Protective film production monitoring method and system based on industrial internet

Abstract

The application discloses a protective film production monitoring method and system based on an industrial internet, relates to the technical field of industrial internet and intelligent manufacturing, and comprises the following steps: collecting process parameters, film layer thickness distribution and surface defect distribution; constructing the process parameters into a time sequence tensor, and constructing the film layer thickness distribution and the surface defect distribution into spatial tensors; and extracting space-time coupling factors by coupling tensor decomposition; reconstructing a prediction distribution based on the space-time coupling factors, identifying an abnormal work station by residual back projection, calculating a constraint change amount of the abnormal work station adjustment on a downstream work station, solving a multi-work station collaborative adjustment scheme in a process feasible region and executing the scheme; updating the space-time coupling factors according to distribution data after execution; and when a coating work station mutation is detected, substituting the mutation into the space-time coupling factors for forward propagation to predict a downstream work station distortion area, and issuing a pre-compensation instruction before distortion formation. The application realizes quality defect cross-work station tracing, multi-work station collaborative adjustment and feedforward pre-compensation.

Description

Technical Field

[0001] This invention relates to the field of industrial internet and intelligent manufacturing technology, specifically to a method and system for monitoring the production of protective films based on the industrial internet. Background Technology

[0002] Traditional protective film production processes rely on manual experience to set process parameters. When an anomaly occurs at a certain station, adjustments are often made only to that station, ignoring the process coupling effects between multiple stations such as coating, leveling, and curing, which arise from the dynamics of adhesive flow and curing. This isolated adjustment method causes parameter changes at upstream stations to trigger film thickness fluctuations and surface defect propagation at downstream stations, forming a cross-station propagation chain of quality defects, ultimately resulting in the scrapping of the entire batch of products.

[0003] Existing quality monitoring methods based on statistical process control identify anomalies by setting upper and lower limits for process parameters. However, these methods can only identify parameter deviations at a single workstation, failing to reveal the causal relationship between the temporal evolution of process parameters and the formation of the film's spatial distribution, and further unable to quantify the constraining impact of adjustments at abnormal workstations on the leveling time and curing rate of the adhesive at downstream workstations. Some literature proposes using machine learning models to establish a mapping relationship between process parameters and quality indicators; however, these black-box models lack an explanation of the physical mechanisms underlying the adhesive leveling and curing kinetics, making it difficult to support collaborative adjustment decisions across multiple workstations. Existing methods generally employ a post-event feedback adjustment model; by the time a quality anomaly is detected, defects have already formed, making it impossible to achieve feedforward pre-compensation for abrupt changes at the coating workstation, resulting in insufficient adjustment timeliness.

[0004] Industrial Internet technology provides infrastructure support for data collection and real-time control of the protective film production process. However, how to extract the spatiotemporal coupling factor that characterizes the flow-leveling curing dynamics of the adhesive and the formation mechanism of the spatial distribution of the film from massive process parameter time series data and film spatial distribution data, and how to achieve accurate identification of abnormal workstations, rapid solution of multi-workstation collaborative adjustment schemes, and feedforward pre-compensation for coating mutations based on the spatiotemporal coupling factor are still technical problems that need to be solved. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for monitoring the production of protective films based on the Industrial Internet, aiming to solve at least one of the technical problems existing in the prior art.

[0006] The technical solution of this invention is: a protective film production monitoring method based on the Industrial Internet, comprising the following steps:

[0007] Collect process parameters, film thickness distribution, and surface defect distribution of the protective film production line;

[0008] The process parameters were constructed as temporal tensors, and the film thickness distribution and surface defect distribution were constructed as spatial tensors. The spatiotemporal coupling factors characterizing the flow-leveling curing dynamics of the adhesive and the formation mechanism of the film spatial distribution were extracted by coupling tensor decomposition.

[0009] Based on the spatiotemporal coupling factor, the predicted distribution is reconstructed, and the residual between the actual distribution and the predicted distribution is back-projected onto the time tensor to identify the abnormal process time segments that generate the residual and the corresponding abnormal workstations.

[0010] The constraint changes on the leveling time and curing rate of the adhesive at downstream stations are calculated based on the spatiotemporal coupling factor. Within the process feasibility domain limited by the constraint changes, a multi-station collaborative adjustment scheme is solved and executed.

[0011] The thickness distribution and surface defect distribution of the film layer after execution are collected to construct the spatial tensor after execution. The spatiotemporal coupling factor is updated based on the difference between the spatial tensor after execution and the spatial tensor.

[0012] When a sudden change in the process parameters of the coating station is detected, the change is substituted into the updated spatiotemporal coupling factor for forward propagation. The changes are then used to predict the film thickness distortion area and surface defect expansion area caused by the change in the downstream station. Before the distortion occurs, a pre-compensation adjustment command is issued to the downstream station.

[0013] Process parameters were constructed as temporal tensors, and film thickness distribution and surface defect distribution were constructed as spatial tensors. Spatiotemporal coupling factors characterizing the adhesive leveling and curing kinetics and the film spatial distribution formation mechanism were extracted through coupled tensor decomposition.

[0014] The process parameters are expanded into a time dimension by time series and into a station dimension by station sequence. The time dimension and station dimension are combined to construct a time series tensor.

[0015] The film thickness distribution and surface defect distribution are unfolded according to the film surface position to form a spatial position dimension, and the spatial position dimension is used to construct a spatial tensor;

[0016] Establish the product association constraint between the time series tensor and the spatial tensor in the same batch. Based on the product association constraint in the same batch, perform coupled tensor decomposition on the time series tensor and the spatial tensor. Obtain the time factor matrix and the workstation factor matrix from the time series tensor, and obtain the spatial location factor matrix from the spatial tensor.

[0017] The spatiotemporal coupling factors characterizing the adhesive leveling and curing kinetics and the formation mechanism of film spatial distribution are extracted by coupling and mapping the time factor matrix, work station factor matrix, and spatial location factor matrix.

[0018] Based on the spatiotemporal coupling factor, the predicted distribution is reconstructed, and the residuals between the actual and predicted distributions are back-projected onto the time series tensor to identify anomalous process time segments and corresponding anomalous workstations that generate residuals.

[0019] The spatial tensor is mapped back to the temporal tensor space by the spatiotemporal coupling factor to generate the reconstructed temporal tensor. The predicted film thickness distribution and the predicted surface defect distribution are obtained by forward deduction based on the reconstructed temporal tensor. The predicted film thickness distribution and the predicted surface defect distribution are used as the predicted distribution.

[0020] The deviations between the actual film thickness distribution and the predicted film thickness distribution, as well as the deviations between the actual surface defect distribution and the predicted surface defect distribution, are constructed as residual distributions.

[0021] Spatial gradient analysis is performed on the residual distribution to extract the residual concentration region. The residual concentration region is then back-projected onto the time series tensor through the spatiotemporal coupling factor to obtain the residual distribution on the time series tensor.

[0022] The process parameter fluctuation period with the highest coupling strength is located from the distribution of residuals on the time tensor. The process parameter fluctuation period is identified as the abnormal process time segment that generates residuals, and the station where the abnormal process time segment is located is identified as the corresponding abnormal station.

[0023] Based on the spatiotemporal coupling factor, the constraint changes on the leveling time and curing rate of the adhesive at downstream stations are calculated by adjusting abnormal workstations. Within the process feasibility domain defined by the constraint changes, a multi-workstation collaborative adjustment scheme is solved and implemented, including:

[0024] Based on the coupling relationship between the abnormal workstation and the downstream workstation in the spatiotemporal coupling factor, the influence of the abnormal workstation adjustment on the adhesive leveling time and the curing rate of the downstream workstation through the adhesive leveling and curing dynamics are calculated. The influence on the adhesive leveling time and the curing rate of the downstream workstation are used as constraint change quantities.

[0025] The constraint change is applied to the original process constraint of the downstream station to obtain the updated process constraint of the downstream station, and the updated process constraint is used as the process feasible domain.

[0026] Within the feasible process domain, the process parameter adjustment amounts of abnormal stations and downstream stations are used as optimization variables to solve the problem, and the combination of process parameter adjustment amounts of abnormal stations and downstream stations is obtained. The combination of process parameter adjustment amounts is used as a multi-station collaborative adjustment scheme.

[0027] The multi-station collaborative adjustment plan is issued to the process equipment corresponding to the abnormal station and downstream station for execution.

[0028] The post-execution spatial tensor is constructed by collecting the film thickness distribution and surface defect distribution. The spatiotemporal coupling factor is updated based on the difference between the post-execution spatial tensor and the spatial tensor, including:

[0029] The film thickness distribution and surface defect distribution after the multi-station collaborative adjustment scheme is executed are collected, and the film thickness distribution and surface defect distribution after execution are stacked into a spatial tensor after execution.

[0030] The difference between the spatial tensors after calculation is used to obtain the actual spatial response. The expected spatial response is calculated based on the multi-station collaborative adjustment scheme and the spatiotemporal coupling factor. The deviation between the actual spatial response and the expected spatial response is used as the coupling deviation.

[0031] Spatial decomposition of coupling deviation is performed to extract the spatial components associated with abnormal workstations and the spatial components associated with downstream workstations in the coupling deviation. The coupling correction coefficients of abnormal workstations and downstream workstations are calculated respectively.

[0032] The coupling correction coefficients of the abnormal workstation and the downstream workstation are applied to the coupling strengths corresponding to the abnormal workstation and the downstream workstation in the spatiotemporal coupling factor, respectively, to obtain the updated spatiotemporal coupling factor.

[0033] When a sudden change in process parameters at the coating station is detected, the change is substituted into the updated spatiotemporal coupling factor for forward propagation. This predicts the film thickness distortion region and surface defect propagation region caused by the change at downstream stations. Before the distortion occurs, a pre-compensation adjustment instruction is issued to the downstream stations, including:

[0034] Real-time monitoring of process parameter changes at the coating station; when the change in process parameters exceeds the preset fluctuation range, it is identified as a sudden change in process parameters at the coating station, and the magnitude and timing of the sudden change in process parameters at the coating station are obtained.

[0035] The mutation amplitude is substituted into the updated spatiotemporal coupling factor for forward propagation to obtain the propagation response of the mutation in the film thickness distribution and surface defect distribution of the downstream station. The propagation time of the mutation to the downstream station is calculated based on the updated spatiotemporal coupling factor.

[0036] Spatial localization is performed on the propagation response of the film thickness distribution and the propagation response of the surface defect distribution at the downstream station. The film thickness distortion region and the surface defect expansion region caused by the sudden change at the downstream station are extracted. The pre-compensation process parameter adjustment amount of the downstream station is calculated based on the film thickness distortion region and the surface defect expansion region.

[0037] The time of distortion formation is determined based on the time of mutation and the propagation time. Before the time of distortion formation, the pre-compensation process parameter adjustment amount is issued to the corresponding process equipment of the downstream station as a pre-compensation adjustment instruction.

[0038] Spatially locate the propagation response on the film thickness distribution and the propagation response on the surface defect distribution at the downstream station, extract the film thickness distortion region and surface defect propagation region caused by the abrupt change at the downstream station, and calculate the pre-compensation process parameter adjustment amount for the downstream station based on the film thickness distortion region and the surface defect propagation region, including:

[0039] Spatial gradient analysis is performed on the propagation response of the membrane thickness distribution at the downstream workstation. Continuous regions where the spatial gradient exceeds the preset fluctuation threshold are spatially located and marked as membrane thickness distortion regions. The spatial location and distortion amplitude of the membrane thickness distortion regions are extracted.

[0040] Spatial boundary identification is performed on the propagation response of surface defects in downstream workstations. The newly added area that expands outward from the original surface defect boundary is spatially located and marked as the surface defect expansion area. The spatial location and expansion scale of the surface defect expansion area are extracted.

[0041] The spatial location and distortion amplitude of the film thickness distortion region are mapped to the downstream station process parameter adjustment space to obtain the process parameter compensation amount used to offset the distortion amplitude. The spatial location and expansion scale of the surface defect expansion region are mapped to the downstream station process parameter adjustment space to obtain the process parameter compensation amount used to suppress the expansion scale.

[0042] The pre-compensation process parameter adjustment amount for downstream stations is obtained by spatially superimposing the process parameter compensation amount used to offset the distortion amplitude and the process parameter compensation amount used to suppress the expansion scale.

[0043] This invention provides a protective film production monitoring system based on the Industrial Internet, the system comprising:

[0044] The data acquisition module is used to collect process parameters, film thickness distribution, and surface defect distribution of the protective film production line.

[0045] The coupling factor extraction module is used to construct process parameters as temporal tensors and film thickness distribution and surface defect distribution as spatial tensors. It extracts spatiotemporal coupling factors that characterize the flow curing dynamics of adhesive and the formation mechanism of film spatial distribution through coupling tensor decomposition.

[0046] The anomaly identification module is used to reconstruct the predicted distribution based on the spatiotemporal coupling factor, back-project the residual between the actual distribution and the predicted distribution onto the time tensor, and identify the abnormal process time segment that generates the residual and the corresponding abnormal work station.

[0047] The collaborative adjustment module is used to calculate the constraint changes of abnormal station adjustments on the leveling time and curing rate of adhesive in downstream stations based on the spatiotemporal coupling factor, and to solve and execute multi-station collaborative adjustment schemes within the process feasible domain limited by the constraint changes.

[0048] The factor update module is used to collect the film thickness distribution and surface defect distribution after execution to construct the spatial tensor after execution, and update the spatiotemporal coupling factor based on the difference between the spatial tensor after execution and the spatial tensor.

[0049] The feedforward compensation module is used to detect sudden changes in process parameters at the coating station, substitute the sudden change into the updated spatiotemporal coupling factor for forward propagation, predict the film thickness distortion area and surface defect expansion area caused by the sudden change at the downstream station, and issue a pre-compensation adjustment command to the downstream station before the distortion occurs.

[0050] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0051] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the steps in any of the aforementioned methods.

[0052] This invention extracts spatiotemporal coupling factors through coupled tensor decomposition, revealing the causal relationship between the temporal evolution of process parameters and the formation of film spatial distribution, providing an interpretable theoretical basis for abnormal station identification and adjustment decisions. The residual back-projection method based on spatiotemporal coupling factors enables precise cross-station tracing of quality defects, accurately locating the abnormal process time segments that generate residuals and their corresponding stations. By calculating the constraint changes in the leveling time and curing rate of the adhesive at downstream stations caused by adjustments at abnormal stations, a multi-station collaborative adjustment scheme is solved within the process feasible domain, avoiding the cross-station propagation problem of quality defects caused by traditional isolated adjustments. The spatiotemporal coupling factors are continuously updated using post-execution distribution data, enabling them to adapt to changes in production conditions. When a sudden change is detected at the coating station, the distortion area of ​​the downstream station is predicted through forward propagation of the spatiotemporal coupling factors, and a pre-compensation command is issued in advance, realizing a shift from post-feedback adjustment to feedforward pre-compensation, significantly improving the timeliness of quality control. Attached Figure Description

[0053] Figure 1 A flowchart illustrating the protective film production monitoring method based on the Industrial Internet provided in this embodiment of the invention;

[0054] Figure 2 This is a schematic diagram of the structure of the protective film production monitoring system based on the Industrial Internet, according to an embodiment of the present invention. Detailed Implementation

[0055] like Figure 1 As shown, Figure 1A flowchart of a protective film production monitoring method based on the Industrial Internet provided in this embodiment of the invention, the method comprising the following steps:

[0056] Step 101: Collect process parameters, film thickness distribution, and surface defect distribution of the protective film production line.

[0057] Step 102: The process parameters are constructed as temporal tensors, and the film thickness distribution and surface defect distribution are constructed as spatial tensors. The spatiotemporal coupling factors characterizing the flow curing dynamics of the adhesive and the formation mechanism of the film spatial distribution are extracted by coupling tensor decomposition.

[0058] In some embodiments of the present invention, step 102 may specifically include the following sub-steps:

[0059] Sub-step 1021: Expand the process parameters according to the time series to form the time dimension, expand them according to the sequence of workstations to form the workstation dimension, and combine the time dimension and the workstation dimension to construct a time series tensor.

[0060] Sub-step 1022: Unfold the film thickness distribution and surface defect distribution according to the film surface position to form a spatial position dimension, and construct a spatial tensor from the spatial position dimension;

[0061] Sub-step 1023: Establish the product association constraint between the time series tensor and the spatial tensor in the same batch. Based on the product association constraint in the same batch, perform coupled tensor decomposition on the time series tensor and the spatial tensor. Obtain the time factor matrix and the workstation factor matrix from the time series tensor, and obtain the spatial location factor matrix from the spatial tensor.

[0062] Sub-step 1024 involves coupling and mapping the time factor matrix, workstation factor matrix, and spatial location factor matrix to extract spatiotemporal coupling factors that characterize the flow leveling and curing dynamics of the adhesive and the formation mechanism of the spatial distribution of the film.

[0063] A sensor network is installed on the protective film production line to collect process parameter data, including key production parameters such as temperature, humidity, pressure, flow rate, and tension. Temperature sensors are placed in the coating, drying, and curing areas to monitor temperature changes in each area in real time; humidity sensors are placed in key locations in the production workshop; pressure sensors monitor the pressure of the coating rollers; flow rate sensors monitor the flow rate of the adhesive; and tension sensors monitor the tension of the film material. A high-precision laser scanner is used to scan and measure the film thickness, with 20 measurement points evenly distributed laterally and one measurement point every 0.5 meters longitudinally, forming a spatial distribution map of the film thickness. An industrial camera and image processing system are used to detect surface defects in the film, including bubbles, impurities, and scratches, and the defect type, location, size, and distribution density are recorded.

[0064] In the time dimension, production data was continuously collected for 3600 seconds at 1-second sampling intervals to form a time series. In the workstation dimension, the data was arranged according to the sequence of five key workstations on the production line: coating, drying, curing, cooling, and winding. Combining these two dimensions, a third-order tensor of shape 3600×5×z was constructed, where z represents the number of process parameters collected. For example, if 10 process parameters were collected, the time series tensor would have a shape of 3600×5×10.

[0065] The membrane material was divided into 20 equal regions horizontally, and data was collected every 0.5 meters vertically, resulting in 2000 vertical sampling points for a standard 1000-meter membrane. The membrane thickness spatial tensor has a shape of 20×2000×1, representing the thickness value at each spatial location. The surface defect spatial tensor has a shape of 20×2000×m, where m represents the number of defect features, including defect type, area, depth, and other characteristics.

[0066] Each batch of products is assigned a unique batch number, and the corresponding batch information is recorded in the temporal tensor and spatial tensor. For the same batch of products, a one-to-one mapping relationship is established between the time period in the temporal tensor and the spatial region in the spatial tensor. This correlation constraint ensures that the intrinsic relationship between process parameter changes and film quality characteristics can be accurately captured in the subsequent tensor decomposition process.

[0067] Coupled tensor decomposition (CFD) is performed on the temporal and spatial tensors. The temporal tensor yields a time factor matrix and a workstation factor matrix, while the spatial tensor yields a spatial location factor matrix. The core of CFD is minimizing the following objective function:

[0068] Where X represents the temporal tensor, Y represents the spatial tensor, A represents the core tensor, B represents the time factor matrix, C represents the workstation factor matrix, D represents the spatial location factor matrix, ×n represents the n-mode product of the tensors, and ‖·‖ F Let λ denote the Frobenius norm, λ be the regularization parameter, and Ω be the regularization term used to control model complexity.

[0069] Solving the above optimization problem using the alternating least squares method yields the time factor matrix B, the station factor matrix C, and the spatial location factor matrix D. The time factor matrix B reflects the variation pattern of process parameters over time, with a size of 3600×r, where r is the number of selected potential factors. The station factor matrix C reflects the influence characteristics of each station on the process parameters, with a size of 5×r. The spatial location factor matrix D reflects the spatial distribution pattern of the film properties, with a size of (20×2000)×r.

[0070] By coupling and mapping the time factor matrix, workstation factor matrix, and spatial location factor matrix, a spatiotemporal coupling factor characterizing the adhesive leveling and curing kinetics and the film spatial distribution formation mechanism is extracted. The coupling mapping is achieved by calculating the correlation matrix between the three factor matrices.

[0071] , among which, S ijk Let S represent the coupling strength between the i-th time factor, the j-th workstation factor, and the k-th spatial location factor, and let corr represent the correlation coefficient. Based on the coupling strength matrix S, the combination of factors with the highest coupling strength is selected as the spatiotemporal coupling factor characterizing the flow-leveling curing kinetics and the formation mechanism of the film spatial distribution.

[0072] This invention establishes a spatiotemporal relationship between process parameters and product quality characteristics in the protective film production process, revealing the kinetics of adhesive leveling and curing and the film formation mechanism. Through spatiotemporal coupling factor analysis, it enables real-time monitoring of production process anomalies, prediction of product quality fluctuations, optimization of process parameter settings, and improvement of production efficiency and product yield. This method can also identify key process parameters and workstations affecting product quality, providing data support for process improvement and product innovation, and achieving refined management and intelligent control of protective film production.

[0073] Step 103: Reconstruct the predicted distribution based on the spatiotemporal coupling factor, back-project the residuals of the actual distribution and the predicted distribution onto the time series tensor, and identify the abnormal process time series segments that generate residuals and the corresponding abnormal workstations.

[0074] In some embodiments of the present invention, step 103 may specifically include the following sub-steps:

[0075] Sub-step 1031: Map the spatial tensor back to the temporal tensor space through the spatiotemporal coupling factor to generate the reconstructed temporal tensor. Based on the reconstructed temporal tensor, the predicted film thickness distribution and the predicted surface defect distribution are obtained through forward deduction. The predicted film thickness distribution and the predicted surface defect distribution are used as the predicted distribution.

[0076] Sub-step 1032: The deviations between the actual film thickness distribution and the predicted film thickness distribution, as well as the deviations between the actual surface defect distribution and the predicted surface defect distribution, are constructed into a residual distribution;

[0077] Sub-step 1033: Perform spatial gradient analysis on the residual distribution to extract the residual concentration region, and back-project the residual concentration region onto the time series tensor through the spatiotemporal coupling factor to obtain the distribution of the residual on the time series tensor;

[0078] Sub-step 1034: Locate the time period of process parameter fluctuation with the highest coupling strength from the distribution of residuals on the time tensor, identify the time period of process parameter fluctuation as the abnormal process time segment that generates residuals, and identify the station where the abnormal process time segment is located as the corresponding abnormal station.

[0079] In the protective film production process, the spatial tensor is mapped back to the temporal tensor space through a spatiotemporal coupling factor, generating a reconstructed temporal tensor. This mapping process utilizes the coupling relationship between the previously obtained time factor matrix, workstation factor matrix, and spatial location factor matrix to establish a conversion channel between spatial and temporal characteristics. Specifically, the reconstructed temporal tensor is obtained by multiplying the spatiotemporal coupling factor by the spatial tensor. For the protective film production line, the calculation of the reconstructed temporal tensor is as follows: the spatial location factor matrix is ​​multiplied by the spatiotemporal coupling matrix, and then combined with the time factor matrix and workstation factor matrix to form the reconstructed temporal tensor.

[0080] Based on the reconstructed time-series tensor, the predicted film thickness distribution and predicted surface defect distribution are obtained through forward inference. During the forward inference process, the time-series information of process parameters contained in the reconstructed time-series tensor is transformed into spatial distribution predictions of film thickness and surface defects through a pre-established mapping relationship between process parameters and product quality characteristics. For film thickness, the predicted distribution is represented as a 20×2000 matrix, corresponding to the predicted thickness values ​​of 20 regions laterally and 2000 sampling points longitudinally. For surface defects, the predicted distribution is represented as a 20×2000×m tensor, where m is the number of defect features, including defect type, area, depth, etc.

[0081] The deviations between the actual and predicted film thickness distributions, as well as the deviations between the actual and predicted surface defect distributions, are constructed as residual distributions. The film thickness residual distribution is calculated by subtracting the predicted thickness value from the actual thickness value, resulting in a 20×2000 residual matrix. The surface defect residual distribution is calculated by subtracting the predicted defect characteristic value from the actual defect characteristic value, resulting in a 20×2000×m residual tensor. Larger residual values ​​indicate greater discrepancies between the actual and predicted results, potentially suggesting anomalies.

[0082] Spatial gradient analysis identifies regions of drastic residual value changes by calculating the rate of change of residuals in the horizontal and vertical directions. For the film thickness residual distribution, the horizontal gradient is calculated as the difference in residual values ​​between adjacent regions, and the vertical gradient is calculated as the difference in residual values ​​between adjacent sampling points. Regions with large gradient values ​​indicate drastic residual changes and may be boundary areas where anomalies occur. Residual concentration regions are extracted using a threshold segmentation method, marking regions where the absolute value of the residual exceeds a preset threshold as residual concentration regions. The preset threshold can be determined according to production process requirements and product quality standards, and is typically set to three times the standard deviation.

[0083] The residual concentration region is back-projected onto the time series tensor using a spatiotemporal coupling factor to obtain the residual distribution on the time series tensor. The back-projection process is the inverse operation of the mapping process, transforming the residual information from the spatial domain back into the time and workstation domains. Specifically, the correlation between spatial location and time / workstation established using the spatiotemporal coupling factor is used to calculate the time period and workstation corresponding to the residual concentration region. The back-projection calculation formula is as follows: , where R t，p S represents the distribution of residuals on the time series tensor. ijk R represents the spatiotemporal coupling factor. s (i, j) represents the residual value at spatial location (i, j). This formula converts the residual information in the spatial domain into residual distributions in the time domain and workstation domain, forming a 3600×5 matrix that represents the residual distribution at 5 workstations within a 3600-second time range.

[0084] The time period with the highest coupling strength for process parameter fluctuations is located from the distribution of residuals on the time series tensor. The coupling strength is calculated based on the product of the residual value and the rate of change of the process parameter, reflecting the degree of correlation between process parameter fluctuations and product quality anomalies. For each time point t and workstation p, its coupling strength is calculated: Where C(t, p) represents the coupling strength at workstation p at time point t, and R t，p (t, p) represents the residual value at that point, X(t, p) represents the process parameter value at that point, and Δt represents the time interval. Time periods where the coupling strength exceeds a threshold are identified as process parameter fluctuation periods, with the threshold set to 5 times the average coupling strength. For protective film production lines, common process parameter fluctuation periods include temperature abrupt changes, pressure fluctuations, and unstable flow rates.

[0085] The time periods of process parameter fluctuations are identified as abnormal process sequence segments that generate residuals, and the workstations where these abnormal process sequence segments are located are identified as corresponding abnormal workstations. For each abnormal process sequence segment, its start time, duration, process parameter fluctuation amplitude, and impact range are recorded. For each abnormal workstation, its location in the production line, the associated process parameters, and their normal value ranges are recorded. An abnormal workstation may be one or more of the following: coating station, drying station, curing station, cooling station, or winding station. By identifying abnormal process sequence segments and abnormal workstations, the source of quality problems can be accurately located, providing a basis for subsequent adjustments and optimizations.

[0086] This invention enables precise location and early warning of abnormal states during the production of protective films. By establishing a correlation mechanism between process parameter fluctuations and product quality defects through spatiotemporal coupling analysis, it transforms the process from "discovering problems" to "finding the root cause." This method not only improves the accuracy and efficiency of production monitoring but also reduces the risk of quality fluctuations, minimizes material waste and production downtime, and provides strong support for the intelligent and refined management of protective film production.

[0087] Step 104: Calculate the constraint change of abnormal station adjustment on the leveling time and curing rate of adhesive in downstream stations based on the spatiotemporal coupling factor, solve the multi-station collaborative adjustment scheme within the process feasible domain limited by the constraint change amount, and execute it.

[0088] In some embodiments of the present invention, step 104 may specifically include the following sub-steps:

[0089] Sub-step 1041: Based on the coupling relationship between the abnormal workstation and the downstream workstation in the spatiotemporal coupling factor, calculate the influence of the abnormal workstation adjustment on the adhesive leveling time and the curing rate of the downstream workstation through the adhesive leveling and curing dynamics, and use the influence on the adhesive leveling time and the curing rate of the downstream workstation as constraint change quantities.

[0090] Sub-step 1042: Apply the constraint change to the original process constraint of the downstream station to obtain the updated process constraint of the downstream station, and use the updated process constraint as the process feasible domain.

[0091] Sub-step 1043: Within the process feasible domain, the process parameter adjustment amounts of abnormal stations and downstream stations are used as optimization variables to solve the problem and obtain the combination of process parameter adjustment amounts of abnormal stations and downstream stations. The combination of process parameter adjustment amounts is used as a multi-station collaborative adjustment scheme.

[0092] Sub-step 1044: The multi-station collaborative adjustment plan is sent to the process equipment corresponding to the abnormal station and the downstream station for execution.

[0093] Based on the coupling relationship between abnormal workstations and downstream workstations in the spatiotemporal coupling factor, the impact of abnormal workstation adjustments on downstream workstations through adhesive leveling and curing kinetics is calculated. The spatiotemporal coupling factor reflects the transmission relationship of process parameter changes between different workstations, especially the degree of influence of upstream workstations on downstream workstations. In the protective film production line, the adhesive leveling and curing process runs through multiple workstations, and parameter changes in upstream workstations are transmitted to downstream workstations through changes in adhesive properties. The impact calculation is divided into two parts: the impact on adhesive leveling time and the impact on curing rate.

[0094] The calculation of the impact on the leveling time of the adhesive is based on the product of the process parameter adjustment and the leveling time sensitivity coefficient. When adjusting the temperature at the coating station, the leveling time changes by approximately 5 seconds for every 1°C adjustment; when adjusting the pressure, the leveling time changes by approximately 3 seconds for every 0.1 MPa adjustment; and when adjusting the adhesive viscosity, the leveling time changes by approximately 8 seconds for every 10 mPa·s adjustment. The formula for calculating the impact on leveling time is: , where ΔT flow α represents the effect of leveling time. y ΔP represents the sensitivity coefficient of the y-th process parameter to the leveling time. y This represents the adjustment amount of the y-th process parameter, and N represents the total number of process parameters that affect the leveling time.

[0095] The calculation of the impact on curing rate is based on the product of the process parameter adjustment and the curing rate sensitivity coefficient. When adjusting the temperature at the drying station, the curing rate changes by approximately 0.02 min for every 1°C adjustment. -1 When adjusting the humidity, the curing rate changes by approximately 0.01 min for every 5% adjustment. -1 When adjusting the ultraviolet light intensity, for every 10mW / cm² adjustment... 2 The curing rate changed by approximately 0.03 min. -1 The formula for calculating the effect of curing rate is: , where ΔR cure β represents the effect of curing rate. u ΔQ represents the sensitivity coefficient of the u-th process parameter to the curing rate. u This represents the adjustment amount of the u-th process parameter, and M represents the total number of process parameters that affect the curing rate.

[0096] The effects of upstream station adjustments on adhesive leveling time and curing rate at downstream stations were used as constraint variables. These constraint variables describe the impact of upstream station adjustments on the adjustable range of downstream station process parameters, providing a basis for multi-station coordinated adjustments. For example, a 2°C increase in the coating station temperature may shorten the adhesive leveling time by 10 seconds and increase the curing rate by 0.04 min at the downstream drying station. -1 This change requires downstream workstations to adjust their process parameters accordingly to maintain stable product quality.

[0097] The changes in constraints are applied to the existing process constraints of downstream stations to obtain updated process constraints for those stations. The original process constraints were allowable ranges of process parameters determined based on product quality requirements and equipment capabilities. For the drying station, the temperature process constraint is 80℃ to 100℃, the humidity process constraint is 30% to 50%, and the air velocity process constraint is 0.5m / s to 2m / s. If an adjustment in the temperature of the upstream coating station causes a 10-second reduction in leveling time, the drying station needs to adjust its temperature process constraint to 78℃ to 98℃ to compensate for the change in leveling time. The updated process constraints constitute the process feasible region, which is the solution space for multi-station collaborative adjustment schemes.

[0098] Within the feasible process domain, the optimization variables are the process parameter adjustments at abnormal and downstream stations. The optimization objective is to minimize product quality fluctuations, while considering factors such as energy consumption, equipment load, and adjustment magnitude. The optimization problem is formulated as: finding the combination of process parameter adjustments that minimizes the objective function while satisfying process constraints. The objective function can be designed as a weighted sum of quality fluctuation and energy consumption indicators, with the weights determined based on production management strategies. For abnormal temperatures at the coating station, possible coordinated adjustment schemes include: reducing the coating station temperature by 1.5℃, increasing the drying station temperature by 2℃, and reducing the UV intensity at the curing station by 5mW / cm². 2 .

[0099] The solution process employs gradient descent or simulated annealing. Starting from an initial solution, the process parameters at each station are gradually adjusted until the solution that satisfies the constraints and minimizes the objective function value is found. The initial solution can be the current process parameters or historically optimal process parameters. For each iteration, the gradient of the objective function is calculated, and the process parameters are adjusted along the gradient direction until convergence or the maximum number of iterations is reached. In practical applications, an upper limit of 100 iterations and a convergence threshold of 0.01 can be set to ensure that the algorithm obtains an effective solution within a reasonable time.

[0100] The multi-station collaborative adjustment plan is distributed to the corresponding process equipment at the abnormal workstations and downstream workstations for execution. The adjustment plan is transmitted to the programmable logic controllers or distributed control systems at each workstation via the industrial internet platform. Temperature parameter adjustments are achieved by regulating heater power; pressure parameter adjustments are achieved by regulating pneumatic or hydraulic system pressure; and speed parameter adjustments are achieved by regulating motor speed. The adjustment process adopts a step-by-step, gradual approach to avoid production fluctuations caused by sudden parameter changes. After each adjustment step is completed, the monitoring system provides real-time feedback on parameter changes to confirm the adjustment effect. Fine-tuning is performed as necessary until the expected target is achieved.

[0101] This invention establishes a coupling relationship between process parameters at different workstations, enabling intelligent management of multi-workstation collaborative adjustments in a protective film production line. Compared to traditional independent adjustments at a single workstation, multi-workstation collaborative adjustments can more comprehensively consider the chain reactions of process parameter changes, reduce adjustment conflicts, and lower the risk of product quality fluctuations. This method not only improves the efficiency of anomaly handling and shortens adjustment time but also enhances the stability and controllability of the production process, providing an effective means for refined management and intelligent control of protective film production, and achieving simultaneous improvement in product quality and production efficiency.

[0102] Step 105: Collect the film thickness distribution and surface defect distribution after execution to construct the spatial tensor after execution, and update the spatiotemporal coupling factor based on the difference between the spatial tensor after execution and the spatial tensor.

[0103] In some embodiments of the present invention, step 105 may specifically include the following sub-steps:

[0104] Sub-step 1051: Collect the film thickness distribution and surface defect distribution after the execution of the multi-station collaborative adjustment scheme, and stack the film thickness distribution and surface defect distribution after the execution into a spatial tensor after the execution.

[0105] Sub-step 1052: Calculate the difference between the spatial tensor and the actual spatial response after execution. Calculate the expected spatial response based on the multi-station collaborative adjustment scheme and the spatiotemporal coupling factor. Use the deviation between the actual spatial response and the expected spatial response as the coupling deviation.

[0106] Sub-step 1053: Spatial decomposition of coupling deviation, extraction of spatial components associated with abnormal workstations and spatial components associated with downstream workstations in coupling deviation, and calculation of coupling correction coefficients for abnormal workstations and downstream workstations respectively.

[0107] Sub-step 1054: Apply the coupling correction coefficients of the abnormal workstation and the downstream workstation to the coupling strengths corresponding to the abnormal workstation and the downstream workstation in the spatiotemporal coupling factor, respectively, to obtain the updated spatiotemporal coupling factor.

[0108] The film thickness distribution and surface defect distribution after the multi-station collaborative adjustment scheme were collected and stacked into a post-execution spatial tensor. Film thickness distribution was acquired using an optical thickness measurement device, with 20 measurement points set horizontally and 2000 sampling positions arranged vertically, forming a 20×2000 thickness data matrix. Surface defect distribution was acquired using a high-speed industrial camera with a light source array to capture surface scratches, bubbles, impurities, and other defect information, including defect type, size, and location, forming a 20×2000×6 defect feature tensor, where 6 represents the defect feature dimension, including defect type, area, depth, edge sharpness, contrast, and texture features. The post-execution spatial tensor was formed by stacking the thickness distribution matrix and the defect feature tensor along the feature dimensions, resulting in a 20×2000×7 tensor structure, where 7 represents the total feature dimension, including the thickness value and the 6 defect feature dimensions.

[0109] The actual spatial response is obtained by calculating the difference between the spatial tensor after execution and the actual spatial tensor. The spatial tensor refers to the tensor formed by the stacking of the film thickness distribution and surface defect distribution before the adjustment scheme was implemented. The actual spatial response is calculated by subtracting the actual spatial tensor from the spatial tensor after execution, resulting in a 20×2000×7 difference tensor, which represents the actual quality change caused by the adjustment scheme. For thickness characteristics, positive values ​​indicate an increase in thickness, and negative values ​​indicate a decrease in thickness; for defect characteristics, positive values ​​indicate enhanced defect properties, and negative values ​​indicate weakened defect properties.

[0110] The expected spatial response is calculated based on a multi-station collaborative adjustment scheme and a spatiotemporal coupling factor. The expected spatial response represents the predicted quality change that the multi-station collaborative adjustment scheme should produce based on known spatiotemporal coupling relationships. The calculation method is the sum of the products of the adjustment amount of each process parameter at each station in the multi-station collaborative adjustment scheme and the corresponding spatiotemporal coupling factor. For a production line with 5 stations, each station adjusts 10 process parameters. The expected spatial response calculation requires multiplying the adjustment amount of each process parameter at each station by the coupling strength of that parameter to each feature at each spatial location, and then summing the effects of all process parameters. This calculation yields an expected spatial response tensor of 20×2000×7.

[0111] The deviation between the actual and expected spatial responses is used as the coupling deviation. This deviation reflects the difference between the spatiotemporal coupling relationship in actual production and the model's prediction, serving as the basis for updating the spatiotemporal coupling factor. The calculation method involves subtracting the expected spatial response from the actual spatial response, resulting in a 20×2000×7 coupling deviation tensor. A large coupling deviation indicates that the current spatiotemporal coupling factor is not accurate enough and requires significant adjustment; a small coupling deviation indicates that the current spatiotemporal coupling factor is relatively accurate and only requires fine-tuning.

[0112] The coupling deviation is spatially decomposed to extract spatial components associated with abnormal workstations and those associated with downstream workstations. Principal component analysis (PCA) is used for spatial decomposition, projecting the coupling deviation onto a pre-defined workstation feature space. The workstation feature space refers to the characteristic response pattern corresponding to each workstation, obtained through historical data analysis. The spatial component extraction process includes calculating the projection coefficients between the coupling deviation and the feature space of each workstation, and then multiplying these projection coefficients by the corresponding eigenvectors to obtain the spatial components of each workstation.

[0113] The coupling correction coefficients for the abnormal workstation and the downstream workstation are calculated separately. The coupling correction coefficient represents the degree of adjustment required to the spatiotemporal coupling factor, calculated based on the amplitude and direction of the spatial component. For the abnormal workstation, the calculation of the coupling correction coefficient needs to consider the relationship between the value of the spatial component and the adjustment amount of the process parameters, reflecting the degree of response deviation caused by a unit adjustment. The coupling correction coefficient is calculated as follows: , where α abn D represents the coupling correction coefficient for abnormal workstations. abn Represents the spatial components of the abnormal workstation; ΔP abn Indicates the adjustment amount of process parameters at abnormal workstations; ||ΔP abn ‖2 represents the Euclidean norm of the process parameter adjustment. The calculation method for the coupling correction coefficient of the downstream station is similar; simply replace the relevant parameters of the abnormal station with the relevant parameters of the downstream station.

[0114] The coupling correction coefficients for the abnormal workstation and the downstream workstation are applied to the corresponding coupling strengths of the abnormal and downstream workstations in the spatiotemporal coupling factor, respectively, to obtain the updated spatiotemporal coupling factor. The update method is to add the product of the coupling correction coefficient and the learning rate to the original coupling strength. The learning rate is a parameter that controls the update speed and is usually set to a value between 0.1 and 0.3. For each process parameter of each workstation, the updated coupling strength is calculated by adding the corresponding correction term to the original coupling strength. The coupling strength update adopts a smooth transition method to avoid drastic fluctuations during the update process. For abnormal workstations, the update amplitude of their coupling strength is usually greater than that of downstream workstations, reflecting the more significant direct impact of abnormal workstation adjustments on quality.

[0115] The updated spatiotemporal coupling factor is saved to the industrial internet platform's database, serving as the basis for the next multi-workstation collaborative adjustment. The update frequency is determined based on production stability and the frequency of anomalies, typically once per hour or once per batch of products. Continuous updates to the spatiotemporal coupling factor ensure that the model can adapt to factors such as process changes, equipment aging, and raw material fluctuations, maintaining predictive accuracy. When production conditions are stable, the update frequency can be reduced to decrease computational burden; when production conditions fluctuate significantly, the update frequency can be increased to accelerate model adaptation. The industrial internet platform also saves historical versions of the spatiotemporal coupling factor, facilitating retrospective analysis or comparison of changes in coupling relationships at different times when needed.

[0116] This invention implements an adaptive update mechanism for spatiotemporal coupling factors based on execution results, enabling the protective film production monitoring method to possess continuous learning and self-optimization capabilities. This closed-loop optimization mechanism effectively addresses dynamic changes in the production environment, enhances the robustness and adaptability of the monitoring method, provides sustainable technical support for the intelligent management of the protective film production process, realizes continuous improvement in production quality and automatic accumulation of process knowledge, and promotes the development of protective film manufacturing towards higher quality and higher efficiency.

[0117] Step 106: When a sudden change in the process parameters of the coating station is detected, the sudden change is substituted into the updated spatiotemporal coupling factor for forward propagation. The film thickness distortion area and surface defect expansion area caused by the sudden change in the downstream station are predicted. Before the distortion is formed, a pre-compensation adjustment command is issued to the downstream station.

[0118] In some embodiments of the present invention, step 106 may specifically include the following sub-steps:

[0119] Sub-step 1061: Monitor the changes in process parameters at the coating station in real time. When the change in process parameters exceeds the preset fluctuation range, it is identified as a sudden change in the process parameters at the coating station. Obtain the magnitude and timing of the sudden change in the process parameters at the coating station.

[0120] Sub-step 1062: Substitute the mutation amplitude into the updated spatiotemporal coupling factor for forward propagation to obtain the propagation response of the mutation in the film thickness distribution and surface defect distribution of the downstream station. Calculate the propagation time of the mutation to the downstream station based on the updated spatiotemporal coupling factor.

[0121] Sub-step 1063: Spatial localization of the propagation response on the film thickness distribution and the propagation response on the surface defect distribution at the downstream station; extraction of the film thickness distortion region and surface defect expansion region caused by the abrupt change at the downstream station; calculation of the pre-compensation process parameter adjustment amount at the downstream station based on the film thickness distortion region and the surface defect expansion region.

[0122] Sub-step 1064: Determine the time of distortion formation based on the time of mutation and the propagation time. Before the time of distortion formation, issue the pre-compensation process parameter adjustment amount to the corresponding process equipment of the downstream station as a pre-compensation adjustment instruction.

[0123] The protective film production line monitors real-time changes in process parameters at the coating station, including key parameters such as coating speed, coating pressure, coating adhesive temperature, coating adhesive flow rate, and coating blade height. High-precision sensors with a sampling frequency of 10Hz ensure the capture of millisecond-level parameter fluctuations. Changes in process parameters exceeding a preset fluctuation range are identified as sudden changes in the coating station's process parameters. The preset fluctuation range is determined based on statistical analysis of historical process data and is typically set to three times the standard deviation of normal fluctuations. For example, if the normal standard deviation of coating speed fluctuations is 0.5 m / min, the preset fluctuation range is 1.5 m / min; if the normal standard deviation of coating pressure fluctuations is 0.2 kPa, the preset fluctuation range is 0.6 kPa. When a parameter change is detected to exceed the preset fluctuation range, the industrial internet platform records the magnitude and timing of the sudden change. The magnitude of the sudden change is calculated by subtracting the parameter value before the change from the value after the change, and the timing of the change is accurate to the millisecond level.

[0124] The abrupt change magnitude is substituted into the updated spatiotemporal coupling factor for forward propagation, yielding the propagation response of the abrupt change in film thickness distribution and surface defect distribution at downstream stations. The forward propagation calculation uses the spatiotemporal coupling factor as a mapping matrix to map the abrupt change in process parameters at the coating station to the quality response at each location in the downstream stations. For each process parameter at the coating station, the impact of its abrupt change on the quality characteristics of the downstream stations is calculated, and the effects of all parameters are summed to obtain the total propagation response. This calculation yields a 20×2000×7 propagation response tensor, including the film thickness response and surface defect response.

[0125] The propagation time of mutations to downstream stations is calculated based on the updated spatiotemporal coupling factor. This propagation time is related to production line speed, distance between stations, and material properties. The calculation considers the conveying speed of the protective film on the production line and the physical propagation characteristics of quality defects. The propagation time is calculated as the physical distance between the downstream station and the coating station divided by the protective film conveying speed, plus the material response delay time. For example, if the distance between the coating station and the drying station is 5m, the protective film conveying speed is 10m / min, and the material response delay time is 0.1min, then the propagation time is 0.6min. The propagation time varies for different downstream stations and must be calculated separately.

[0126] Spatial localization was performed on the propagation responses of the film thickness distribution and surface defect distribution at downstream workstations to extract regions of film thickness distortion and surface defect propagation caused by abrupt changes at downstream workstations. Spatial localization employed a threshold segmentation method, applying a threshold to the propagation response tensor to identify regions exceeding the threshold as distorted or defect-propagated regions. The thickness distortion threshold was set to twice the normal thickness variation range, typically ±2 μm; the surface defect propagation threshold was set to twice the normal defect variation range. The spatial localization results were represented as a binary mask, with regions of 1 representing distorted or defect-propagated regions and regions of 0 representing normal regions.

[0127] The pre-compensation process parameter adjustments for downstream stations are calculated based on the film thickness distortion region and the surface defect propagation region. The calculation employs the inverse mapping of the spatiotemporal coupling factor, mapping the desired quality correction value back to the process parameter adjustment amount. During the calculation, the quality influence value to be eliminated is first determined, which is equivalent to the negative of the predicted propagation response value. Then, through the inverse mapping relationship of the spatiotemporal coupling factor, these desired quality corrections are mapped to the process parameter adjustments for each downstream station. The importance of quality characteristics is considered during the calculation, with higher weights given to the responses of important features and critical regions. Through this calculation, the pre-compensation adjustment amount for each process parameter at each downstream station is obtained.

[0128] The distortion initiation time is calculated as the abrupt change time plus the propagation time. Before the distortion initiation time, the pre-compensation process parameter adjustment amount is sent to the corresponding process equipment at the downstream station as a pre-compensation adjustment instruction. The pre-compensation instruction is typically sent 30 seconds in advance to allow sufficient time for the equipment to execute the adjustment. The pre-compensation adjustment instruction includes information such as adjustment parameters, adjustment amount, and execution time. After receiving the pre-compensation adjustment instruction, the process equipment executes the parameter adjustment at the specified time to achieve precise compensation for the impact of the abrupt change.

[0129] The pre-compensation adjustment employs a smooth transition approach to avoid introducing new fluctuations. The adjustment process is divided into three phases: a warm-up phase, a main adjustment phase, and a recovery phase. In the warm-up phase, parameters are gradually adjusted to 80% of the target value; in the main adjustment phase, parameters are maintained at the target adjustment value; and in the recovery phase, parameters are gradually restored to their normal values. The time allocation for these three phases is determined based on the duration of the abrupt change's impact, typically in a 1:3:1 ratio. This phased adjustment ensures a smooth and controllable pre-compensation process.

[0130] The industrial internet platform records the execution status and effects of each pre-compensation adjustment, including the predicted distortion value before adjustment, the actual measured value after adjustment, and the evaluation of the adjustment effect. These records are used to subsequently optimize the pre-compensation strategy and improve the accuracy of pre-compensation. At the same time, the platform analyzes historical data on pre-compensation adjustments, identifies cases with good and poor adjustment effects, extracts key factors affecting the pre-compensation effect, and continuously optimizes the pre-compensation model and strategy.

[0131] For accurate prediction of the propagation response, the impact of sudden changes in process parameters at the coating station on the quality characteristics of downstream stations can be expressed by the following formula: , where Y d (i, j, l) represents the response value of the l-th quality feature at the downstream workstation location (i, j), X c (c) represents the abrupt change amplitude of the c-th process parameter at the coating station, S c，d (m, i, j, l) represents the coupling strength between the c-th process parameter at the coating station and the l-th quality feature at the downstream station position (i, j) in the spatiotemporal coupling factor, F c This indicates the total number of process parameters for the coating station.

[0132] This invention enables early prediction and proactive compensation for the impact of sudden changes in process parameters at the coating station, transforming traditional passive response into proactive prevention. Through forward propagation calculation of the spatiotemporal coupling factor, it accurately predicts the area and extent of the impact of sudden changes on the quality of downstream stations, achieving early detection of quality problems. Based on a pre-compensation strategy using inverse mapping calculation, it precisely calculates the required adjustment amount of process parameters, ensuring the targetedness and effectiveness of the adjustment. A phased smooth adjustment method is adopted to avoid introducing new fluctuations during the adjustment process, ensuring the stability of the production process.

[0133] Sub-step 1063 involves spatially locating the propagation response on the film thickness distribution and the propagation response on the surface defect distribution at the downstream station, extracting the film thickness distortion region and surface defect expansion region caused by the abrupt change at the downstream station, and calculating the pre-compensation process parameter adjustment amount for the downstream station based on the film thickness distortion region and the surface defect expansion region. This also includes:

[0134] Spatial gradient analysis is performed on the propagation response of the membrane thickness distribution at the downstream workstation. Continuous regions where the spatial gradient exceeds the preset fluctuation threshold are spatially located and marked as membrane thickness distortion regions. The spatial location and distortion amplitude of the membrane thickness distortion regions are extracted.

[0135] Spatial boundary identification is performed on the propagation response of surface defects in downstream workstations. The newly added area that expands outward from the original surface defect boundary is spatially located and marked as the surface defect expansion area. The spatial location and expansion scale of the surface defect expansion area are extracted.

[0136] The spatial location and distortion amplitude of the film thickness distortion region are mapped to the downstream station process parameter adjustment space to obtain the process parameter compensation amount used to offset the distortion amplitude. The spatial location and expansion scale of the surface defect expansion region are mapped to the downstream station process parameter adjustment space to obtain the process parameter compensation amount used to suppress the expansion scale.

[0137] The pre-compensation process parameter adjustment amount for downstream stations is obtained by spatially superimposing the process parameter compensation amount used to offset the distortion amplitude and the process parameter compensation amount used to suppress the expansion scale.

[0138] Spatial gradient analysis was performed on the propagation response of the membrane thickness distribution at downstream workstations. The propagation response data was represented as a two-dimensional matrix with 2000 sampling points in the membrane width direction and 20 sampling points in the membrane length direction. Spatial gradient analysis was achieved by calculating the thickness difference between adjacent sampling points, with gradient values ​​calculated in both the transverse and longitudinal directions. The transverse gradient was calculated by dividing the difference between two adjacent points by the sampling interval, and the longitudinal gradient was calculated similarly. The normal protective membrane thickness gradient value generally does not exceed 0.05 μm / mm; therefore, the preset fluctuation threshold was set to 0.15 μm / mm. When the spatial gradient exceeds the preset fluctuation threshold and at least five adjacent points within the same region exceed the threshold, the area formed by these points is marked as a membrane thickness distortion region. During spatial positioning, the starting and ending coordinates of the distortion region were recorded to form a spatial location description of the distortion region. Simultaneously, the thickness deviation value of each point within the distortion region was calculated, with the maximum deviation value used as the distortion amplitude. In practical applications, thickness distortion regions typically exhibit patchy or strip-like shapes, with areas ranging from a few square centimeters to hundreds of square centimeters. The distortion range is generally between 1 and 5 μm. Distortions exceeding 5 μm may lead to a decrease in the functional performance of the protective film.

[0139] Spatial boundary identification is performed on the propagation response of surface defects at downstream workstations, primarily targeting various types of surface defects such as point-like, line-like, and mesh-like defects. The propagation response data is also represented in a two-dimensional matrix, using the same sampling point settings. Spatial boundary identification first extracts the boundary contour of the original surface defects using an edge detection algorithm, processing the defect response matrix. Edge detection employs a gradient threshold method, calculating the rate of change of the defect exponent between adjacent points; points exceeding a set threshold are identified as edge points. After determining the original surface defect boundary, the difference between the defect boundary after propagation response and the original boundary is compared to identify the newly added set of edge points. The region formed by connecting these newly added edge points is marked as the surface defect expansion region. During spatial positioning, the boundary coordinates of the expansion region are recorded, forming a spatial location description of the expansion region. The expansion scale is calculated as the average distance from the original defect boundary to the expansion region boundary, reflecting the degree of defect expansion. In practical applications, surface defect expansion regions typically exhibit a ring-like or irregular shape surrounding the original defect, with an expansion scale generally ranging from 0.5 to 3 mm. Expansions exceeding 3 mm may lead to interconnected defects forming large-area defects.

[0140] Mapping the spatial location and amplitude of the film thickness distortion region to the downstream process parameter adjustment space is the core step in calculating the process parameter compensation. The mapping process requires establishing the correspondence between the spatial location of the distortion region and the influence range of the process parameters, as well as the quantitative relationship between the distortion amplitude and the adjustment amount of the process parameters. For the downstream drying station, its temperature distribution corresponds to different positions along the film width direction and can be divided into several temperature intervals, each of which can be adjusted independently. During mapping, the temperature interval where the distortion region is located is first determined, and then the temperature adjustment amount required to compensate for the distortion is calculated based on the temperature response curve of that interval. The temperature response curve describes the influence coefficient of temperature change on film thickness, with a typical value of 0.5 μm / °C. For a region with a thickness distortion amplitude of 3 μm, the required temperature adjustment amount is -6°C. For the downstream tension control station, tension adjustment corresponds to different positions along the film length direction. The mapping method is similar, and the required tension adjustment amount is calculated based on the tension response curve. The tension response curve describes the influence coefficient of tension change on film thickness, with a typical value of 0.15 μm / N. For a region with a thickness distortion amplitude of 3 μm, the required tension adjustment amount is -20 N. This mapping relationship converts the thickness distortion in spatial location into a compensation amount for process parameters.

[0141] The process of mapping the spatial location and scale of surface defect propagation areas to the downstream process parameter adjustment space is similar to that of mapping thickness distortion areas, but the focus is on suppressing defect propagation. For the downstream cooling station, the cooling rate corresponds to the sensitivity of defect propagation and can be divided into several cooling intervals, each of which can be adjusted independently. During mapping, the cooling interval where the propagation area is located is determined, and then the required cooling rate adjustment to suppress propagation is calculated based on the cooling response curve. The cooling response curve describes the influence coefficient of cooling rate changes on defect propagation, with a typical value of 1.2 mm / (°C / min). For an area with a propagation scale of 2.4 mm, the required cooling rate adjustment is -2°C / min. For the downstream traction station, traction speed adjustment also affects the degree of defect propagation. The mapping method is similar, calculating the required speed adjustment based on the traction speed response curve. The traction speed response curve describes the influence coefficient of speed changes on defect propagation, with a typical value of 0.8 mm / (m / min). For an area with a propagation scale of 2.4 mm, the required speed adjustment is -3 m / min. Through this mapping relationship, the defect propagation area is converted into a compensation amount for process parameters.

[0142] The process parameter compensation amounts used to offset the distortion amplitude and those used to suppress the expansion scale are spatially superimposed to obtain the pre-compensation process parameter adjustment amount for the downstream station. Spatial superposition considers the overlap of different regions and the mutual influence of process parameters. For completely overlapping regions, the compensation amounts are directly added; for partially overlapping regions, a weighted sum is calculated according to the overlap ratio; for non-overlapping regions, their respective compensation amounts are retained. During the superposition process, the adjustment range limitations of the process parameters must also be considered to ensure that the superimposed adjustment amount does not exceed the equipment's allowable adjustment range. Taking the temperature adjustment of the downstream drying station as an example, the temperature compensation amount to offset thickness distortion is -6°C, the temperature compensation amount to suppress defect expansion is -0.8°C, and the superimposed temperature adjustment amount is -6.8°C. Considering that the allowable adjustment range of drying temperature is ±7°C, this adjustment amount is within the allowable range and can be directly used as the pre-compensation process parameter adjustment amount. For the downstream tension control station, the tension compensation amount to offset thickness distortion is -20N, the tension compensation amount to suppress defect expansion is -15N, and the superimposed tension adjustment amount is -35N. Considering that the allowable adjustment range for tension is ±30N, this adjustment exceeds the allowable range and needs to be trimmed. The final pre-compensation process parameter adjustment is -30N. Through spatial superposition and range constraint processing, the final pre-compensation process parameter adjustment for each downstream station is obtained.

[0143] The adjustment amounts of the pre-compensation process parameters after spatial overlay need further optimization to ensure the stability and effectiveness of the adjustment. A smooth transition strategy is adopted for optimization to avoid new problems caused by abrupt parameter changes. For temperature parameters, a gradient approach is used, gradually reaching the target adjustment amount at a rate of 0.5°C every 15 seconds; for tension parameters, a step-wise adjustment is used, gradually adjusting at a step size of 5N every 10 seconds; for velocity parameters, a gradual adjustment is used, gradually adjusting at a rate of 0.5m / min every 20 seconds. This smooth transition strategy makes the pre-compensation adjustment process more stable and avoids secondary fluctuations. Before issuing the adjustment amount, the physical characteristics and response time of the downstream processing stations must be considered, and the adjustment is executed a certain time in advance. Depending on the response characteristics of different stations, the advance time is generally set to range from 30 seconds to 2 minutes.

[0144] This invention achieves proactive prevention of quality problems in downstream processes by accurately analyzing and predicting the impact of sudden changes in process parameters at the coating station. Spatial gradient analysis and boundary recognition technologies accurately locate regions of film thickness distortion and surface defect propagation, extracting key spatial location and amplitude information. Spatial mapping and overlay technologies precisely convert quality problems into process parameter adjustment amounts, realizing a mapping from quality space to process space. A smooth transition strategy ensures the stability of pre-compensation adjustments and avoids secondary fluctuations.

[0145] like Figure 2 As shown, Figure 2 This is a schematic diagram of a protective film production monitoring system based on the Industrial Internet provided in an embodiment of the present invention. The system includes:

[0146] The data acquisition module 201 is used to collect process parameters, film thickness distribution, and surface defect distribution of the protective film production line.

[0147] The coupling factor extraction module 202 is used to construct process parameters as temporal tensors, construct film thickness distribution and surface defect distribution as spatial tensors, and extract spatiotemporal coupling factors that characterize the flow curing dynamics of adhesive liquid and the formation mechanism of film spatial distribution through coupling tensor decomposition.

[0148] The anomaly identification module 203 is used to reconstruct the predicted distribution based on the spatiotemporal coupling factor, back-project the residual between the actual distribution and the predicted distribution onto the time tensor, and identify the abnormal process time segment that generates the residual and the corresponding abnormal work station.

[0149] The collaborative adjustment module 204 is used to calculate the constraint change of abnormal station adjustment on the leveling time and curing rate of adhesive in downstream stations based on the spatiotemporal coupling factor, and solve and execute the multi-station collaborative adjustment scheme within the process feasible domain limited by the constraint change.

[0150] The factor update module 205 is used to collect the film thickness distribution and surface defect distribution after execution to construct the spatial tensor after execution, and update the spatiotemporal coupling factor according to the difference between the spatial tensor after execution and the spatial tensor.

[0151] The feedforward compensation module 206 is used to, when a sudden change in the process parameters of the coating station is detected, substitute the sudden change into the updated spatiotemporal coupling factor for forward propagation, predict the film thickness distortion area and surface defect expansion area caused by the sudden change in the downstream station, and issue a pre-compensation adjustment command to the downstream station before the distortion is formed.

[0152] One technical solution provided in this embodiment of the invention is an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0153] One technical solution provided in this embodiment of the invention is a computer-readable storage medium storing a computer program, wherein the processor executes the computer program to implement the steps in any of the aforementioned methods.

[0154] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.

Claims

1. A method for monitoring the production of protective films based on the Industrial Internet, characterized in that, Includes the following steps: Collect process parameters, film thickness distribution, and surface defect distribution of the protective film production line; The process parameters were constructed as temporal tensors, and the film thickness distribution and surface defect distribution were constructed as spatial tensors. The spatiotemporal coupling factors characterizing the flow-leveling curing dynamics of the adhesive and the formation mechanism of the film spatial distribution were extracted by coupling tensor decomposition. Based on the spatiotemporal coupling factor, the predicted distribution is reconstructed, and the residual between the actual distribution and the predicted distribution is back-projected onto the time tensor to identify the abnormal process time segments that generate residuals and the corresponding abnormal workstations. The constraint changes on the leveling time and curing rate of the adhesive at downstream stations are calculated based on the spatiotemporal coupling factor. Within the process feasibility domain limited by the constraint changes, a multi-station collaborative adjustment scheme is solved and executed. The thickness distribution and surface defect distribution of the film layer after execution are collected to construct the spatial tensor after execution. The spatiotemporal coupling factor is updated based on the difference between the spatial tensor after execution and the spatial tensor. When a sudden change in the process parameters of the coating station is detected, the change is substituted into the updated spatiotemporal coupling factor for forward propagation. The changes are then used to predict the film thickness distortion area and surface defect expansion area caused by the change in the downstream station. Before the distortion occurs, a pre-compensation adjustment command is issued to the downstream station.

2. The method according to claim 1, characterized in that, Process parameters were constructed as temporal tensors, and film thickness distribution and surface defect distribution were constructed as spatial tensors. Spatiotemporal coupling factors characterizing the adhesive leveling and curing kinetics and the film spatial distribution formation mechanism were extracted through coupled tensor decomposition. The process parameters are expanded into a time dimension by time series and into a station dimension by station sequence. The time dimension and station dimension are combined to construct a time series tensor. The film thickness distribution and surface defect distribution are unfolded according to the film surface position to form a spatial position dimension, and the spatial position dimension is used to construct a spatial tensor; Establish the product association constraint between the time series tensor and the spatial tensor in the same batch. Based on the product association constraint in the same batch, perform coupled tensor decomposition on the time series tensor and the spatial tensor. Obtain the time factor matrix and the workstation factor matrix from the time series tensor, and obtain the spatial location factor matrix from the spatial tensor. The time factor matrix, workstation factor matrix, and spatial location factor matrix are coupled and mapped to extract the spatiotemporal coupling factor that characterizes the curing kinetics of adhesive flow and the formation mechanism of film spatial distribution.

3. The method according to claim 1, characterized in that, Based on the spatiotemporal coupling factor, the predicted distribution is reconstructed, and the residuals between the actual and predicted distributions are back-projected onto the time series tensor to identify anomalous process time segments and corresponding anomalous workstations that generate residuals. The spatial tensor is mapped back to the temporal tensor space by the spatiotemporal coupling factor to generate the reconstructed temporal tensor. The predicted film thickness distribution and the predicted surface defect distribution are obtained by forward deduction based on the reconstructed temporal tensor. The predicted film thickness distribution and the predicted surface defect distribution are used as the predicted distribution. The deviations between the actual film thickness distribution and the predicted film thickness distribution, as well as the deviations between the actual surface defect distribution and the predicted surface defect distribution, are constructed as residual distributions. Spatial gradient analysis is performed on the residual distribution to extract the residual concentration region. The residual concentration region is then back-projected onto the time series tensor through the spatiotemporal coupling factor to obtain the residual distribution on the time series tensor. The process parameter fluctuation period with the highest coupling strength is located from the distribution of residuals on the time tensor. The process parameter fluctuation period is identified as the abnormal process time segment that generates residuals, and the station where the abnormal process time segment is located is identified as the corresponding abnormal station.

4. The method according to claim 1, characterized in that, Based on the spatiotemporal coupling factor, the constraint changes on the leveling time and curing rate of the adhesive at downstream stations are calculated by adjusting abnormal workstations. Within the process feasibility domain defined by the constraint changes, a multi-workstation collaborative adjustment scheme is solved and implemented, including: Based on the coupling relationship between the abnormal workstation and the downstream workstation in the spatiotemporal coupling factor, the influence of the abnormal workstation adjustment on the adhesive leveling time and the curing rate of the downstream workstation through the adhesive leveling and curing dynamics are calculated. The influence on the adhesive leveling time and the curing rate of the downstream workstation are used as constraint change quantities. The constraint change is applied to the original process constraint of the downstream station to obtain the updated process constraint of the downstream station, and the updated process constraint is used as the process feasible domain. Within the feasible process domain, the process parameter adjustment amounts of abnormal stations and downstream stations are used as optimization variables to solve the problem, and the combination of process parameter adjustment amounts of abnormal stations and downstream stations is obtained. The combination of process parameter adjustment amounts is used as a multi-station collaborative adjustment scheme. The multi-station collaborative adjustment plan is issued to the process equipment corresponding to the abnormal station and downstream station for execution.

5. The method according to claim 1, characterized in that, The post-execution spatial tensor is constructed by collecting the film thickness distribution and surface defect distribution. The spatiotemporal coupling factor is updated based on the difference between the post-execution spatial tensor and the spatial tensor, including: The film thickness distribution and surface defect distribution after the multi-station collaborative adjustment scheme is executed are collected, and the film thickness distribution and surface defect distribution after execution are stacked into a spatial tensor after execution. The difference between the spatial tensors after calculation is used to obtain the actual spatial response. The expected spatial response is calculated based on the multi-station collaborative adjustment scheme and the spatiotemporal coupling factor. The deviation between the actual spatial response and the expected spatial response is used as the coupling deviation. Spatial decomposition of coupling deviation is performed to extract the spatial components associated with abnormal workstations and the spatial components associated with downstream workstations in the coupling deviation. The coupling correction coefficients of abnormal workstations and downstream workstations are calculated respectively. The coupling correction coefficients of the abnormal workstation and the downstream workstation are applied to the coupling strengths corresponding to the abnormal workstation and the downstream workstation in the spatiotemporal coupling factor, respectively, to obtain the updated spatiotemporal coupling factor.

6. The method according to claim 1, characterized in that, When a sudden change in process parameters at the coating station is detected, the change is substituted into the updated spatiotemporal coupling factor for forward propagation. This predicts the film thickness distortion region and surface defect propagation region caused by the change at downstream stations. Before the distortion occurs, a pre-compensation adjustment instruction is issued to the downstream stations, including: Real-time monitoring of process parameter changes at the coating station; when the change in process parameters exceeds the preset fluctuation range, it is identified as a sudden change in process parameters at the coating station, and the magnitude and timing of the sudden change in process parameters at the coating station are obtained. The mutation amplitude is substituted into the updated spatiotemporal coupling factor for forward propagation to obtain the propagation response of the mutation in the film thickness distribution and surface defect distribution of the downstream station. The propagation time of the mutation to the downstream station is calculated based on the updated spatiotemporal coupling factor. Spatial localization is performed on the propagation response of the film thickness distribution and the propagation response of the surface defect distribution at the downstream station. The film thickness distortion region and the surface defect expansion region caused by the sudden change at the downstream station are extracted. The pre-compensation process parameter adjustment amount of the downstream station is calculated based on the film thickness distortion region and the surface defect expansion region. The time of distortion formation is determined based on the time of mutation and the propagation time. Before the time of distortion formation, the pre-compensation process parameter adjustment amount is issued to the corresponding process equipment of the downstream station as a pre-compensation adjustment instruction.

7. The method according to claim 6, characterized in that, Spatially locate the propagation response on the film thickness distribution and the propagation response on the surface defect distribution at the downstream station, extract the film thickness distortion region and surface defect propagation region caused by the abrupt change at the downstream station, and calculate the pre-compensation process parameter adjustment amount for the downstream station based on the film thickness distortion region and the surface defect propagation region, including: Spatial gradient analysis is performed on the propagation response of the membrane thickness distribution at the downstream workstation. Continuous regions where the spatial gradient exceeds the preset fluctuation threshold are spatially located and marked as membrane thickness distortion regions. The spatial location and distortion amplitude of the membrane thickness distortion regions are extracted. Spatial boundary identification is performed on the propagation response of surface defects in downstream workstations. The newly added area that expands outward from the original surface defect boundary is spatially located and marked as the surface defect expansion area. The spatial location and expansion scale of the surface defect expansion area are extracted. The spatial location and distortion amplitude of the film thickness distortion region are mapped to the downstream station process parameter adjustment space to obtain the process parameter compensation amount used to offset the distortion amplitude. The spatial location and expansion scale of the surface defect expansion region are mapped to the downstream station process parameter adjustment space to obtain the process parameter compensation amount used to suppress the expansion scale. The pre-compensation process parameter adjustment amount for downstream stations is obtained by spatially superimposing the process parameter compensation amount used to offset the distortion amplitude and the process parameter compensation amount used to suppress the expansion scale.

8. A protective film production monitoring system based on the Industrial Internet, used to implement the method described in any one of claims 1-7, characterized in that, The system includes: The data acquisition module is used to collect process parameters, film thickness distribution, and surface defect distribution of the protective film production line. The coupling factor extraction module is used to construct process parameters as temporal tensors and film thickness distribution and surface defect distribution as spatial tensors. It extracts spatiotemporal coupling factors that characterize the flow curing dynamics of adhesive and the formation mechanism of film spatial distribution through coupling tensor decomposition. The anomaly identification module is used to reconstruct the predicted distribution based on the spatiotemporal coupling factor, back-project the residual between the actual distribution and the predicted distribution onto the time tensor, and identify the abnormal process time segment that generates the residual and the corresponding abnormal work station. The collaborative adjustment module is used to calculate the constraint changes of abnormal station adjustments on the leveling time and curing rate of adhesive in downstream stations based on the spatiotemporal coupling factor, and to solve and execute multi-station collaborative adjustment schemes within the process feasible domain limited by the constraint changes. The factor update module is used to collect the film thickness distribution and surface defect distribution after execution to construct the spatial tensor after execution, and update the spatiotemporal coupling factor based on the difference between the spatial tensor after execution and the spatial tensor. The feedforward compensation module is used to detect sudden changes in process parameters at the coating station, substitute the sudden change into the updated spatiotemporal coupling factor for forward propagation, predict the film thickness distortion area and surface defect expansion area caused by the sudden change at the downstream station, and issue a pre-compensation adjustment command to the downstream station before the distortion occurs.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions that, when executed by a processor, implement the steps of the method as described in any one of claims 1 to 7.

Images