Data monitoring-based pressure-sensitive adhesive online production analysis method and system

Abstract

The present application relates to the technical field of pressure sensitive adhesive production monitoring, and particularly relates to a pressure sensitive adhesive online production analysis method and system based on data monitoring. The method comprises the following steps: deploying a coating data monitoring network on a coating production line; synchronously obtaining coating real-time monitoring data containing coating thickness data and coating pressure data online through the coating data monitoring network; performing horizontal and vertical decoupling analysis and pressure disturbance feature based on the coating thickness data in the coating real-time monitoring data, then determining the defect source type, and generating corresponding vertical control instructions, horizontal control instructions or step-by-step cooperative control instructions according to the defect source type, so as to realize real-time automatic control of the horizontal and vertical thickness of the wide-width pressure sensitive adhesive coating. The present application realizes accurate determination of the defect source through online correlation diagnosis of the coating thickness and process pressure data, so as to complete real-time regulation of the horizontal and vertical thickness.

Description

Technical Field

[0001] This invention relates to the field of pressure-sensitive adhesive production monitoring technology, and in particular to a method and system for online production analysis of pressure-sensitive adhesives based on data monitoring. Background Technology

[0002] Throughout the entire lifecycle of pressure-sensitive adhesive products, the thickness and uniformity of the adhesive coating are core physical indicators that determine their final performance, such as bond strength, peel force, holding power, weather resistance, and optical properties. If the adhesive layer is too thin, it will directly lead to insufficient adhesion, causing quality issues such as detachment and edge curling during application, seriously affecting the reliability and safety of the end product. Conversely, if the adhesive layer is too thick, it will not only cause excessive consumption of adhesive materials, significantly increasing production costs, but also lead to a series of process problems such as adhesive overflow, residual adhesive, die-cutting difficulties, and incomplete curing / drying. However, traditional pressure-sensitive adhesive coating layer inspection involves sampling at specific locations after a roll of product is produced, and the operator uses this offline sampling method to measure the thickness. This method cannot capture in real time typical lateral distribution defects such as "center bulge," "edge thickening," or "stripe-like" defects caused by localized die blockage, uneven temperature, or bolt deformation. This method has a significant time lag and cannot achieve real-time control during the process. Summary of the Invention

[0003] Based on this, the present invention provides a method and system for online production analysis of pressure-sensitive adhesives based on data monitoring, in order to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above objectives, a data-monitoring-based online production analysis method for pressure-sensitive adhesives is applied to a coating production line for pressure-sensitive adhesives. The coating production line includes an adhesive supply pump and a coating die. The data-monitoring-based online production analysis method for pressure-sensitive adhesives includes the following steps:

[0005] Step S1: Deploy a coating data monitoring network on the coating production line; acquire real-time coating monitoring data, including coating thickness data and coating pressure data, online synchronously through the coating data monitoring network;

[0006] Step S2: Perform horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data to obtain the horizontal morphological characteristics and vertical fluctuation characteristics respectively; analyze the pressure main frequency data and die head steady-state pressure data based on the coating pressure data in the real-time coating monitoring data.

[0007] Step S3: Use the pressure main frequency data and the steady-state pressure data of the mold head as pressure disturbance characteristics, and then use them with the lateral morphology characteristics and longitudinal fluctuation characteristics to determine the defect source type, and generate corresponding longitudinal control commands, lateral control commands or step-by-step coordinated control commands according to the defect source type.

[0008] Step S4: Send the longitudinal control command, lateral control command, or step-by-step collaborative control command to the production line programmable logic controller for execution, so as to realize real-time automatic control of the longitudinal and lateral thickness of the wide pressure-sensitive adhesive coating.

[0009] This invention also provides a data-monitoring-based online production analysis system for pressure-sensitive adhesives, which executes the data-monitoring-based online production analysis method for pressure-sensitive adhesives as described above. The data-monitoring-based online production analysis system for pressure-sensitive adhesives includes:

[0010] The online synchronous monitoring module is used to deploy a coating data monitoring network on the coating production line; it acquires real-time coating monitoring data, including coating thickness data and coating pressure data, online synchronously through the coating data monitoring network.

[0011] The horizontal and vertical decoupling analysis module is used to perform horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data, and obtain the horizontal morphological characteristics and vertical fluctuation characteristics respectively; it analyzes the pressure main frequency data and die head steady-state pressure data based on the coating pressure data in the real-time coating monitoring data.

[0012] The defect source judgment module is used to use the pressure main frequency data and the steady-state pressure data of the mold head as pressure disturbance characteristics, and then use them with the lateral morphology characteristics and longitudinal fluctuation characteristics to determine the defect source type, and generate corresponding longitudinal control instructions, lateral control instructions or step-by-step coordinated control instructions according to the defect source type.

[0013] The instruction control module is used to send longitudinal control instructions, lateral control instructions, or step-by-step collaborative control instructions to the production line programmable logic controller for execution, so as to realize real-time automatic control of the longitudinal and lateral thickness of the wide pressure-sensitive adhesive coating.

[0014] The beneficial effects of this invention are as follows:

[0015] On the one hand, traditional offline sampling or single-point online monitoring methods cannot perceive the overall morphology of the coating, and easily confuse stable defects such as lateral "center bulges" and "edge thickenings" with longitudinal periodic fluctuations, resulting in unclear defect identification. This invention, by constructing a thickness data matrix and performing decoupled calculations, can effectively separate defects from different sources and with different manifestations. It can not only identify the type of lateral defects (such as center bulges and edge thickenings), but also accurately calculate their influence range (lateral defect span) and severity (local extreme point amplitude). At the same time, it captures the dominant frequency and energy distribution of longitudinal fluctuations, greatly improving the process perception and control capability of lateral thickness uniformity.

[0016] On the other hand, based on accurate defect source determination results, this invention generates and automatically executes targeted control instructions, significantly improving the quality of pressure-sensitive adhesive products, material utilization, and production efficiency. Traditional manual adjustment suffers from lag and limited precision, failing to meet the production requirements of high-speed, wide-width, and precision coating. After determining the defect source type, this invention can automatically calculate the optimal adjustment scheme—for longitudinal sources, it precisely adjusts the glue supply pump speed; for transverse sources, it locates the target bolt and calculates the adjustment sequence; for mixed sources, it executes a step-by-step collaborative control strategy. These instructions are directly sent to the production line PLC via the industrial network, achieving real-time, rapid, and closed-loop automatic control of the coating's longitudinal and transverse thicknesses. This system not only effectively suppresses various thickness unevenness defects, significantly improving product quality consistency and pass rate, but also reduces unnecessary material redundancy by precisely controlling the lower thickness limit, reducing material waste caused by thickness deviations, and improving the overall operating efficiency of the production line by reducing manual intervention and downtime for debugging. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the steps in the online production analysis method for pressure-sensitive adhesives based on data monitoring according to the present invention;

[0018] Figure 2 This is a schematic diagram of the modules of the pressure-sensitive adhesive online production analysis system based on data monitoring according to the present invention;

[0019] Figure 3 This is an example diagram of a coating machine model for the pressure-sensitive adhesive in this invention;

[0020] Figure 4 This is a diagram illustrating a production example of the pressure-sensitive adhesive in this invention.

[0021] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0023] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0024] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0025] To achieve the above objectives, please refer to Figures 1 to 4 This invention provides a data-monitoring-based online production analysis method for pressure-sensitive adhesives, applied to a coating production line for pressure-sensitive adhesives. The coating production line includes an adhesive supply pump and a coating die. The data-monitoring-based online production analysis method for pressure-sensitive adhesives includes the following steps:

[0026] Step S1: Deploy a coating data monitoring network on the coating production line; acquire real-time coating monitoring data, including coating thickness data and coating pressure data, online synchronously through the coating data monitoring network;

[0027] In this embodiment of the invention, a sensor network is first deployed at key locations in the pressure-sensitive adhesive coating production line. Specifically, an array containing 64 laser thickness sensors is deployed at equal intervals along the width direction of the coating die outlet. A high-frequency pressure sensor is installed on the outlet pipe of the adhesive supply pump. Simultaneously, eight pressure sensors are distributed and installed inside the coating die along its width direction in the melt distribution manifold. In addition, a rotary encoder is installed on the substrate traction drive roller of the coating production line to measure the real-time linear velocity. All sensors are connected to a data acquisition server via an industrial Ethernet network. The server registers the physical location, measurement dimension, and data format of each sensor, forming a structured coating data monitoring network.

[0028] In one implementation of this invention, when the production line programmable logic controller sends a production start signal, the data acquisition server receives the signal as a trigger instruction for data acquisition. The server first obtains the real-time coating speed through the rotary encoder. Assuming the preset sampling speed threshold is 100 meters per minute, if the real-time coating speed is greater than 100 meters per minute, the server uniformly adjusts the sampling frequency of all sensors to a high-frequency sampling mode of 200 Hz; otherwise, it adopts the standard sampling mode of 50 Hz. The action cycle of the laser thickness sensor array completing one full-width data acquisition, i.e., 5 milliseconds, is used as the reference acquisition cycle. Within each reference acquisition cycle, the server synchronously captures and packages the data of all sensors through the network according to the current sampling mode, forming a timestamp data frame containing thickness, pressure, and speed, which serves as the real-time coating monitoring data.

[0029] Step S2: Perform horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data to obtain the horizontal morphological characteristics and vertical fluctuation characteristics respectively; analyze the pressure main frequency data and die head steady-state pressure data based on the coating pressure data in the real-time coating monitoring data.

[0030] In this embodiment of the invention, a thickness data matrix of 1000 rows by 64 columns is constructed based on thickness data within 1000 consecutive reference acquisition cycles, i.e., 5 seconds. The data in each column of the thickness data matrix is ​​averaged to obtain a horizontal thickness baseline containing 64 elements. The residual matrix is ​​obtained by subtracting the average value of its column from each value in the horizontal thickness baseline.

[0031] In one implementation of this invention, the first and second derivatives of the transverse thickness baseline are calculated to identify its shape. Assuming a significant local maximum is found at the 30th to 34th column measuring points, with an amplitude 5 micrometers higher than the average thickness, and inflection points are located at the 25th and 38th column measuring points on both sides of it, this shape is classified as a central convex type. The transverse influence range parameter of the defect, i.e., the transverse defect span, is the physical distance of 130 millimeters between the 38th and 25th measuring points. Finally, the central convex type, amplitude of 5 micrometers, and span of 130 millimeters are packaged into transverse morphological features.

[0032] In another implementation of this invention, a fast Fourier transform is performed on each of the 64 longitudinal residual sequences in the residual matrix to calculate their power spectral density in the range of 0.1 Hz to 100 Hz. It is assumed that in the power spectra of multiple sequences, a peak frequency point exceeding the preset noise floor is identified near 15 Hz, with an average amplitude of -25 dB. All identified peak frequency points at 15 Hz and their corresponding amplitudes of -25 dB are associated as longitudinal fluctuation features.

[0033] In another implementation of this invention, the pump outlet pressure sequence measured by the glue supply pump outlet pressure sensor and the internal pressure matrix of the die head measured by eight pressure sensors inside the die head are separated from the real-time coating monitoring data. A fast Fourier transform is performed on the pump outlet pressure sequence, and a significant peak is found near 15 Hz, which is used as the pressure main frequency data. The internal pressure matrix of the die head is averaged in the time dimension to obtain a vector containing eight pressure values, which represents the lateral distribution of steady-state pressure inside the die head and is used as the steady-state pressure data of the die head.

[0034] Step S3: Use the pressure main frequency data and the steady-state pressure data of the mold head as pressure disturbance characteristics, and then use them with the lateral morphology characteristics and longitudinal fluctuation characteristics to determine the defect source type, and generate corresponding longitudinal control commands, lateral control commands or step-by-step coordinated control commands according to the defect source type.

[0035] In this embodiment of the invention, the peak frequency of 15 Hz in the longitudinal fluctuation feature is compared with the pressure main frequency data of 15 Hz. Assuming that the preset frequency matching threshold is 1 Hz, since the frequency difference between the two is 0 Hz which is less than the threshold, a longitudinal source association identifier is generated. Then, the local extreme point position in the transverse morphological feature, namely the 30th to 34th column measuring points, is spatially matched with the steady-state pressure data of the die head. Assuming that the positions of the 4th and 5th pressure sensors in the die head correspond exactly to the 28th to 36th column measuring point area on the coating width, and the steady-state pressure values ​​of these two sensors are significantly higher than those of other sensors, the positions are determined to be corresponding, and a transverse source association identifier is generated.

[0036] In one implementation of this invention, since both the longitudinal source association identifier and the transverse source association identifier are generated, the current defect source is determined to be a mixed source. At this time, the longitudinal and transverse control calculations will be started simultaneously. The longitudinal calculation module calculates that the glue supply pump speed needs to be reduced by 0.5 percent based on the amplitude of the longitudinal fluctuation feature of -25 dB through a preset control model. The transverse calculation module locates the mold head bolt that needs to be adjusted as bolt No. 16 corresponding to the 32nd measuring point based on the transverse morphological features of a central convex shape, an amplitude of 5 micrometers, and a span of 130 mm. Based on the model, it calculates that bolt No. 16 needs to be tightened by one-eighth of a turn, and its adjacent bolts No. 15 and No. 17 need to be tightened by one-sixteenth of a turn. These two independent calculation results are integrated into a step-by-step collaborative control command.

[0037] Step S4: Send the longitudinal control command, lateral control command, or step-by-step collaborative control command to the production line programmable logic controller for execution, so as to realize real-time automatic control of the longitudinal and lateral thickness of the wide pressure-sensitive adhesive coating.

[0038] In this embodiment of the invention, since the defect source is determined to be a mixed source, a step-by-step collaborative control instruction will be executed. This instruction is encapsulated into a set of logical instructions containing the execution order and conditions, and sent to the production line programmable logic controller.

[0039] In one implementation of this invention, after receiving an instruction, the programmable logic controller (PLC) first executes the first step, which is to reduce the speed setting of the glue supply pump by 0.5%. After executing the longitudinal adjustment instruction, the PLC enters a preset stabilization waiting period of 10 seconds. During this period, it continuously monitors the longitudinal fluctuation characteristics. When the peak frequency amplitude of 15 Hz is detected to drop below the noise floor or the waiting period ends, the PLC automatically triggers the second step of the instruction, which is to drive the adjustment motor of the No. 16 die head bolt to tighten by one-eighth of a turn, and at the same time drive the motors of the No. 15 and No. 17 bolts to tighten by one-sixteenth of a turn. The whole process realizes that the global longitudinal fluctuation is first processed, and the local lateral distribution is fine-tuned after stabilization, thereby completing a precise, stable and automated closed-loop control.

[0040] Preferably, in step S1, the real-time coating monitoring data, which includes coating thickness data and coating pressure data, is acquired online synchronously through the coating data monitoring network, including:

[0041] Receive the production batch start signal from the coating production line controller as the trigger command for data acquisition;

[0042] Based on trigger commands, coating speed data is acquired in real time through the coating speed encoder in the coating data monitoring network;

[0043] Determine if the coating speed data is greater than the preset sampling speed threshold; if so, adjust the sampling frequency of all sensors in the coating data monitoring network to the preset high-frequency sampling mode; otherwise, adjust to the standard sampling mode of the standard frequency.

[0044] The time it takes for the laser thickness sensor array in the coating data monitoring network to complete one full-width scan is used as the benchmark acquisition time.

[0045] Within each baseline acquisition cycle, based on the adjusted sampling frequency mode, real-time coating monitoring data is acquired online synchronously through the coating data monitoring network; among which, the real-time coating monitoring data includes coating thickness data measured by the laser thickness sensor array and coating pressure data from the pressure sensor.

[0046] In one implementation of this invention, when the programmable logic controller on the coating production line starts to execute a new production task, it sends a digital start signal containing a unique batch number to the data acquisition server. After receiving this signal, the monitoring program inside the data acquisition server parses it into a start trigger instruction for the data acquisition process and immediately starts the subsequent data acquisition and analysis modules.

[0047] Specifically, coating speed data is acquired in real time through the coating speed encoder in the coating data monitoring network based on trigger commands.

[0048] In another implementation of this invention, a coating speed encoder installed on the drive roller of the coating production line continuously generates pulse signals as the roller rotates. A data acquisition server collects these pulse signals in real time and calculates the current real-time coating speed according to the following formula:

[0049] ;

[0050] in, This represents the real-time coating speed, measured in meters per minute. Pi; The diameter of the drive roller is a pre-measured fixed parameter of the equipment, for example, 0.5 meters; This is the rotational speed per minute of the drive roller calculated using the pulse frequency.

[0051] It should be noted that if the coating speed data is greater than the preset sampling speed threshold, the sampling frequency of all sensors in the coating data monitoring network will be adjusted to the preset high-frequency sampling mode; otherwise, it will be adjusted to the standard sampling mode of the standard frequency.

[0052] In one implementation of this invention, the preset sampling speed threshold is determined based on historical production data and expert experience in a product quality control standard library. For example, analysis reveals that when the coating speed exceeds 120 meters per minute, the impact of high-frequency thickness fluctuations in the coating caused by minor equipment vibrations on product quality increases significantly. Therefore, the threshold is set to 120 meters per minute. When the real-time coating speed calculated by the server, for example, 150 meters per minute, is greater than this threshold, the server broadcasts instructions to all sensors in the monitoring network to uniformly set their sampling frequency to a high-frequency sampling mode of 200 Hz. If the calculated speed is 80 meters per minute, the sampling frequency is set to a standard sampling mode of 50 Hz.

[0053] It should be noted that the time it takes for the laser thickness sensor array in the coating data monitoring network to complete one full-width scan is used as the benchmark acquisition period.

[0054] In another implementation of this invention, the deployed laser thickness sensor array is a fixed array, and its internal controller is set to a fixed electronic acquisition cycle, such as 5 milliseconds. That is, every 5 milliseconds, all 64 sensors in the array will simultaneously complete a thickness measurement and output. The data acquisition server defines this 5 milliseconds as the benchmark acquisition cycle of the entire data monitoring network, and the subsequent data synchronous acquisition of all sensors will strictly follow this time rhythm.

[0055] Specifically, within each baseline acquisition cycle, based on the adjusted sampling frequency mode, real-time coating monitoring data is acquired online synchronously through the coating data monitoring network; among which, the real-time coating monitoring data includes coating thickness data measured by the laser thickness sensor array and coating pressure data from the pressure sensor.

[0056] In one implementation of this invention, at the arrival of each 5-millisecond baseline acquisition cycle, the data acquisition server, based on the current sampling mode (high frequency or standard mode), sends a synchronous acquisition trigger signal via industrial Ethernet to all registered sensors, including 64 laser thickness sensors, 1 glue pump outlet pressure sensor, and 8 die head internal pressure sensors. Upon receiving the signal, each sensor immediately performs measurement and transmits the data back. The server packages all data received within the same cycle into a data frame. This data frame contains a timestamp accurate to milliseconds, as well as 64 coating thickness values, 9 coating pressure values, and real-time coating speed values ​​corresponding to that moment. This complete data frame constitutes the real-time coating monitoring data.

[0057] Of particular importance is the deployment of a coating data monitoring network on the coating production line, including:

[0058] A laser thickness sensor array, a pressure sensor, and a coating speed encoder are deployed on the coating production line. The laser thickness sensor array is deployed at equal intervals along the coating width direction, the pressure sensors are installed at the glue supply pump outlet and distributed along the melt flow channel inside the coating die, and the coating speed encoder is installed on the drive roller of the coating production line.

[0059] The laser thickness sensor array, pressure sensor, and coating speed encoder are connected to the data acquisition server via industrial Ethernet.

[0060] The data acquisition server registers the physical node location, measurement dimension, and data format of each sensor to form a coating data monitoring network containing a network topology.

[0061] In one implementation of this invention, assuming a pressure-sensitive adhesive coating production line with a width of 1600 mm, a fixed crossbeam spanning the entire width is installed directly above the substrate. Sixteen non-contact laser thickness sensor probes are then installed on this crossbeam at 100 mm intervals, starting from the operating side of the production line, forming a sensor array capable of covering the entire width. Simultaneously, a high-temperature melt pressure sensor is installed at the outlet of the glue supply gear pump on the glue supply pipeline between the extruder and the coating die. Furthermore, five miniature high-temperature melt pressure sensors are drilled and installed at five key locations along the width direction in the melt distribution channel inside the coating die: near the two side edges, at one-quarter of the width, and at the center. Finally, an incremental rotary encoder is connected to the shaft end of the main drive roller of the production line via a coupling to ensure its synchronous rotation with the follower roller.

[0062] In one implementation of this invention, to achieve high-speed and reliable communication between each sensor and the server, an independent industrial Ethernet switch is deployed on-site. All 16 laser thickness sensors, 6 pressure sensors, and coating speed encoders installed are connected to different ports of the switch via shielded industrial-grade Ethernet cables. The data acquisition server is also connected to the switch via a network cable, thereby physically constructing a star-shaped local area network topology.

[0063] It should be noted that this network uses Time-Sensitive Networking (TSN) technology to ensure low latency and high synchronization accuracy in data transmission, which is crucial for the synchronous acquisition and correlation analysis of multi-source data in subsequent steps.

[0064] In one implementation of this invention, during the initial system configuration, a unique device identifier is created for each sensor and its detailed attributes are entered using dedicated configuration software running on the data acquisition server. Specifically, for example, the 32nd laser thickness sensor in the width direction is registered with a physical coordinate of 800 mm in the lateral position, a measurement dimension of "thickness", a data format of "32-bit single-precision floating-point number", and a unit of "micrometer"; the pressure sensor installed at the center of the die head is registered with a physical coordinate of 800 mm in the lateral position, a measurement dimension of "pressure", a data format of "16-bit unsigned integer", and a unit of "kilopascal". After completing the registration of information for all 74 sensors, a device registry is automatically generated. This registry logically constitutes the topology of the coating data monitoring network.

[0065] Preferably, before performing the horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data in step S2, the method further includes:

[0066] Based on the coating thickness data in the real-time coating monitoring data, a thickness data matrix is ​​constructed with the lateral measuring points of the laser thickness sensor array as columns and the acquisition time as rows;

[0067] The coating thickness data in each column of the thickness data matrix is ​​averaged to obtain the steady-state thickness value;

[0068] The steady-state thickness values ​​are arranged in order of the physical location of the transverse measuring points, and used as the transverse thickness baseline to characterize the transverse stability non-uniformity of the coating.

[0069] In one implementation of this invention, a time window, such as 100 seconds, is set, corresponding to 1000 reference acquisition cycles. Within this time window, the one-dimensional array containing the thickness values ​​of 16 measurement points obtained in each acquisition cycle is filled into a two-dimensional array structure row by row in chronological order, ultimately forming a thickness data matrix with a dimension of 1000 rows by 16 columns. Each row of the matrix represents a specific acquisition time, and each column uniquely corresponds to a measurement point of a laser thickness sensor with a fixed physical location. This transforms discrete time-series data into a structured data entity that can simultaneously characterize time and spatial distribution.

[0070] Specifically, the 16 columns of data in the matrix are traversed, and an arithmetic mean is calculated for the 1000 thickness data points in each column to filter out random noise and high-frequency longitudinal fluctuations during the production process, thereby extracting the relatively stable average thickness value at that lateral position, i.e., the steady-state thickness value. Specifically, for the matrix... The formula for calculating the steady-state thickness of the column is as follows:

[0071] ;

[0072] in, Representing the Steady-state thickness values ​​at each measuring point; This represents the total number of collection periods within the time window, which is 1000 in this case. The thickness data matrix represents the first... line, number The thickness value of the column; This indicates summing all elements in the column. The value ranges from 1 to For example, if the sum of the 1000 thickness data points in the column corresponding to the first measurement point is 25150 micrometers, then the steady-state thickness value of that measurement point is 25150 / 1000 = 25.15 micrometers.

[0073] In one implementation of this invention, the 16 steady-state thickness values ​​calculated in the previous step are arranged strictly according to the physical installation order of their corresponding laser thickness sensors from the production line operation side to the drive side, forming a one-dimensional array containing 16 elements, which is the transverse thickness baseline.

[0074] Preferably, the horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data in step S2 includes:

[0075] Calculate the first and second derivatives of the transverse thickness baseline to identify local extrema and inflection points;

[0076] Based on the location and magnitude of local extreme points, the morphology of the lateral thickness baseline is classified into one of the following: central convex type, edge thickening type, or periodic concave type, thus obtaining the lateral morphology classification.

[0077] Calculate the span of lateral defects based on their lateral morphology classification;

[0078] The classification of lateral morphology, the magnitude of local extreme points, and the span of lateral defects are correlated as lateral morphological features.

[0079] In one implementation of this invention, since the transverse thickness baseline is composed of 16 discrete measurement points, the central difference method is used to approximate its first and second derivatives. Specifically, for the first derivative of the transverse thickness baseline... The formula for calculating the first derivative at each measurement point is: First derivative The formula for calculating its second derivative is: Second derivative ;in, Representing the first Steady-state thickness values ​​at each measuring point This represents the physical distance between adjacent measuring points, which is 100 mm in this invention. Local extreme points are identified by determining whether the first derivative changes sign. For example, if the first derivative... For positive, the first derivative If it is negative, then the first The measurement point was identified as a local maximum point.

[0080] In one implementation of this invention, the morphology classification is based on a set of preset rule bases. These rule bases are derived from a large amount of historical production data and process experience. For example, the rules are defined as follows: if a local maximum point with an amplitude exceeding a preset threshold (e.g., 5% of the average thickness) is identified only in the central area of ​​the coating width (defined as between 25% and 75% of the total width), the morphology is classified as "central convex type"; if significant local maximum points are identified in the areas near the two side edges (e.g., between 0% and 15% and 85% and 100% of the total width), the morphology is classified as "edge thickening type".

[0081] In another implementation of the present invention, the method for calculating the transverse defect span is to search outwards from the identified local extreme point as the center until the boundary point where the thickness value returns to the preset stable range is found. The physical distance between these two boundary points is the transverse defect span. The preset stable range is defined as the average thickness value of the transverse thickness baseline plus or minus a small tolerance, such as 0.2 micrometers.

[0082] Specifically, the current morphology is classified as "central convex type," with its extreme point located at the 8th measuring point. Starting from the 8th measuring point, a traversal search is performed towards both the operating side (decreasing sequence number) and the driving side (increasing sequence number). Assuming that the thickness value first enters the stable range of 25.8 to 26.2 micrometers when searching to the 4th measuring point on the operating side, and the thickness value first enters this stable range when searching to the 12th measuring point on the driving side, then the transverse defect span is calculated as (12-4). 100 mm = 800 mm.

[0083] In one implementation of this invention, the precise amplitude of a local extreme point is calculated, which is defined as the difference between the steady-state thickness value of the extreme point and the average value of the entire transverse thickness baseline. For example, 27.2 micrometers - 26.0 micrometers = 1.2 micrometers. Then, the four pieces of information, namely transverse morphology classification (central convex type), extreme point location (8th measuring point), extreme point amplitude (1.2 micrometers), and transverse defect span (800 mm), are integrated into a data structure as transverse morphological features.

[0084] Preferably, the horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data in step S2 includes:

[0085] Centered on the local extreme point on which the horizontal morphology classification is based, a pair of adjacent inflection points are retrieved and located on both sides of it.

[0086] Calculate the lateral physical distance between the pair of inflection points, and use this distance as a parameter of the lateral influence range of the defect corresponding to the lateral morphology classification, as the lateral defect span.

[0087] In one implementation of this invention, the specific operation of this step is as follows: after identifying the local extreme point in the transverse thickness baseline (e.g., the maximum point located at the 8th measurement point), starting from this point, a point-by-point search is performed on the operation side (in the direction of decreasing sequence number) and the driving side (in the direction of increasing sequence number) to locate the inflection point. The inflection point is identified by judging whether the sign of the second derivative has changed. The second derivative has been calculated based on the central difference method in the previous step.

[0088] Specifically, starting from the 7th measurement point, the process traverses towards the operating side, checking the second derivative value of each measurement point. Assuming that the second derivative is positive at the 4th measurement point and becomes negative at the 3rd measurement point, it is determined that there is an inflection point between the 4th and 3rd measurement points, and its position is approximately recorded as the 4th measurement point. Similarly, starting from the 9th measurement point, the process traverses towards the driving side. If it is found that the sign of the second derivative changes between the 11th and 12th measurement points, the position of another inflection point is recorded as the 12th measurement point, thereby locating a pair of inflection points adjacent to the local extreme point.

[0089] In one implementation of this invention, the calculation of the transverse defect span is based on the positions of the two inflection points located in the previous step. The measurement point numbers corresponding to these two inflection points are read, and the total distance between adjacent measurement points is calculated using the following mathematical expression based on the known physical distance between them:

[0090] ;

[0091] in, For the span of the lateral defect; The measurement point number of the inflection point on the operating side; The measurement point number is the inflection point on the driving side; This is the physical distance between adjacent measuring points, which is 100 millimeters in this invention.

[0092] It should be noted that, mathematically, an inflection point represents the point where the concavity or convexity of a curve changes. Physically, the region between two inflection points can be considered the core defect region dominated by that local extremum. Therefore, the distance between them can effectively and quantitatively characterize the main influence range of the transverse defect in the width direction, providing a precise spatial basis for generating subsequent control instructions (e.g., determining the number of bolts to be adjusted). For example, if the two inflection points are located at the 4th and 12th measuring points respectively, the transverse defect span is calculated as (12-4). 100 mm = 800 mm.

[0093] Preferably, step S2, which involves performing horizontal and vertical decoupling analysis based on the coating thickness data from the real-time coating monitoring data, further includes:

[0094] Subtract the steady-state thickness value of the column corresponding to the horizontal thickness baseline from each data point in the thickness data matrix to generate a residual matrix that eliminates the influence of the inherent horizontal distribution.

[0095] The residual values ​​of each column in the residual matrix are arranged in chronological order to form a vertical residual sequence;

[0096] Perform a Fast Fourier Transform on each longitudinal residual sequence to calculate the power spectral density within a preset frequency range;

[0097] Identify the peak frequency points in the power spectral density where the amplitude exceeds the preset noise floor;

[0098] All identified peak frequency points and their corresponding amplitudes are used as longitudinal fluctuation features.

[0099] In one implementation of this invention, the specific calculation for this step is as follows: traverse each element in the aforementioned 1000-row by 16-column thickness data matrix and subtract the steady-state thickness value corresponding to the column containing that element from its value. This steady-state thickness value has been obtained in the calculation step of the transverse thickness baseline. By performing this subtraction operation on all 16000 data points in the matrix, a new matrix with the same dimensions of 1000 rows by 16 columns is generated, which is the residual matrix.

[0100] Specifically, the th in the residual matrix line, number Column elements The calculation formula is as follows:

[0101] ;

[0102] in, Represents the first element in the original thickness data matrix. line, number The thickness value of the column; Representing the The steady-state thickness values ​​at each measuring point; it should be noted that the values ​​in the residual matrix physically represent the instantaneous deviation of each measuring point from its own stable average level at each moment, thus effectively eliminating the static influence of uneven lateral distribution, allowing subsequent analysis to focus entirely on dynamic fluctuations along the time axis.

[0103] In one implementation of this invention, the residual matrix is ​​extracted column by column. Since each row of the residual matrix corresponds to a collection time point, the data in each column (a total of 1000 residual values) are naturally arranged in chronological order to form a time series representing the thickness fluctuation of the transverse measuring point, i.e., the longitudinal residual sequence. Through this operation, 16 independent longitudinal residual sequences are finally obtained, and each sequence uniquely corresponds to a measuring point of a laser thickness sensor.

[0104] Specifically, the Fast Fourier Transform algorithm is applied to each sequence containing 1000 data points to transform it from the time domain to the frequency domain, resulting in a series of complex spectral coefficients. Then, the square of the modulus of these spectral coefficients is calculated and normalized to obtain the power spectral density curve of the sequence within a preset frequency range (e.g., 0.1 Hz to 50 Hz). This curve describes the distribution of thickness fluctuation energy at different frequencies.

[0105] In one implementation of this invention, the noise base is pre-calibrated based on data collected from the production line under no-load or stable operating conditions. It represents the inherent random noise level of the system without a specific physical source. Peak search is performed on each power spectral density curve, and only those local maxima points with amplitudes significantly higher than the noise base are identified as valid peak frequency points.

[0106] Specifically, assuming the calibrated noise floor is 0.1, when analyzing the power spectral density curve at the 8th measurement point, a peak at 15 Hz was found with a power spectral density amplitude of 0.8, which is much greater than 0.1. Therefore, 15 Hz was identified as an effective peak frequency point. Another peak at 35 Hz had an amplitude of only 0.12, which is close to the noise floor, so it was judged as random noise and ignored.

[0107] In one implementation of this invention, all valid peak frequency points and their corresponding amplitudes identified from the power spectral density analysis of 16 longitudinal residual sequences are summarized and integrated. If multiple measurement points show significant peaks at the same or very close frequencies, this is considered a global fluctuation, and its average frequency and average amplitude are calculated. These filtered and integrated frequency points and amplitudes are encapsulated into a set as the final result characterizing the longitudinal fluctuation characteristics of the coating thickness during the current production process. For example, if all 16 measurement points detect significant peaks near 15 Hz, the final longitudinal fluctuation characteristics are recorded as: {peak frequency: 15.0 Hz, average amplitude: 0.8 Hz}.

[0108] Preferably, step S2, which involves analyzing the pressure frequency data and the die head steady-state pressure data based on the coating pressure data in the real-time coating monitoring data, includes:

[0109] Based on the coating pressure data in the real-time coating monitoring data, the pump outlet pressure sequence measured by the pressure sensor installed at the glue supply pump outlet and the internal pressure matrix of the die head measured by the pressure sensor installed inside the coating die head are separated.

[0110] Perform a fast Fourier transform on the pump outlet pressure sequence to calculate the power spectral density within a preset frequency range, and identify the peak frequency points whose amplitude exceeds the preset noise floor as the pressure main frequency data.

[0111] The pressure matrix inside the die head is averaged over time to obtain steady-state pressure data of the die head that characterizes the lateral distribution of pressure inside the coating die head.

[0112] In one implementation of this invention, data is separated from the synchronously acquired data frames based on the sensor identification registered in step S1. Specifically, 1,000 pressure values ​​measured by the pressure sensor identified as "glue pump outlet" within 1,000 consecutive reference acquisition cycles are arranged in chronological order to form a one-dimensional pump outlet pressure sequence. At the same time, the data measured by the eight pressure sensors identified as "mold head interior" are used to construct a 1,000-row by 8-column mold head interior pressure matrix, where each row represents a synchronous measurement value at a time point and each column represents a pressure time sequence at a fixed position.

[0113] It should be noted that a fast Fourier transform is performed on the pump outlet pressure sequence to calculate the power spectral density within a preset frequency range, and the peak frequency point whose amplitude exceeds the preset noise floor is identified as the pressure master frequency data.

[0114] In one implementation of this invention, a fast Fourier transform algorithm is applied to the aforementioned pump outlet pressure sequence containing 1000 data points to calculate its power spectral density within a preset frequency range, such as 0.1 Hz to 100 Hz. The frequency range is set based on the operating frequency and harmonic characteristics of the glue pump motor. The preset noise floor, such as -50 dB, is the background noise level of the pressure signal measured when the equipment is running stably. If a peak with an amplitude of -30 dB is found at 15 Hz, since its amplitude is higher than the preset noise floor, the 15 Hz frequency point is identified and recorded as the pressure main frequency data.

[0115] It is important to note that the pressure matrix inside the die head is averaged over time to obtain steady-state pressure data of the die head that characterizes the lateral distribution of pressure inside the coating die head.

[0116] In one implementation of this invention, an independent arithmetic mean is calculated for each column of the 1000-row by 8-column internal pressure matrix of the mold head to eliminate instantaneous fluctuations and extract steady-state features. The calculation formula is as follows:

[0117] ;

[0118] in, Representing the The steady-state pressure value of the pressure sensor inside the mold head. The value range is from 1 to 8; This represents the length of the time series, which is the number of rows in the matrix, 1000. The first element in the internal pressure matrix of the die head line, number The pressure value of the column, specifically, taking the calculation of the steady-state pressure of the 4th sensor as an example, is obtained by adding the 1000 pressure values ​​in the column and dividing by 1000. The result, for example, 5.2 MPa, is the steady-state pressure value at that location. Finally, a one-dimensional array containing 8 steady-state pressure values ​​is obtained, which serves as the steady-state pressure data of the mold head.

[0119] Preferably, in step S3, the pressure main frequency data and the mold head steady-state pressure data are used as pressure disturbance characteristics, and then the defect source type is determined by combining them with the lateral morphological characteristics and longitudinal fluctuation characteristics, including:

[0120] The peak frequency points in the longitudinal fluctuation characteristics are compared with the pressure main frequency data. If the frequency difference between the two is less than the preset frequency matching threshold, a longitudinal source association identifier is generated.

[0121] Spatial location matching is performed between the local extreme point locations in the lateral morphological features and the pressure anomaly points in the steady-state pressure data of the mold head. If the locations correspond, a lateral source association identifier is generated.

[0122] Determine the generation status of the vertical source association identifier and the horizontal source association identifier;

[0123] If only vertical source association identifiers are generated, the defect source is determined to be a vertical source; if only horizontal source association identifiers are generated, the defect source is determined to be a horizontal source; if both are generated, the defect source is determined to be a mixed source.

[0124] In one implementation of this invention, the peak frequency (e.g., 15.0 Hz) and pressure main frequency data (e.g., 14.95 Hz) in the longitudinal fluctuation characteristics are read respectively. The preset frequency matching threshold is set based on the resolution of the spectrum analysis and the measurement uncertainty, for example, 0.1 Hz, and the absolute difference between the two frequency values ​​is calculated.

[0125] Specifically, the calculation process is as follows: Frequency difference = |15.0-14.95| = ​​0.05 Hz; Since the calculated frequency difference of 0.05 Hz is less than the preset frequency matching threshold of 0.1 Hz, it is determined that the longitudinal fluctuation of the coating thickness and the pressure fluctuation at the outlet of the glue pump are highly correlated in frequency and have the same origin. Therefore, a Boolean flag bit, namely "Longitudinal Source Association Identifier", is generated in memory and its state is set to "True".

[0126] In one implementation of this invention, before spatial location matching, pressure anomalies are identified from the steady-state pressure data of the die head. Specifically, the average value of five steady-state pressure values ​​is calculated, and points that deviate from the average value by more than a preset threshold (e.g., 2% of the average value) are identified as pressure anomalies. Assuming the steady-state pressure data of the die head is [10.1, 10.3, 10.5, 10.2, 10.0] MPa, its average value is 10.22 MPa, and the threshold is 0.20 MPa, the third pressure value of 10.5 MPa located at the center deviates from the average value by 0.28 MPa, exceeding the threshold, and is therefore identified as a pressure anomaly.

[0127] Next, the location of the local extreme point in the lateral morphological features (e.g., the 8th measuring point) is matched with the location of the pressure anomaly point (the 3rd pressure measuring point) in physical space. According to the sensor registration information, the 8th thickness measuring point and the 3rd pressure measuring point are both located at the center of the coating die head. The two correspond completely in physical space. Therefore, it is determined that there is a direct spatial correlation between the lateral unevenness of the thickness and the uneven pressure distribution inside the die head. Another Boolean flag bit, namely "lateral source association identifier", is generated and its state is set to "true".

[0128] In one implementation of this invention, this step is a logical judgment process that checks the status of the two association identifiers generated in the previous two steps. In this example, the status of the "vertical source association identifier" is detected as "true", and the status of the "horizontal source association identifier" is also "true".

[0129] In one implementation of this invention, the final defect source type is determined based on the combined state of two associated identifiers and according to a preset determination rule. The determination rule is as follows: If the "vertical source associated identifier" is "true" and the "lateral source associated identifier" is "false", the defect source type is determined to be "vertical source". If the "vertical source associated identifier" is "false" and the "lateral source associated identifier" is "true", the defect source type is determined to be "lateral source". If the "vertical source associated identifier" is "true" and the "lateral source associated identifier" is also "true", the defect source type is determined to be "mixed source". It should be noted that in this example scenario, since both identifiers are "true", the coating quality problem in the current production process is ultimately determined to be a "mixed source". This clear diagnostic result will directly guide the next step in generating what type of control instruction.

[0130] Preferably, step S3, which generates corresponding vertical control instructions, horizontal control instructions, or step-by-step collaborative control instructions based on the defect source type, includes:

[0131] If the defect source is determined to be a longitudinal source, the adjustment amount of the glue supply pump speed is calculated based on the longitudinal fluctuation characteristics, and a longitudinal control command is generated.

[0132] If the defect source is a transverse source, the target bolt is located based on the position of the local extreme point in the transverse morphological characteristics, and the adjustment sequence of the target bolt and its adjacent bolts is calculated based on the amplitude of the local extreme point and the transverse defect span to generate transverse control instructions.

[0133] If the defect source is determined to be a mixed source, the calculations for both the vertical and horizontal sources are performed simultaneously to generate step-by-step coordinated control instructions.

[0134] In one implementation of this invention, a preset proportional-integral-derivative (PID) controller is used to calculate the adjustment amount. The input error e(t) of this controller is defined as the peak frequency amplitude in the longitudinal fluctuation characteristic. The output ΔN of the controller is the adjustment amount of the glue supply pump speed, and its calculation formula is as follows:

[0135] ;

[0136] in, For controller output at time The change , , These are the proportional, integral, and derivative coefficients, which are pre-tuned during the system debugging phase after step response testing and model identification on the production line. Specifically, assuming the peak amplitude of the currently identified longitudinal fluctuation feature is -25 dB, the PID controller calculates, based on this input error and historical error data, that the glue pump speed needs to be reduced by 0.5 percent. Then, this adjustment is encapsulated into a digital instruction containing the target device address "glue pump controller" and the adjustment value "-0.5 percent", which serves as the longitudinal control instruction.

[0137] In one implementation of this invention, a pre-configured physical location mapping table is queried. This table records in detail the correspondence between the serial numbers of 64 laser thickness sensors and the unique numbers of all adjustable bolts on the coating die. For example, when the local extreme point in the lateral morphological feature is located at the 32nd measuring point, the corresponding target bolt is located as bolt number 16 by looking up the table. Then, a bolt adjustment response model is invoked. This model is constructed based on finite element simulation and a large amount of experimental data. The input is the amplitude of the defect (5.2 micrometers) and the span (325 millimeters). The output is a set of adjustment sequences. For example, the model calculates that target bolt number 16 needs to be tightened by one-eighth of a turn, while its adjacent bolts number 15 and 17 are co-adjusting bolts and each needs to be tightened by one-sixteenth of a turn. This set of sequences containing bolt numbers and adjustment turns is generated as a lateral control command.

[0138] In one implementation of this invention, when the defect source is determined to be a mixed source, its internal calculation module will launch two independent calculation tasks in parallel, namely, simultaneously executing the aforementioned longitudinal source PID control calculation and lateral source bolt adjustment response model calculation. Specifically, the longitudinal fluctuation characteristics are input into the PID controller to obtain the speed adjustment amount of the glue supply pump, and the lateral morphological characteristics are input into the bolt adjustment model to obtain the adjustment sequence of the target bolt and its adjacent bolts. Then, the two calculation results are integrated into a step-by-step collaborative control instruction with internal execution logic. This instruction is designed as a sequence containing two steps: the first step is the longitudinal control instruction, and the second step is the lateral control instruction. A conditional waiting loop is inserted between the two steps to ensure the stability of the control process.

[0139] Of particular importance, if the defect source is determined to be a longitudinal source, generating the longitudinal control instruction includes the following steps:

[0140] Phase analysis was performed on the longitudinal fluctuation characteristics to determine the range of rotation angles of the glue supply pump corresponding to the peaks and troughs of the fluctuations.

[0141] Based on the amplitude information in the longitudinal fluctuation characteristics, the instantaneous drive current increment of the servo motor required to offset the thickness decrease within the rotation angle range corresponding to the trough is calculated.

[0142] Based on the amplitude information in the longitudinal fluctuation characteristics, the instantaneous drive current reduction of the servo motor required to offset the increase in thickness is calculated within the rotation angle range corresponding to the wave peak.

[0143] The instantaneous drive current increment and instantaneous drive current decrement are encapsulated into a dynamic current compensation sequence that is strictly synchronized with the rotation angle of the glue supply pump, and this sequence is used as the longitudinal control command.

[0144] In one implementation of this invention, a high-precision rotary encoder signal from the servo motor of the glue supply pump is invoked. This signal provides the real-time rotation angle of the glue supply pump from 0 to 359 degrees. Specifically, the two signals are precisely aligned by calculating the cross-correlation function between the longitudinal residual sequence of the coating thickness and the encoder angle sequence, and considering the transmission delay time of the melt from the pump outlet to the die lip. The phase relationship between the thickness fluctuation and the pump rotation angle is determined by analyzing the aligned data. For example, the analysis results show that the thickness trough mainly occurs in the range of 90 to 120 degrees when the pump rotates, while the thickness peak corresponds to the rotation range of 270 to 300 degrees.

[0145] It should be noted that, based on the amplitude information in the longitudinal fluctuation characteristics, the instantaneous drive current increment of the servo motor required to offset the thickness decrease is calculated within the rotation angle range corresponding to the trough.

[0146] In another implementation of this invention, a pre-established pump dynamic response model is used to calculate the required current increment. This model is established by applying a series of small current pulses of known amplitude to the glue supply pump during the commissioning phase and measuring its outlet pressure response. Its core is a compensation coefficient, and the specific calculation formula is as follows:

[0147] ;

[0148] in, This is the instantaneous increase in drive current that needs to be calculated, in amperes; It is the compensation coefficient of the model, representing the amount of current that needs to be compensated for per unit thickness deviation, for example, 0.1 amperes per micrometer. This coefficient is obtained from the above model calibration. This refers to the amplitude of the longitudinal fluctuation characteristic. For example, the equivalent time-domain amplitude obtained through power spectrum analysis is 2 micrometers. Calculated according to this formula... =0.1 2 = 0.2 amperes.

[0149] In one implementation of this invention, the calculation process is similar to that for calculating the current increment, also using the pump dynamic response model, but with the adjustment direction reversed. The calculation formula is as follows:

[0150] ;

[0151] Using the same parameters as described above, we can calculate =-0.1 2 = -0.2 Amperes. This negative value means that the original drive current of the servo motor needs to be reduced by 0.2 Amperes within the rotation angle range corresponding to the peak.

[0152] In one implementation of this invention, a data lookup table or function curve is generated. This table finely divides the 360-degree rotation cycle of the glue supply pump and maps the calculated current adjustment amount to the corresponding angle range. For example, the generated sequence specifies that: in the rotation angle range of 0 to 89 degrees, the current compensation is 0 amps; in the range of 90 to 120 degrees, the current compensation is +0.2 amps; in the range of 121 to 269 degrees, the compensation is 0 amps; in the range of 270 to 300 degrees, the compensation is -0.2 amps; and in the range of 301 to 359 degrees, the compensation returns to 0 amps. This dynamic current compensation sequence, which is strictly synchronized with the pump rotation angle, is sent to the servo driver of the glue supply pump as a feedforward compensation signal superimposed on the original speed loop control, thereby constituting the final longitudinal control command that can realize fine-grained flow regulation within a single rotation of the pump.

[0153] This invention also provides a data-monitoring-based online production analysis system for pressure-sensitive adhesives, which executes the data-monitoring-based online production analysis method for pressure-sensitive adhesives as described above. The data-monitoring-based online production analysis system for pressure-sensitive adhesives includes:

[0154] The online synchronous monitoring module S101 is used to deploy a coating data monitoring network on the coating production line; it acquires real-time coating monitoring data, including coating thickness data and coating pressure data, online synchronously through the coating data monitoring network.

[0155] The horizontal and vertical decoupling analysis module S102 is used to perform horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data, and obtain the horizontal morphological characteristics and vertical fluctuation characteristics respectively; and to analyze the pressure main frequency data and die head steady-state pressure data based on the coating pressure data in the real-time coating monitoring data.

[0156] The defect source judgment module S103 is used to use the pressure main frequency data and the mold head steady-state pressure data as pressure disturbance characteristics, and then use them with the lateral morphology characteristics and longitudinal fluctuation characteristics to determine the defect source type, and generate corresponding longitudinal control instructions, lateral control instructions or step-by-step collaborative control instructions according to the defect source type.

[0157] The instruction control module S104 is used to send longitudinal control instructions, lateral control instructions, or step-by-step collaborative control instructions to the production line programmable logic controller for execution, so as to realize real-time automatic control of the longitudinal and lateral thickness of the wide pressure-sensitive adhesive coating.

[0158] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0159] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for online production analysis of pressure-sensitive adhesives based on data monitoring, characterized in that, A coating production line for pressure-sensitive adhesives, comprising an adhesive supply pump and a coating die, includes the following steps in a data-monitored online production analysis method for pressure-sensitive adhesives: Step S1: Deploy a coating data monitoring network on the coating production line; acquire real-time coating monitoring data, including coating thickness data and coating pressure data, online synchronously through the coating data monitoring network; Step S2: Perform horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data to obtain the horizontal morphological characteristics and vertical fluctuation characteristics respectively; analyze the pressure main frequency data and die head steady-state pressure data based on the coating pressure data in the real-time coating monitoring data. Step S3: Use the pressure main frequency data and the steady-state pressure data of the mold head as pressure disturbance characteristics, and then use them with the lateral morphology characteristics and longitudinal fluctuation characteristics to determine the defect source type, and generate corresponding longitudinal control commands, lateral control commands or step-by-step coordinated control commands according to the defect source type. Step S4: Send the longitudinal control command, lateral control command, or step-by-step collaborative control command to the production line programmable logic controller for execution, so as to realize real-time automatic control of the longitudinal and lateral thickness of the wide pressure-sensitive adhesive coating.

2. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 1, characterized in that, In step S1, real-time coating monitoring data, including coating thickness data and coating pressure data, is acquired online synchronously through the coating data monitoring network. Receive the production batch start signal from the coating production line controller as the trigger command for data acquisition; Based on trigger commands, coating speed data is acquired in real time through the coating speed encoder in the coating data monitoring network; Determine if the coating speed data is greater than the preset sampling speed threshold; if so, adjust the sampling frequency of all sensors in the coating data monitoring network to the preset high-frequency sampling mode; otherwise, adjust to the standard sampling mode of the standard frequency. The time it takes for the laser thickness sensor array in the coating data monitoring network to complete one full-width scan is used as the benchmark acquisition time. Within each baseline acquisition cycle, based on the adjusted sampling frequency mode, real-time coating monitoring data is acquired online synchronously through the coating data monitoring network; among which, the real-time coating monitoring data includes coating thickness data measured by the laser thickness sensor array and coating pressure data from the pressure sensor.

3. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 1, characterized in that, Before performing the horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data in step S2, the following steps are also included: Based on the coating thickness data in the real-time coating monitoring data, a thickness data matrix is ​​constructed with the lateral measuring points of the laser thickness sensor array as columns and the acquisition time as rows; The coating thickness data in each column of the thickness data matrix is ​​averaged to obtain the steady-state thickness value; The steady-state thickness values ​​are arranged in order of the physical location of the transverse measuring points, and used as the transverse thickness baseline to characterize the transverse stability non-uniformity of the coating.

4. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 3, characterized in that, Step S2, which involves decoupling the horizontal and vertical axes based on the coating thickness data from the real-time coating monitoring data, includes: Calculate the first and second derivatives of the transverse thickness baseline to identify local extrema and inflection points; Based on the location and magnitude of local extreme points, the morphology of the lateral thickness baseline is classified into one of the following: central convex type, edge thickening type, or periodic concave type, thus obtaining the lateral morphology classification. Calculate the span of lateral defects based on their lateral morphology classification; The classification of lateral morphology, the magnitude of local extreme points, and the span of lateral defects are correlated as lateral morphological features.

5. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 4, characterized in that, Step S2, which involves decoupling the horizontal and vertical axes based on the coating thickness data from the real-time coating monitoring data, includes: Centered on the local extreme point on which the horizontal morphology classification is based, a pair of adjacent inflection points are retrieved and located on both sides of it. Calculate the lateral physical distance between the pair of inflection points, and use this distance as a parameter of the lateral influence range of the defect corresponding to the lateral morphology classification, as the lateral defect span.

6. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 4, characterized in that, Step S2, which involves decoupling the horizontal and vertical axes based on the coating thickness data from the real-time coating monitoring data, further includes: Subtract the steady-state thickness value of the column corresponding to the horizontal thickness baseline from each data point in the thickness data matrix to generate a residual matrix that eliminates the influence of the inherent horizontal distribution. The residual values ​​of each column in the residual matrix are arranged in chronological order to form a vertical residual sequence; Perform a Fast Fourier Transform on each longitudinal residual sequence to calculate the power spectral density within a preset frequency range; Identify the peak frequency points in the power spectral density where the amplitude exceeds the preset noise floor; All identified peak frequency points and their corresponding amplitudes are used as longitudinal fluctuation features.

7. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 6, characterized in that, Step S2 involves analyzing the pressure frequency data and die head steady-state pressure data based on the coating pressure data from the real-time coating monitoring data, including: Based on the coating pressure data in the real-time coating monitoring data, the pump outlet pressure sequence measured by the pressure sensor installed at the glue supply pump outlet and the internal pressure matrix of the die head measured by the pressure sensor installed inside the coating die head are separated. Perform a fast Fourier transform on the pump outlet pressure sequence to calculate the power spectral density within a preset frequency range, and identify the peak frequency points whose amplitude exceeds the preset noise floor as the pressure main frequency data. The pressure matrix inside the die head is averaged over time to obtain steady-state pressure data of the die head that characterizes the lateral distribution of pressure inside the coating die head.

8. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 1, characterized in that, In step S3, the pressure main frequency data and the mold head steady-state pressure data are used as pressure disturbance characteristics, and then combined with the lateral morphological characteristics and longitudinal fluctuation characteristics to determine the defect source type, including: The peak frequency points in the longitudinal fluctuation characteristics are compared with the pressure main frequency data. If the frequency difference between the two is less than the preset frequency matching threshold, a longitudinal source association identifier is generated. Spatial location matching is performed between the local extreme point locations in the lateral morphological features and the pressure anomaly points in the steady-state pressure data of the mold head. If the locations correspond, a lateral source association identifier is generated. Determine the generation status of the vertical source association identifier and the horizontal source association identifier; If only vertical source association identifiers are generated, the defect source is determined to be a vertical source; if only horizontal source association identifiers are generated, the defect source is determined to be a horizontal source; if both are generated, the defect source is determined to be a mixed source.

9. The online production analysis method for pressure-sensitive adhesives based on data monitoring according to claim 1, characterized in that, Step S3, which generates corresponding vertical control instructions, horizontal control instructions, or step-by-step collaborative control instructions based on the defect source type, includes: If the defect source is determined to be a longitudinal source, the adjustment amount of the glue supply pump speed is calculated based on the longitudinal fluctuation characteristics, and a longitudinal control command is generated. If the defect source is a transverse source, the target bolt is located based on the position of the local extreme point in the transverse morphological characteristics, and the adjustment sequence of the target bolt and its adjacent bolts is calculated based on the amplitude of the local extreme point and the transverse defect span to generate transverse control instructions. If the defect source is determined to be a mixed source, the calculations for both the vertical and horizontal sources are performed simultaneously to generate step-by-step coordinated control instructions.

10. A data-monitoring-based online production analysis system for pressure-sensitive adhesives, characterized in that, For executing the data monitoring-based online production analysis method for pressure-sensitive adhesives as described in claim 1, the data monitoring-based online production analysis system for pressure-sensitive adhesives includes: The online synchronous monitoring module is used to deploy a coating data monitoring network on the coating production line; it acquires real-time coating monitoring data, including coating thickness data and coating pressure data, online synchronously through the coating data monitoring network. The horizontal and vertical decoupling analysis module is used to perform horizontal and vertical decoupling analysis based on the coating thickness data in the real-time coating monitoring data, and obtain the horizontal morphological characteristics and vertical fluctuation characteristics respectively; it analyzes the pressure main frequency data and die head steady-state pressure data based on the coating pressure data in the real-time coating monitoring data. The defect source judgment module is used to use the pressure main frequency data and the steady-state pressure data of the mold head as pressure disturbance characteristics, and then use them with the lateral morphology characteristics and longitudinal fluctuation characteristics to determine the defect source type, and generate corresponding longitudinal control instructions, lateral control instructions or step-by-step coordinated control instructions according to the defect source type. The instruction control module is used to send longitudinal control instructions, lateral control instructions, or step-by-step collaborative control instructions to the production line programmable logic controller for execution, so as to realize real-time automatic control of the longitudinal and lateral thickness of the wide pressure-sensitive adhesive coating.

Images