Intelligent film laminating machine control method and system

Abstract

According to the intelligent film laminating machine control method and system provided by the invention, the defect phase coordinates are established through film edge period marks, the elongation is estimated and the defect backtracking time delay is corrected in combination with the difference of upstream and downstream odometer rollers, and accurate attribution of defects in a process section is realized. According to the system, an online micro-stripping module is arranged in a rim charge recovery channel, a stripping mean value P, a fluctuation coefficient CV and stripping work W are obtained in real time, an adhesion allowable domain Omega is constructed, and an adhesion margin S is calculated; and solving a parameter adjustment increment delta u by adopting tangential constraint under omega constraint, performing safe self-adaptive adjustment on actuators in a temperature zone, pressing pressure, linear speed and the like, and avoiding transboundary failure through continuous verification and a step-down backspacing mechanism. According to the scheme, the first piece passing rate and stability are improved, the defect rate of bubbles, wrinkles, delamination and the like is reduced, batch tracing and controlled updating are supported, and the method is suitable for a roll-to-roll film laminating production line.

Description

Technical Field

[0001] This invention relates to the field of laminating machine control, and more particularly to an intelligent laminating machine control method and system. Background Technology

[0002] Lamination (film application, composite, and coating), as a typical roll-to-roll (R2R) continuous processing technology, is widely used in the preparation of functional films, adhesive products, optical films, lithium battery separators and electrode protective films, decorative films, and packaging composite materials. Laminating machines typically include unwinding, guiding and flattening, tension adjustment, heating / preheating, pressing (roll pressing / hot pressing), web guiding, cooling and shaping, and rewinding stages, along with speed control, tension closed-loop control, and web guiding control to ensure the film maintains a stable tape track and acceptable bonding quality at high speeds. Due to the viscoelasticity, compressibility, and heat sensitivity of the materials themselves, and the coupled influence of temperature, pressure, speed, tension, and substrate surface conditions (dust, static electricity, roughness, moisture content, etc.), quality problems such as bubbles, wrinkles, edge lifting, tunneling, whitening, delamination, poor adhesion, uneven adhesive transfer, and edge defects caused by misalignment are prone to occur in actual production. These defects not only reduce yield but may also be amplified during subsequent slitting, punching, die-cutting, and final assembly processes, leading to batch rework or scrap. Therefore, the industry generally adopts a combination of online inspection and closed-loop control to improve coating stability.

[0003] In the prior art, there are already solutions that combine the lamination process with defect detection. For example, patent document CN102991750A (“Lamination Machine and Lamination Method”) discloses lamination equipment and process flow, and introduces detection-related devices / steps in the lamination process to monitor and handle the lamination quality. Its overall approach belongs to the typical route of “lamination processing + quality inspection / disposal”. Such solutions can achieve defect detection and screening to a certain extent, providing a basis for subsequent control. However, their focus is usually on the result-oriented approach of “whether defects exist / whether they are qualified”, and they are more about classification, processing or simple linkage control after detection. It is difficult to directly solve the problem of hidden defects caused by “adhesion state fluctuating near the process window boundary” during the lamination process. Especially when defects have not yet manifested significantly in appearance, relying solely on visual inspection often has lag and uncertainty.

[0004] Meanwhile, at the roll-to-roll equipment level, constant tension and web correction control are mature fundamental technologies. Solutions represented by CN103241575A / B (“Automatic Web Correction Constant Speed ​​Constant Tension Rewinding Machine”) disclose unwinding / rewinding tension control structures, photoelectric position sensors and web correction mechanisms, encoder rollers, PLCs, and frequency converters, achieving constant speed and tension in film transport and automatic web correction, thereby reducing problems such as web deviation, stress deformation, and uneven adhesion. These solutions play a crucial role in the stability of roll material operation, but they primarily address the fundamental mechanical and control issues of “belt path and tension stability.” They typically lack online quantifiable process quality indicators and targeted constraint control strategies for the more sensitive aspects of the lamination process, such as “adhesion mechanism changes caused by temperature-pressure-speed combinations, adhesion window boundary failures, and intermittent defects caused by adhesion fluctuations.” In other words, even if tension and correction control are well done, defects such as bubbles, delamination or edge lifting may still occur due to factors such as the activation level of the adhesive system, interface wetting and degassing efficiency, and hot pressing contact state. Moreover, the causes of these defects are often not explained by a single control quantity deviation.

[0005] To further improve process quality, some adjacent processes (such as coating and lamination) have proposed a closed-loop approach: "locating and mapping abnormal areas to adjustable components, and then making targeted adjustments." For example, CN116748081B ("Coating System and Method") proposes that when an abnormality is detected in the target coating area, the corresponding flow regulating block can be quickly located and adjusted to restore the coating quality. This approach embodies the engineering concept of "quality abnormality—location—precise parameter adjustment," which is significant for reducing blind parameter adjustments and improving recovery speed. However, the quality indicators of this type of approach are usually parameters such as coating weight / deviation, and the adjustment object is the coating feeding component, which differs from the mechanism of "interfacial adhesion formation and stabilization" in the lamination process. In the coating process, defects are not only related to the set values ​​of a certain section, but also highly coupled with material elongation, thermal history, pressure contact state, edge heat dissipation, and the microscopic venting / flow state of the adhesive layer. Relying solely on the framework of "abnormal coordinate mapping - adjusting the corresponding actuator" may still result in: inaccurate positioning leading to incorrect parameter adjustment; or although parameter adjustment restores the appearance, it pushes the working condition outside the adhesion window boundary, causing potential risks such as subsequent delamination / adhesion attenuation.

[0006] In summary, existing technologies related to lamination and roll material control generally have the following shortcomings:

[0007] 1) Lack of online quantifiable adhesion quality process indicators. Existing solutions mostly rely on appearance / defect detection results or indirect quantities such as tension and deviation, which are difficult to reflect the "interfacial adhesion strength and its stability (fluctuation)" in real time. Although peel strength (e.g., 90° peel) is widely used in material testing to evaluate adhesion performance, traditional peel tests are usually off-line sampling or laboratory tests, making it difficult to synchronize with continuous coating conditions and constrain process parameter combinations in real time.

[0008] 2) Lack of a strong constraint mechanism for "adhesion window cross-boundary failure". In coating production, the combination of temperature, pressure, and speed has an acceptable process window (allowable range). Once the boundary is crossed, irreversible delamination, blistering, or edge lifting may occur. Most existing control methods rely on single-variable or empirical rule-based parameter tuning. Even with the addition of anomaly localization, it is difficult to ensure that the parameter tuning trajectory remains within the safe allowable range. Furthermore, there is a lack of an implementable closed-loop mechanism of "verification-backtracking-step reduction".

[0009] 3) Defect backtracking and localization are susceptible to elongation and time delay drift. Coil materials elongate and shrink in different tension and thermal zones, resulting in time delay and distance mapping errors between the detected and causal locations of defects. Relying solely on encoder mileage or theoretical speed estimations can easily lead to attribution bias and subsequent parameter mistuning. Although the coil material industry has developed methods to address elongation through measurement and control, in laminated applications, this correction mechanism still needs to be coupled with the defect coordinate system, process attribution, and subsequent parameter tuning strategies to form a reproducible closed-loop chain. Summary of the Invention

[0010] The technical objective of this invention is to provide an intelligent coating machine control method and system. By implementing online micro-peeling tests in the edge material recycling channel to obtain adhesion quality indicators such as peeling average, fluctuation, and peeling work, an adhesion allowable domain of temperature-pressure-speed is constructed, and tangential constraints and backtracking verification mechanisms are applied to the parameter adjustment process. At the same time, by combining the phase coordinates of the film edge marker and the differential elongation correction of the dual odometer, defect backtracking and intra-segment attribution are realized. Thus, in continuous coating production, the quality fluctuations caused by adhesion cross-boundary failure and mis-adjustment of parameters are suppressed, defects such as bubbles, wrinkles, and delamination are reduced, and the yield and process stability are improved.

[0011] To achieve the objectives of this invention, the following technical solution is adopted:

[0012] An intelligent control method for a roll-to-roll laminating machine is disclosed. This method collects data on deviation, tension / floating roller displacement, linear speed, pressing pressure, and zone temperature. Online vision is used to obtain the defect type and its phase coordinates relative to the film edge periodic marker. The elongation rate is estimated by the length difference of the odometer rollers before and after the pressing section to correct for defect backtracking delay, and the backtracking zero point is reset by marking phase abrupt changes. Sample strips are cut in the edge material recovery channel and subjected to online micro-peeling of a fixed length L at a peel angle of 90±5° and a peeling speed proportional to the linear speed, yielding the unit width peeling average P, fluctuation CV, and peeling work W. A temperature-pressure-speed allowable domain Ω is constructed from P, CV, and W, and the adhesion margin S is calculated. Intra-segment attribution is performed based on defect and process signals. Only for the attribution segment actuator, a parameter adjustment increment Δu is generated within Ω, and Δu satisfies the requirement of increasing S monotonically along the tangential direction of the Ω boundary. If S does not increase after M consecutive samplings, the step size is reduced according to the backtracking factor, and the process reverts to the previous stable setting.

[0013] As a further improvement, during the initialization phase, the product standard parameters Pmin, Pmax, Cmax, Wmin, sampling length L, strip width w, peeling speed proportional coefficient k_v, and safety upper limits for tension / deviation / temperature / pressure / torque and parameter adjustment change rate are loaded to limit the allowable domain Ω and parameter adjustment search boundary.

[0014] As a further improvement, a unified clock synchronization and buffer queue management is implemented for deviation amount, tension, speed, pressure, temperature, torque and defect phase coordinates, so that defect backtracking delay correction and online micro-peeling measurement are associated under the same time reference.

[0015] As a further improvement, the periodic mark is a printed mark, a laser-engraved mark, or an embossed mark; the online visual output shows the phase value of the defect relative to adjacent marks and the longitudinal coordinate within the mark spacing, and triggers a zero-point reset when a sudden change in the mark phase caused by film splicing, roll changing, or missing marks is detected.

[0016] As a further improvement, the elongation is obtained by filtering the difference between the cumulative length of the upstream odometer roller and the cumulative length of the downstream odometer roller in the pressing section, and this elongation is used simultaneously for the joint correction of defect backtracking distance and backtracking time delay.

[0017] As a further improvement, the online micro-peeling is performed in an independently guided edge material path; the peeling speed is controlled by a servo drive and satisfies v_peel=k_v·v; the peeling work W is the integral of the peeling force-displacement curve over a fixed length L; the allowable domain Ω is at least limited by P∈[Pmin,Pmax], CV≤Cmax, and W≥Wmin; the "tangential along the Ω boundary" includes estimating the Ω boundary normal n at the current working point and letting Δu satisfy n·Δu=0 or |n·Δu|≤δ and not crossing Ω; the backtracking step size includes shrinking Δu by the coefficient β (0<β<1) and backtracking to the previous stable setting before verification.

[0018] Another aspect of the present invention provides an intelligent laminating machine control system, characterized in that it includes: a data acquisition unit, an online visual inspection unit, a film edge marking unit, an upstream odometer roller and a downstream odometer roller, an elongation and backtracking correction unit, an edge material sampling and online micro-peeling unit, an allowable domain Ω construction unit, an intra-segment attribution unit, a tangential constraint parameter tuning unit, and a backtracking verification and rollback unit; wherein the edge material sampling and online micro-peeling unit outputs P, CV, and W; the allowable domain Ω construction unit forms a temperature-pressure-speed allowable domain Ω based on P, CV, and W and calculates the adhesion margin S; the tangential constraint parameter tuning unit generates Δu and applies tangential constraints only to the attribution segment actuator within Ω; the backtracking verification and rollback unit performs step reduction and rollback when S is not satisfied for M consecutive times.

[0019] As a further improvement, the edge material sampling and online micro-peeling unit is located in the edge material recycling channel and includes an edge trimming and cutting mechanism, an independent guiding mechanism, a 90±5° peeling angle guiding structure, a servo peeling drive, and a force sensor, used to achieve zero-interference sampling of the main strip and fixed-length L micro-peeling measurement; the tangential constraint parameter tuning unit includes a boundary normal estimation module, used to estimate the boundary normal n according to the constraint relationship of the allowable domain Ω, and apply n·Δu=0 or |n·Δu|≤δ and a maximum rate of change limit to Δu; the backtracking verification rollback unit includes a backtracking factor β configuration module and a stable setting storage module, used to implement step size reduction and parameter rollback.

[0020] A third aspect of the present invention provides a computer device including a processor, a graphics processing unit (GPU), and a memory, wherein the memory stores a computer program that, when executed by the processor and the GPU, causes the computer device to perform the method described thereon.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer device, causes the computer device to perform the method described thereon.

[0022] This invention achieves comprehensive technical effects that are difficult to obtain by traditional laminating machines that rely solely on tension / correction / appearance inspection through a closed-loop synergy of "online micro-peeling feedback of edge material - adhesion allowable domain Ω constraint - tangential adjustment and retrospective verification backtracking - defect retrospective attribution":

[0023] 1. Online micro-peeling outputs the average peeling value P, fluctuation CV, and peeling work W in real time, transforming the adhesion strength and stability from "post-event sampling / experience judgment" to "process measurable and controllable". Based on this, the allowable range Ω and adhesion margin S of temperature-pressure-speed are constructed, which can maintain stable adhesion performance under the conditions of increased speed, environmental fluctuations, or material batch differences, and significantly reduce hidden defects such as delamination, edge lifting, and whitening.

[0024] 2. The parameter adjustment increment Δu is restricted to tangential movement along the Ω boundary and S is required to increase monotonically. Combined with the release criterion of continuous M sampling, backtracking step size reduction and backtracking stability setting mechanism, adhesion failure and quality collapse caused by "cross-boundary parameter adjustment" can be effectively avoided, and downtime, waste and energy consumption caused by repeated trials can be reduced.

[0025] 3. The phase coordinate of the membrane edge periodic mark and the differential elongation correction of the dual odometers before and after the pressing section are used together for defect backtracking delay correction and zero point reset, which significantly reduces mispositioning and misattribution caused by elongation drift. This makes it possible to fine-tune only the actuator of the attribution section, thereby reducing multi-segment coupling interference, shortening defect convergence time and improving the overall stability of the machine.

[0026] Therefore, this invention can simultaneously reduce defect rate, converge adhesion fluctuations, significantly reduce cross-boundary failure risk, accelerate recovery speed and improve batch consistency in continuous production, ultimately improving coating yield and process controllability, and is suitable for high-speed, high-requirement functional film and composite material coating scenarios. Attached Figure Description

[0027] Figure 1 This is a block diagram of the overall structure of the intelligent film coating machine control system of the present invention.

[0028] Figure 2 This is a schematic diagram of the data flow and control flow of the present invention.

[0029] Figure 3 This is a flowchart illustrating the intelligent film-coating machine control method of the present invention. Detailed Implementation

[0030] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of the present invention.

[0031] I. Explanation of Terms and Symbols

[0032] (1) Laminating machine: refers to the production equipment that continuously laminates the first roll of material (substrate) and the second roll of material (coating) in the unwinding, guiding, flattening, preheating / heating, pressing, correction, cooling and rewinding stations. It can be used for PET / PE / PI / PP film, optical film, protective film, metal foil composite film, etc.

[0033] (2) Periodic Marks (Membrane Edge Marks): Periodic identification patterns or structures (printed lines, dot matrix, embossing, laser engraving, barcode fragments, etc.) set in the membrane edge area to provide a stable "phase coordinate" reference and serve as anchor points for defect backtracking. Periodic marks are preferably placed in the trimming safety zone and do not enter the effective width of the finished product.

[0034] (3) Phase coordinates : Normalized position coordinates with the distance (or time) between two adjacent periodic markers set to 1. Phase coordinates are used to compensate for the effects of roll diameter variations, velocity perturbations, and material elongation on defect localization.

[0035] (4) Odometer Roller: The length measuring encoder roller (or length measuring wheel) is set upstream / downstream of the pressing section to output the cumulative length. , It is used to estimate the equivalent elongation and transport delay of materials in critical process stages.

[0036] (5) Equivalent elongation The length difference between the two odometers reflects the equivalent elongation ratio of the material in the critical section of the pressing process, and is used to correct the defect backtracking distance and time delay.

[0037] (6) Online micro-peeling: Perform online peeling tests on the trimmed strip at a fixed angle and length in the edge material recycling channel, synchronized with the main line speed, obtain the peeling force curve in real time, and calculate the adhesion strength and stability index.

[0038] (7) Stripping the mean Volatility coefficient Peeling function : : Average peel force per unit width, characterizing the level of adhesion strength; Peel force fluctuation coefficient, characterizing adhesion stability and intermittent delamination risk; Stripping work (energy) characterizes the overall toughness / energy dissipation capacity of the interface and is used to identify potential problems such as "average values ​​are acceptable but the interface is fragile".

[0039] (8) Allowed domains :Depend on , , The feasible set of operating conditions is jointly defined by adhesion indicators, and the operating condition variable is usually temperature. ,pressure ,speed (Tension settings can also be expanded). In Internal operation indicates that adhesion meets the threshold and has a certain safety margin.

[0040] (9) Adhesion margin : Indicates the current operating point relative to The minimum safety margin at the boundary is used as the release criterion and the adaptive parameter tuning objective.

[0041] (10) Boundary normal With tangential constraints: To allow the domain boundary to be approximated by the outward normal at the current operating point, the tangential constraint requires parameter tuning increments. Moving tangentially (or approximately tangentially) along the boundary reduces the probability of adhesion failure due to crossing the boundary.

[0042] (11) Parameter adjustment increment : For the working condition vector Incremental commands (or their components) are applied for the combined adjustment of temperature zone setting, pressing pressure, and linear speed.

[0043] (12) Retrospective factor Number of verifications : Used for backtracking and shrinking the step size; The size of the continuous verification window (by number of times or by length) is used for the release / retrogression decision.

[0044] II. System Structure

[0045] See Figure 1 , Figure 2 The intelligent laminating machine control system of the present invention can be integrated into the laminating machine body or added as a modification module. Overall, it can be divided into an electromechanical execution layer, a sensing layer, a computing control layer, and a human-machine and data layer.

[0046] 1) Electromechanical Actuation Layer: This layer includes at least an unwinding mechanism, a guide roller flattening mechanism, a preheating / heating zone, a pressing roller assembly, a web guiding mechanism, a cooling and shaping mechanism, and a winding mechanism. The pressing roller assembly can employ servo pressure or pneumatic-hydraulic pressurization methods and be equipped with a pressure sensor closed-loop system. The web guiding mechanism can employ a swing frame type or a linear module type web guiding mechanism, possessing sufficient dynamic response to suppress web deviation.

[0047] 2) Sensing layer: including at least a film edge position detection sensor (photoelectric / CCD), tension sensor or force sensing roller, zone temperature sensor (thermocouple / PT100), pressing pressure sensor, drive current / torque acquisition, dual odometer roller encoder, and online visual inspection camera and light source.

[0048] 3) Online visual inspection module: Detects and classifies defects after bonding, and outputs the defect type. Confidence level Occurrence timestamp Phase coordinates Horizontal position The defect size, etc., are recorded and the defect record is written to the defect queue.

[0049] 4) Periodic Marking Module: Used to generate / identify periodic marks on the membrane edge. It can be used in conjunction with a vision camera or a separate marking camera can be set up. The mark recognition results are used to construct phase coordinates and serve as backtracking anchor points.

[0050] 5) Edge material sampling and online micro-peeling module: A cutting mechanism is set in the edge material recycling channel to form a strip of fixed width. The strip is guided by a peeling angle guide structure to form a stable peeling angle; the peeling servo drive is synchronized with the main line speed; a force sensor collects the peeling force curve and calculates it. , , The module is preferably configured with strip breakage detection, angle deviation detection, sensor self-test, and overload protection.

[0051] 6) Computational Control Layer: A PLC / motion controller + industrial computer architecture can be adopted. The PLC is responsible for correction, tension, speed, pressure and temperature closed loop; the industrial computer is responsible for defect fusion, elongation backtracking correction, allowable domain construction, tangential constraint parameter tuning, verification backtracking, and data traceability.

[0052] 7) Human-Machine and Data Layer: The HMI provides recipe management, threshold configuration, alarm prompts, reports and traceability queries; and can report key process quality data through the MES / SCADA interface.

[0053] III. Overall Technical Route

[0054] See Figure 3 This invention, centered on "online micro-peeling providing realistic adhesion feedback," expands coating control from traditional tension / correction / appearance inspection to "safe adaptive parameter tuning under adhesion allowable domain constraints." The overall process is as follows: S1 Initialization and parameter loading → S2 Multi-source synchronous acquisition → S3 Phase coordinate construction → S4 Elongation and backtracking correction → S5 Online micro-peeling measurement → S6 Allowable domain With margin Construction → S7 Defect Attribution and Actuator Selection → S8 Tangential Constraint Solution and Parameter Tuning → S9 Continuous verification and backtracking → S10 Batch traceability and controlled update.

[0055] IV. Specific Implementation Methods of Steps S1–S10

[0056] S1 Initialization and Parameter Loading

[0057] (1) Formulation parameter package: After the system is powered on, the parameter package is read from the process database according to the index "Product No. / Material Combination / Adhesive Type / Batch No.". The parameter package includes at least: Adhesion threshold: , , , Sampling parameters: strip width Peeling length Peeling angle Peeling speed proportional coefficient Adaptive parameter: Tangential tolerance Retrospective Factors Validate window size Maximum number of rollbacks Maximum number of invalid measurements Safety constraints: upper and lower limits for temperature / pressure / speed / tension and upper limit for rate of change.

[0058] (2) Safety interlock and boundary condition initialization: , , , , Constraints are written into the PLC's interlock area to form hard protection. For example: when or Prioritize deceleration and enter steady state; when the temperature control circuit is abnormal, prohibit adaptive parameter adjustment and maintain a safe temperature range.

[0059] (3) Stable setting Initialization: Establish a stable buffer. This can be taken as the recommended process. Or the value of a successfully cured batch from the previous batch. To avoid incorrect curing due to accidental measurements, the system is set to a "continuous pass counter," which only counts when consecutive passes occur. Only when all windows meet the release criteria will the candidate setting be upgraded to the new one. .

[0060] (4) Self-test and calibration: Perform self-tests sequentially for temperature, pressure, tension / force, encoder, camera trigger, and peel force sensor. Typical strategies include: checking zero drift, noise RMS, saturation, and frame drop rate; performing no-load baseline sampling and recording zero-point compensation for force sensors; and performing low-speed belt running calibration to verify the consistency of "length count - actual length" for odometer rollers. If the self-test fails, the system only allows manual / semi-automatic mode and does not allow fully automatic adaptive mode.

[0061] (5) Mode Management: At least manual mode, semi-automatic mode, and fully automatic mode should be provided. Mode switching conditions should include: key sensors are continuously effective, stripping measurement channels are available, camera frame drop rate is below the threshold, and there are no major alarms.

[0062] S2 Multi-source signal acquisition and unified time reference

[0063] (1) Unified clock and triggering strategy: The PLC master clock or PTP is selected as the unified time reference. Camera triggering, encoder sampling and analog sampling are all recorded with the same timestamp. If a line scan camera is used, it is preferable to use encoder frequency division triggering to achieve "sampling by length", so that the image coordinates correspond naturally with the material length and reduce the scale error caused by speed perturbation.

[0064] (2) Sampling frequency configuration: deviation ,tension ,pressure ,speed Recommended 50–500Hz; Temperature range Recommended Hz: 5–50 Hz; Torque / Current 1. Swing roller displacement: Recommended 50–200 Hz; Peel force curve : ≥500 Hz is recommended (to capture unstable fluctuations at the interface).

[0065] (3) Filtering and delay compensation: for , Equal-height noise signals can be filtered using first-order / second-order low-pass filtering or moving average filtering; however, the delay introduced by the filtering must be recorded (the delay amount is stored as a system parameter) and compensated for during S4 backtracking alignment to avoid "misalignment between signal peak and defect event".

[0066] (4) Circular buffer and data quality flags: establish coverage for maximum transport delay A circular cache is used. Each data entry carries a quality flag; invalid data must not be included. Construction and parameter tuning decisions. Interpolation can be used for short-term packet loss, but the weight of the interpolated segment needs to be reduced or it should be directly removed.

[0067] (5) Visual defect record structure: The defect record shall include at least the following: .when When the value is below the threshold, it can be counted in the statistics without triggering parameter tuning, so as to reduce the risk of false detection leading to incorrect parameter tuning.

[0068] S3 Periodic Marker Phase Coordinate Construction and Defect Representation

[0069] (1) Periodic marker design: periodic marker spacing The width can be selected from 50 to 300 mm, and the width / area must be guaranteed to ensure stable recognition at the target speed (e.g., 5–20 pixel coverage). The marking should avoid the effective area of ​​the finished product and should not adversely affect subsequent processes (rewinding, die-cutting).

[0070] (2) Implementation of recognition algorithm: Gray-level normalization, edge enhancement and threshold segmentation are performed within the ROI at the membrane edge. Template matching / correlation peak detection is used to obtain the marker center position. For the linear array, sampling mode is based on length. Periodic peaks can be detected on the length axis to output the marker sequence number. With arrival time Simultaneously, a "periodic consistency check" is performed; if the interval between adjacent markers deviates from the set value by more than [a certain amount], [the check will be performed]. If the percentage is 5%–20%, it is considered a risk of missing labels / damage / frame loss and an event will occur.

[0071] (3) Phase calculation: for the time of defect occurrence Find the nearest neighboring marker time ,calculate:

[0072] ;

[0073] in: For defect phase; For the arrival time of adjacent markers; This is the defect timestamp. If length triggering is used, it can also be calculated equivalently using cumulative length coordinates.

[0074] (4) Phase abrupt change and anchor point events: When splicing, film changing, marker contamination, or camera frame loss occur, marker sequence number breakage or... Abrupt change. The system records this type of exception as an anchor event. It also records the time and sequence number range for S4 backtracking zero point reset, avoiding cumulative errors across segments.

[0075] (5) Phase bin statistics: Divided into several phase boxes (e.g., 20–100 boxes), the distribution of defect density as a function of phase is statistically analyzed. If a certain phase box has a consistently high incidence of defects, it can indicate the presence of periodic mechanical disturbances (local defects on the roller surface, bearing eccentricity, guide roller contamination, etc.), providing a basis for maintenance.

[0076] S4 Dual Odometer Differential Elongation Estimation and Backtracking Correction

[0077] (1) Odometer Roller Arrangement and Anti-Slippage: The upstream odometer roller is arranged in the stable tension zone before pressing, and the downstream odometer roller is arranged in the stable zone before cooling after pressing; the rubber coating on the rollers and the pressing mechanism ensure reliable friction. The system monitors the slippage risk by comparing the difference between the odometer speed and the main drive speed. Once the slippage exceeds the limit, the odometer weight is reduced or a maintenance prompt is triggered.

[0078] (2) Length difference and elongation estimation: Real-time calculation of length difference:

[0079] ;

[0080] in: This represents the cumulative length difference between upstream and downstream. , These represent the cumulative lengths of the upstream and downstream odometers, respectively. Let the equivalent reference length of the critical pressing section be... The equivalent elongation is:

[0081] ;

[0082] in: Equivalent elongation; This can be determined by the equipment's geometric path or obtained through calibration. To suppress noise, [the following can be done / adjusted / adjusted]: , Perform low-pass filtering, with a filtering time constant of 0.1–1.0 s, to make it a slow variable for backtracking correction.

[0083] (3) Calculation of transportation delay and backtracking time window: The time when the defect is observed at the detection point is The causes typically occur in the upstream pressing / temperature zone / tension section, etc. Let the equivalent path length from the pressing section to the detection section be... The linear velocity is The time delay estimate after introducing elongation correction is:

[0084] ;

[0085] in: For backtracking delay; This is the equivalent path length; Let be the linear velocity. Then the estimated time of origin is... The system retrieves from the circular cache. Features such as tension, deviation, temperature, pressure, and torque within the time window are used for S7 attribution.

[0086] (4) Phase anchor point reset: When S3 gives an anchor point event In this process, backtracking mapping is prioritized to be completed within the same marker sequence range to avoid distance uncertainties caused by cross-stripping. This mechanism is particularly critical for operations such as roll changing, film splicing, and re-threading.

[0087] (5) Abnormal handling: If the odometer slips or the encoder loses pulses, the phase coordinate will be used as the main reference for backtracking, with the odometer as the secondary reference; and the "roller surface cleaning / tightening check" maintenance prompt will be triggered to prevent incorrect backtracking from causing misadjustment of parameters.

[0088] S5 Edge Material Online Micro-Peeling Measurement

[0089] (1) Decoupling of edge cutting and sampling: After trimming, the edge material enters the recycling channel. The system is equipped with a cutting mechanism to form a fixed-width strip. A thickness of 5–15 mm is acceptable. The strip enters an independent guide path and is decoupled from the main belt via a buffer roller or independent tension control, avoiding disturbance to the main belt tension and correction caused by sampling.

[0090] (2) Peel corner structure and stability: adopting Peeling angle. In engineering implementation, a stable angle can be formed using a fixed guide roller radius and a limiting plate; alternatively, an adjustable angle mechanism (with graduations and locking) can be used. To prevent angle drift from causing incomparable measurements, an angle positioning detection function can be set; if the angle deviates beyond a threshold, the measurement is marked as invalid.

[0091] (3) Peeling speed synchronized with the main line: In order to ensure that the online adhesion measurement is consistent with the actual working conditions, the peeling speed is synchronized with the main line. With the main line speed satisfy:

[0092] ;

[0093] in: The speed proportionality coefficient is preferably 0.95–1.05. The system can be specified to operate only in the speed stability region (e.g., If the value is less than the threshold, stripping sampling is triggered to avoid force drift caused by acceleration and deceleration.

[0094] (4) Fixed-length peeling and force curve acquisition: Fixed-length peeling is performed each time. (e.g., 50–200 mm). A force sensor acquires the peeling force curve. A sampling frequency of ≥500 Hz is recommended to capture high-frequency fluctuations caused by "microbubbling-microdelamination". The system... Perform denoising and abnormal spike removal (stuck / impact spikes), and calculate within the effective range. and . It can be defined as standard deviation / mean or peak-to-peak value / mean to suit different noise characteristics.

[0095] (5) Peeling function Calculation and significance: The stripping function characterizes interfacial toughness and energy dissipation.

[0096] ;

[0097] in: For stripping work; This refers to the peeling length; Peeling force; For displacement.

[0098] (6) Invalid measurement judgment and protection: When situations such as strip breakage, angle deviation, force sensor saturation, peeling servo overload, or guide jamming occur, the system will mark the measurement as invalid and will not participate. Construction and release judgment. The number of consecutive invalid entries exceeds [a certain threshold]. When this happens, the system freezes the adaptive parameter adjustment and prompts you to check the cutter, guide roller, force sensor, and strip path.

[0099] (7) Measurement package packaging and time alignment: Each stripping generates a measurement package. ( (For measurement timestamps), written to the cache and aligned with the backtracking time window of S4, so that "defect-condition-adhesion" forms a traceable closed-loop link.

[0100] S6 Allowed Domains Construction and Adhesion Margin calculate.

[0101] (1) Engineering meaning of allowable domain: allowable domain It's not a single threshold judgment, but rather a judgment based on temperature. ,pressure ,speed A set of combined constraints is used to ensure that the adhesion failure zone is not crossed during the parameter tuning process. This idea solves the hidden risk caused by the traditional empirical parameter tuning method where "defects are reduced, but adhesion is pulled to the critical edge".

[0102] (2) Threshold source and determination method: This can be derived from customer standards, process validation tests, or historical batch statistics. The preferred method is to determine this using a combination of quantiles and a safety factor: for example, using historically successful batches. Using the 5th percentile as a benchmark, and then applying a conservative coefficient, we obtain... ;Pick The 95th percentile as When there are large environmental fluctuations or material batch variations, a more conservative setting should be used.

[0103] (3) Internal and external determination and representation methods: The system will use the operating point With measurement package Binding, determine if the condition is met:

[0104] ;

[0105] ;

[0106] .

[0107] If satisfied, then it is Interior point; otherwise, External points trigger conservative strategies (such as prioritizing speed reduction, then boosting, and then fine-tuning the temperature range). The engineering representation can be achieved using: three-dimensional grid lookup table, piecewise linear inequality, or local regression model.

[0108] (4) Margin Calculation: To achieve "quantification of safety margin", the following definition is made:

[0109] ;

[0110] in: This represents the adhesion margin. Indicates the current point is at Inside and at least a distance from the boundary ; This indicates a risk of crossing the boundary or touching the edge. To suppress noise-induced misjudgments, [the following can be done / adjusted]. Perform a sliding average, but the window should not be too large to avoid reducing the response speed.

[0111] (5) Boundary normal The estimate: due to From measurements, analytical gradients are difficult to obtain; the system can use the nearest... Perform local linear fitting on subsamples, for example:

[0112] ;

[0113] ;

[0114] .

[0115] When the tightest constraint is At that time, with Normalization as When the tightest constraint is At that time, with Normalization as When multiple constraints are simultaneously tight, they can be weighted and synthesized according to their urgency. This method allows for online updates without the need for complex models.

[0116] (6) Early warning mechanism: When Approaching 0 or the indicator is close to the threshold (e.g.) When the value is less than the warning value, the system reduces the step size and increases the verification window in advance. Limit the speed limit to avoid irreversible delamination caused by cross-limit parameter adjustments.

[0117] S7 Intra-segment Attribution and Actuator Channel Selection

[0118] (1) Attribution goal: Correspond the defects to the possible cause process segments, so that parameter tuning is only open to the "related actuator channels" to avoid blind tuning of the whole segment and causing coupling disturbance.

[0119] (2) Input characteristics: Attribution inputs should include at least: defect type Confidence level Phase Horizontal position Retrospective correction cause window And the signal characteristics within the window (deviation peak, tension gradient, temperature difference, pressure fluctuation, torque anomaly) and adhesion deviation ( ).

[0120] (3) Rule base / scoring mechanism: Establish a rule base for "defect-signal-process segment" and output segment confidence level. Example: Edge-class defects follow Drift: High-weighted directional correction section; Wrinkles and tension gradient exceeding limits, abnormal oscillating rollers: directional tension section / flattening section; Accompanying bubbles / poor bonding. Decline or Increased temperature and abnormal temperature difference: indicates a temperature zone / pressing section; periodic defects concentrated in a specific phase box: indicates localized defects on the roller surface or periodic mechanical disturbances. When At that time, Duan Inclusion in attribution set .

[0121] (4) Actuator channel open: Only open Corresponding actuator channel For example, temperature zone settings (single or multiple zones), pressing pressure, and linear velocity; unattributed segments remain unchanged. If dual attribution exists (both segments are high confidence), it can be set as follows: The adjustment range can be proportionally allocated or a "primary first, secondary second" order can be adopted to reduce the cost of misjudgment.

[0122] (5) Maintenance linkage: If the phase statistics indicate periodic mechanical problems, the system generates maintenance suggestions (cleaning / replacing the roller surface, checking the bearing eccentricity), so that quality control and maintenance form a closed loop.

[0123] S8 Tangential constraint parameter tuning increment Solving and distributing

[0124] (1) Definition of parameter adjustment variables: Let the working condition vector be... (Temperature region can be a vector) ), adjust the parameters to The system is based on the attribution set. Determine the dimensions of the variables involved in the solution (e.g., only call) Or adjust only certain temperature zones).

[0125] (2) Setting tangential constraints: to avoid crossing Boundary constraints cause adhesion failure; tangential constraints are used instead.

[0126] ;

[0127] in: This is an approximation of the boundary normal. This is the tangential tolerance. The engineering significance of this constraint is to move along the boundary tangentially as much as possible, eliminating the risk of defects without sacrificing the adhesion safety margin.

[0128] (3) Construction of the objective function: Under the condition of satisfying hard constraints (upper and lower limits of variables, upper limit of rate of change, ... Minimize, under constraints (including tangential constraints):

[0129] ;

[0130] in: The defect risk indicator (composed of defect count, confidence accumulation, and weighted critical defects); Suppress excessive adjustments; Indicates the margin increase; These are the weighting coefficients.

[0131] (4) Solution method: The variable dimension is usually low (3-10 dimensions), so the following method can be used: Projection method: First reduce the dimension. The direction gives the candidate Then project it onto the tangent plane and truncate it within the safety constraints; Small-scale quadratic programming (QP): Form a QP with the objective and linear / quadratic constraints, and solve it in real time on an industrial computer. After solving, it needs to be verified again: the quantization step size still satisfies the requirements. With tangential constraints, cross-boundary quantization is avoided.

[0132] (5) Issue Interlock: When a safety interlock is triggered (over-temperature, over-pressure, over-tension, major alarm), issue the interlock immediately and revert to the previous state. At the same time, this sample is marked as a failure sample and will not participate in the self-learning update to avoid "abnormal data polluting the model".

[0133] (6) Step size quantization: To match the resolution of the field actuator, Quantization can be performed in fixed steps (e.g., 0.5–2℃, 0.5–5% pressure, 0.5–3% speed). After quantization, the measurement must be repeated. Internal validation.

[0134] S9 Continuous verification release and backtracking step size rollback

[0135] (1) Verification window definition: Send Then the verification phase begins. The verification window can be defined as "continuous". "Effective peel measurement" or "film length window covering propagation delay". To avoid the effects of acceleration and deceleration, it is recommended to only include window samples in the velocity stable region.

[0136] (2) Release criteria: Release requires the simultaneous fulfillment of: a) Adhesion safety: a) Does not decrease within the verification window, and shows an overall increasing trend (allowing for minor noise); b) Defect improvement: target defect risk The parameter is significantly reduced in the latter half of the window (considering propagation delay, the latter half has a higher weight). If this condition is met, the parameter tuning is deemed effective and a stable candidate is generated. .

[0137] (3) Backtracking step size: When any of the following situations occur: Instead of increasing, it decreased, and appeared External point If the price rises instead of falling, a backtracking process is triggered:

[0138] ;

[0139] And revert to the previous stable setting. Then execute S8 again to resolve for the smaller step size.

[0140] (4) Maximum rollback count and conservative mode: When the rollback count exceeds If the number of consecutive invalid strippings exceeds the threshold, the system enters conservative mode: freezes adaptive parameter tuning, limits the maximum speed, operates with conservative temperature and pressure settings, and prompts for manual inspection to prevent repeated attempts and losses caused by equipment malfunctions.

[0141] (5) Suppression of propagation delay misjudgment: Verification judgment needs to be combined with S4 delay correction to avoid misjudgment of "parameter tuning has taken effect but the defect has not yet reached the detection point". The weight of the first half of the window can be reduced and the weight of the second half can be increased.

[0142] (6) Curing rule: Only when continuous Only when all windows are successfully opened and the defect rate is consistently below the threshold will the process be... Solidify to the new stable settings It also allows writing initial values ​​recommended by the recipe.

[0143] S10 Batch Traceability, Stable Setting Fixing and Controlled Updates

[0144] (1) Batch data record fields: at least record: recipe version, Trajectory, defect records (type / phase box / location / confidence), anchor events, Curve, stripping measurement sequence , Internal operation ratio, number of rollbacks, number of releases, final The logs are stored in a structured format (database / JSON / CSV) and can be linked to the original image index for traceability.

[0145] (2) Stable setting, curing, and first article pass: When a material combination operates stably in multiple batches and When the internal ratio reaches the set threshold, The initial value of the formula is fixed and applied directly when the machine is started for the next batch, which improves the first-piece pass rate and shortens the ramp-up time.

[0146] (3) Threshold-controlled updates: allow for Small, controlled updates can be performed, but must meet the following conditions: the update is performed during a stable operating phase; and the update must pass at least [a certain threshold] afterward. The system verifies the data in a single window; otherwise, it automatically rolls back. This mechanism ensures that it adapts to batch and differential testing without compromising safety standards.

[0147] (4) Maintenance tips and health assessment: Based on indicators such as phase box defect distribution, odometer slippage frequency, number of ineffective peeling, and retraction frequency, maintenance suggestions are generated, such as roller surface cleaning, tool replacement, and inspection of clamping and bearings, forming a long-term closed loop of "control-maintenance".

[0148] (5) External interface: Report key indicators (defect rate, etc.) to MES / SCADA. Features include internal percentage, rollback times, stability settings, etc., supporting enterprise-level quality statistics and traceability.

[0149] The foregoing description of embodiments of the present invention, through which those skilled in the art are able to implement or use the present invention, will be readily apparent to those skilled in the art. Various modifications to these embodiments will be readily apparent to those skilled in the art. 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 novelty disclosed herein.

Claims

1. An intelligent control method for a laminating machine, applied to a roll-to-roll laminating machine, characterized in that: Data collection includes deviation, tension / floating roller displacement, linear velocity, pressing pressure, and zone temperature. Online vision is used to obtain defect types and their phase coordinates relative to the film edge periodic markers. The elongation rate is estimated by the length difference of the odometer rollers before and after the pressing section to correct for defect backtracking delay, and the backtracking zero point is reset by marking phase abrupt changes. Sample strips are cut from the edge material recycling channel and subjected to online micro-peeling of a fixed length L at a peel angle of 90±5° and a peeling speed proportional to the linear velocity, yielding the average peeling value per unit width P, fluctuation CV, and peeling work W. A temperature-pressure-speed allowable domain Ω is constructed from P, CV, and W, and the adhesion margin S is calculated. Based on the defect and process signal, the segment attribution is completed. Only the actuator of the attribution segment generates the parameter adjustment increment Δu within Ω, and Δu satisfies the requirement of increasing S monotonically along the tangential direction of the Ω boundary. If the increase of S is not satisfied after M consecutive samplings, the step size is reduced according to the backtracking factor and the step size is returned to the previous stable setting.

2. The method according to claim 1, characterized in that: During the initialization phase, the product standard parameters Pmin, Pmax, Cmax, Wmin, sampling length L, strip width w, peeling speed proportional coefficient k_v, and safety upper limits for tension / deviation / temperature / pressure / torque and parameter adjustment change rate are loaded to limit the allowable domain Ω and parameter adjustment search boundary.

3. The method according to claim 1, characterized in that: A unified clock synchronization and buffer queue management system is implemented for deviation amount, tension, speed, pressure, temperature, torque, and defect phase coordinates, so that defect backtracking delay correction and online micro-peeling measurement are correlated under the same time reference.

4. The method according to claim 1, characterized in that: The periodic markers are printed markers, laser-engraved markers, or embossed markers; the online vision outputs the phase value of the defect relative to adjacent markers and the longitudinal coordinates within the marker spacing, and triggers a zero-point reset when a sudden change in marker phase caused by film splicing, roll changing, or missing markers is detected.

5. The method according to claim 1, characterized in that: The elongation rate is obtained by filtering the difference between the cumulative length of the upstream odometer roller and the cumulative length of the downstream odometer roller in the pressing section, and this elongation rate is used simultaneously for the joint correction of defect backtracking distance and backtracking time delay.

6. The method according to claim 1, characterized in that: The online micro-peeling is performed in an independently guided edge material path; the peeling speed is controlled by a servo drive and satisfies v_peel=k_v·v; the peeling work W is the integral of the peeling force-displacement curve over a fixed length L; the allowable domain Ω is at least limited by P∈[Pmin,Pmax], CV≤Cmax, and W≥Wmin; the "tangential along the Ω boundary" includes estimating the Ω boundary normal n at the current working point and letting Δu satisfy n·Δu=0 or |n·Δu|≤δ and not crossing Ω; the backtracking step size includes shrinking Δu by the coefficient β (0<β<1) and backtracking to the previous stable setting before verification.

7. An intelligent laminating machine control system, characterized in that, include: The system includes a data acquisition unit, an online visual inspection unit, a film edge marking unit, upstream and downstream odometer rollers, an elongation and backtracking correction unit, an edge material sampling and online micro-peeling unit, an allowable domain Ω construction unit, an intra-segment attribution unit, a tangential constraint parameter tuning unit, and a backtracking verification and rollback unit. The edge material sampling and online micro-peeling unit outputs P, CV, and W, and the allowable domain Ω construction unit forms the temperature-pressure-speed allowable domain Ω based on P, CV, and W and calculates the adhesion margin S. The tangential constraint parameter tuning unit generates Δu and applies tangential constraints only within Ω to the attribution segment actuator; the backtracking verification rollback unit performs step size reduction and rollback when S is not satisfied for M consecutive times.

8. The system according to claim 7, characterized in that: The edge material sampling and online micro-peeling unit is located in the edge material recycling channel and includes an edge trimming and cutting mechanism, an independent guiding mechanism, a 90±5° peeling angle guiding structure, a servo peeling drive, and a force sensor. It is used to achieve zero-interference sampling of the main strip and fixed-length L micro-peeling measurement. The tangential constraint parameter adjustment unit includes a boundary normal estimation module, which is used to estimate the boundary normal n based on the constraint relationship of the allowable domain Ω, and apply n·Δu=0 or |n·Δu|≤δ and a maximum rate of change limit to Δu. The backtracking verification and rollback unit includes a backtracking factor β configuration module and a stable setting storage module, which are used to implement step size reduction and parameter rollback.

9. A computer device, characterized in that, The computer device includes a processor, a graphics processing unit (GPU), and a memory, wherein the memory stores a computer program that, when executed by the processor and the GPU, causes the computer device to perform the method as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a computer device, causing the computer device to perform the method as described in any one of claims 1 to 6.

Images