film gelation, shear stress along the film surface primarily affects the in-plane orientation of liquid crystal molecules, while negative pressure suction along the film thickness direction mainly influences the solvent migration path along the thickness direction. Their mechanisms of action on gel network anisotropy are entirely different. Existing technologies employ single control methods or simple superposition approaches, failing to achieve decoupled control of anisotropy along the film surface and thickness directions. This results in mutually coupled control effects, leading to a trade-off between the two.

[0005] Third, the inconsistent solvent evaporation rate is particularly prominent in large-size dimming films due to factors such as uneven tension distribution and temperature field differences along the width direction. Existing negative pressure suction devices use a uniformly distributed suction hole structure, which cannot precisely control the differentiated evaporation requirements of different regions along the width direction of the film. This results in significant differences in optical performance between the edge and center areas of the finished film, making it difficult to meet the stringent requirements for large-format uniformity in applications such as automotive sunroofs and building curtain walls.

[0006] Therefore, there is an urgent need for an intelligent control method that can actively regulate the anisotropy of solvent evaporation and achieve precise pre-programming of gel network anisotropy. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent control system and method for the production process of dimming films, so as to solve the technical problems of the prior art that ignores the essential difference in solvent evaporation rate between the film thickness direction and the film surface direction of the wet film coating of dimming films, and cannot accurately control the degree of anisotropy of the gel network, resulting in poor three-dimensional optical uniformity, high side haze and poor large-format consistency of the final product.

[0008] To solve the above-mentioned technical problems, the present invention specifically provides the following technical solution: A smart control method for the dimming film production process includes the following steps: S1. Obtain rheological response data and solvent concentration distribution data of the coated wet film in the film thickness direction and film surface direction, identify the rheological anisotropic critical transition feature vector and construct the spatiotemporal distribution matrix of residual solvent anisotropy; S2. Input the rheological anisotropic critical transition feature vector and the spatiotemporal distribution matrix of residual solvent anisotropicity into the prediction model, and output the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction; S3. During the critical relaxation window, based on the time-varying shear stress waveform parameter set and the negative pressure gradient sequence in the transmembrane thickness direction, an in-plane shear stress field is applied through an array-type airflow shearing device, and a transmembrane negative pressure suction force is applied through a porous negative pressure suction cup array to form a gel network structure with a preset degree of anisotropy.

[0009] As a preferred embodiment of the present invention, S1 specifically includes: S11. Dynamic rheological response data is acquired by an online rheometer array positioned after the coating roller. The online rheometer array includes a multilayer microprobe arranged along the film thickness direction and a rotating shear probe arranged along the film surface direction. The multilayer microprobe collects the rheological modulus changes at different depths along the film thickness direction and fits them to form a rheological modulus change curve along the film thickness direction. The rotating shear probe collects the rheological modulus changes along the film surface direction and fits them to form a rheological modulus change curve along the film surface direction. S12. Residual solvent concentration distribution data are collected by combining an infrared thermal imaging spectrometer and a terahertz time-domain spectrometer. The infrared thermal imaging spectrometer obtains the planar distribution of solvent concentration along the film surface, and the terahertz time-domain spectrometer obtains the gradient distribution of solvent concentration at different depths along the film thickness. S13. Extract the gel point time difference between the rheological modulus change curve in the thickness direction and the rheological modulus change curve in the surface direction, and use the gel point time difference as the characteristic vector of the critical transition of rheological anisotropy. S14. Extract the ratio of the evaporation rate constant between the solvent concentration gradient distribution along the film thickness direction and the solvent concentration plane distribution along the film surface direction, and use the ratio of the evaporation rate constant as the characteristic value of the spatiotemporal distribution matrix of residual solvent anisotropy.

[0010] As a preferred embodiment of the present invention, S13 specifically includes: S131. The rheological modulus variation curves in the thickness direction and the rheological modulus variation curves in the surface direction are respectively subjected to smoothing filtering to eliminate high-frequency noise interference, so as to obtain the rheological modulus variation curves in the thickness direction and the surface direction after noise reduction. S132. Calculate the first derivative of the rheological modulus change curve in the thickness direction and the first derivative of the rheological modulus change curve in the surface direction after denoising, respectively, and mark the time when the first extreme point of the first derivative appears as the gel point time in the thickness direction and the gel point time in the surface direction, respectively. S133. The gel point time difference is calculated by subtracting the gel point time in the film thickness direction from the gel point time in the film surface direction. S134. Extract the ratio of the rate of change of modulus of the film thickness direction rheological modulus change curve to the rate of change of modulus of the film surface direction before and after the gel point, and perform feature fusion of the time difference at the gel point and the ratio of the rate of change of modulus to form a multi-dimensional feature vector of the critical transition of rheological anisotropy.

[0011] As a preferred embodiment of the present invention, S14 specifically includes: S141. The planar distribution of solvent concentration in the film surface direction is divided into grids along the film width direction and the travel direction. The planar decay curve of solvent concentration decaying with time in each grid cell is obtained. The planar decay curve is subjected to exponential fitting. The evaporation rate constant in the film surface direction of each grid cell is extracted to form a distribution matrix of evaporation rate constant in the film surface direction. S142. The solvent concentration gradient distribution along the film thickness direction is divided into equal layers, and the depth decay curve of the solvent concentration in each depth layer with time decay is obtained. The depth decay curve is fitted with Fick's second law, and the film thickness direction diffusion coefficient of each depth layer is extracted to form the film thickness direction diffusion coefficient distribution vector. S143. Calculate the matrix ratio of the evaporation rate constant distribution matrix in the film surface direction to the diffusion coefficient distribution vector in the film thickness direction, and use the matrix ratio as the basic characteristic value of the spatiotemporal distribution matrix of the residual solvent anisotropy; S144. The basic eigenvalues ​​of the residual solvent anisotropic spatiotemporal distribution matrix are compared with a preset evaporation uniformity threshold. When the basic eigenvalues ​​deviate from the evaporation uniformity threshold, the basic eigenvalues ​​are normalized and corrected to generate a residual solvent anisotropic spatiotemporal distribution matrix that characterizes the difference in evaporation rates of the solvent in the film thickness direction and the film surface direction.

[0012] As a preferred embodiment of the present invention, S2 specifically includes: S21. Input the rheological anisotropic critical transition feature vector and the residual solvent anisotropic spatiotemporal distribution matrix into the phase separation-solidification coupling prediction model, extract spatial and temporal features, and output the anisotropic index of the dual continuous phase structure and the anisotropic index of the orientation degree of microdroplets in the thickness direction. S22. Input the anisotropy index of the dual continuous phase structure and the anisotropy index of the microdroplet orientation degree in the thickness direction into the anisotropy collaborative control strategy generator. The generator calculates the deviation based on the preset ideal anisotropy target value, and determines the film surface direction control amount and the film thickness direction control amount according to the deviation amount. S23. Generate a time-varying shear stress waveform parameter set according to the membrane surface direction control amount, and generate a negative pressure gradient sequence across the membrane thickness direction according to the membrane thickness direction control amount; S24. Configure a preset phase difference between the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction in the time domain, with the shear stress leading in the first segment, the middle segment in phase, and the negative pressure leading in the last segment, to achieve decoupling control of the anisotropy of the gel network in the membrane surface direction and the membrane thickness direction.

[0013] As a preferred embodiment of the present invention, S21 specifically includes: S211. Normalize the rheological anisotropic critical transition feature vector and convert it into a temporal feature input vector; perform two-dimensional interpolation on the residual solvent anisotropic spatiotemporal distribution matrix and convert it into a spatial feature input matrix. S212. The phase separation-curing coupling prediction model adopts a dual-branch architecture of convolutional neural network-long short-term memory network. The parameters of the phase separation-curing coupling prediction model are initialized by pre-training. The pre-training is supervised learning based on the mapping relationship between different process parameters and corresponding anisotropy indices in historical production data. S213. The spatial feature input matrix is ​​input into a convolutional neural network branch, and the convolutional neural network branch extracts spatial features by sequentially passing through a two-dimensional convolutional layer, a batch normalization layer, and a pooling layer, and outputs a spatial feature vector; S214. The temporal feature input vector is input into a branch of a long short-term memory network. The long short-term memory network branch extracts temporal dependency features through memory units with forget gates, input gates, and output gates, and outputs a temporal feature vector. S215. The spatial feature vector and the temporal feature vector are concatenated and fused in the feature fusion layer to form a fused feature vector. The fused feature vector is input into the fully connected layer and output as the anisotropy index of the dual continuous phase structure and the anisotropy index of the orientation degree of the droplets in the thickness direction after being mapped by the activation function.

[0014] As a preferred embodiment of the present invention, S24 specifically includes: S241. Obtain the total duration of the critical relaxation window period, and divide the total duration into a first control phase, a second control phase, and a third control phase according to the time ratio; S242. In the first regulation stage, the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are configured such that the phase of the shear stress waveform leads the phase of the negative pressure gradient waveform, with a leading phase angle of 45° to 90°, so as to preferentially establish an anisotropic gel network skeleton in the membrane surface direction. S243. In the second regulation stage, the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are configured such that the shear stress waveform and the negative pressure gradient waveform are in phase with a phase difference of 0° to 10°, so that the gel network in the membrane surface direction and the membrane thickness direction grow synchronously and collaboratively. S244. In the third regulation stage, the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are configured such that the phase of the negative pressure gradient waveform leads the phase of the shear stress waveform, with a leading phase angle of 30° to 60°, in order to enhance the structural shaping of the gel network in the membrane thickness direction. S245. The phase difference configuration parameters of the first regulation stage, the second regulation stage and the third regulation stage are sequentially spliced ​​to form a complete time-domain phase difference configuration sequence, which is then output to the actuator to achieve decoupling control of the anisotropy of the gel network in the film surface direction and the film thickness direction.

[0015] As a preferred embodiment of the present invention, S3 specifically includes: S31. During the critical relaxation window period before the coated wet film enters the curing phase lock, the time-varying shear stress waveform parameter set is input into an array-type airflow shearing device set above the coated wet film. The array-type airflow shearing device applies an in-plane shear stress field with time-varying characteristics to the surface of the coated wet film according to the time-varying shear stress waveform parameter set. S32. Input the negative pressure gradient sequence across the membrane thickness direction into the porous negative pressure suction cup array, and apply a dynamically changing negative pressure suction force along the membrane thickness direction according to the sequence. The suction hole density of the porous negative pressure suction cup array is gradient distributed along the membrane surface direction. S33. The in-plane shear stress field and the transmembrane negative pressure suction force form a spatiotemporal synergy within the critical relaxation window period, jointly driving the solvent to complete directional mass transfer in the membrane surface direction and membrane thickness direction according to the preset anisotropic evaporation path, inducing the polymer matrix to form a gel network structure with a preset degree of anisotropy in three-dimensional space. S34. The anisotropy index of the solidified phase-locked dual continuous phase structure and the anisotropy index of the microdroplet orientation in the thickness direction are monitored in real time by a multi-channel dielectric spectroscopy analyzer. The monitoring results are fed back to step S1 to update the rheological anisotropy critical transition eigenvector and the spatiotemporal distribution matrix of residual solvent anisotropy, forming a closed-loop adaptive control.

[0016] As a preferred embodiment of the present invention, S32 specifically includes: S321. The porous negative pressure suction cup array is divided into N independently controlled negative pressure regions along the width direction of the film. Each negative pressure region corresponds to a control unit along the width direction of the film. The pore density of the suction holes in each negative pressure region is distributed in a gradient increasing or gradient decreasing manner along the width direction of the film. S322. Obtain the regional negative pressure time series curve corresponding to each negative pressure region in the negative pressure gradient sequence across the membrane thickness direction, and input the regional negative pressure time series curve into the pressure controller of the corresponding negative pressure region; S323. The pressure controller of each negative pressure region adjusts the negative pressure suction force amplitude of each suction hole in the corresponding negative pressure region according to the negative pressure time sequence curve of the region, so that the negative pressure suction force generated by different suction holes in the same negative pressure region forms a pressure gradient distribution that matches the suction hole density gradient along the film width direction. S324. The actual negative pressure suction force value of each negative pressure area is collected in real time by a pressure sensor array set below the multi-hole negative pressure suction cup array. The actual negative pressure suction force value is compared with the target negative pressure value in the negative pressure time series curve of the area, and the pressure deviation value is calculated. S325. The pressure deviation value is input into the PID controller, and the PID controller outputs a correction signal to adjust the output power of the negative pressure source in each negative pressure zone, so that the actual negative pressure suction force value approaches the target negative pressure value, thus forming a closed-loop pressure control for each negative pressure zone.

[0017] An intelligent control system for a dimming film production process, comprising: An anisotropic sensing module is used to acquire online rheological response data and solvent concentration distribution data of the coated wet film in the film thickness direction and film surface direction, identify the rheological anisotropic critical transition feature vector, and construct the spatiotemporal distribution matrix of residual solvent anisotropy; An anisotropic pre-programming module is used to input the rheological anisotropic critical transition feature vector and the residual solvent anisotropic spatiotemporal distribution matrix into the phase separation-curing coupling prediction model, and output the anisotropic index of the dual continuous phase structure and the anisotropic index of the microdroplet orientation degree in the thickness direction. Based on the two, a time-varying shear stress waveform parameter set and a negative pressure gradient sequence in the transmembrane thickness direction are generated. An anisotropic collaborative execution module is used to apply an in-plane shear stress field according to the time-varying shear stress waveform parameter set during the critical relaxation window period, and at the same time apply a transmembrane negative pressure suction force according to the transmembrane thickness direction negative pressure gradient sequence to form a gel network structure with a preset anisotropy.

[0018] Compared with the prior art, the present invention has the following advantages: 1. By identifying the ratio of the gel point time difference and evaporation rate constant between the film thickness direction and the film surface direction, the three-dimensional anisotropy of solvent evaporation is quantitatively characterized, transforming uncontrollable anisotropic evaporation into a preprogrammable anisotropic mass transfer process, thereby achieving precise design of the three-dimensional optical properties of the dimming film.

[0019] 2. A dual-branch architecture prediction model was constructed to decouple and characterize the three-dimensional anisotropic state. By configuring the phase difference in the time domain, the shear stress field and the negative pressure suction force were coordinated in stages within the critical relaxation window. The anisotropy of the gel network in the membrane surface direction and the membrane thickness direction was independently and precisely controlled, thus realizing the decoupled control of the anisotropy in the membrane surface and membrane thickness directions.

[0020] 3. A porous negative pressure suction cup array with a gradient distribution of suction hole density is adopted. Combined with independent zone control and closed-loop feedback, a pressure gradient field matching the evaporation demand in the width direction is formed, which significantly improves the full-width optical uniformity of large-size dimming films and solves the problem of consistency in the width direction of large-size films. Attached Figure Description

[0021] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0022] Figure 1 This is a flowchart illustrating the method described in Embodiment 1 of the present invention.

[0023] Figure 2 This is a framework diagram of the system described in Embodiment 2 of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The concepts involved in this application will first be described with reference to the accompanying drawings. It should be noted that the following descriptions of various concepts are only for the purpose of making the content of this application easier to understand and do not constitute a limitation on the scope of protection of this application; furthermore, the embodiments and features in the embodiments of this application can be combined with each other unless otherwise specified. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] Example 1 like Figure 1 As shown, the present invention provides an intelligent control method for the dimming film production process, comprising the following steps: S1. Obtain rheological response data and solvent concentration distribution data of the coated wet film in the film thickness direction and film surface direction, identify the rheological anisotropic critical transition feature vector, and construct the spatiotemporal distribution matrix of residual solvent anisotropy; specifically including: S11. Online acquisition of dynamic rheological response data, specifically: An online rheometer array is mounted on a frame immediately behind the coating roller, near the wet film exit, using a precision mechanical fixing device. This array consists of two core functional modules: a multilayer microprobe array arranged along the film thickness direction and a rotating shear probe arranged along the film surface direction.

[0027] S111. The multilayer microprobe array employs puncture-type microrheological measurement technology, consisting of five to seven vertically arranged, equally spaced micro shear stress sensors. These sensors are manufactured using microelectromechanical systems (MEMS) technology and feature micron-sized probe tips. Each microprobe is vertically embedded into the coated wet film at a preset penetration depth. The penetration depth is set proportionally to the total thickness of the wet film, corresponding to different depth positions such as the surface, near-surface, intermediate, near-bottom, and bottom layers of the coated wet film, ensuring complete depth coverage from the film surface to the bottom. These microprobes acquire real-time analog signals of the storage modulus and loss modulus of the polymer slurry at each depth layer as a function of time. The sampling frequency is set to the millisecond level, and the sampling accuracy reaches the micropascal level, to accurately capture the subtle mechanical changes during the rapidly occurring sol-gel transition.

[0028] The acquired raw signals are first amplified by a preamplifier, then filtered by a low-pass filter to remove high-frequency electromagnetic interference noise, and finally converted into digital signals by a high-precision analog-to-digital converter before being transmitted to the central data processing unit. In the data processing unit, a weighted least squares fitting algorithm is used to smooth the discrete modulus data points acquired at each depth layer. During the fitting process, cubic spline interpolation is used to fill the data gaps, connecting the discrete spatiotemporal data points to form a continuous and smooth rheological modulus variation curve along the film thickness direction. This curve fully characterizes the mechanical response evolution process of the coated wet film as it transitions from a liquid to a viscoelastic state along the thickness direction, including the characteristics of the rapid modulus growth region and the plateau region.

[0029] S112. The rotating shear probe arranged along the film surface adopts the parallel plate rotating shear rheological measurement principle. The probe body consists of two parallel disks, the upper disk is fixed, and the lower disk is in contact with or maintains a micron-level air gap with the surface of the coated wet film, applying controllable shear strain to the wet film surface in a non-contact or micro-contact manner. The probe is equipped with a high-precision torque sensor and a photoelectric encoder angular displacement sensor. By rotating the lower disk and measuring the generated torque response, the rheological modulus change in the film surface direction is calculated. This rotating shear probe is mounted on a precision linear module and can reciprocate along the film width direction with a set scanning step size, which is usually set to 10 mm to 50 mm. While the film is moving, rheological modulus data at different lateral positions in the width direction are collected.

[0030] The collected spatial discrete data is processed by spatial interpolation algorithm. Kriging interpolation or bicubic spline interpolation method is used to expand the discrete measurement point data into a continuous two-dimensional distributed data field. Then, through numerical fitting, a rheological modulus change curve in the film surface direction is formed. This curve reflects the mechanical response characteristics of the coated wet film during anisotropic flow and curing process in the film surface plane, including the non-uniform distribution of modulus in the width direction.

[0031] S12. Joint acquisition of residual solvent concentration distribution data, specifically: An infrared thermal imaging spectrometer and a terahertz time-domain spectrometer are respectively arranged above and below the wet film coating transmission path to form a multimodal spectral joint acquisition system.

[0032] S121. An infrared thermal imaging spectrometer is fixedly positioned directly above the coated wet film at a fixed height from the film surface. It employs mid-infrared thermal imaging spectroscopy with a wavelength range of 2.5 to 5 micrometers. This wavelength range corresponds to the strong absorption peaks of characteristic functional groups of common organic solvent molecules, such as carbonyl and hydroxyl groups. The spectrometer is equipped with a cooled area array detector with a detector unit count of 320 x 256 or higher resolution. Its detection field of view covers the entire width of the film, and it continuously acquires spectral image data of the film surface at a frame rate of more than 30 frames per second.

[0033] By analyzing the spectral absorption intensity at a specific wavelength and combining it with Beer-Lambert's law to calculate the relationship between optical path and absorbance, a real-time inversion calculation is performed to generate a plane distribution map of solvent concentration along the film surface. This distribution map records the spatial concentration difference of solvent in the film width direction and the travel direction in the form of a two-dimensional pixel matrix. Each pixel corresponds to a specific horizontal and vertical coordinate position and its normalized solvent concentration value, forming complete visualization data of film surface mass transfer.

[0034] The S122. Terahertz time-domain spectrometer is positioned below the coated wet film, employing a transmission measurement configuration. It utilizes a femtosecond laser to excite a photoconductive antenna, generating a broadband terahertz pulse wave with a frequency range of 0.1 to 10 terahertz. This wavelength exhibits characteristic absorption by organic solvent molecules and can penetrate the polymer substrate. The spectrometer is equipped with a precision electrically controlled displacement stage and a parabolic focusing lens. By adjusting the focusing depth of the terahertz beam, it performs layered scanning detection along the film thickness direction, typically dividing the film into five to seven equidistant depth layers.

[0035] During the detection process, terahertz waves penetrate the polymer substrate from bottom to top and interact with the vibrational-rotational energy levels of solvent molecules at different depths. By measuring the time-domain waveform changes of the transmitted signal, Fourier transform is used to extract the frequency-domain absorption spectrum, and the solvent concentration values ​​at each depth layer are calculated. The concentration data obtained from different depth layers are arranged vertically in depth order to form a sequence of solvent concentration gradient distribution data at different depths along the film thickness direction. This data, in the form of depth coordinates corresponding to concentration values, characterizes the non-uniform distribution of solvent in the film thickness direction, and completely records the concentration gradient information and mass transfer kinetics characteristics from the film surface to the film bottom.

[0036] S13. Extraction of the eigenvectors of the critical transition of rheological anisotropy, specifically: S131. Curve smoothing and noise reduction processing, specifically: Savitzky-Golay smoothing filtering algorithm or Gaussian smoothing filtering algorithm is applied to the rheological modulus variation curves in the thickness direction and the surface direction, respectively. An appropriate window width is set, usually 11 to 21 points, to eliminate high-frequency noise interference caused by mechanical vibration, electromagnetic interference or sensor noise. The denoised rheological modulus variation curves in the thickness direction and the surface direction are obtained, ensuring that the subsequent feature extraction is based on the real physical evolution trend rather than measurement noise.

[0037] S132. Gel spot time identification, specifically: Numerical differentiation was performed on the denoised rheological modulus variation curves along the film thickness and the film surface. The central difference method was used to calculate the first derivative value for each data point. For any point on the curve, the modulus difference between its immediate and adjacent data points was divided by the time interval to obtain the instantaneous rate of change at that point. The first derivative sequence of the entire curve was traversed, and extreme points were identified by comparing the first derivative values ​​of three adjacent points. When the first derivative value of a point is greater than that of its immediate and adjacent points, that point was determined to be a local maximum. The first local maximum point was selected as the characteristic point of the gel transition, and the corresponding time coordinates were recorded as the gel point time along the film thickness and the gel point time along the film surface, respectively.

[0038] S133. Calculation of gel point time difference, specifically: A unified time coordinate system is established, mapping the gel point time in the film thickness direction to the gel point time in the film surface direction onto the same time axis for comparison. A time difference calculation is performed, subtracting the gel point time in the film surface direction from the gel point time in the film thickness direction to obtain the gel point time difference. A positive difference indicates that the gelation process in the film thickness direction lags behind that in the film surface direction; a negative difference indicates that the gelation process in the film thickness direction leads that in the film surface direction. The absolute value and sign of this time difference are recorded together as a temporal characteristic parameter to quantify the asynchronous nature of the gelation transition of the coated wet film in three-dimensional space, used subsequently to characterize the time dimension component of the rheological anisotropy critical transition characteristic vector.

[0039] S134. Multi-dimensional feature fusion, specifically: On the rheological modulus change curve along the film thickness direction, curve segments within one time unit before and after the gel point time along the film thickness direction are extracted, and the average slope of these curve segments is calculated as the rate of change of modulus along the film thickness direction. Using the same method, curve segments within one time unit before and after the gel point time along the film surface direction are extracted from the rheological modulus change curve, and the rate of change of modulus along the film surface direction is calculated.

[0040] The modulus change rate in the thickness direction is divided by the modulus change rate in the surface direction to obtain the modulus change rate ratio. This modulus change rate ratio is then vector-concatenated with the gel point time difference obtained in S133 to form a multi-dimensional rheological anisotropic critical transition feature vector containing both time-asynchronous and kinetic differences.

[0041] S14. Construction of the spatiotemporal distribution matrix of residual solvent anisotropy, specifically as follows: S141. Extraction of the evaporation rate constant along the membrane surface, specifically: The planar distribution of solvent concentration along the film surface, acquired by an infrared thermal imaging spectroscopy, was discretized and gridded. Transverse grid lines were evenly spaced along the film width, and longitudinal grid lines were evenly spaced along the film's travel direction, forming a rectangular grid cell array. Each grid cell corresponds to a specific spatial region on the film surface. Time-series data continuously acquired within each grid cell were extracted to obtain a discrete data point sequence showing the solvent concentration change over time within that cell. Connecting these data points formed a planar decay curve, which describes the concentration decay kinetics of solvent evaporation from the film surface corresponding to that grid cell into the bulk gas phase.

[0042] For each grid cell, a nonlinear least-squares exponential fitting is performed on the planar decay curve. The fitting model adopts a single exponential decay function, assuming that the solvent concentration decays exponentially with time. By minimizing the sum of squared residuals between the measured concentration value and the model prediction value, the evaporation rate constant in the film surface direction of that grid cell is obtained through iterative optimization. This constant characterizes the surface mass transfer coefficient of the solvent at that location. The above fitting process is repeated for all grid cells, and the evaporation rate constants in the film surface direction calculated for each grid cell are arranged according to their spatial position to form a two-dimensional array-like distribution matrix of the evaporation rate constants in the film surface direction.

[0043] S142. Extraction of diffusion coefficient in the film thickness direction, specifically: The solvent concentration gradient distribution along the film thickness direction, acquired by terahertz time-domain spectrometry, was equidistantly layered along the film thickness direction. Based on the total wet film thickness, it was divided into 5-7 depth layers, each with equal thickness and corresponding to a specific depth range. The time-series data of continuously monitored solvent concentration changes within each depth layer were extracted to form a depth decay curve describing the decrease in solvent content over time within that depth layer.

[0044] For each depth layer, the depth decay curve is fitted using a nonsteady-state diffusion process based on Fick's second law, which describes the concentration-time-location relationship during one-dimensional nonsteady-state diffusion. The fitting process considers the series mass transfer resistance of solvent diffusion from the membrane interior to the membrane surface and then evaporation. The diffusion equation is solved analytically or using numerical finite difference methods. By minimizing the deviation between the theoretical concentration distribution and the measured concentration value, the thickness-direction diffusion coefficient for that depth layer is extracted. This coefficient characterizes the ability of solvent molecules to diffuse and migrate along the thickness direction within that depth layer. The above fitting calculation is performed sequentially for all depth layers. The thickness-direction diffusion coefficients extracted from each depth layer are then arranged vertically in depth order to form a one-dimensional array-like distribution vector of the thickness-direction diffusion coefficients.

[0045] S143. Calculation of matrix ratios, specifically: The coupled calculation of mass transfer parameters in the film surface direction and film thickness direction is performed. First, the one-dimensional diffusion coefficient distribution vector in the film thickness direction is extended into a two-dimensional matrix with the same row and column dimensions as the evaporation rate constant distribution matrix in the film surface direction through linear interpolation or nearest neighbor interpolation. This ensures that each film surface grid cell corresponds to a specific diffusion coefficient value in the film thickness direction, achieving spatial dimension matching of mass transfer parameters in the two directions.

[0046] Then, element-wise division is performed on the expanded thickness-direction diffusion coefficient matrix and the surface-direction evaporation rate constant distribution matrix, i.e., the ratio of the surface-direction evaporation rate constant to the thickness-direction diffusion coefficient is calculated for each spatial location. This ratio quantifies the relative strength of the solvent's evaporation capacity in the horizontal surface direction and its diffusion capacity in the vertical thickness direction, reflecting the dominant direction and anisotropy of the solvent mass transfer process. All calculated ratios are organized according to their original spatial locations to form a matrix ratio with the same dimension as the surface grid. This matrix ratio serves as the fundamental characteristic value of the spatiotemporal distribution matrix of residual solvent anisotropy, directly characterizing the difference in solvent evaporation rates in the thickness and surface directions.

[0047] S144. Normalization correction and matrix generation, specifically: An anisotropy judgment and correction mechanism is established. First, the basic characteristic value calculated in S143 is compared element-by-element with a preset volatility uniformity threshold. This threshold is typically set as the characteristic ratio under theoretical isotropic conditions or an allowable fluctuation range determined based on process experience. The absolute or relative deviation between the basic characteristic value and the volatility uniformity threshold is calculated. When this deviation exceeds the preset allowable tolerance range, it is determined that there is significant mass transfer anisotropy at that location, requiring correction.

[0048] For the basic eigenvalues ​​identified as deviating, a normalization correction algorithm is applied. The Z-score standardization method converts the original values ​​into standard scores with a mean of 0 and a standard deviation of 1, or the Min-Max normalization method linearly maps the values ​​to a unified interval of zero to one, eliminating dimensional and numerical scale differences between different batches or locations. The normalized eigenvalues ​​are then reorganized according to their original spatiotemporal coordinates and filled into a three-dimensional tensor structure to generate the final spatiotemporal distribution matrix of residual solvent anisotropy. This matrix fully preserves the spatial distribution information and temporal evolution trend of the difference in solvent evaporation rates along the film thickness and film surface directions.

[0049] S2. Input the rheological anisotropic critical transition eigenvector and the spatiotemporal distribution matrix of the residual solvent anisotropically into the prediction model, and output the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the film thickness direction; specifically including: S21. Feature extraction and exponential output of the phase separation-solidification coupling prediction model are as follows: S211. Data preprocessing and input transformation, specifically: Normalization is performed on the rheological anisotropic critical transition feature vector. The max-min normalization algorithm is used to traverse the two components of the feature vector, namely the gel point time difference and the ratio of the modulus change rate. The maximum and minimum values ​​of each component are calculated respectively. The original values ​​are mapped to a closed interval of zero to one through linear transformation to eliminate the numerical scale difference caused by different physical dimensions and form a standardized time-series feature input vector.

[0050] Two-dimensional interpolation is performed on the spatiotemporal distribution matrix of the residual solvent anisotropy. A bicubic interpolation algorithm is used to subdivide the original matrix into grids along the rows and columns. The values ​​of newly inserted nodes are determined by weighted calculations of the 16 neighboring nodes. The matrix dimension is expanded or compressed to a fixed size that matches the specifications of the input layer of the convolutional neural network, forming a normalized spatial feature input matrix. This preprocessing ensures that the temporal feature input vector has uniform dimensions and a stable numerical distribution, and that the spatial feature input matrix has standard dimensions and continuous spatial information, laying the foundation for the standardized input of the subsequent two-branch neural network.

[0051] S212. Model architecture construction and pre-training parameter initialization, specifically: The phase separation-solidification coupling prediction model employs a dual-branch composite architecture combining a convolutional neural network (CNN) and a long short-term memory (LSTM) network. The CNN branch is responsible for spatial feature extraction, while the LSM branch is responsible for temporal feature extraction. Before online prediction, the model requires pre-training to initialize its parameters. This pre-training process is based on a sample set of historical production data. This sample set includes multiple sets of rheological anisotropic critical transition feature vectors collected under different process conditions, residual solvent anisotropic spatiotemporal distribution matrices, and measured labels for the corresponding batches' bicontinuous phase structure anisotropy index and thickness-direction droplet orientation anisotropy index. Through supervised learning, process parameters are used as input features, and anisotropy indices are used as output labels. The backpropagation algorithm and gradient descent optimizer iteratively adjust the network weight parameters, minimizing the mean squared error loss function between predicted and measured values ​​until the loss converges below a preset threshold, thus completing the initialization and solidification of the model parameters.

[0052] S213. Spatial feature extraction and vector generation, specifically: The spatial feature input matrix is ​​fed into a convolutional neural network branch, which consists of multiple processing layers connected in series. First, a two-dimensional convolutional layer performs local perception. The convolutional kernel slides across the matrix with a set stride, performing dot product operations with local regions to extract the local spatial correlation of residual solvent distribution, generating a primary feature map. Next, a batch normalization layer calculates the mean and variance of each channel of the feature map, performs a subtraction of the mean and division of the variance operation, and introduces learnable scaling and offset parameters to stabilize the data distribution and accelerate network convergence. Then, a pooling layer performs downsampling, using max pooling to select the maximum value within a local region as a representative, reducing the spatial dimensionality of the feature map while preserving salient features. After repeated processing through multiple layers of convolution, normalization, and pooling, a flattening operation finally transforms the multidimensional feature map into a one-dimensional high-dimensional spatial feature vector. This vector encodes the spatial pattern and topological structure information of the solvent distribution.

[0053] S214. Temporal feature extraction and vector generation, specifically: The temporal feature input vector is fed into a branch of the Long Short-Term Memory (LSTM) network, which consists of multiple memory units connected in series. Each memory unit contains three gate structures: a forget gate, an input gate, and an output gate. The forget gate uses a sigmoid function to calculate a coefficient between zero and one, controlling the degree to which historical information from the previous cell state is retained. The input gate also uses a sigmoid function to determine the proportion of current input information to be accepted and multiplies it with candidate memory content to determine new information. The output gate regulates the proportion of cell states transitioning to hidden states. The temporal feature input vector is sequentially input into the memory units at each time step. The cell state is passed and updated along the time axis. The forget gate and the input gate work together to filter and accumulate long-term dependent information, and the output gate controls the feature expression at the current time step. After recursive processing over multiple time steps, the LSM branch outputs a temporal feature vector, which encodes the temporal evolution and long-term dependencies of rheological transformation and solvent mass transfer.

[0054] S215. Feature fusion and exponential output, specifically: Spatial and temporal feature vectors are transmitted to a feature fusion layer, where a lateral concatenation operation is performed. This concatenates the elements of each dimension of the spatial feature vector with those of the temporal feature vector, forming a fused feature vector. This fused feature vector contains both spatial distribution information along the membrane thickness and surface direction, as well as temporal evolution information. The fused feature vector is then input into a fully connected layer, which consists of multiple neurons. Each neuron performs a weighted summation operation on the input vector and adds a bias term; the weight parameters are optimized through pre-training. The result after linear transformation by the fully connected layer is input into a nonlinear activation function, mapped using either ReLU or Sigmoid functions to introduce nonlinear expressive power. The final output layer contains two neurons, corresponding to the anisotropy index of the bicontinuous phase structure and the anisotropy index of droplet orientation along the thickness direction, respectively. Through linear activation or normalization of the output, two quantitative index values ​​characterizing the expected anisotropy of the gel network are obtained.

[0055] S22. Calculation of deviation and determination of control amount for anisotropic collaborative control strategy, specifically as follows: The anisotropy index of the dual continuous phase structure and the anisotropy index of droplet orientation in the thickness direction output by S21 are input into the anisotropy synergistic control strategy generator. This generator internally stores preset ideal anisotropy target values, including the target anisotropy index of the dual continuous phase structure and the target anisotropy index of droplet orientation in the thickness direction. These target values ​​are preset according to product performance requirements. The generator executes a deviation calculation program, performing difference calculations between the currently measured anisotropy index of the dual continuous phase structure and the anisotropy index of droplet orientation in the thickness direction and their corresponding ideal target values, respectively, to obtain the deviation amount of the dual continuous phase structure and the deviation amount of droplet orientation in the thickness direction.

[0056] Based on the magnitude and sign of the deviation, and the preset control gain coefficient, piecewise linear mapping or nonlinear sigmoid mapping algorithms are used to calculate and determine the control amounts in the film surface direction and the film thickness direction, respectively. The control amount in the film surface direction represents the time-varying shear stress intensity level to be applied in the film surface direction, and the control amount in the film thickness direction represents the negative pressure gradient intensity level to be applied in the film thickness direction. The determination of the control amounts follows the principle of dynamic compensation. When the deviation deviates positively and exceeds the set threshold, the control amount in the corresponding direction is increased to suppress excessive anisotropy. When the deviation deviates negatively, the control amount is decreased or the control direction is changed to promote anisotropy development, ensuring that the control strategy accurately corresponds to the defect type of the gel network structure.

[0057] S23. Generation of the time-varying shear stress waveform parameter set and negative pressure gradient sequence, specifically: Based on the membrane orientation control amount determined in S22, a time-varying shear stress waveform parameter set is generated through a preset parameter mapping algorithm. This parameter set specifically includes four key parameters: shear frequency, shear amplitude, shear waveform type, and initial phase angle. The shear frequency is dynamically selected within the range of 0.1Hz to 10Hz according to the control amount; the larger the control amount, the higher the frequency to accelerate the response. The shear amplitude is set proportionally to the control amount, and the waveform type is selected from sine wave, triangular wave, or sawtooth wave according to process requirements.

[0058] Based on the control parameters along the film thickness direction, a negative pressure gradient sequence across the film thickness direction is generated using a gradient calculation algorithm. This sequence includes the negative pressure intensity amplitude, the duration of the negative pressure application, and the initial phase angle. The negative pressure intensity amplitude exhibits a non-linear gradient distribution along the film thickness direction from the film surface to the film bottom, with the maximum negative pressure value at the film bottom to enhance solvent removal from the underlying layer, and the minimum negative pressure value at the film surface to prevent premature surface crusting, forming a directional suction gradient pointing towards the film bottom. Two parameter sets encapsulate the physical field control requirements along the film surface direction and the film thickness direction, respectively, constituting a complete process control instruction set, providing executable quantitative instructions for subsequent spatiotemporal synergistic effects.

[0059] S24. Implementation of time-domain phase difference configuration and decoupling control, specifically as follows: S241. By analyzing the gel point time difference and modulus change rate in the rheological anisotropic critical transition characteristic vector obtained in step S1, and combining the preset gelation kinetic model or real-time gelation process data monitored online, the total duration of the critical relaxation window period from the viscous flow state to the gel state until complete phase lock-in is determined.

[0060] Based on process requirements, the total time is divided into three control stages: the first control stage, the second control stage, and the third control stage. The first control stage corresponds to the network nucleation and framework formation stage in the early stage of gelation, the second control stage corresponds to the rapid growth and expansion stage of the network in the middle stage of gelation, and the third control stage corresponds to the network stabilization and phase-locked solidification stage in the later stage of gelation. The time allocation ratio of the three stages is usually set to 1:2:1 or equally divided according to the actual production cycle, ensuring that the sum of the duration of each stage is strictly equal to the total duration of the critical relaxation window, thus forming a complete time-domain control framework.

[0061] S242. In the first regulation stage, the shear stress waveform corresponding to the time-varying shear stress waveform parameter set generated in step S23 is configured with phase coupling with the negative pressure gradient waveform corresponding to the negative pressure gradient sequence across the film thickness direction, so that the phase angle of the shear stress waveform leads the phase angle of the negative pressure gradient waveform. The leading phase angle value is set between 45° and 90°. This phase leading configuration is achieved by adjusting the start time of the piezoelectric ceramic micro-motion platform relative to the start time of the vacuum negative pressure cavity, or by adjusting the initial phase offset of the driving signals of the two, to ensure that in the early stage of gelation, when the polymer chain segments have not yet entangled in large quantities, the polymer matrix is ​​preferentially induced to form a pre-oriented anisotropic gel network skeleton in the film surface direction by the in-plane shear stress field, while the delayed negative pressure gradient only provides the basic solvent removal function, avoiding premature mass transfer in the film thickness direction from interfering with the establishment of molecular orientation in the film surface direction.

[0062] S243. In the second regulation stage, the time-varying shear stress waveform parameter set and the transmembrane thickness direction negative pressure gradient sequence are configured to be in phase or quasi-in phase, so that the phase difference between the shear stress waveform and the negative pressure gradient waveform is controlled within a very small range of 0° to 10°. This configuration is achieved by precisely synchronizing the vibration period of the piezoelectric ceramic micro-motion platform with the suction period of the vacuum negative pressure cavity, or by finely adjusting the phase of the driving signals of the two to make them coincide in time. This ensures that the in-plane shear stress field and the transmembrane negative pressure suction force act simultaneously on the coated wet film during the rapid network growth stage in the middle stage of gelation, so that the shear-induced orientation in the film surface direction and the gradient mass transfer-induced concentration in the film thickness direction occur synchronously in time and space, promoting the balanced and coordinated growth of the polymer network in three-dimensional space, avoiding structural defects caused by excessive development in one direction, and realizing the synchronous strengthening of the gel network skeleton in the plane and thickness directions.

[0063] S244. In the third regulation stage, the relative phase relationship between the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the film thickness direction is adjusted so that the phase angle of the negative pressure gradient waveform leads the phase angle of the shear stress waveform. The leading phase angle value is set between 30° and 60°. This configuration is achieved by delaying the vibration start time of the piezoelectric ceramic micro-motion platform or advancing the suction start time of the vacuum negative pressure chamber. This ensures that during the critical period when the network is close to phase lock-in and solidification in the later stage of gelation, the directional mass transfer and structural compaction in the film thickness direction are preferentially enhanced. The residual solvent is completely removed and the microdroplet orientation in the film thickness direction is solidified by high-intensity negative pressure gradient suction. The delayed shear stress only maintains slight vibration disturbance to prevent interface defects, thereby ensuring that the gel network structure in the film thickness direction is fully solidified while completing the final densification.

[0064] S245. The independent phase difference configuration parameters of the first, second, and third control stages are sequentially spliced ​​together to form a continuous temporal phase difference configuration sequence spanning the entire critical relaxation window. This sequence, indexed by timestamps, records the expected phase relationship and intensity ratio between the surface shear stress field and the transmembrane negative pressure suction force at each time point, constituting a complete dynamic control instruction set. This temporal phase difference configuration sequence is output to a distributed actuator, including a piezoelectric ceramic micro-motion platform controller and a vacuum negative pressure chamber solenoid valve controller, driving the two physical fields to perform spatiotemporal synergy according to a preset phase evolution law within the critical relaxation window before solidification and phase locking, ultimately achieving decoupling control of the anisotropy of the gel network in the membrane surface and thickness directions in the temporal domain.

[0065] S3. Within the critical relaxation window, based on the time-varying shear stress waveform parameter set and the negative pressure gradient sequence in the transmembrane thickness direction, an in-plane shear stress field is applied through an array-type airflow shearing device, and a transmembrane negative pressure suction force is applied through a porous negative pressure suction cup array, forming a gel network structure with a preset degree of anisotropy; specifically including: S31. The application of the in-plane shear stress field is as follows: During the critical relaxation window before the coated wet film enters the curing phase-locking stage, the time-varying shear stress waveform parameter set generated in step S24 is transmitted to an array-type airflow shearing device positioned above the coated wet film. This device consists of multiple rows of parallel micro-airflow nozzle arrays, with the nozzle arrays arranged at equal intervals along the film width direction. Each row of nozzles corresponds to a control zone in the film width direction, and the nozzle outlets face the surface of the coated wet film and maintain a constant air gap distance from the film surface.

[0066] The shear frequency, shear amplitude, and phase information from the time-varying shear stress waveform parameter set are converted into corresponding pulse width modulation signals by a signal generator, driving the high-speed solenoid valves or piezoelectric actuators inside each nozzle to open and close at a set frequency. When the high-speed airflow is ejected from the nozzle and impacts the wet film surface, a frictional shear force is generated at the gas-solid interface along the film surface direction. By adjusting the jet pressure and pulse frequency of each nozzle, an in-plane shear stress field with time-varying characteristics is formed on the coated wet film surface. This stress field changes periodically according to the time-domain evolution law of the time-varying shear stress waveform parameter set, inducing the polymer chain segments to undergo controlled orientation in the film surface direction.

[0067] S32. The application and closed-loop control of transmembrane negative pressure suction force are as follows: S321. The porous negative pressure suction cup array is physically divided into N independent and airtight negative pressure regions along the width of the film. Each negative pressure region corresponds to a specific control unit along the width of the film, undertaking the independent suction control function of that width segment. Suction holes are processed and arranged on the suction cup body of each negative pressure region. The opening density of each suction hole shows a gradient increasing or decreasing distribution along the width of the film, forming a non-uniform spatial arrangement pattern. That is, by changing the number of holes per unit area, differentiated air permeability is constructed, so that different positions in the width of the film bear differentiated basic suction resistance, laying the structural foundation for the subsequent formation of a pressure gradient distribution in the width direction.

[0068] S322. Extract the regional negative pressure time-series curves corresponding to each negative pressure region from the negative pressure gradient sequence across the membrane thickness direction. These curves describe the target trajectory of the negative pressure intensity to be applied by the corresponding control unit during the critical relaxation window period, evolving over time. The extracted regional negative pressure time-series curve data is imported into the pressure controller of the corresponding negative pressure region via an industrial bus or analog transmission line. This pressure controller, acting as the execution decision unit for that region, receives and parses the target negative pressure value sequence from the time-series curves, converting it into executable valve opening commands or pump speed commands, providing real-time setpoints for independent pressure regulation in each region.

[0069] S323. The pressure controller for each negative pressure zone adjusts the actual negative pressure suction force amplitude of each suction orifice in the corresponding negative pressure zone in real time based on the received negative pressure time sequence curve. By changing the opening of the proportional valve or the speed of the variable frequency vacuum pump, the gas flow rate through each suction orifice is controlled, thereby adjusting the vacuum level at the orifice. Since each suction orifice in the same negative pressure zone has a density gradient distribution along the width of the film, combined with the differential suction force amplitude adjustment, the actual negative pressure suction force generated by suction orifices at different positions forms a pressure gradient distribution along the width of the film that matches the density gradient of the suction orifice, thus realizing the construction of a non-uniform but controllable negative pressure field in the width direction.

[0070] S324. A pressure sensor array arranged below the porous negative pressure suction cup array and corresponding to each negative pressure area is used to collect the actual negative pressure suction force value established in each negative pressure area in real time. The collected actual negative pressure suction force value is compared with the target negative pressure value corresponding to the current moment in the area negative pressure time series curve. A subtraction operation is performed to obtain the pressure deviation value between the two. This deviation value reflects the degree of deviation between the actual suction effect and the process set value in real time, providing a quantitative error signal for subsequent closed-loop correction control.

[0071] S325. The calculated pressure deviation value is input into a proportional-integral-derivative (PID) controller. This controller performs a linear combination calculation on the pressure deviation value based on preset proportional coefficients, integral time constants, and derivative time constants, generating an adjustment signal to correct the control output. This correction signal is output to the negative pressure source power adjustment unit in each negative pressure zone. By changing the speed of the vacuum pump motor or adjusting the opening of the suction pipe valve, the output power of the negative pressure source is adjusted in real time, so that the actual negative pressure suction force dynamically tracks and approaches the target negative pressure value, forming a real-time closed-loop pressure control of the negative pressure suction force in each negative pressure zone.

[0072] S33. Spatiotemporal synergy and gel network induction, specifically: Within the critical relaxation window, the in-plane shear stress field applied by the array-type airflow shearing device and the transmembrane negative pressure suction force applied by the porous negative pressure suction cup array work together in a precise spatiotemporal coordination according to the temporal phase difference configuration sequence configured in step S24. Specifically: S331. In the first regulation stage, the in-plane shear stress field is established first and plays a dominant role. The periodically changing shear force induces the polymer chain segments to undergo preferential orientation in the film surface direction, which promotes the liquid crystal droplets to initially arrange along the shear direction in the film surface plane, forming a pregel network skeleton with anisotropic characteristics. At this time, the transmembrane negative pressure suction force maintains a low background level and only performs the basic solvent removal function, avoiding premature mass transfer in the film thickness direction to disturb the molecular orientation process in the film surface direction.

[0073] S332. Entering the second regulation stage, the transmembrane negative pressure suction force is synchronously enhanced to an intensity level comparable to the in-plane shear stress field according to the phase configuration. The two work synergistically, with the in-plane shear stress field maintaining and reinforcing the molecular orientation order along the membrane surface, while the transmembrane negative pressure suction force establishes a significant directional mass transfer driving force along the membrane thickness direction. This forces residual solvent molecules to preferentially migrate from the membrane surface to the membrane bottom along a direction perpendicular to the membrane surface, while simultaneously inhibiting lateral diffusion of the solvent along the membrane surface direction, resulting in a controlled gradient distribution of solvent concentration along the membrane thickness direction. During this stage, the distribution of liquid crystal droplets along the membrane thickness direction is affected by the negative pressure-driven solvent flow, resulting in directional migration and spatial rearrangement.

[0074] S333. Transitioning to the third regulation stage, the transmembrane negative pressure suction force becomes the dominant physical field according to the phase-leading configuration. Its intensity is further increased and maintained until the end of the critical relaxation window period. High-intensity directional suction completely removes the deep residual solvent and completes the final shaping of the liquid crystal droplet distribution in the film thickness direction. Meanwhile, the in-plane shear stress field is weakened to an auxiliary level, providing only slight vibration disturbance to prevent the generation of interface defects.

[0075] Through the aforementioned phased spatiotemporal synergistic effect, solvent molecules complete directional mass transfer along a predetermined anisotropic evaporation path in both the film surface and thickness directions. Specifically, the solvent preferentially diffuses vertically outward along the film thickness direction and is then drawn off. Simultaneously, a non-uniform but controllable concentration distribution field is formed in the film surface direction. This directional mass transfer process induces asymmetric phase separation kinetics in the polymer matrix, prompting the polymer phase and liquid crystal phase to form a bicontinuous phase gel network structure with a predetermined degree of anisotropy in three-dimensional space. The polymer network skeleton exhibits shear-induced preferred orientation in the film surface direction and negative pressure-driven gradient densification in the film thickness direction. The spatial distribution of liquid crystal droplets simultaneously exhibits ordered arrangement in the film surface direction and gradient distribution characteristics in the film thickness direction.

[0076] S34. Closed-loop monitoring and adaptive feedback, specifically: After the wet film is coated, cured, and exits the curing oven, the film is immediately subjected to online non-destructive monitoring using a multi-channel dielectric spectroscopy analyzer located at the exit of the curing oven. This analyzer employs broadband dielectric spectroscopy measurement technology, applying an alternating electric field to the film within a frequency range of 10 Hz to 10 MHz while simultaneously measuring its complex dielectric constant response. By analyzing the spectral curves of the real and imaginary parts of the complex dielectric constant as a function of frequency, characteristic dielectric relaxation peaks generated by the polarization of the liquid crystal microdroplet interface and the relaxation process of the polymer network are identified. Since the cured bicontinuous phase gel network structure exhibits differentiated dielectric response characteristics in the film surface direction and film thickness direction, the anisotropy characteristics of the intensity, frequency position, and relaxation time distribution of the dielectric relaxation peaks are analyzed. Real-time inversion calculations are then performed to determine the anisotropy index of the cured bicontinuous phase structure and the anisotropy index of the microdroplet orientation in the thickness direction. The former quantifies the degree of asymmetry in the continuous phase structure of the polymer network framework and the liquid crystal phase in three-dimensional space, while the latter quantifies the orientational order of the liquid crystal microdroplets along the film thickness direction.

[0077] The two anisotropy indices obtained from real-time monitoring are numerically compared with the predicted values ​​output by the phase separation-solidification coupling prediction model in step S2. The model prediction deviation is calculated, and the deviation value, together with the measured anisotropy index value, is fed back to the data acquisition and processing unit in step S1 via industrial Ethernet or fieldbus.

[0078] In step S1, the feedback information is used to execute an adaptive update algorithm, which specifically includes updating the identification threshold parameters of the rheological anisotropy critical transition feature vector, such as adjusting the extraction weight of the ratio of the determination sensitivity of the gel point time difference to the modulus change rate, and correcting the calculation weight coefficient of the spatiotemporal distribution matrix of residual solvent anisotropy, such as adjusting the normalized mapping relationship between the evaporation rate constant in the film surface direction and the diffusion coefficient in the film thickness direction, so that the initial feature identification of the next production batch more accurately reflects the actual process state, thereby forming a closed-loop adaptive control from process execution to quality monitoring to model optimization, and realizing continuous and precise control of the anisotropy of the dimming film gel network.

[0079] Example 2 like Figure 2 As shown, an intelligent control system for a dimming film production process is used to implement an intelligent control method for the dimming film production process, comprising: A. Anisotropy sensing module, used to acquire online rheological response data and solvent concentration distribution data of the coated wet film in the film thickness direction and film surface direction, identify the rheological anisotropy critical transition feature vector, and construct the spatiotemporal distribution matrix of residual solvent anisotropy; specifically including: An online rheometer array, comprising multilayer microprobes arranged along the film thickness direction and a rotating shear probe arranged along the film surface direction, is used to acquire the rheological modulus change curves along the film thickness direction and the rheological modulus change curves along the film surface direction. Infrared thermal imaging spectrometer and terahertz time-domain spectrometer are used to jointly acquire the planar distribution of solvent concentration along the film surface and the gradient distribution of solvent concentration along the film thickness. The feature extraction unit is used to extract the time difference between gel points in the thickness direction and the surface direction as the feature vector of the critical transition of rheological anisotropy, and to extract the ratio of the evaporation rate constant in the thickness direction and the surface direction as the feature value of the spatiotemporal distribution matrix of residual solvent anisotropy.

[0080] B. Anisotropic pre-programming module, used to input the rheological anisotropic critical transition eigenvector and the residual solvent anisotropic spatiotemporal distribution matrix into the phase separation-solidification coupling prediction model, outputting the bicontinuous phase structure anisotropy index and the thickness-direction droplet orientation anisotropy index, and based on these, generating a time-varying shear stress waveform parameter set and a transmembrane thickness-direction negative pressure gradient sequence; specifically including: The phase separation-solidification coupling prediction model adopts a dual-branch architecture of convolutional neural network and long short-term memory network to extract spatial and temporal features and output the anisotropy index of the dual continuous phase structure and the anisotropy index of droplet orientation in the thickness direction. An anisotropic collaborative control strategy generator is used to generate a time-varying shear stress waveform parameter set and a transmembrane thickness direction negative pressure gradient sequence based on the anisotropic index of the dual continuous phase structure and the anisotropic index of the microdroplet orientation degree in the thickness direction. The phase difference configuration unit is used to configure a preset phase difference between the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction in the time domain. The shear stress leads in the first segment, the middle segment is in phase, and the negative pressure leads in the last segment, thereby achieving decoupled control of the anisotropy of the gel network in the membrane surface direction and the membrane thickness direction.

[0081] C. An anisotropic collaborative execution module, used to apply an in-plane shear stress field according to a time-varying shear stress waveform parameter set during the critical relaxation window, and simultaneously apply a transmembrane negative pressure suction force according to a negative pressure gradient sequence in the transmembrane thickness direction, forming a gel network structure with a preset anisotropy; specifically including: An array-type airflow shearing device is positioned above the coated wet film to apply an in-plane shear stress field with time-varying characteristics to the surface of the coated wet film based on a set of time-varying shear stress waveform parameters. A porous negative pressure suction cup array is set below the coated wet film to apply a dynamically changing negative pressure suction force along the film thickness direction to the coated wet film according to the negative pressure gradient sequence across the film thickness direction. The suction hole density of the porous negative pressure suction cup array is gradient distributed along the film surface direction. A pressure sensor array is positioned below the porous negative pressure suction cup array to collect the actual negative pressure suction force value in real time. A PID controller is used to compare the actual negative pressure suction force with the target negative pressure value and output a correction signal to form a closed-loop pressure control. A multi-channel dielectric spectrum analyzer is used to monitor the anisotropy index and the thickness direction droplet orientation anisotropy index of the dual continuous phase structure after solidification and phase locking in real time, and feeds the monitoring results back to the anisotropy sensing module to form a closed-loop adaptive control.

[0082] As can be seen from the above description, the embodiments of the present invention achieve the following technical effects: Existing technologies generally treat solvent evaporation during the coating of dimming films as a uniform, isotropic process, focusing only on the total residual solvent or average evaporation rate, neglecting the fundamental difference in evaporation rates between the film thickness direction and the film surface direction and its decisive influence on the anisotropy of the gel network. This invention achieves precise quantitative characterization of the anisotropy of the gelation process in the three-dimensional space of the wet film by online identification of the gelation point time difference and the ratio of the evaporation rate constant between the film thickness direction and the film surface direction. Based on this, a phase separation-curing coupling prediction model is used to predict the degree of anisotropy of the gel network and generate differentiated synergistic control parameters, transforming the originally uncontrollable anisotropic evaporation into a designable and pre-programmable anisotropic mass transfer process. This fundamentally solves the technical problem of the inability of traditional processes to accurately control the three-dimensional optical properties of dimming films, providing a new technical path for the preparation of dimming films with low side-view haze and wide-viewing-angle consistent properties.

[0083] Existing technologies employ a single control method for overall control of the wet film, failing to differentiate the varying anisotropic requirements of the gel network along the film surface and thickness directions. This results in mutually coupled control effects, leading to compromises in one aspect. This invention constructs a phase separation-curing coupling prediction model with a convolutional neural network-long short-term memory (LSTM) dual-branch architecture. The convolutional neural network branch extracts the spatial features of the spatiotemporal distribution matrix of residual solvent anisotropy to characterize anisotropy along the film surface, while the LSTM branch extracts the temporal features of the critical transition feature vector of rheological anisotropy to characterize anisotropy along the film thickness, achieving decoupled characterization of the three-dimensional anisotropic state. Furthermore, this invention introduces a temporal phase difference configuration mechanism. In the early stage of the critical relaxation window, shear stress precedes negative pressure to preferentially establish the gel network framework along the film surface; in the middle stage, both are configured to be in phase to achieve synchronous three-dimensional growth; and in the later stage, negative pressure precedes shear stress to strengthen the structural shaping along the film thickness. This achieves independent decoupled control of the anisotropy of the gel network along the film surface and thickness, enabling the precise formation of the three-dimensional optical structure of the dimming film along a preset trajectory.

[0084] Existing negative pressure suction devices employ a uniformly distributed suction hole structure, which cannot address the inconsistent solvent evaporation rates in large-size dimming films due to uneven tension distribution and temperature field differences along the width direction. This results in significant differences in optical performance between the edge and center regions of the finished film. This invention pioneers a porous negative pressure suction cup array with a gradient distribution of suction hole density along the film surface. Combined with independently controlled negative pressure regions divided along the width direction, differentiated negative pressure suction forces can be applied to different regions based on the negative pressure gradient sequence across the film thickness, forming a pressure gradient field that matches the evaporation requirements along the film's width. Simultaneously, a pressure sensor array collects actual negative pressure values ​​in real time, and a PID controller forms a closed-loop pressure control, ensuring precise execution of the negative pressure suction force in each region. Finally, the anisotropy index after solidification and phase-locking is monitored in real time by a multi-channel dielectric spectrum analyzer and fed back to the front-end feature extraction, forming a complete closed-loop adaptive control system from anisotropic feature identification to anisotropic collaborative regulation and then to anisotropic effect verification. This significantly improves the full-area optical consistency of large-size dimming films in mass production, meeting the stringent requirements of large-area uniformity for high-end application scenarios such as automotive sunroofs and building curtain walls.

[0085] The embodiments and / or implementation methods described above are merely preferred embodiments and / or implementation methods for implementing the technology of the present invention, and are not intended to limit the implementation methods of the technology of the present invention in any way. Any person skilled in the art may make some modifications or alterations to other equivalent embodiments without departing from the scope of the technical means disclosed in the present invention, but these should still be regarded as the technology or embodiments that are substantially the same as the present invention. This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. The above descriptions are only preferred embodiments of this application. It should be noted that due to the limitations of written expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A smart control method for the dimming film production process, characterized in that, include: The rheological response data and solvent concentration distribution data of the coated wet film in the film thickness direction and film surface direction are obtained, the rheological anisotropic critical transition feature vector is identified, and the spatiotemporal distribution matrix of residual solvent anisotropy is constructed. The rheological anisotropic critical transition feature vector and the spatiotemporal distribution matrix of residual solvent anisotropy are input into the prediction model, and the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are output. Within the critical relaxation window, based on the time-varying shear stress waveform parameter set and the negative pressure gradient sequence in the transmembrane thickness direction, an in-plane shear stress field is applied by an array-type airflow shearing device, and a transmembrane negative pressure suction force is applied by a porous negative pressure suction cup array to form a gel network structure with a preset degree of anisotropy.

2. The intelligent control method for a dimming film production process according to claim 1, characterized in that, The process of acquiring rheological response data and solvent concentration distribution data of the coated wet film in the film thickness and film surface directions, identifying the rheological anisotropic critical transition feature vector, and constructing the spatiotemporal distribution matrix of residual solvent anisotropy specifically includes: Dynamic rheological response data is acquired by an online rheometer array positioned after the coating roller. The online rheometer array includes a multilayer microprobe and a rotating shear probe. The multilayer microprobe collects the rheological modulus changes at different depths in the film thickness direction and fits them to form a rheological modulus change curve in the film thickness direction. The rotating shear probe collects the rheological modulus changes in the film surface direction and fits them to form a rheological modulus change curve in the film surface direction. Residual solvent concentration distribution data were acquired by combining an infrared thermal imaging spectrometer and a terahertz time-domain spectrometer. The infrared thermal imaging spectrometer obtained the planar distribution of solvent concentration along the film surface, and the terahertz time-domain spectrometer obtained the gradient distribution of solvent concentration at different depths along the film thickness. Extract the gel point time difference between the rheological modulus change curve in the thickness direction and the rheological modulus change curve in the surface direction, and use the gel point time difference as the characteristic vector of critical rheological anisotropy transition. The ratio of the evaporation rate constant between the solvent concentration gradient distribution along the film thickness direction and the solvent concentration plane distribution along the film surface direction is extracted, and the ratio of the evaporation rate constant is used as the characteristic value of the spatiotemporal distribution matrix of residual solvent anisotropy.

3. The intelligent control method for the dimming film production process according to claim 2, characterized in that, Extract the gel point time difference between the rheological modulus change curve along the film thickness direction and the rheological modulus change curve along the film surface direction, and use the gel point time difference as the characteristic vector of the critical transition of rheological anisotropy, specifically including: The rheological modulus variation curves in the thickness direction and the surface direction are smoothed and filtered to obtain the denoised rheological modulus variation curves in the thickness direction and the surface direction. Calculate the first derivative of the rheological modulus change curve in the thickness direction and the first derivative of the rheological modulus change curve in the surface direction after denoising, respectively, and mark the time when the first extreme point of the first derivative appears as the gel point time in the thickness direction and the gel point time in the surface direction, respectively. The gel point time difference is calculated by subtracting the gel point time in the film thickness direction from the gel point time in the film surface direction. The ratio of the rate of modulus change before and after the gel point is extracted from the rheological modulus change curve in the thickness direction and the rheological modulus change curve in the surface direction. The time difference at the gel point and the ratio of the rate of modulus change are fused to form a multi-dimensional rheological anisotropic critical transition feature vector.

4. The intelligent control method for the dimming film production process according to claim 3, characterized in that, The ratio of the evaporation rate constant between the solvent concentration gradient distribution along the film thickness direction and the solvent concentration plane distribution along the film surface direction is extracted, and this ratio is used as the characteristic value of the spatiotemporal distribution matrix of residual solvent anisotropy. Specifically, this includes: The planar distribution of solvent concentration in the film surface direction is divided into grids along the film width direction and the travel direction. The planar decay curve of solvent concentration in each grid cell is obtained. The planar decay curve is subjected to exponential fitting to extract the evaporation rate constant in the film surface direction of each grid cell, forming a distribution matrix of evaporation rate constant in the film surface direction. The solvent concentration gradient distribution along the film thickness direction is divided into equal layers, and the depth decay curve of the solvent concentration in each depth layer is obtained. The depth decay curve is fitted with Fick's second law, and the film thickness direction diffusion coefficient of each depth layer is extracted to form the film thickness direction diffusion coefficient distribution vector. Calculate the matrix ratio of the evaporation rate constant distribution matrix in the film surface direction to the diffusion coefficient distribution vector in the film thickness direction, and use the matrix ratio as the basic characteristic value of the spatiotemporal distribution matrix of residual solvent anisotropy; The basic eigenvalues ​​of the residual solvent anisotropic spatiotemporal distribution matrix are compared with a preset evaporation uniformity threshold. When the basic eigenvalues ​​deviate from the evaporation uniformity threshold, the basic eigenvalues ​​are normalized and corrected to generate the residual solvent anisotropic spatiotemporal distribution matrix.

5. The intelligent control method for a dimming film production process according to claim 4, characterized in that, The rheological anisotropic critical transition eigenvector and the spatiotemporal distribution matrix of the residual solvent anisotropy are input into the prediction model, and the output time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are specifically included: The rheological anisotropic critical transition feature vector and the spatiotemporal distribution matrix of residual solvent anisotropy are input into the phase separation-solidification coupling prediction model to extract spatial and temporal features and output the anisotropy index of the dual continuous phase structure and the anisotropy index of the orientation degree of microdroplets in the thickness direction. The anisotropy index of the dual continuous phase structure and the anisotropy index of droplet orientation in the thickness direction are input into the anisotropy collaborative control strategy generator. The generator calculates the deviation based on the preset ideal anisotropy target value and determines the film surface direction control amount and film thickness direction control amount according to the deviation amount. A time-varying shear stress waveform parameter set is generated based on the membrane surface direction control amount, and a negative pressure gradient sequence across the membrane thickness direction is generated based on the membrane thickness direction control amount. By configuring a preset phase difference between the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction in the time domain, the shear stress in the first segment leads, the middle segment is in phase, and the negative pressure in the last segment leads, thereby achieving decoupled control of the anisotropy of the gel network in the membrane surface direction and the membrane thickness direction.

6. The intelligent control method for a dimming film production process according to claim 5, characterized in that, The rheological anisotropic critical transition feature vector and the spatiotemporal distribution matrix of residual solvent anisotropy are input into the phase separation-solidification coupling prediction model to extract spatial and temporal features, and output the anisotropy index of the bicontinuous phase structure and the anisotropy index of droplet orientation in the thickness direction, specifically including: The rheological anisotropic critical transition feature vector is normalized and transformed into a temporal feature input vector; the residual solvent anisotropic spatiotemporal distribution matrix is ​​subjected to two-dimensional interpolation and transformed into a spatial feature input matrix. The phase separation-curing coupling prediction model adopts a dual-branch architecture of convolutional neural network-long short-term memory network. The parameters of the phase separation-curing coupling prediction model are initialized by pre-training. The pre-training is supervised learning based on the mapping relationship between different process parameters and corresponding anisotropy indices in historical production data. The spatial feature input matrix is ​​input into a branch of a convolutional neural network. The convolutional neural network branch extracts spatial features by sequentially passing through a two-dimensional convolutional layer, a batch normalization layer, and a pooling layer, and outputs a spatial feature vector. The temporal feature input vector is input into a branch of a long short-term memory network. The long short-term memory network branch extracts temporal dependency features through memory units with forget gates, input gates, and output gates, and outputs a temporal feature vector. The spatial feature vector and the temporal feature vector are concatenated and fused in the feature fusion layer to form a fused feature vector. The fused feature vector is then input into a fully connected layer, and after being mapped by an activation function, the anisotropy index of the dual continuous phase structure and the anisotropy index of the orientation degree of droplets in the thickness direction are output.

7. The intelligent control method for a dimming film production process according to claim 6, characterized in that, By configuring a preset phase difference in the time domain between the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction, with the initial shear stress leading, the middle section in phase, and the final section negative pressure leading, decoupling control of the anisotropy of the gel network in the membrane surface direction and the membrane thickness direction is achieved. Specifically, this includes: The total duration of the critical relaxation window period is obtained, and the total duration is divided into a first control phase, a second control phase, and a third control phase according to the time ratio. In the first regulation stage, the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are configured such that the phase of the shear stress waveform leads the phase of the negative pressure gradient waveform, so as to preferentially establish an anisotropic gel network skeleton in the membrane surface direction. In the second regulation stage, the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are configured such that the shear stress waveform and the negative pressure gradient waveform are in phase, so that the gel network in the membrane surface direction and the membrane thickness direction grow synchronously and collaboratively. In the third regulation stage, the time-varying shear stress waveform parameter set and the negative pressure gradient sequence across the membrane thickness direction are configured such that the phase of the negative pressure gradient waveform leads the phase of the shear stress waveform, so as to enhance the structural shaping of the gel network in the membrane thickness direction. The phase difference configuration parameters of the first, second, and third control stages are sequentially spliced ​​together to form a complete temporal phase difference configuration sequence, which is then output to the actuator to achieve decoupled control of the anisotropy of the gel network in the membrane surface direction and membrane thickness direction.

8. The intelligent control method for a dimming film production process according to claim 7, characterized in that, Within the critical relaxation window, based on the time-varying shear stress waveform parameter set and the negative pressure gradient sequence in the transmembrane thickness direction, an in-plane shear stress field is applied through an array-type airflow shearing device, and a transmembrane negative pressure suction force is applied through a porous negative pressure suction cup array, forming a gel network structure with a preset degree of anisotropy, specifically including: During the critical relaxation window period before the coated wet film enters the curing phase lock, the time-varying shear stress waveform parameter set is input into an array-type airflow shearing device set above the coated wet film. The array-type airflow shearing device applies an in-plane shear stress field with time-varying characteristics to the surface of the coated wet film according to the time-varying shear stress waveform parameter set. The negative pressure gradient sequence across the membrane thickness direction is input into a porous negative pressure suction cup array, and a negative pressure suction force that dynamically changes along the membrane thickness direction is applied according to the sequence. The in-plane shear stress field and the transmembrane negative pressure suction force form a spatiotemporal synergy within the critical relaxation window period, jointly driving the solvent to complete directional mass transfer in the membrane surface and thickness directions according to a preset anisotropic evaporation path, and inducing the polymer matrix to form a gel network structure with a preset degree of anisotropy in three-dimensional space. The anisotropy index of the solidified phase-locked bicontinuous phase structure and the anisotropy index of droplet orientation in the thickness direction are monitored in real time using a multi-channel dielectric spectroscopy analyzer. The monitoring results are fed back to update the rheological anisotropy critical transition eigenvector and the spatiotemporal distribution matrix of residual solvent anisotropy.

9. The intelligent control method for a dimming film production process according to claim 8, characterized in that, The negative pressure gradient sequence across the membrane thickness direction is input into a porous negative pressure suction cup array, and a dynamically changing negative pressure suction force along the membrane thickness direction is applied according to the sequence, specifically including: The porous negative pressure suction cup array is divided into N independently controlled negative pressure regions along the width direction of the film. Each negative pressure region corresponds to a control unit along the width direction of the film. The pore density of the suction holes in each negative pressure region is distributed in a gradient increasing or gradient decreasing manner along the width direction of the film. Obtain the regional negative pressure time series curve corresponding to each negative pressure region in the negative pressure gradient sequence across the membrane thickness direction, and input the regional negative pressure time series curve into the pressure controller of the corresponding negative pressure region; The pressure controller of each negative pressure region adjusts the negative pressure suction force amplitude of each suction hole in the corresponding negative pressure region according to the negative pressure time sequence curve of the region, so that the negative pressure suction force generated by different suction holes in the same negative pressure region forms a pressure gradient distribution that matches the suction hole density gradient along the film width direction. The actual negative pressure suction force value of each negative pressure area is collected in real time by a pressure sensor array set below the multi-hole negative pressure suction cup array. The actual negative pressure suction force value is compared with the target negative pressure value in the negative pressure time series curve of the area to calculate the pressure deviation value. The pressure deviation value is input into the PID controller, and the PID controller outputs a correction signal to adjust the output power of the negative pressure source in each negative pressure zone, so that the actual negative pressure suction force value approaches the target negative pressure value, thus forming a closed-loop pressure control for each negative pressure zone.

10. An intelligent control system for a dimming film production process, characterized in that, A smart control method for implementing the dimming film production process according to any one of claims 1-9 includes: An anisotropic sensing module is used to acquire online rheological response data and solvent concentration distribution data of the coated wet film in the film thickness direction and film surface direction, identify the rheological anisotropic critical transition feature vector, and construct the spatiotemporal distribution matrix of residual solvent anisotropy; An anisotropic pre-programming module is used to input the rheological anisotropic critical transition feature vector and the residual solvent anisotropic spatiotemporal distribution matrix into the phase separation-curing coupling prediction model, and output the anisotropic index of the dual continuous phase structure and the anisotropic index of the microdroplet orientation degree in the thickness direction. Based on the two, a time-varying shear stress waveform parameter set and a negative pressure gradient sequence in the transmembrane thickness direction are generated. An anisotropic collaborative execution module is used to apply an in-plane shear stress field according to the time-varying shear stress waveform parameter set during the critical relaxation window period, and at the same time apply a transmembrane negative pressure suction force according to the transmembrane thickness direction negative pressure gradient sequence to form a gel network structure with a preset anisotropy.

Images