Method and system for precise adjustment of filter membrane coating thickness based on adaptive control

Abstract

The application provides a filter membrane coating thickness accurate regulation preparation method and system based on adaptive control, relates to the technical field of adaptive control, and comprises the following steps: acquiring multi-point coating thickness measurement values, scraper vibration spectrum data and coating liquid temperature distribution data of a filter membrane base material; generating a global continuous coating thickness distribution field through spatial interpolation reconstruction; calculating a second derivative by using a gradient operator to identify thickness mutation positions and amplitudes; performing frequency domain decomposition on the scraper vibration spectrum to extract dominant frequency components, establishing a mapping relationship between the dominant frequency components and the thickness mutation positions to determine abnormal scraper contact regions; calculating a viscosity correction coefficient in combination with the coating liquid temperature distribution to generate a regional thickness deviation compensation amount; designing an adaptive control architecture combining model predictive control and incremental PID; solving an optimal sequence of scraper control parameters through rolling optimization; dynamically adjusting PID gain coefficients according to integral and differential items of the thickness deviation compensation amount; and outputting scraper angle, gap and coating liquid supply rate regulation amounts.

Description

Technical Field

[0001] This invention relates to the field of adaptive control technology, and in particular to a method and system for precisely adjusting the thickness of filter membrane coatings based on adaptive control. Background Technology

[0002] The filter membrane coating process is a core step in the production of membrane separation materials. The uniformity and accuracy of the coating thickness directly affect the pore size distribution, flux characteristics, and service life of the filter membrane. Traditional filter membrane coating methods mainly employ blade coating technology, adjusting the coating thickness by controlling the gap between the blade and the substrate, the coating speed, and the supply of coating liquid. In actual production, operators typically adjust process parameters manually based on offline sampling and testing results, or use a simple feedback control system to maintain the set thickness. This method can meet the basic quality requirements of conventional products and is widely used in the low-to-mid-range filter membrane manufacturing field. Some advanced production lines have introduced online thickness gauges for real-time monitoring, coupled with fixed-gain PID controllers to automatically adjust the blade gap, improving production stability compared to purely manual operation.

[0003] Existing coating preparation methods have significant shortcomings when dealing with complex working conditions. On the one hand, traditional control strategies can only obtain thickness information from a limited number of measurement points, failing to accurately grasp the spatial distribution characteristics of the coating thickness across the entire area. When there are local unevennesses on the substrate surface or fluctuations in the rheological properties of the coating solution, it is difficult to identify the specific location and cause of thickness anomalies, resulting in a lack of targeted control response. On the other hand, the doctor blade vibrates during high-speed operation, and the resulting changes in contact state cause periodic fluctuations in coating thickness. However, existing methods do not incorporate vibration characteristics into control decisions, relying solely on a single thickness deviation signal for adjustment, thus limiting control accuracy. Furthermore, the viscosity of the coating solution significantly affects the coating leveling process, and temperature fluctuations alter the flow characteristics of the coating solution. Controllers with fixed parameters cannot adaptively adjust control strategies according to changes in process conditions, exhibiting poor adaptability when handling multi-variety, small-batch production tasks. This easily leads to problems such as coating thickness deviations and large batch-to-batch quality fluctuations, hindering the stable manufacturing of high-performance filter membrane products. Summary of the Invention

[0004] The present invention provides a method and system for precisely adjusting the thickness of filter membrane coatings based on adaptive control, which can solve the problems in the prior art.

[0005] A first aspect of the present invention provides a method for precisely adjusting the thickness of a filter membrane coating based on adaptive control, comprising: The system acquires multi-point coating thickness measurements of the filter membrane substrate, doctor blade vibration spectrum data of the coating equipment, and coating liquid temperature distribution data; it then performs spatial interpolation reconstruction on the multi-point coating thickness measurements to generate a global continuous coating thickness distribution field, and calculates the second derivatives of the coating thickness distribution field in the conveying direction and transverse direction using a gradient operator to identify the locations and magnitudes of thickness abrupt changes. Frequency domain decomposition is performed based on the scraper vibration spectrum data to extract the dominant frequency components related to the uneven contact of the scraper, and a mapping relationship between the dominant frequency components and the thickness abrupt change location is established to determine the abnormal contact area of ​​the scraper. Based on the temperature distribution data of the coating liquid, the viscosity correction coefficient of the coating liquid in the abnormal contact area of ​​the scraper is calculated, and the thickness deviation compensation amount of each region is generated according to the viscosity correction coefficient and the abrupt change amplitude. The design incorporates an adaptive control architecture combining model predictive control and incremental PID control. By using rolling optimization to find the optimal sequence of scraper control parameters in the future prediction time domain, the PID gain coefficient is dynamically adjusted based on the integral and derivative terms of the thickness deviation compensation. The adaptive control architecture outputs scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

[0006] Spatial interpolation reconstruction is performed on multi-point coating thickness measurements to generate a globally continuous coating thickness distribution field. The second derivatives of this coating thickness distribution field in the transport direction and transverse direction are calculated using a gradient operator to identify the locations and magnitudes of thickness abrupt changes, including: The multi-point coating thickness measurement values ​​are mapped in a grid according to their spatial coordinates on the filter membrane substrate to establish the correspondence between discrete measurement points and spatial positions. Based on the radial basis function, continuous interpolation is performed on the thickness values ​​between adjacent measurement points to generate a coating thickness distribution field covering the entire filter membrane substrate. A first-order partial derivative operator is applied to the coating thickness distribution field along the conveying direction to obtain the thickness change rate field in the conveying direction. Then, a first-order partial derivative operator is applied to the thickness change rate field in the conveying direction to obtain the second-order derivative field in the conveying direction. A first-order partial derivative operator is applied to the coating thickness distribution field along the transverse direction to obtain the transverse thickness change rate field. Then, a first-order partial derivative operator is applied to the transverse thickness change rate field to obtain the transverse second-order derivative field. Spatial locations in the second derivative field of the conveying direction and the second derivative field of the transverse direction whose absolute values ​​exceed a preset curvature threshold are extracted as thickness abrupt change locations, and the difference between the thickness value at the thickness abrupt change location and the average thickness of its neighborhood is calculated as the abrupt change amplitude.

[0007] Extract the dominant frequency components related to uneven scraper contact, establish a mapping relationship between the dominant frequency components and the locations of abrupt thickness changes, and determine the abnormal scraper contact areas, including: The vibration spectrum data of the scraper is subjected to a fast Fourier transform to convert the time-domain vibration signal into a frequency-domain amplitude spectrum; In the frequency domain amplitude spectrum, identify frequency components whose amplitude peaks exceed the background noise level, and extract the frequency and amplitude values ​​corresponding to the frequency components. Based on the relative motion speed between the scraper and the filter membrane substrate and the periodicity of the scraper structure, a spatial mapping function between the frequency value and the contact position of the scraper is established. The frequency value is converted into the contact position coordinates of the scraper on the filter membrane substrate through the spatial mapping function, thus obtaining the spatial distribution of uneven scraper contact. Calculate the spatial correlation coefficient between the uneven spatial distribution of the scraper contact and the location of the thickness abrupt change, and screen out the scraper contact locations with correlation coefficients exceeding a preset correlation threshold as scraper contact abnormal areas. The scraper contact abnormal areas correspond to areas where there is uneven distribution of contact force or gap fluctuation between the scraper and the filter membrane substrate.

[0008] Based on the relative motion speed between the doctor blade and the filter membrane substrate and the periodicity of the doctor blade structure, a spatial mapping function between the frequency value and the doctor blade contact position is established, including: Obtain the structural parameters of the scraper, including the spacing of the periodic micro-convex structures on the scraper blade edge; Based on the conveying speed of the filter membrane substrate and the relative motion relationship between the scraper and the filter membrane substrate, calculate the displacement of the filter membrane substrate through the scraper blade per unit time. The characteristic frequency caused by the periodic structure of the scraper blade is obtained by calculating the ratio of the spacing of the periodic micro-convex structure to the displacement. A linear mapping relationship between frequency values ​​and the spatial position of the filter membrane substrate is established, and the frequency values ​​in the frequency domain amplitude spectrum are converted into the position coordinates of the filter membrane substrate in the conveying direction. For each frequency component in the frequency domain amplitude spectrum, the corresponding scraper contact position coordinates are determined according to the linear mapping relationship, forming a spatial mapping function from frequency domain characteristics to spatial position. The spatial mapping function directly associates the abnormal frequency components in the scraper vibration spectrum with the specific contact area on the filter membrane substrate.

[0009] Based on the coating temperature distribution data, a viscosity correction factor for the coating in the abnormal contact area with the scraper is calculated. Then, based on the viscosity correction factor and the abrupt change amplitude, a thickness deviation compensation amount for each region is generated, including: Extract the temperature measurement values ​​corresponding to the spatial location of the abnormal contact area of ​​the scraper from the temperature distribution data of the coating liquid, and establish a local temperature field in the abnormal contact area of ​​the scraper; according to the exponential decay characteristics of the coating liquid material, perform an exponential function operation on the difference between the temperature measurement value in the local temperature field and the reference temperature, and then multiply it by the temperature sensitivity coefficient to obtain the viscosity change factor caused by temperature; Obtain the standard viscosity value of the coating liquid under the standard temperature conditions of the coating process, multiply the standard viscosity value by the viscosity change factor, and calculate the actual viscosity value of the coating liquid at each spatial position in the abnormal contact area of ​​the scraper. The viscosity of the coating liquid is compared with the standard viscosity value to obtain the viscosity correction coefficient for each spatial position within the abnormal contact area of ​​the scraper. For each thickness change position identified within the abnormal contact area of ​​the scraper, the change amplitude and viscosity correction coefficient corresponding to the thickness change position are extracted. The change amplitude and the viscosity correction coefficient are multiplied to obtain the thickness deviation value after temperature compensation at the thickness change position. Based on the spatial boundary of the abnormal contact area of ​​the scraper, the thickness deviation values ​​after temperature compensation at all locations of thickness abrupt change in the area are weighted and averaged to generate the regional thickness deviation compensation amount for the abnormal contact area of ​​the scraper.

[0010] An adaptive control architecture based on a combination of model predictive control and incremental PID is designed. This architecture solves for the optimal sequence of scraper control parameters in the future prediction time domain through rolling optimization. Simultaneously, the PID gain coefficient is dynamically adjusted based on the integral and derivative terms of the thickness deviation compensation. The adaptive control architecture outputs scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment, including: A state-space model is established to describe the dynamic influence of the scraper angle, scraper gap and coating liquid supply rate on the coating thickness. Based on the state space model, a prediction time domain and a control time domain are set, and the coating thickness evolution trajectory at multiple future moments is predicted within the prediction time domain. A target function containing a thickness tracking error term and a control input change rate penalty term is constructed, and the optimal sequence of scraper control parameters that minimizes the target function is solved in the control time domain by a quadratic programming algorithm. The first control cycle value of the optimal sequence of scraper control parameters is extracted as the model predictive control output; The difference between the current value and the historical value of the thickness deviation compensation is calculated as the differential term, and the cumulative sum of the thickness deviation compensation is calculated as the integral term. Based on the magnitudes of the integral and derivative terms, the proportional gain coefficient, integral gain coefficient, and derivative gain coefficient of the incremental PID are dynamically updated using gain scheduling rules. Calculate the incremental control output of the incremental PID using the updated gain coefficient; The model predictive control output and the control increment output are weighted and fused to generate the final scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

[0011] A target function is constructed that includes a thickness tracking error term and a control input change rate penalty term. The optimal sequence of scraper control parameters that minimizes the target function is obtained by solving a quadratic programming algorithm within the control time domain. This sequence includes: Calculate the deviation between the predicted coating thickness value and the target thickness set value at each prediction time in the prediction time domain, square the deviation and multiply it by the thickness tracking weight coefficient to obtain the thickness tracking error term; Calculate the differences in the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment between adjacent control cycles in the control time domain. After squaring the differences, multiply them by the corresponding control input change rate weighting coefficients to obtain the control input change rate penalty term. The thickness tracking error term and the control input change rate penalty term are summed to construct a quadratic objective function; Based on the physical constraints of the scraper control parameters, an inequality constraint set is established, which includes the adjustable range constraint of the scraper angle, the mechanical limit constraint of the scraper gap, and the flow boundary constraint of the coating liquid supply rate. Transform the quadratic objective function and the set of inequality constraints into a matrix representation of a standard quadratic programming problem; By solving the optimization conditions of the standard quadratic programming problem, the sequence of scraper control parameters that minimizes the quadratic objective function is obtained. Extract the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment corresponding to each control cycle within the control time domain from the scraper control parameter sequence to form the optimal scraper control parameter sequence.

[0012] A second aspect of the present invention provides a filter membrane coating thickness precision adjustment preparation system based on adaptive control, comprising: The data acquisition unit is used to acquire multi-point coating thickness measurements of the filter membrane substrate, blade vibration spectrum data of the coating equipment, and coating liquid temperature distribution data. The thickness analysis unit is used to perform spatial interpolation reconstruction on the multi-point coating thickness measurement values ​​to generate a global continuous coating thickness distribution field, and to calculate the second derivative of the coating thickness distribution field in the conveying direction and the transverse direction through the gradient operator to identify the location and magnitude of thickness abrupt changes. The frequency domain mapping unit is used to perform frequency domain decomposition based on the scraper vibration spectrum data, extract the dominant frequency component related to the uneven contact of the scraper, establish the mapping relationship between the dominant frequency component and the thickness change position, and determine the abnormal contact area of ​​the scraper. The viscosity compensation unit is used to calculate the viscosity correction coefficient of the coating liquid in the abnormal contact area of ​​the scraper based on the coating liquid temperature distribution data, and to generate the thickness deviation compensation amount of each region based on the viscosity correction coefficient and the abrupt change amplitude. An adaptive control unit is used to design an adaptive control architecture based on a combination of model predictive control and incremental PID. It solves for the optimal sequence of scraper control parameters in the future prediction time domain through rolling optimization. At the same time, it dynamically adjusts the PID gain coefficient according to the integral and derivative terms of the thickness deviation compensation. Based on the adaptive control architecture, it outputs the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

[0013] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0014] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0015] The beneficial effects of this invention are as follows: By establishing a mapping relationship between the vibration spectrum data of the scraper and the location of thickness abrupt changes, a closed-loop diagnostic mechanism that traces the result back to the cause is realized. This mechanism can simultaneously identify specific areas of uneven scraper contact during the coating formation process. Compared with post-processing detection, it can intervene in the control process in advance, effectively shortening the response time of quality anomalies and reducing the scrap rate.

[0016] By introducing dynamic correction of viscosity based on coating temperature distribution, the problem of compensation deviation caused by neglecting the non-uniformity of temperature field in traditional methods is solved. Especially when the temperature gradient can reach 5-10℃ during large-area coating, viscosity correction can improve the thickness control accuracy by more than 15% and ensure the uniformity of coating in different temperature areas.

[0017] An adaptive architecture combining model predictive control and incremental PID is adopted. Multi-step prediction and constraint handling are achieved through rolling optimization. At the same time, the PID gain is dynamically adjusted according to the deviation characteristics. It combines the speed of feedforward compensation with the robustness of feedback regulation. When dealing with time-varying disturbances and parameter drift, the overshoot can be controlled within 2%, and the steady-state error is less than 0.5 micrometers. Attached Figure Description

[0018] Figure 1 This is a schematic flowchart of the preparation method for precisely adjusting the thickness of the filter membrane coating based on adaptive control, according to an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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.

[0020] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0021] Figure 1 This is a schematic flowchart of the preparation method for precisely adjusting the thickness of a filter membrane coating based on adaptive control, as described in an embodiment of the present invention. Figure 1 As shown, the method includes: Acquire multi-point coating thickness measurements of the filter membrane substrate, vibration spectrum data of the doctor blade of the coating equipment, and temperature distribution data of the coating liquid; Spatial interpolation is performed on the multi-point coating thickness measurement values ​​to reconstruct a global continuous coating thickness distribution field. The second derivative of the coating thickness distribution field in the conveying direction and the transverse direction is calculated by the gradient operator to identify the location and magnitude of thickness abrupt changes. Frequency domain decomposition is performed based on the scraper vibration spectrum data to extract the dominant frequency components related to the uneven contact of the scraper, and a mapping relationship between the dominant frequency components and the thickness abrupt change location is established to determine the abnormal contact area of ​​the scraper. Based on the temperature distribution data of the coating liquid, the viscosity correction coefficient of the coating liquid in the abnormal contact area of ​​the scraper is calculated, and the thickness deviation compensation amount of each region is generated according to the viscosity correction coefficient and the abrupt change amplitude. The design incorporates an adaptive control architecture combining model predictive control and incremental PID control. By using rolling optimization to find the optimal sequence of scraper control parameters in the future prediction time domain, the PID gain coefficient is dynamically adjusted based on the integral and derivative terms of the thickness deviation compensation. The adaptive control architecture outputs scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

[0022] In one optional implementation, spatial interpolation reconstruction is performed on the multi-point coating thickness measurements to generate a globally continuous coating thickness distribution field. The second derivatives of the coating thickness distribution field in the transport direction and lateral direction are calculated using a gradient operator to identify the locations and magnitudes of thickness abrupt changes, including: The multi-point coating thickness measurement values ​​are mapped in a grid according to their spatial coordinates on the filter membrane substrate to establish the correspondence between discrete measurement points and spatial positions. Based on the radial basis function, continuous interpolation is performed on the thickness values ​​between adjacent measurement points to generate a coating thickness distribution field covering the entire filter membrane substrate. A first-order partial derivative operator is applied to the coating thickness distribution field along the conveying direction to obtain the thickness change rate field in the conveying direction. Then, a first-order partial derivative operator is applied to the thickness change rate field in the conveying direction to obtain the second-order derivative field in the conveying direction. A first-order partial derivative operator is applied to the coating thickness distribution field along the transverse direction to obtain the transverse thickness change rate field. Then, a first-order partial derivative operator is applied to the transverse thickness change rate field to obtain the transverse second-order derivative field. Spatial locations in the second derivative field of the conveying direction and the second derivative field of the transverse direction whose absolute values ​​exceed a preset curvature threshold are extracted as thickness abrupt change locations, and the difference between the thickness value at the thickness abrupt change location and the average thickness of its neighborhood is calculated as the abrupt change amplitude.

[0023] In practice, a laser thickness sensor array installed on the coating production line collects multi-point coating thickness measurements. The sensor array is arranged laterally with a measurement interval of 20-50 mm, and the sampling frequency along the conveying direction is 10-100 Hz. The collected thickness data is correlated with the two-dimensional coordinates (x, y) on the filter membrane substrate, where x represents the position coordinate in the conveying direction and y represents the lateral position coordinate. The measurement area is divided into a uniform grid, with the grid size determined based on the sensor spacing, typically set to 10-30 mm. The coordinates of the actual measurement points are mapped to the nearest grid node, establishing a correspondence matrix between the discrete measurement point set and the grid nodes.

[0024] A Gaussian radial basis function is chosen as the interpolation kernel, expressed as a negative exponential function of the squared distance. The smoothness of the interpolation is controlled by adjusting the shape parameter, which typically ranges from 0.5 to 2.0. For nodes in the mesh that are not directly measured, the Euclidean distance from that location to all known measurement points is calculated. This distance is then substituted into the radial basis function to calculate the weighting coefficients, and the thickness interpolation result for that location is obtained through weighted summation. Considering boundary effects, boundary constraints are applied to areas less than 50 mm from the filter membrane edge to ensure the physical validity of the interpolation results. After interpolation, a two-dimensional array containing the thickness values ​​of all mesh nodes is generated, which represents the globally continuous coating thickness distribution field.

[0025] The first-order partial derivative is calculated using a central difference scheme. For the conveying direction, the thickness values ​​of two adjacent grid points before and after the target node are selected. The thickness of the preceding node is subtracted from the thickness of the following node, and then divided by twice the grid spacing to obtain the thickness change rate in the conveying direction at that node. This operation is performed on all grid nodes to form a thickness change rate field in the conveying direction. Based on this, the same central difference operator is applied again to the thickness change rate field to obtain the second-order derivative field in the conveying direction. The physical meaning of this second-order derivative field is the curvature of the thickness curve, reflecting the acceleration characteristics of the thickness change.

[0026] Following the same principle, the lateral data is processed. For each grid node, the thickness values ​​of its two adjacent nodes in the lateral direction are extracted. The central difference formula is then applied to calculate the first-order partial derivative in the lateral direction, generating a lateral thickness change rate field. The central difference operator is then applied to this rate field to obtain the second-order derivative field in the lateral direction. For boundary nodes, a one-sided difference scheme is used instead of the central difference to avoid computational overflow.

[0027] A curvature threshold of 0.05-0.2 micrometers per square millimeter is set, determined based on the statistical characteristics of normal coating thickness fluctuations. The second derivative fields in the conveying direction and transverse direction are traversed, and the coordinates of grid nodes with absolute second derivative values ​​greater than the curvature threshold are marked. These locations correspond to regions of abrupt changes in coating thickness. For each marked location, the thickness values ​​of neighboring nodes within a radius of 3-5 grid cells are extracted, and the arithmetic mean of the neighboring thicknesses is calculated. The difference between the actual thickness value at the abrupt change location and the mean neighboring thickness is the abrupt change amplitude; a positive value indicates a bulge, and a negative value indicates a depression. The absolute value of the abrupt change amplitude is typically in the range of 2-20 micrometers. Abrupt changes exceeding 20 micrometers are marked as severe defects, requiring immediate triggering of equipment maintenance procedures.

[0028] In one optional implementation, the dominant frequency component related to uneven scraper contact is extracted, and a mapping relationship is established between the dominant frequency component and the location of the thickness abrupt change to determine the abnormal scraper contact area, including: The vibration spectrum data of the scraper is subjected to a fast Fourier transform to convert the time-domain vibration signal into a frequency-domain amplitude spectrum; In the frequency domain amplitude spectrum, identify frequency components whose amplitude peaks exceed the background noise level, and extract the frequency and amplitude values ​​corresponding to the frequency components. Based on the relative motion speed between the scraper and the filter membrane substrate and the periodicity of the scraper structure, a spatial mapping function between the frequency value and the contact position of the scraper is established. The frequency value is converted into the contact position coordinates of the scraper on the filter membrane substrate through the spatial mapping function, thus obtaining the spatial distribution of uneven scraper contact. Calculate the spatial correlation coefficient between the uneven spatial distribution of the scraper contact and the location of the thickness abrupt change, and screen out the scraper contact locations with correlation coefficients exceeding a preset correlation threshold as scraper contact abnormal areas. The scraper contact abnormal areas correspond to areas where there is uneven distribution of contact force or gap fluctuation between the scraper and the filter membrane substrate.

[0029] A Fast Fourier Transform (FFT) was performed on the scraper vibration spectrum data to convert the acquired time-domain vibration acceleration signal x(t) into a frequency-domain amplitude spectrum X(f). The sampling frequency was set to 2048 Hz, and the sampling duration was 10 seconds to ensure a frequency resolution of 0.1 Hz. The transformed frequency-domain amplitude distribution was obtained in the range of 0 to 1024 Hz. The low-frequency band (0 to 50 Hz) mainly reflects the overall vibration of the scraper, the mid-frequency band (50 to 300 Hz) reflects the local contact state of the scraper, and the high-frequency band above 300 Hz represents the inherent noise of the equipment.

[0030] When identifying amplitude peaks in the frequency domain amplitude spectrum, the background noise level of the spectrum is first calculated. The entire spectrum is then divided into segments with a 10Hz bandwidth, and the median amplitude of each segment is calculated. The average of all medians is taken as the background noise benchmark. An identification threshold is set to three times the background noise benchmark, and the spectrum is scanned to extract all frequency components exceeding this threshold. The center frequency f_i, peak amplitude A_i, and full width at half maximum (FWHM) of each peak frequency are recorded. The half-width at half-maximum (WHM) reflects the degree of energy concentration of the frequency component.

[0031] When establishing the spatial mapping function between the frequency value and the contact position of the scraper, the motion velocity v of the scraper relative to the filter membrane substrate and the structural periodicity parameter of the scraper are introduced. If the scraper adopts a multi-segment structure, the length of each segment is... Then the contact position The mapping relationship with frequency f is as follows For cases where the scraper surface has periodic microstructures, when the structural period is p, the contact position corresponding to the frequency f_k is... , where n is a multiple of the structural period. Frequency-position mapping parameters of the scraper at different operating speeds were obtained through calibration experiments, and a complete spatial mapping function library was established.

[0032] The extracted dominant frequencies f_i are substituted into the spatial mapping function to convert them into contact position coordinates of the scraper on the filter membrane substrate. Where x_i represents the conveying direction position and y_i represents the lateral position. Since there are multiple contact points in the lateral direction of the scraper, the lateral position is inferred based on the relative magnitude of the amplitude value A_i. The frequency component with the largest amplitude corresponds to the lateral center position, and the component with the second largest amplitude corresponds to the edge position. The discrete contact position coordinates are extended into a continuous spatial distribution field D(x, y) through an interpolation algorithm. The numerical magnitude of this distribution field characterizes the severity of contact non-uniformity.

[0033] Calculate the spatial correlation coefficient between the spatial distribution of uneven scraper contact D(x, y) and the locations of abrupt thickness changes. Represent the locations of abrupt thickness changes as a binary distribution function T(x, y), with a value of 1 at the abrupt change location and 0 at other locations. Calculate the local Pearson correlation coefficient within a sliding window. The window size is set to 0.2 times the scraper width. The preset association threshold is set to 0.6 to filter out... The area is designated as the abnormal contact area of ​​the scraper. The boundaries of these abnormal areas are smoothed using morphological closing operations to eliminate isolated noise points, ultimately forming several continuous abnormal area polygons. The center coordinates, area, and peak correlation coefficient of each area are recorded, providing a spatial positioning reference for subsequent viscosity correction and control parameter adjustment.

[0034] In one optional embodiment, based on the relative motion speed between the scraper and the filter membrane substrate and the periodicity of the scraper's structure, a spatial mapping function between the frequency value and the scraper contact position is established, including: Obtain the structural parameters of the scraper, including the spacing of the periodic micro-convex structures on the scraper blade edge; Based on the conveying speed of the filter membrane substrate and the relative motion relationship between the scraper and the filter membrane substrate, calculate the displacement of the filter membrane substrate through the scraper blade per unit time. The characteristic frequency caused by the periodic structure of the scraper blade is obtained by calculating the ratio of the spacing of the periodic micro-convex structure to the displacement. A linear mapping relationship between frequency values ​​and the spatial position of the filter membrane substrate is established, and the frequency values ​​in the frequency domain amplitude spectrum are converted into the position coordinates of the filter membrane substrate in the conveying direction. For each frequency component in the frequency domain amplitude spectrum, the corresponding scraper contact position coordinates are determined according to the linear mapping relationship, forming a spatial mapping function from frequency domain characteristics to spatial position. The spatial mapping function directly associates the abnormal frequency components in the scraper vibration spectrum with the specific contact area on the filter membrane substrate.

[0035] When obtaining the structural parameters of the scraper, a laser confocal microscope is used to perform a three-dimensional morphological scan of the scraper's cutting edge, extracting the distribution characteristics of the micro-protrusion structure on the cutting edge surface. The spacing of the periodic micro-protrusion structures is determined by performing a Fourier transform on the scanned profile, identifying the regular texture period caused by the cutting edge machining process. This spacing is typically in the range of 0.05 to 0.3 mm. For scrapers processed by grinding, the micro-protrusion structure originates from the grinding marks of the grinding wheel; for laser-processed scrapers, the micro-protrusion structure comes from the overlap effect of laser pulses.

[0036] When calculating the displacement of the filter membrane substrate over the scraper edge per unit time, the encoder signal of the filter membrane conveying system is directly read to obtain the real-time conveying speed. The relative speed between the filter membrane substrate and the scraper is the conveying speed, because the scraper maintains a fixed spatial position during operation. Taking a typical operating condition with a conveying speed of 5 meters per second as an example, the displacement of the filter membrane substrate relative to the scraper edge is 5 meters within a 1-second time window.

[0037] When performing the ratio calculation, the filter membrane delivery speed is divided by the spacing of the periodic micro-protrusions to obtain the characteristic frequency caused by the periodic structure of the scraper blade. Assuming the micro-protrusion spacing is 0.1 mm and the delivery speed is 5 m / s, the characteristic frequency is calculated as 5000 mm / s divided by 0.1 mm, which equals 50000 Hz. This characteristic frequency corresponds to the inherent excitation frequency in the scraper vibration spectrum generated by the interaction between the blade microstructure and the coating liquid.

[0038] When establishing a linear mapping relationship between frequency values ​​and the spatial position of the filter membrane substrate, the conveying speed is used as the conversion coefficient. For a component with frequency f in the frequency domain amplitude spectrum, its corresponding spatial wavelength is the conveying speed v divided by the frequency f. This spatial wavelength is used as the position resolution on the filter membrane substrate. Combined with the absolute position of the filter membrane substrate at the moment of vibration signal acquisition, the coordinates of the scraper contact position corresponding to this frequency component are calculated. In specific operation, the position of the filter membrane front end at the start of signal acquisition is taken as the origin of the coordinate system, and the position coordinate x at any time t is determined.

[0039] For each frequency component in the frequency domain amplitude spectrum, coordinate transformation is performed one by one. Frequency components with amplitudes exceeding a preset threshold are identified as anomalous frequencies. The spatial location of the anomalous frequency components is directly calculated using a linear mapping relationship. For example, when an anomalous frequency component of 800 Hz is detected, under the condition of a conveying speed of 5 meters per second, its corresponding spatial wavelength is 6.25 millimeters, indicating that a scraper contact anomaly occurs every 6.25 millimeters on the filter membrane substrate.

[0040] The resulting spatial mapping function transforms the frequency axis of the vibration spectrum into the spatial position axis of the filter membrane substrate. This function achieves batch conversion through matrix operations, taking as input a discrete frequency point array of the frequency domain amplitude spectrum and outputting as a position coordinate array on the filter membrane substrate. The function incorporates a real-time update mechanism for the filter membrane conveying speed; when the conveying speed fluctuates, the mapping coefficients are automatically recalculated to ensure the accurate correspondence between frequency components and spatial positions. Through this spatial mapping function, abnormal frequency components in the scraper vibration spectrum are directly marked on the specific contact area of ​​the filter membrane substrate, providing precise spatial positioning data for subsequent thickness deviation compensation.

[0041] In one optional implementation, the viscosity correction coefficient of the coating liquid in the abnormal contact area of ​​the scraper is calculated based on the coating liquid temperature distribution data, and a thickness deviation compensation amount for each region is generated based on the viscosity correction coefficient and the abrupt change amplitude, including: Extract the temperature measurement values ​​corresponding to the spatial location of the abnormal contact area of ​​the scraper from the temperature distribution data of the coating liquid, and establish the local temperature field within the abnormal contact area of ​​the scraper. Based on the exponential decay characteristics of the coating material, the difference between the measured temperature value in the local temperature field and the reference temperature is calculated using an exponential function, and then multiplied by the temperature sensitivity coefficient to obtain the viscosity change factor caused by temperature. Obtain the standard viscosity value of the coating liquid under the standard temperature conditions of the coating process, multiply the standard viscosity value by the viscosity change factor, and calculate the actual viscosity value of the coating liquid at each spatial position in the abnormal contact area of ​​the scraper. The viscosity correction coefficient for each spatial location within the abnormal contact area of ​​the scraper is obtained by comparing the actual viscosity value of the coating liquid with the standard viscosity value. For each thickness abrupt change location identified within the abnormal contact area of ​​the scraper, the abrupt change amplitude and viscosity correction coefficient corresponding to that thickness abrupt change location are extracted; The product of the abrupt change amplitude and the viscosity correction coefficient is used to obtain the temperature-compensated thickness deviation value at the location of the thickness abrupt change. Based on the spatial boundary of the abnormal contact area of ​​the scraper, the thickness deviation values ​​after temperature compensation at all locations of thickness abrupt change in the area are weighted and averaged to generate the regional thickness deviation compensation amount for the abnormal contact area of ​​the scraper.

[0042] After locating the abnormal doctor blade contact area, the thickness deviation compensation amount needs to be accurately calculated by combining the real-time correction of the effect of coating temperature distribution on viscosity. Coating temperature distribution data is collected through an infrared temperature sensor array positioned above the coating platform, with a spatial resolution of 5 mm × 5 mm and a sampling frequency of 10 Hz. For the identified abnormal doctor blade contact area, all temperature measurements within the spatial boundary of the abnormal area are extracted from the coating temperature distribution data to establish a local temperature field. The local temperature field is stored in discrete matrix form, with each element of the matrix corresponding to the temperature value of a spatial grid.

[0043] The viscosity of the coating solution decreases exponentially with temperature. The reference temperature for the coating process is set to 25 degrees Celsius. The temperature values ​​T_i at each measurement point in the current local temperature field are obtained, and the temperature deviation is calculated. Based on the rheological test data of the coating material, the temperature sensitivity coefficient k is determined, typically ranging from 0.02 to 0.05. The temperature deviation is then substituted into an exponential function. This yields the viscosity change factor caused by temperature. This exponential function characterizes the physical process by which the viscosity of the coating decreases as the temperature increases.

[0044] Read the standard viscosity value of the coating liquid under standard temperature conditions from the coating process parameter library. This value was measured and stored using an offline viscometer. The actual viscosity of the coating liquid at each location within the abnormal blade contact area was calculated by multiplying the standard viscosity value by the viscosity change factor at each spatial location. The actual viscosity value constitutes the actual viscosity distribution field, reflecting the influence of temperature non-uniformity on the rheological properties of the coating liquid.

[0045] Perform viscosity correction factor calculation, and convert the actual viscosity value at each spatial location. Compared with standard viscosity value Perform ratio calculations to obtain the viscosity correction factor. A viscosity correction factor less than 1 indicates that the viscosity of the coating at that location is lower than the standard value, resulting in enhanced fluidity; a factor greater than 1 indicates increased viscosity and decreased fluidity. The viscosity correction factor field is spatially registered with the location of abrupt changes in thickness.

[0046] For each abrupt change point in the set of thickness abrupt change locations, the magnitude of the abrupt change corresponding to that point is extracted using spatial coordinate indexing. and viscosity correction factor The abrupt change amplitude is the thickness jump value calculated from the second derivative of the coating thickness distribution field. The abrupt change amplitude is then multiplied by the viscosity correction factor. The thickness deviation value after temperature compensation is obtained. This calculation process quantifies the influence of temperature on the flow behavior of the coating liquid into the thickness deviation, achieving a coupled correction of the physical mechanism.

[0047] Based on the spatial boundary definition of the abnormal contact area of ​​the scraper, a set of temperature-compensated thickness deviation values ​​is collected for all thickness abrupt change locations within this area. A weighted average is then performed, with weighting coefficients determined by the absolute value of the deviation at each abrupt change location; the larger the absolute value of the deviation, the higher the weight. The weighted average formula is: the thickness deviation compensation for each region equals the sum of the products of each temperature-compensated thickness deviation value and its weighting coefficient, divided by the sum of the weighting coefficients. The generated thickness deviation compensation for each region represents the comprehensive deviation characteristics of the abnormal region and serves as feedback input for model predictive control in the adaptive control architecture, used to calculate the adjustments for scraper angle, gap, and coating supply rate.

[0048] In one optional implementation, an adaptive control architecture based on a combination of model predictive control and incremental PID is designed. The optimal sequence of scraper control parameters in the future prediction time domain is solved through rolling optimization. Simultaneously, the PID gain coefficient is dynamically adjusted based on the integral and derivative terms of the thickness deviation compensation. The adaptive control architecture outputs scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment, including: A state-space model is established to describe the dynamic influence of the scraper angle, scraper gap and coating liquid supply rate on the coating thickness. Based on the state space model, a prediction time domain and a control time domain are set, and the coating thickness evolution trajectory at multiple future moments is predicted within the prediction time domain. A target function containing a thickness tracking error term and a control input change rate penalty term is constructed, and the optimal sequence of scraper control parameters that minimizes the target function is solved in the control time domain by a quadratic programming algorithm. The first control cycle value of the optimal sequence of scraper control parameters is extracted as the model predictive control output; The difference between the current value and the historical value of the thickness deviation compensation is calculated as the differential term, and the cumulative sum of the thickness deviation compensation is calculated as the integral term. Based on the magnitudes of the integral and derivative terms, the proportional gain coefficient, integral gain coefficient, and derivative gain coefficient of the incremental PID are dynamically updated using gain scheduling rules. Calculate the incremental control output of the incremental PID using the updated gain coefficient; The model predictive control output and the control increment output are weighted and fused to generate the final scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

[0049] To address the need for controlling filter membrane coating thickness, a state-space model is established to describe the dynamic response relationship between the scraper control parameters and the coating thickness. The scraper angle is selected. Using the scraper gap d and coating supply rate v as control variables, and coating thickness h as a state variable, a discrete-time state equation is constructed. ,in The coefficient matrices A and B were obtained through offline identification. The step response test method was used to record the time series data of the coating thickness after the control variables changed under different working conditions. The model parameters were obtained by fitting the data using the least squares method.

[0050] Setting the prediction time domain in the model predictive control framework Each sampling period, control time domain Each sampling period. The coating thickness evolution trajectory in the predicted time domain is obtained through recursive calculation, expressed as: ; Where i = 1, 2, ..., N_p.

[0051] The objective function consists of two parts: a thickness tracking error penalty term. and control change rate penalty term The weighting coefficients are Q = 5.0 and R = 0.2. An effective set quadratic programming algorithm is used to solve for the optimal control sequence, while imposing constraints. , and Ensure that the implementing agency's response capabilities meet the requirements.

[0052] The model predictive control output is the first element of the optimal sequence. The incremental PID control section first calculates the difference in thickness deviation compensation as the differential term. Where e_k is the current thickness deviation compensation amount, and the cumulative sum is used as the integration term. The gain scheduling rule is dynamically adjusted based on the magnitude of the deviation: when When the proportional gain K_p = 0.3, integral gain K_i = 0.08, and differential gain K_d = 0.12 are set; when When the proportional effect is increased, K_p = 0.6, Ki = 0.15, and K_d = 0.18; when When the fast response is enhanced, K_p = 1.0, Ki = 0.25, and K_d = 0.22. The incremental calculation formula for incremental PID control is: .

[0053] The final control output is generated through weighted fusion: ; The fusion coefficient α is adaptively determined based on the confidence level of the prediction model. When the model prediction error is less than 100%, the fusion coefficient α is determined by the model confidence level. When α is set to 0.8, model predictive control results are prioritized; when the prediction error exceeds... The α value is adjusted to 0.5 to enhance the PID feedback correction effect. The output of the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment is directly sent to the actuator of the coating equipment to achieve real-time closed-loop control.

[0054] A target function is constructed that includes a thickness tracking error term and a control input change rate penalty term. The optimal sequence of scraper control parameters that minimizes the target function is obtained by solving a quadratic programming algorithm within the control time domain. This sequence includes: Calculate the deviation between the predicted coating thickness value and the target thickness set value at each prediction time in the prediction time domain, square the deviation and multiply it by the thickness tracking weight coefficient to obtain the thickness tracking error term; Calculate the differences in the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment between adjacent control cycles in the control time domain. After squaring the differences, multiply them by the corresponding control input change rate weighting coefficients to obtain the control input change rate penalty term. The thickness tracking error term and the control input change rate penalty term are summed to construct a quadratic objective function; Based on the physical constraints of the scraper control parameters, an inequality constraint set is established, which includes the adjustable range constraint of the scraper angle, the mechanical limit constraint of the scraper gap, and the flow boundary constraint of the coating liquid supply rate. Transform the quadratic objective function and the set of inequality constraints into a matrix representation of a standard quadratic programming problem; By solving the optimization conditions of the standard quadratic programming problem, the sequence of scraper control parameters that minimizes the quadratic objective function is obtained. Extract the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment corresponding to each control cycle within the control time domain from the scraper control parameter sequence to form the optimal scraper control parameter sequence.

[0055] In the coating preparation process, precise tracking and control of coating thickness is a core requirement for achieving high-quality filter membranes. The optimization solution mechanism of model predictive control needs to complete the construction of the objective function, the establishment of constraints, and the solution of the optimal control sequence within each control cycle.

[0056] For evaluating thickness tracking performance in the prediction time domain, within a prediction time domain of length N_p, the prediction times starting from the current time k are calculated. The predicted coating thickness is calculated. The predicted thickness at each prediction time is subtracted from the target thickness value set by the process, resulting in a thickness deviation sequence. Each element in this deviation sequence is squared and then multiplied by a thickness tracking weighting coefficient Q, which reflects the importance of thickness control accuracy in the objective function. The weighted squared deviations at all times within the prediction time domain are summed to form a thickness tracking error term; the smaller this term, the closer the predicted trajectory is to the target thickness curve.

[0057] To suppress drastic changes in control input, protect the scraper's mechanical structure, and prevent severe disturbances in the coating fluid flow field, a penalty needs to be applied to the rate of change of the control input. Within the control time domain N_c, the increment of the scraper angle adjustment between adjacent control cycles is calculated, i.e., the difference between the angle adjustment in the current cycle and the angle adjustment in the previous cycle. This difference is then squared and multiplied by the angle change rate weighting coefficient. Using the same processing method, the incremental square term of the scraper gap adjustment is calculated and multiplied by the gap change rate weighting coefficient R_g, and the incremental square term of the coating liquid supply rate adjustment is multiplied by the rate change rate weighting coefficient R_v. The above three penalty values ​​for all control cycles in the control time domain are summed to form the control input change rate penalty term.

[0058] A quadratic objective function is constructed by scalar summation of the thickness tracking error term and the control input change rate penalty term. This function is the sum of the quadratic forms of the error vector and the control increment vector, where the weight matrix consists of Q and... It consists of a diagonal matrix.

[0059] Based on the physical structural limitations of the coating equipment, a set of inequality constraints for the control parameters is established. The adjustable range of the scraper angle is limited by the rotational limit of the scraper support mechanism, typically between 15 and 75 degrees. The scraper gap is constrained by the mechanical limiting device; the minimum gap must be greater than the maximum particle size in the coating liquid, while the maximum gap is limited by the surface tension stability of the coating liquid. The coating liquid supply rate is constrained by the flow boundary of the metering pump; the minimum flow rate must maintain a continuous supply of coating liquid, while the maximum flow rate is limited by the pipeline pressure carrying capacity. These physical constraints are transformed into upper and lower bound constraint vectors for the control variables.

[0060] Expanding the error vector and control increment vector in the quadratic objective function into matrix form, and using the state-space representation of the prediction model, the objective function is rewritten as a standard quadratic programming problem of control variables, i.e., a combination of the quadratic coefficient matrix and the linear coefficient vector. The set of inequality constraints is then expressed as inequality relationships between the constraint matrix and the constraint vector. This transformation forms the matrix representation of the standard quadratic programming problem.

[0061] The standard quadratic programming problem is solved using the interior-point method or the effective set method. The optimal solution that minimizes the gradient of the objective function is found by iteratively searching the feasible region that satisfies the constraints. The solution process outputs a sequence of scraper control parameters in the control time domain. This sequence contains numerical combinations of angle, gap, and rate adjustments for N_c control cycles. The three adjustments for each control cycle are extracted from this sequence to form the optimal scraper control parameter sequence, which serves as the execution instruction for the current control cycle in the rolling optimization mechanism.

[0062] A second aspect of the present invention provides a filter membrane coating thickness precision adjustment preparation system based on adaptive control, comprising: The data acquisition unit is used to acquire multi-point coating thickness measurements of the filter membrane substrate, blade vibration spectrum data of the coating equipment, and coating liquid temperature distribution data. The thickness analysis unit is used to perform spatial interpolation reconstruction on the multi-point coating thickness measurement values ​​to generate a global continuous coating thickness distribution field, and to calculate the second derivative of the coating thickness distribution field in the conveying direction and the transverse direction through the gradient operator to identify the location and magnitude of thickness abrupt changes. The frequency domain mapping unit is used to perform frequency domain decomposition based on the scraper vibration spectrum data, extract the dominant frequency component related to the uneven contact of the scraper, establish the mapping relationship between the dominant frequency component and the thickness change position, and determine the abnormal contact area of ​​the scraper. The viscosity compensation unit is used to calculate the viscosity correction coefficient of the coating liquid in the abnormal contact area of ​​the scraper based on the coating liquid temperature distribution data, and to generate the thickness deviation compensation amount of each region based on the viscosity correction coefficient and the abrupt change amplitude. An adaptive control unit is used to design an adaptive control architecture based on a combination of model predictive control and incremental PID. It solves for the optimal sequence of scraper control parameters in the future prediction time domain through rolling optimization. At the same time, it dynamically adjusts the PID gain coefficient according to the integral and derivative terms of the thickness deviation compensation. Based on the adaptive control architecture, it outputs the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

[0063] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0064] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0065] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0066] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for precisely adjusting the thickness of a filter membrane coating based on adaptive control, characterized in that, include: Acquire multi-point coating thickness measurements of the filter membrane substrate, vibration spectrum data of the doctor blade of the coating equipment, and temperature distribution data of the coating liquid; Spatial interpolation is performed on the multi-point coating thickness measurement values ​​to reconstruct a global continuous coating thickness distribution field. The second derivative of the coating thickness distribution field in the conveying direction and the transverse direction is calculated by the gradient operator to identify the location and magnitude of thickness abrupt changes. Frequency domain decomposition is performed based on the scraper vibration spectrum data to extract the dominant frequency components related to the uneven contact of the scraper, and a mapping relationship between the dominant frequency components and the thickness abrupt change location is established to determine the abnormal contact area of ​​the scraper. Based on the temperature distribution data of the coating liquid, the viscosity correction coefficient of the coating liquid in the abnormal contact area of ​​the scraper is calculated, and the thickness deviation compensation amount of each region is generated according to the viscosity correction coefficient and the abrupt change amplitude. The design incorporates an adaptive control architecture combining model predictive control and incremental PID control. By using rolling optimization to find the optimal sequence of scraper control parameters in the future prediction time domain, the PID gain coefficient is dynamically adjusted based on the integral and derivative terms of the thickness deviation compensation. The adaptive control architecture outputs scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

2. The method according to claim 1, characterized in that, Spatial interpolation reconstruction is performed on multi-point coating thickness measurements to generate a globally continuous coating thickness distribution field. The second derivatives of this coating thickness distribution field in the transport direction and transverse direction are calculated using a gradient operator to identify the locations and magnitudes of thickness abrupt changes, including: The multi-point coating thickness measurement values ​​are mapped in a grid according to their spatial coordinates on the filter membrane substrate to establish the correspondence between discrete measurement points and spatial positions. Based on the radial basis function, continuous interpolation is performed on the thickness values ​​between adjacent measurement points to generate a coating thickness distribution field covering the entire filter membrane substrate. A first-order partial derivative operator is applied to the coating thickness distribution field along the conveying direction to obtain the thickness change rate field in the conveying direction. Then, a first-order partial derivative operator is applied to the thickness change rate field in the conveying direction to obtain the second-order derivative field in the conveying direction. A first-order partial derivative operator is applied to the coating thickness distribution field along the transverse direction to obtain the transverse thickness change rate field. Then, a first-order partial derivative operator is applied to the transverse thickness change rate field to obtain the transverse second-order derivative field. Spatial locations in the second derivative field of the conveying direction and the second derivative field of the transverse direction whose absolute values ​​exceed a preset curvature threshold are extracted as thickness abrupt change locations, and the difference between the thickness value at the thickness abrupt change location and the average thickness of its neighborhood is calculated as the abrupt change amplitude.

3. The method according to claim 1, characterized in that, Extract the dominant frequency components related to uneven scraper contact, establish a mapping relationship between the dominant frequency components and the locations of abrupt thickness changes, and determine the abnormal scraper contact areas, including: The vibration spectrum data of the scraper is subjected to a fast Fourier transform to convert the time-domain vibration signal into a frequency-domain amplitude spectrum. In the frequency domain amplitude spectrum, identify frequency components whose amplitude peaks exceed the background noise level, and extract the frequency and amplitude values ​​corresponding to the frequency components. Based on the relative motion speed between the scraper and the filter membrane substrate and the periodicity of the scraper structure, a spatial mapping function between the frequency value and the contact position of the scraper is established. The frequency value is converted into the contact position coordinates of the scraper on the filter membrane substrate through the spatial mapping function, thus obtaining the spatial distribution of uneven scraper contact. Calculate the spatial correlation coefficient between the uneven spatial distribution of the scraper contact and the location of the thickness abrupt change, and screen out the scraper contact locations with correlation coefficients exceeding a preset correlation threshold as scraper contact abnormal areas. The scraper contact abnormal areas correspond to areas where there is uneven distribution of contact force or gap fluctuation between the scraper and the filter membrane substrate.

4. The method according to claim 3, characterized in that, Based on the relative motion speed between the doctor blade and the filter membrane substrate and the periodicity of the doctor blade structure, a spatial mapping function between the frequency value and the doctor blade contact position is established, including: Obtain the structural parameters of the scraper, including the spacing of the periodic micro-convex structures on the scraper blade edge; Based on the conveying speed of the filter membrane substrate and the relative motion relationship between the scraper and the filter membrane substrate, calculate the displacement of the filter membrane substrate through the scraper blade per unit time. The characteristic frequency caused by the periodic structure of the scraper blade is obtained by calculating the ratio of the spacing of the periodic micro-convex structure to the displacement. A linear mapping relationship between frequency values ​​and the spatial position of the filter membrane substrate is established, and the frequency values ​​in the frequency domain amplitude spectrum are converted into the position coordinates of the filter membrane substrate in the conveying direction. For each frequency component in the frequency domain amplitude spectrum, the corresponding scraper contact position coordinates are determined according to the linear mapping relationship, forming a spatial mapping function from frequency domain characteristics to spatial position. The spatial mapping function directly associates the abnormal frequency components in the scraper vibration spectrum with the specific contact area on the filter membrane substrate.

5. The method according to claim 1, characterized in that, Based on the coating temperature distribution data, a viscosity correction factor for the coating in the abnormal contact area with the scraper is calculated. Then, based on the viscosity correction factor and the abrupt change amplitude, a thickness deviation compensation amount for each region is generated, including: Extract the temperature measurement values ​​corresponding to the spatial location of the abnormal contact area of ​​the scraper from the temperature distribution data of the coating liquid, and establish a local temperature field in the abnormal contact area of ​​the scraper; according to the exponential decay characteristics of the coating liquid material, perform an exponential function operation on the difference between the temperature measurement value in the local temperature field and the reference temperature, and then multiply it by the temperature sensitivity coefficient to obtain the viscosity change factor caused by temperature; Obtain the standard viscosity value of the coating liquid under the standard temperature conditions of the coating process, multiply the standard viscosity value by the viscosity change factor, and calculate the actual viscosity value of the coating liquid at each spatial position in the abnormal contact area of ​​the scraper. The viscosity of the coating liquid is compared with the standard viscosity to obtain the viscosity correction coefficient for each spatial position within the abnormal contact area of ​​the scraper. For each thickness change position identified within the abnormal contact area of ​​the scraper, the change amplitude and viscosity correction coefficient corresponding to the thickness change position are extracted. The change amplitude and the viscosity correction coefficient are multiplied to obtain the temperature-compensated thickness deviation value at the thickness change position. Based on the spatial boundary of the abnormal contact area of ​​the scraper, the thickness deviation values ​​after temperature compensation at all locations of thickness abrupt change in the area are weighted and averaged to generate the regional thickness deviation compensation amount for the abnormal contact area of ​​the scraper.

6. The method according to claim 1, characterized in that, An adaptive control architecture based on a combination of model predictive control and incremental PID is designed. This architecture solves for the optimal sequence of scraper control parameters in the future prediction time domain through rolling optimization. Simultaneously, the PID gain coefficient is dynamically adjusted based on the integral and derivative terms of the thickness deviation compensation. The adaptive control architecture outputs scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment, including: A state-space model is established to describe the dynamic influence of the scraper angle, scraper gap and coating liquid supply rate on the coating thickness. Based on the state space model, a prediction time domain and a control time domain are set, and the coating thickness evolution trajectory at multiple future moments is predicted within the prediction time domain. A target function containing a thickness tracking error term and a control input change rate penalty term is constructed, and the optimal sequence of scraper control parameters that minimizes the target function is solved in the control time domain by a quadratic programming algorithm. The first control cycle value of the optimal sequence of scraper control parameters is extracted as the model predictive control output; The difference between the current value and the historical value of the thickness deviation compensation is calculated as the differential term, and the cumulative sum of the thickness deviation compensation is calculated as the integral term. Based on the magnitudes of the integral and derivative terms, the proportional gain coefficient, integral gain coefficient, and derivative gain coefficient of the incremental PID are dynamically updated using gain scheduling rules. Calculate the incremental control output of the incremental PID using the updated gain coefficient; The model predictive control output and the control increment output are weighted and fused to generate the final scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

7. The method according to claim 6, characterized in that, A target function is constructed that includes a thickness tracking error term and a control input change rate penalty term. The optimal sequence of scraper control parameters that minimizes the target function is obtained by solving a quadratic programming algorithm within the control time domain. This sequence includes: Calculate the deviation between the predicted coating thickness value and the target thickness set value at each prediction time in the prediction time domain, square the deviation and multiply it by the thickness tracking weight coefficient to obtain the thickness tracking error term; Calculate the differences in the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment between adjacent control cycles in the control time domain. After squaring the differences, multiply them by the corresponding control input change rate weighting coefficients to obtain the control input change rate penalty term. The thickness tracking error term and the control input change rate penalty term are summed to construct a quadratic objective function; Based on the physical constraints of the scraper control parameters, an inequality constraint set is established, which includes the adjustable range constraint of the scraper angle, the mechanical limit constraint of the scraper gap, and the flow boundary constraint of the coating liquid supply rate. Transform the quadratic objective function and the set of inequality constraints into a matrix representation of a standard quadratic programming problem; By solving the optimization conditions of the standard quadratic programming problem, the sequence of scraper control parameters that minimizes the quadratic objective function is obtained. Extract the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment corresponding to each control cycle within the control time domain from the scraper control parameter sequence to form the optimal scraper control parameter sequence.

8. A filter membrane coating thickness precision adjustment preparation system based on adaptive control, used to implement the method as described in any one of claims 1-7, characterized in that, include: The data acquisition unit is used to acquire multi-point coating thickness measurements of the filter membrane substrate, blade vibration spectrum data of the coating equipment, and coating liquid temperature distribution data. The thickness analysis unit is used to perform spatial interpolation reconstruction on the multi-point coating thickness measurement values ​​to generate a global continuous coating thickness distribution field, and to calculate the second derivative of the coating thickness distribution field in the conveying direction and the transverse direction through the gradient operator to identify the location and magnitude of thickness abrupt changes. The frequency domain mapping unit is used to perform frequency domain decomposition based on the scraper vibration spectrum data, extract the dominant frequency component related to the uneven contact of the scraper, establish the mapping relationship between the dominant frequency component and the thickness change position, and determine the abnormal contact area of ​​the scraper. The viscosity compensation unit is used to calculate the viscosity correction coefficient of the coating liquid in the abnormal contact area of ​​the scraper based on the coating liquid temperature distribution data, and to generate the thickness deviation compensation amount of each region based on the viscosity correction coefficient and the abrupt change amplitude. An adaptive control unit is used to design an adaptive control architecture based on a combination of model predictive control and incremental PID. It solves for the optimal sequence of scraper control parameters in the future prediction time domain through rolling optimization. At the same time, it dynamically adjusts the PID gain coefficient according to the integral and derivative terms of the thickness deviation compensation. Based on the adaptive control architecture, it outputs the scraper angle adjustment, scraper gap adjustment, and coating liquid supply rate adjustment.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images