Search

CN-121979107-A - Die cutting phase dynamic compensation method for interlayer sliding of composite optical film

CN121979107ACN 121979107 ACN121979107 ACN 121979107ACN-121979107-A

Abstract

The application discloses a die cutting phase dynamic compensation method for compound optical film interlayer slippage, which is applied to a die cutting machine and comprises the steps of obtaining a real geometric center of a surface alignment mark in an actual image collected by a camera, obtaining a target tension signal collected by a tension sensor, performing time stamp alignment and optimal error estimation fusion on the real geometric center and the target tension signal to obtain a comprehensive observation vector, wherein the comprehensive observation vector is used for representing a multi-mode physical state of a film strip section under the current high-speed travelling, performing functional separation and linearization prediction of microscopic slippage trend on the comprehensive observation vector to obtain a theoretical compensation offset, performing line self-adaptive correction on the theoretical compensation offset to obtain an accurate compensation offset, and controlling a circular knife servo motor of the die cutting machine to work according to the accurate compensation offset. By the mode, unidirectional microscopic interlayer slip of the adhesive layer, which is induced by tension difference caused by roll diameter change, is solved.

Inventors

  • Lu Fafa
  • ZHANG GONGMIN
  • Sun Peishan

Assignees

  • 深圳市宇辉光学科技有限公司

Dates

Publication Date
20260505
Application Date
20260409

Claims (10)

  1. 1. The die cutting phase dynamic compensation method for the interlayer slip of the composite optical film is characterized by being applied to a die cutting machine, and comprises the following steps of: acquiring a real geometric center of a surface alignment mark in an actual image acquired by a camera, and acquiring a target tension signal acquired by a tension sensor; Performing time stamp alignment and optimal error estimation fusion on the real geometric center and the target tension signal to obtain a comprehensive observation vector, wherein the comprehensive observation vector is used for representing the multi-mode physical state of the membrane belt section under the current high-speed running; Performing functional separation and linearization prediction of microscopic sliding trend on the comprehensive observation vector to obtain theoretical compensation offset; Performing line self-adaptive correction on the theoretical compensation offset to obtain an accurate compensation offset; and controlling the circular knife servo motor of the die cutting machine to work according to the accurate compensation offset.
  2. 2. The method for dynamically compensating for die-cutting phase according to claim 1, wherein the step of obtaining the true geometric center of the surface alignment mark in the actual image acquired by the camera comprises: The method comprises the steps of obtaining a time-space displacement model, obtaining a physical displacement quantity of a film strip in the exposure time window by obtaining an exposure time window of single imaging of a camera, and calculating the physical displacement quantity of the film strip in the exposure time window according to the real-time linear speed of a current servo spindle; Simulating a smear forming process on a preset surface alignment mark template by using the time-space displacement model to obtain a simulated integral image; Comparing the structural similarity of the analog integral image and the actual image to obtain a pixel-level confidence weight of each pixel point in the actual image; and inversely solving the real geometric center of the surface alignment mark in the actual image according to the pixel-level confidence weight.
  3. 3. The die cutting phase dynamic compensation method according to claim 2, wherein the obtaining the target tension signal collected by the tension sensor comprises: Collecting an original tension signal according to a preset frequency, and generating an observation mask sequence in real time, wherein effective data in the observation mask sequence is marked as 1, and abnormal jump or lost data is marked as 0; Extracting features of the observation mask sequence by utilizing a plurality of groups of one-dimensional sliding time window filters with different scales to obtain local features, wherein different filtering channels respectively capture fluctuation features of different physical properties in the original tension signals; When missing data deduction is carried out, in order to prevent the mutual interference of the characteristics of different physical attributes, a single-pair single-channel hard binding is established, so that each independent local characteristic extraction channel is only physically connected with one exclusive cross-period association weight distribution unit; when data is missing in a certain period, the cross-period associated weight distribution unit of each channel only searches similar fluctuation rules in the context history time window of the similar characteristic channels to perform interpolation calculation, and the target tension signal is obtained.
  4. 4. The die-cut phase dynamic compensation method according to claim 3, wherein the performing time stamp alignment and optimal error estimation fusion on the real geometric center and the target tension signal to obtain a comprehensive observation vector comprises: The theoretical position deviation and deviation change rate of the die-cut section in the running direction at the current moment are predicted in real time by forward recursion of a nonlinear state transfer equation, wherein the nonlinear state transfer equation is built according to the target tension signal, the elastic modulus of the film belt and a kinematic model; Acquiring an accurate exposure trigger time stamp of the actual image according to the real geometric center, acquiring a predicted coordinate corresponding to the exposure trigger time stamp from a state prediction queue of the nonlinear state transfer equation, and calculating a difference value between the predicted coordinate and the coordinate of the real geometric center; Dynamically updating the observed noise covariance matrix in real time according to uncertainty evaluation in the interpolation calculation process and the pixel-level confidence weight; and the comprehensive observation vector is obtained through the closed loop alternating execution of the prediction and the correction.
  5. 5. The die-cutting phase dynamic compensation method according to claim 1, wherein the performing functional separation and linearization prediction of microscopic sliding trend on the integrated observation vector to obtain a theoretical compensation offset comprises: Acquiring the comprehensive observation vectors of N die cutting periods to form a section of time sequence containing machine high-frequency oscillation; Extracting an accumulated displacement curve showing low-frequency and unidirectional smooth drift from the time sequence, wherein the accumulated displacement curve is used for representing the true microscopic interlayer dislocation trend of the adhesive layer in the composite film; Acquiring current working condition characteristics acquired by a micro-tension floating roller at the current moment, wherein the current working condition characteristics comprise an instantaneous tension value, a tension change rate and the environmental temperature of the surface of the film material read by an infrared temperature sensor; Constructing a multidimensional feature state matrix according to the current working condition features; and obtaining the theoretical compensation offset according to the multidimensional characteristic state matrix and the accumulated displacement curve.
  6. 6. The die cut phase dynamic compensation method of claim 5, wherein said deriving said theoretical compensation bias amount from said multi-dimensional feature state matrix and said cumulative displacement curve comprises: Mapping the multi-dimensional feature state matrix into a potential state space by using a dynamic adjustable activator; Collecting front and rear historical state sequences in the potential state space in a continuous limited time window to obtain a global linear state transition matrix, wherein the global linear state transition matrix is used for representing the evolution rule of the adhesive layer slip in a high-dimensional space under the current working condition; Acquiring a high-dimensional state vector at the current moment according to the accumulated displacement curve, and utilizing the high-dimensional state vector and the global linear state transition matrix to estimate the high-dimensional state vector of the next die cutting period; and inversely mapping the high-dimensional state vector of the next die cutting period back to the original three-dimensional physical space, extracting a numerical value representing space displacement, and taking the numerical value as the theoretical compensation offset.
  7. 7. The die-cutting phase dynamic compensation method according to claim 6, wherein the performing line-adaptive correction on the theoretical compensation offset to obtain an accurate compensation offset comprises: In the running process, continuously recording a running state feature set of the equipment, and dividing the running state feature set into a plurality of independent local working condition feature clusters in a multidimensional feature space; acquiring transient characteristic data of the running equipment, calculating Euclidean distances between the transient characteristic data and each local working condition characteristic cluster, and quantifying working condition deviation according to the Euclidean distances; removing failure features in the running state feature set according to the working condition deviation degree, and updating the global linear state transition matrix; Constructing a real-time reference sequence, wherein the reference sequence comprises M theoretical deduction state vectors before cutting and M coordinate deviation values actually generated after cutting; performing linear vector inner product operation by utilizing a die cutting theoretical deduction state to be performed at the current moment and theoretical deduction state vectors before M cutting occurs in the reference sequence to obtain a similarity coefficient; Weighting and summing the coordinate offset values actually generated after the M pieces of cutting by utilizing the similarity coefficient to obtain a numerical compensation offset term; and superposing the numerical compensation offset term on the theoretical compensation offset to obtain the accurate compensation offset.
  8. 8. The die cutting phase dynamic compensation method according to claim 7, wherein the removing the failure feature in the running state feature set according to the working condition deviation degree comprises: Responding to the working condition deviation exceeding a threshold value, and eliminating failure characteristics corresponding to the working condition deviation; said updating said global linear state transition matrix comprising: And for the remaining features in the running state feature set, re-giving weight according to the inverse of the working condition deviation degree, and updating the global linear state transition matrix in real time according to the re-giving weight.
  9. 9. The method for dynamically compensating die cutting phase according to claim 1, wherein the controlling the operation of the circular knife servo motor of the die cutting machine according to the accurate compensation offset comprises: Residual data between a predicted state and a current actual observed state are extracted, and an error covariance matrix of the current state of the system is constructed by combining fluctuation variances of a plurality of continuous sampling periods; performing eigenvalue decomposition on the error covariance matrix to obtain two mutually orthogonal physical components; intercepting and judging the accurate compensation offset according to the two mutually orthogonal physical components; Reading the running line speed of the current knife roller, the clearance tolerance of the bearing and the maximum safe occlusion depth parameter of the blade and the bottom roller in real time, and constructing a dynamic physical safety envelope domain according to the running line speed of the current knife roller, the clearance tolerance of the bearing and the maximum safe occlusion depth parameter of the blade and the bottom roller; Taking the accurate compensation offset as a target input, and embedding the dynamic physical security envelope domain into a solving step of instruction generation; During operation, the probability that the accurate compensation offset breaks through the safety envelope is forcefully evaluated, and the output position instruction of the circular knife servo motor is cut and locked on an absolute safe physical boundary limit value according to the probability; Equivalent conversion of the offset angle to be compensated into a continuous time integral target value is carried out by combining the constant angular speed of the current spindle, and continuous numerical integral accumulation is carried out on servo following errors under a high-frequency clock; When the internal integral accumulated value exceeds a set time integral target value threshold value, the hardware-level trigger releases a microsecond-level electronic pulse signal in a bottom control loop, wherein the microsecond-level electronic pulse signal is used for forcing a servo driver to advance or retard response time on a time axis; All actions from image acquisition, feature extraction, trend deduction, orthogonal interception and boundary constraint to final pulse distribution are defined as independent calculation nodes in the calculation task dependency relationship tree according to the front-back data dependency sequence; And dynamically monitoring the calculation load of each node according to the calculation task dependency relation tree, and carrying out microsecond time slice forced allocation on the multi-core processor to ensure that parallel nodes calculate simultaneously and serial nodes are connected seamlessly.
  10. 10. The die-cut phase dynamic compensation method according to claim 1, wherein the two mutually orthogonal physical components include a high-frequency random noise index and a continuous trend offset index, the performing interception determination on the accurate compensation offset according to the two mutually orthogonal physical components includes: When the high-frequency random noise index is instantaneously increased, intercepting is implemented, a servo compensation channel is locked, and the original phase operation is maintained; And if and only if the high-frequency random noise index is in a gentle low level, and the continuous trend deviation index steadily breaks through the lower limit of the safety tolerance, confirming that the real material interlayer slip occurs, and releasing a compensation instruction.

Description

Die cutting phase dynamic compensation method for interlayer sliding of composite optical film Technical Field The application relates to the technical field of composite optical films, in particular to a die-cutting phase dynamic compensation method for interlayer slippage of a composite optical film. Background In the field of modern high-end display panel manufacturing, the composite optical film is generally formed by bonding multiple layers of polymer soft materials such as polyethylene terephthalate, polyvinyl alcohol, cellulose triacetate and the like through pressure-sensitive adhesives. At present, a circular knife rotary die-cutting unit is generally adopted for continuous high-speed cutting in industrial production, and control logic mainly depends on a high-precision color code sensor to track alignment marks printed on the surface of a film material, and a command is sent to a servo motor based on the alignment marks to control a knife feeding phase. For high-end optical film products, the geometric accuracy of the die-cutting process must be tightly controlled to within + -0.01 mm. Disclosure of Invention The die-cutting phase dynamic compensation method for the interlayer slippage of the composite optical film can construct a complete method system from multi-source signal acquisition, trend separation prediction and on-line parameter self-adaption to physical boundary constraint servo execution, and solves the single and specific interlayer accumulated slippage problem. The application provides a die cutting phase dynamic compensation method for compound optical film interlayer slippage, which is applied to a die cutting machine and comprises the steps of obtaining a real geometric center of a surface alignment mark in an actual image collected by a camera, obtaining a target tension signal collected by a tension sensor, performing time stamp alignment and optimal error estimation fusion on the real geometric center and the target tension signal to obtain a comprehensive observation vector, wherein the comprehensive observation vector is used for representing a multi-mode physical state of a film strip section under the current high-speed travelling, performing functional separation and linearization prediction of microscopic slippage trend on the comprehensive observation vector to obtain a theoretical compensation offset, performing line self-adaptive correction on the theoretical compensation offset to obtain an accurate compensation offset, and controlling a circular knife servo motor of the die cutting machine to work according to the accurate compensation offset. The method comprises the steps of obtaining a time-space displacement model, calculating physical displacement of a film strip in the exposure time window by the time-space displacement model through obtaining an exposure time window of single imaging of a camera and according to real-time linear speed of a current servo spindle, dispersing the physical displacement into a plurality of tiny time slices along a time axis to be constructed, simulating a smear forming process on a preset surface alignment mark template by using the time-space displacement model to obtain a simulated integral image, comparing structural similarity of the simulated integral image and the actual image to obtain pixel-level confidence weight of each pixel point in the actual image, and reversely solving the real geometric center of the surface alignment mark in the actual image according to the pixel-level confidence weight. The method comprises the steps of acquiring an original tension signal acquired by a tension sensor, generating an observation mask sequence in real time, wherein effective data in the observation mask sequence is marked as 1, abnormal jump or lost data are marked as 0, extracting features of the observation mask sequence by utilizing a plurality of groups of one-dimensional sliding time window filters with different scales to obtain local features, respectively capturing fluctuation features of different physical attributes in the original tension signal by different filtering channels, establishing single-pair single-channel hard binding for preventing the features of different physical attributes from interfering with each other when missing data are deduced, enabling each independent local feature extraction channel to be only physically connected with a dedicated cross-period associated weight distribution unit, and searching similar fluctuation rules by the cross-period associated weight distribution unit of each channel only in a context historical time window of the similar feature channels to conduct interpolation calculation when data of a certain period of time are lost, so as to obtain the target tension signal. The method comprises the steps of performing time stamp alignment and optimal error estimation fusion on a real geometric center and a target tension signal to obtain a comprehensive observation vector, perfor