A method for recommending parameters of a multilayer optical film facing to light transmittance and haze index backstepping

Abstract

This invention discloses a method for recommending parameters of multilayer optical films by back-calculating transmittance and haze indices, belonging to the field of optical thin film data analysis technology. The method includes: obtaining a global deviation sequence characterizing the differences in the optical performance of products; constructing a compressed sensitivity matrix based on the functional division table of film layers; extracting sub-matrices from the compressed sensitivity matrix according to functional groups and calculating contribution scores; sequentially performing cascaded regularized least squares solution, deducting explained deviations and updating residuals at each level until convergence; and concatenating the correction values ​​at each level into a joint parameter correction vector as the basis for recommending process parameters. This method embeds prior knowledge of the physical structure of the film system during the inversion process, suppresses ill-conditioned problems through hierarchical solution, and improves the process feasibility of the parameter recommendation results.

Description

Technical Field

[0001] This invention relates to the field of optical thin film data analysis technology, and more specifically, to a method for recommending parameters of multilayer optical films by back-calculating transmittance and haze indices. Background Technology

[0002] In the mass production of multilayer optical film products, to ensure that the optical performance indicators such as light transmittance and haze meet the design standards, the industry generally uses online spectrometers to collect the spectral data of the products in real time and compares it with the preset standard spectrum. When a deviation is detected between the measured spectrum and the standard spectrum, it is usually necessary to reverse-calculate the required process parameter correction based on the pre-established correlation model between process parameters and optical indicators, and adjust the subsequent production equipment accordingly. To achieve this reverse-calculation process, the mainstream technical route includes the perturbation difference method based on the optical thin film transfer matrix theory. That is, by applying small perturbations to adjustable parameters such as the physical thickness and material concentration of each film layer, the resulting spectral response is calculated to construct a sensitivity matrix, and then a set of parameter corrections that can compensate for the current spectral deviation is solved using numerical methods such as least squares method or singular value decomposition. This method is widely accepted because of its clear physical meaning and high consistency with thin film optical theory.

[0003] However, the current problem is that conventional inversion methods, when dealing with two very different optical parameters—transmittance and haze—usually either simply concatenate the deviations of the two parameters and then perform a single matrix inversion, or mechanically superimpose corrections after independently inverting the two parameters. Neither of these approaches fully considers the fundamental differences in the influence of different functional layers on transmittance and haze in multilayer optical film structures, nor their interlayer coupling effects. Specifically, conventional methods neglect the specific contribution of the scattering-dominant layer to haze, the interference modulation effect of the transmission-dominant layer on transmittance, and the optical transition characteristics of the interface coupling layer between adjacent layers. This leads to the inversion process easily misallocating deviations that should be explained by specific functional layers to other film parameters with no physical response or extremely weak response. This easily causes ill-conditioned problems in the inversion equations, manifesting as drastic numerical oscillations in the correction calculation results, exceeding the physical adjustment limits of the equipment, or introducing new haze deviations while correcting transmittance deviations. Ultimately, the recommended parameter values ​​obtained through inversion lack technological feasibility and are difficult to directly apply to the closed-loop control of production line equipment.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a method for recommending parameters of multilayer optical films by back-calculating transmittance and haze indices, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for recommending parameters of multilayer optical films by back-calculating transmittance and haze indices includes the following steps: collecting the measured transmittance and haze value sequences of current multilayer optical film products, subtracting them from the corresponding transmittance and haze standard value sequences, and splicing them together to obtain a global deviation sequence;

[0008] Based on the current product model, retrieve the membrane layer functional division table, construct an initial sensitivity matrix based on the perturbation differential response of each membrane layer's adjustable parameters, perform physical threshold mask determination and functional linkage mask determination on the initial sensitivity matrix to obtain the threshold mask matrix and functional mask matrix, perform element-wise logical AND operation on the threshold mask matrix and functional mask matrix and multiply them element-wise by the initial sensitivity matrix to obtain the compressed sensitivity matrix;

[0009] Extract each functional submatrix from the compression sensitivity matrix according to the grouping of the membrane functional division table, calculate the contribution score of each functional group to the global deviation sequence, normalize and sort to form an inversion priority sequence, perform regularized least squares solution step by step according to the inversion priority sequence to obtain the parameter correction amount at each level, and after each level of solution, deduct the explained deviation from the current residual and pass it to the next level until the residual converges or all functional groups are processed, and concatenate the parameter correction amounts at each level into a joint parameter correction vector;

[0010] Based on the compression sensitivity matrix and the product-bound optical tolerance limit, the process tolerance adjustment range is calculated for each parameter in the joint parameter correction vector, and then converted into an equipment instruction message and sent to the corresponding workstation controller for execution.

[0011] In a preferred embodiment, the acquisition of a sequence of measured transmittance and a sequence of measured haze values ​​for current multilayer optical film products includes:

[0012] The production line uses an online spectrometer to scan the current product at fixed wavelength sampling intervals within the visible light band, obtaining several wavelength sampling points. For each wavelength sampling point, the arithmetic mean of the number of consecutive sampling integrations is taken as the final reading for that wavelength point. When the standard deviation between the consecutive sampling values ​​of a certain wavelength sampling point exceeds the preset sampling stability threshold, the number of samplings is increased to twice the number of consecutive sampling integrations, and the mean is taken again. The measured transmittance values ​​of all wavelength sampling points are arranged in ascending order of wavelength to form a transmittance measured value sequence, and the measured haze values ​​are arranged in the same order to form a haze measured value sequence.

[0013] In a preferred embodiment, the global deviation sequence is obtained by subtracting from and concatenating the corresponding transmittance standard value sequence and haze standard value sequence, including:

[0014] The transmittance deviation sequence is obtained by subtracting the measured transmittance value sequence from the transmittance standard value sequence element by element. The haze deviation sequence is obtained by subtracting the measured haze value sequence from the haze standard value sequence element by element. The transmittance deviation sequence is arranged in order of wavelength from low to high as the upper half of the column vector, and the haze deviation sequence is arranged in order of the same wavelength as the lower half of the column vector, forming the global deviation sequence.

[0015] In a preferred embodiment, an initial sensitivity matrix is ​​constructed based on the perturbation differential response of the tunable parameters of each film layer, including:

[0016] Using the wavelength sampling point as the row reference, perturbation differential calculation is performed on each adjustable parameter one by one. For each adjustable parameter, a positive perturbation increment is applied to the parameter from its nominal value, and the nominal values ​​of the other parameters are substituted into the optical thin film transfer matrix calculation model. The transmittance sequence and haze sequence under the perturbation state are calculated point by point at each wavelength. Then, all nominal values ​​are substituted into the same model to calculate the transmittance sequence and haze sequence under the nominal state. The perturbation state sequence of each adjustable parameter is subtracted element by element from the nominal state sequence and divided by the positive perturbation increment to obtain the sensitivity elements of the adjustable parameter at each wavelength index position. The elements are arranged in row and column order to form the initial sensitivity matrix.

[0017] In a preferred embodiment, physical threshold masking and functional linkage masking are performed on the initial sensitivity matrix to obtain a threshold mask matrix and a functional mask matrix, including:

[0018] The functional division table of film layers is retrieved from the process database. This table divides all film layers of the current product into four functional groups: Functional group A is the transmission-dominant layer, characterized by film layers that mainly regulate light transmittance through interference effects; Functional group B is the scattering-dominant layer, characterized by film layers that mainly affect haze; Functional group C is the interface coupling layer, characterized by transitional film layers that couple with the optical behavior of adjacent layers between different functional layers; and Functional group D is the functional protection layer, characterized by film layers that have a very weak impact on optical indicators.

[0019] The physical threshold mask determination is as follows: for each element position of the initial sensitivity matrix, check whether the absolute value of the sensitivity is less than the minimum response threshold. If it is less, the threshold mask matrix is ​​marked as 0 at the corresponding position; otherwise, it is marked as 1.

[0020] The functional linkage mask determination is performed on each element of the initial sensitivity matrix based on the functional assignment of each membrane layer and the physical positional relationship between layers in the membrane layer functional division table. Specifically: for the parameters of functional group C, i.e. the interface coupling layer, only when the threshold mask of the upper and lower adjacent layers of the interface coupling layer is 1 in the corresponding index row, the corresponding position in the functional mask matrix is ​​marked as 1; otherwise, the corresponding position in the functional mask matrix is ​​marked as 0.

[0021] For the parameters of functional group B, i.e. the scattering-dominant layer, the corresponding position in the transmittance index row is marked as 1 only if the ratio of the design thickness of the scattering-dominant layer to the corresponding wavelength exceeds the scattering effectiveness threshold; otherwise, the corresponding position in the functional mask matrix is ​​marked as 0. For other positions, the corresponding position in the functional mask matrix is ​​assigned a value according to whether the corresponding film layer in the film layer functional division table is marked as the effective influence layer of the corresponding index.

[0022] In a preferred embodiment, the contribution score of each functional group to the global deviation sequence is calculated and normalized to form an inversion priority sequence, including:

[0023] From the compressed sensitivity matrix, extract the column blocks corresponding to each of the four functional groups in the membrane functional division table to obtain four functional sub-matrices. Multiply the transpose of each functional sub-matrix by the global deviation sequence, and take the L2 norm of the resulting column vector as the contribution score of the functional group. The contribution score reflects the degree of matching between the effective sensitivity retained by the functional group in the compressed sensitivity matrix after double masking and the current deviation. The contribution score of functional groups whose non-zero elements in the functional sub-matrix are largely masked by double masking is reduced accordingly.

[0024] Sum the contribution scores of all functional groups to obtain the total contribution score; divide the contribution score of each functional group by the total contribution score to obtain the normalized contribution score of each functional group.

[0025] Functional groups whose normalized contribution is below the minimum contribution threshold are marked as frozen and not allowed to participate in subsequent inversion; the remaining functional groups are arranged in descending order of normalized contribution to form an inversion priority sequence.

[0026] In a preferred embodiment, regularized least squares solution is performed step-by-step according to the inversion priority sequence to obtain the parameter corrections at each level, including:

[0027] Let the initial residual sequence be equal to the global bias sequence; starting from the functional group with the highest contribution according to the inversion priority sequence, solve the parameter correction amount of the functional group at level i with the current residual sequence as input; where i is the level index in the inversion priority sequence, and i=1,2,…,M, M is the total number of active functional groups after contribution filtering.

[0028] The parameter correction is calculated by multiplying the transpose of the functional group submatrix by itself to obtain a square matrix, then adding the product of the smoothing adjustment coefficient of that level and the identity matrix to form a regularization coefficient matrix. The regularization coefficient matrix is ​​then inverted and multiplied by the product of the transpose of the functional group submatrix and the current residual sequence. The smoothing adjustment coefficient is equal to the smoothing adjustment coefficient benchmark value divided by the normalized contribution of the functional group.

[0029] In a preferred embodiment, after each level of solution, the explained bias is subtracted from the current residual and passed to the next level, until the residual converges or all functional groups have finished processing, including:

[0030] Multiply the parameter correction obtained from the i-th level solution with the functional group submatrix to obtain the explanation of the bias for that level correction. Subtract the explanation from the current residual sequence to obtain the updated residual sequence. Calculate the ratio of the L2 norm of the updated residual sequence to the L2 norm of the initial global bias sequence to obtain the residual convergence ratio.

[0031] When the residual convergence ratio is lower than the preset convergence threshold, the inversion is terminated and the remaining functional groups no longer participate in the inversion; otherwise, the updated residual sequence is used as the input for the next level of inversion and the next level of inversion is continued until all functional groups have been processed.

[0032] In a preferred embodiment, based on the compression sensitivity matrix and the product-bound optical tolerance limits, a process tolerance adjustment range is calculated for each parameter in the joint parameter correction vector, and converted into an equipment instruction message which is then sent to the corresponding workstation controller for execution, including:

[0033] For each parameter component in the joint parameter correction vector, the diagonal sensitivity value of the corresponding column in the compression sensitivity matrix is ​​taken as the parameter's intrinsic sensitivity. The maximum allowable deviation of the parameter from the correction center value is obtained by dividing the optical tolerance limit by the absolute value of the intrinsic sensitivity. The process tolerance adjustment range is equal to the correction center value plus or minus the maximum deviation. The corrected physical parameter values ​​are converted into corresponding electromechanical control parameters through the production line equipment process model, and grouped and encapsulated into equipment instruction messages according to the workstation number and sent to the controller of the corresponding workstation for execution. The production line equipment process model is a pre-established set of parameter mapping relationships stored in the equipment control database, which contains the conversion function from the physical parameters of each workstation to the equipment control quantity.

[0034] In a preferred embodiment, before converting the corrected physical parameter values ​​into corresponding electromechanical control parameters through the production line equipment process model, and grouping and encapsulating them into equipment instruction messages according to workstation numbers and sending them to the controller of the corresponding workstation for execution, the method further includes:

[0035] For each parameter's correction center value, perform equipment limit verification: compare the corrected physical parameter value with the parameter adjustment upper limit and lower limit pre-stored at the workstation. If the corrected value exceeds the parameter adjustment upper limit or falls below the parameter adjustment lower limit, clamp the correction center value to the corresponding limit value and mark the equipment limit clamping triggered in the process parameter recommendation table. After the parameter triggers the equipment limit clamping, automatically include the clamping difference in the residual and start a fast secondary inversion to ensure that the total deviation still meets the optical tolerance limit requirements.

[0036] The technical effects and advantages of this invention, a method for recommending parameters of multilayer optical films based on the back-calculation of transmittance and haze indices, are as follows: This invention constructs a compressed sensitivity matrix by introducing a functional division table of film layers and clearly divides the film layers into different functional groups such as transmission-dominant layers, scattering-dominant layers, and interface coupling layers. This approach embeds prior physical knowledge of the film system structure into the mathematical inversion process, effectively avoiding the problem of incorrectly attributing transmittance deviations to the scattering-dominant layer or haze deviations to the transmission-dominant layer, as is common in conventional methods. The calculated parameter corrections strictly follow the physical functional boundaries of each film layer, and the recommended results are more consistent with the adjustment experience of process engineers, significantly reducing the risk of corrections exceeding equipment limits or causing secondary optical distortions. This scheme does not use a one-time full parameter inversion method, but instead sorts the contribution scores of each functional group to the current deviation, and performs hierarchical regularized least squares solution with a strategy of priority from high to low and deviation explanation deduction at each level. This mechanism decomposes the large-scale underdetermined inversion problem into multiple smaller-scale, conditionally superior problems. The subproblems are solved sequentially, ensuring that each level of calculation focuses only on the most significant deviation component in the current residual. This not only reduces the dependence on the numerical accuracy of the computer but also effectively suppresses the oscillations in the solution caused by the similarity of responses between parameters of different functional groups, ensuring the smoothness and determinism of parameter corrections. This scheme freezes and excludes functional groups with excessively low contributions and sets a residual convergence ratio threshold as a cascade termination condition, enabling timely cessation of inversion after the main components of the deviation have been effectively explained. This avoids overfitting to noise components or minor disturbances and prevents the generation of minor correction suggestions that have no practical technological significance and fluctuate frequently. While ensuring that the product's optical indicators return to the tolerance range, unnecessary equipment operations are reduced, improving the stability of the production line operation. Attached Figure Description

[0037] Figure 1 This is a flowchart illustrating a method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to the present invention.

[0038] Figure 2 This is a schematic diagram illustrating the functional division of the multilayer optical film layers in this invention;

[0039] Figure 3 This is a schematic diagram illustrating the construction of the global deviation sequence and compression sensitivity matrix of this invention;

[0040] Figure 4 This is a schematic diagram of contribution ranking and cascaded regularized least squares inversion in this invention;

[0041] Figure 5 This is a schematic diagram illustrating the calculation of the process tolerance adjustment range and the issuance of equipment commands according to the present invention. Detailed Implementation

[0042] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0043] Example

[0044] Please see Figure 1 As shown, this invention discloses a method for recommending parameters of multilayer optical films by back-calculating transmittance and haze indices, comprising the following steps:

[0045] Step 1: Using an online spectrometer on the production line, collect the measured transmittance and haze values ​​of the current multilayer optical film product point-by-point within the visible light band. Retrieve the standard transmittance and haze value sequences corresponding to the current product model from the product database. Subtract the measured value sequences from the standard value sequences point-by-point and then stitch them together vertically to construct a global deviation sequence that drives all subsequent calculations; specifically, this includes:

[0046] The production line's online spectrometer scans the current product wavelength by wavelength within the visible light band at a fixed wavelength sampling interval; the wavelength sampling interval is denoted as a parameter. In this embodiment, the initial value is set to 5 nanometers. The selection criteria are: if the interval is too large, the spectral details will be lost; if the interval is too small, data redundancy and computational burden will be increased; and a 5-nanometer interval can effectively capture the optical performance change trend in the visible light band and meet the real-time calculation requirements.

[0047] Lower limit wavelength value of the visible light band and upper limit wavelength value Based on the product application scenario, in this embodiment... The initial value is set to 380nm. The initial value was set to 780nm, corresponding to the entire visible light spectrum;

[0048] Based on the above sampling interval, a total of Each wavelength sampling point, The value is equal to the visible light band range divided by the wavelength sampling interval plus 1. In this embodiment... Equals 81; the spectrometer outputs two optical index readings at each wavelength sampling point, including the measured transmittance and the measured haze. The measured transmittance values ​​at all wavelength sampling points are arranged in ascending order of wavelength, forming a sequence of measured transmittance values, denoted as . The length of the sequence is Arrange the measured haze values ​​at all wavelength sampling points in the same order to form a sequence of measured haze values, denoted as . The length of this sequence is also ;

[0049] While collecting measured data, the standard optical parameters corresponding to the current product model are retrieved from the product database based on the current product model code; the standard optical parameters include a sequence of standard transmittance values. and haze standard value sequence Both sequences have a length of 1. Furthermore, the wavelength sampling points of the measured sequence correspond one-to-one; the standard value sequence is determined during the product design stage through a combination of optical thin film theoretical calculations and sample measurement calibration, representing the optical performance that the product should achieve under normal process conditions.

[0050] It should be noted that the standard values ​​in the product model database are pre-calculated and entered by the product design department based on the display performance requirements of end customers. The calculation process is based on the multilayer film simulation results of optical thin film design software such as EssentialMacleod. Each standard value sequence is simultaneously labeled with its corresponding product model code and version number when it is entered to ensure that no confusion occurs during retrieval. The product model database adopts a relational data structure with the product model as the primary key and the standard value sequence is stored in BLOB binary large object format. The query response time does not exceed 100 milliseconds.

[0051] Furthermore, the obtained sequence of measured transmittance values... With transmittance standard value sequence Subtracting element by element yields the transmittance deviation sequence. , its first Each element equals a sequence of measured transmittance values. The Subtract each element The There are elements; where k is the wavelength sampling point index, k is a positive integer, and the value range is k=1,2,… ;

[0052] Similarly, the sequence of measured haze values... Haze Standard Value Sequence Subtracting element by element yields the haze deviation sequence. , its first The elements equal to The Subtract each element The There are 10 elements; where a positive value in the transmittance deviation sequence indicates that the transmittance of the current product at that wavelength is higher than the standard value, and a negative value indicates that it is lower than the standard value; the meaning of the positive and negative signs in the haze deviation sequence is the same.

[0053] transmittance deviation sequence With haze deviation sequence Vertical splicing forms a global deviation sequence Global Deviation Sequence The length is , among which the former Each element represents transmittance deviation, and the subsequent... Each element represents a haze deviation; the vertical splicing specifically involves: first, assembling the transmittance deviation sequence... of The elements are arranged in ascending order of wavelength to form the upper half of the column vector, and then the haze deviation sequence is... of The elements are arranged in order of wavelength to form the lower half of the column vector, ultimately forming a... dimensional global bias sequence ,in In this embodiment dimension;

[0054] The purpose of splicing the two types of index deviations into a single sequence is to enable the sensitivity matrix operation and cascade inversion in subsequent steps to process the deviations of both transmittance and haze in a unified vector space, thereby avoiding the loss of interlayer coupling effect caused by processing the two indicators independently.

[0055] It should be noted that the above wavelength-by-wavelength acquisition method ensures that there is a definite one-to-one correspondence between each element in the deviation sequence and the specific wavelength position. This correspondence is used in step two to map the row position of the deviation sequence to the corresponding row of the sensitivity matrix, which is a prerequisite for the correct execution of the subsequent mask matrix and cascade inversion calculation.

[0056] To ensure the reliability of the measured data during the spectrometer acquisition process, continuous acquisition and integration are performed for each wavelength sampling point. The arithmetic mean is taken as the final reading for that wavelength point. The purpose of this approach is to suppress random measurement noise in a single acquisition of the spectrometer, so that the numerical changes in the deviation sequence truly reflect the process deviation rather than the instrument fluctuation.

[0057] When the standard deviation between consecutively acquired values ​​at a certain wavelength sampling point exceeds the preset acquisition stability threshold... Automatically increase the number of data collections to [number]. Double the result and then take the average again; where the number of consecutive integrals collected is... This is a preset value, initially set to 3 measurements. The calibration basis for this value is: the standard deviation of the repeatability of a single measurement by the online spectrometer on the production line is approximately 0.03 percentage points; after averaging three measurements, the standard deviation decreases to... The error is approximately 0.017 percentage points, which is 0.02 percentage points lower than the spectrometer's nominal accuracy, ensuring data quality without significantly increasing acquisition time; the acquisition stability threshold... Its initial value is set at 0.05 percentage points. The calibration basis of this value is: take 1.5 times the standard deviation of the repeatability of a single measurement of the spectrometer as the judgment limit of abnormal fluctuation. When the standard deviation exceeds the limit, it indicates that there is an abnormal disturbance in the acquisition environment and the number of acquisitions needs to be increased to suppress it.

[0058] It should be noted that the global deviation sequence The dimensions of all elements are percentage points, and the dimensions of transmittance deviation and haze deviation are consistent, so they can be directly spliced ​​and calculated.

[0059] For example, taking a 5-layer optical film product as an example, the production line spectrometer collects 81 wavelength points with measured transmittance and haze values ​​at 5-nanometer intervals within the range of 380 nm to 780 nm, and takes the average of 3 collections for each wavelength point; the standard value sequence for this model is retrieved from the product database; at a wavelength of 550 nm, the measured transmittance value is 91.2%, the standard value is 92.0%, and the deviation is -0.8 percentage points; the measured haze value is 1.8%, the standard value is 1.5%, and the deviation is +0.3 percentage points; at a wavelength of 450 nm, the transmittance deviation is -1.1 percentage points, and the haze deviation is +0.5 percentage points; the 81 transmittance deviation values ​​and 81 haze deviation values ​​are vertically spliced ​​to obtain a global deviation sequence of length 162. The largest absolute deviation in the first 81 elements of the sequence occurred at a wavelength of 420 nm, at -1.4 percentage points, while the largest absolute deviation in the last 81 elements occurred at a wavelength of 400 nm, at +0.8 percentage points.

[0060] This step outputs the global deviation sequence. Pass it on to the next step.

[0061] Step Two: Based on the current product model, retrieve the membrane layer functional division table and nominal values ​​of each membrane layer design parameters from the process database. Using the wavelength sampling point sequence determined in Step One as the row reference, perform perturbation differential calculations on each membrane layer parameter to construct an initial sensitivity matrix. Generate two mask matrices according to the physical threshold determination rules and functional linkage determination rules. Perform a logical AND operation on each element of the two mask matrices and multiply them element-wise with the initial sensitivity matrix to obtain a compressed sensitivity matrix carrying the functional structure information of the membrane system; specifically, including:

[0062] Based on the current product model code, retrieve three data items from the process database: the first item is the functional division table of the film layers, the second item is the set of nominal values ​​of the design parameters of all film layers, and the third item is the table of the dispersion relationship between the refractive index and extinction coefficient of each film layer material in the visible light band and the wavelength.

[0063] Please see Figure 2 As shown, the functional division table of film layers is a structured data table pre-compiled by process engineers during the product design stage based on the material composition, optical function positioning, and interlayer positional relationship of each film layer. The table divides all film layers into four functional groups: functional group A is the transmission-dominant layer, which is the film layer that mainly controls the light transmittance through the interference effect; functional group B is the scattering-dominant layer, which is the film layer that mainly affects the haze; functional group C is the interface coupling layer, which is the transition film layer that has a coupling effect on the optical behavior of adjacent layers; and functional group D is the functional protection layer, which is the film layer that has a very weak effect on optical indicators.

[0064] The nominal set of design parameters includes the values ​​of all adjustable process parameters for all film layers under the design target state, including the physical thickness and coating density of each film layer, etc. Let the total number of adjustable parameters be... In this embodiment, the 5-layer product... It equals 8;

[0065] Please see Figure 3 As shown, subsequently, based on the steps determined in step one... Each wavelength sampling point serves as the row reference. Each adjustable parameter is subjected to perturbation differential calculation to construct the initial sensitivity matrix. The number of rows in the initial sensitivity matrix is ​​equal to the global bias sequence. The length is The number of columns equals the total number of adjustable parameters. The first in the matrix Line number Column elements Indicates the first The adjustable parameter is at the first The optical index response caused by a unit change at each wavelength index position; where m is the adjustable process parameter number, m is a positive integer, and the value range is m=1,2,… Then the initial sensitivity matrix The construction process is as follows:

[0066] For the An adjustable parameter is applied, with a positive perturbation increment applied from its nominal value. , The initial value is set to 0.1% of the nominal value of the parameter. The calibration basis is that the perturbation increment needs to be small enough to ensure that the difference approximates the true derivative, while also being greater than the production line control accuracy resolution to avoid numerical noise. The perturbed parameter value, along with the nominal values ​​of the other parameters, is substituted into the optical thin film transfer matrix calculation model, and the transmittance sequence under the perturbation state is calculated point by point of wavelength. and haze sequence Then, substitute all the nominal values ​​into the same model to calculate the transmittance sequence under the nominal conditions. and haze sequence The calculation results of the nominal state only need to be executed once in the perturbation loop of all parameters;

[0067] For the portion of the initial sensitivity matrix that belongs to the transmittance index row, i.e., the front row... Okay, number Line number The elements of the column are equal to the perturbation transmittance sequence. The Subtract the nominal transmittance sequence from each element The Each element is then divided by the perturbation increment. ;

[0068] For the part belonging to the haze index row, i.e. the latter Okay, number Line number The elements of the column are equal to the perturbation haze sequence. The Subtract the nominal haze sequence from each element The Each element is then divided by the perturbation increment. ; All The perturbation difference results of each parameter are arranged in columns to complete the initial sensitivity matrix. The construction; in this embodiment The dimensions are 162 rows × 8 columns;

[0069] It should be noted that the above-mentioned optical thin film transfer matrix calculation model is a standard theoretical tool in the field of optical thin films. Each film layer is determined by its refractive index, extinction coefficient, physical thickness, and incident light wavelength. The feature matrix description involves multiplying the feature matrices of all layers in stacked order to obtain the total transfer matrix of the film system, from which the transmittance is extracted. Haze calculation is based on the transmittance plus the volume scattering contribution of the dominant scattering layer. An exemplary haze calculation uses a Mie scattering model, with the following formula: ,in Transmitted haze, The intrinsic haze of the scattering-dominant layer, The scattering efficiency factor is determined by the ratio of film thickness to wavelength, and can be directly reproduced by those skilled in the art based on scattering theory.

[0070] The row order of the initial sensitivity matrix and the global bias sequence in step one The row order is strictly consistent, which ensures that the multiplication operation between the submatrix and the deviation sequence in step three corresponds correctly in physical terms;

[0071] After completing the initial sensitivity matrix construction, the position of each element in the matrix is... Perform the following two types of rule determinations, where k is the wavelength-index position corresponding to the row number, and m is the adjustable parameter corresponding to the column number; generate two parameters related to the initial sensitivity matrix. Mask matrices of the same size:

[0072] The first type is physical threshold mask determination: element-by-element inspection. Is it less than the preset minimum response threshold? , The initial value was set to 0.001. The calibration basis was that the measurement noise of the spectrometer after three averages in step one was about 0.017 percentage points. Sensitivity elements with an absolute value lower than 0.001 meant that the parameter response was less than one-tenth of the measurement noise and was physically unobservable.

[0073] when Less than At that time, the threshold mask matrix exist A position marker of 0 indicates masking; when Not less than The time marker is set to 1 to be reserved;

[0074] The second category is functional linkage mask determination: a functional mask matrix is ​​generated based on the functional assignment of each membrane layer in the functional division table and the physical positional relationship between layers. For the parameters of functional group C, i.e., the interface coupling layer, the corresponding position in the functional mask matrix is ​​marked as 1 only when the threshold mask of the upper and lower adjacent layers of the interface coupling layer is 1 in the corresponding index row; otherwise, the corresponding position in the functional mask matrix is ​​marked as 0. The physical basis is that the transition effect of the coupling layer is based on the effective response of the adjacent layer. Among them, the upper and lower adjacent layers of functional group C, i.e., the interface coupling layer, are determined according to the film layer stack order. The upper layer of the nth coupling layer is the (n-1)th layer and the lower layer is the (n+1)th layer. The adjacent layer index corresponds one-to-one with the film layer number in the process database.

[0075] For functional group B, i.e., the parameters of the scattering-dominant layer, at the corresponding position on the transmittance index row, only when the ratio of the designed thickness of the scattering-dominant layer to the corresponding wavelength exceeds the scattering effectiveness threshold... If the function is active, mark the corresponding position in the function mask matrix as 1; otherwise, mark the corresponding position in the function mask matrix as 0. The initial value is set to 0.5, and the calibration basis is that the scattering efficiency increases significantly when the ratio of the feature size to the wavelength exceeds 0.5 according to the Mie scattering theory. For the remaining positions not covered by the above two rules, the film layer is assigned a value according to whether it is marked as an effective influence layer of the corresponding index type in the functional division table. If it is marked as effective, it is assigned 1; otherwise, it is assigned 0.

[0076] Threshold mask matrix With functional mask matrix Perform an element-wise logical AND operation to obtain the synthesized mask matrix; then combine the synthesized mask matrix with the initial sensitivity matrix. Perform element-wise multiplication to obtain the compression sensitivity matrix. The corresponding calculation formula is expressed as:

[0077] In the formula, This indicates element-wise multiplication. This represents an element-wise logical AND operation; The initial sensitivity matrix is ​​constructed through perturbation difference calculation. This is the physical threshold mask matrix. This is a functional linkage mask matrix; the function of this formula is to directly set the sensitivity value of any parameter-index association position that is determined to be invalid or inactive by any mask to zero, so that only elements that have a valid influence relationship in the current product structure are retained in the compressed sensitivity matrix, thereby achieving data dimensionality reduction based on process structure knowledge.

[0078] For example, the initial sensitivity matrix size of a 5-layer product is... The first layer, functional group A, has a nominal thickness of 120 nanometers. The perturbation increment is 0.12 nanometers. After perturbation, the transmittance at a wavelength of 550 nanometers changes from 92.00% to 91.85%, a difference of -1.25 percentage points per nanometer. (Fill in the blank.) Row 35, Column 1; The third layer is the interface coupling layer. The threshold mask of the adjacent layers above and below is 1 at a wavelength of 550 nm, so the functional mask is 1. At a wavelength of 400 nm, the threshold mask of the lower layer is 0, so the functional mask is 0 and is masked. After double masking compression, about 35% of the elements are set to zero.

[0079] Please see Figure 4 As shown, step three: Extract functional sub-matrices from the compressed sensitivity matrix by function group, calculate the contribution score of each functional group to the global bias and normalize and sort them, set an inversion qualification threshold to freeze and exclude functional groups with too low contribution, and perform regularized least squares solution step by step in order of contribution from high to low. After each level of solution, the explained bias is deducted from the current residual and passed to the next level until the residual converges or all active functional groups are processed, and the correction values ​​at each level are concatenated into a joint parameter correction vector; specifically, this includes:

[0080] From the compression sensitivity matrix in step two In the functional division table, the column blocks corresponding to each functional group are extracted according to the grouping information to obtain the functional submatrix. , , , Each submatrix has 10 rows. The number of columns is equal to the number of adjustable parameters contained in each functional group; for each functional group, its contribution score is calculated; the contribution score is calculated by taking the transpose of the functional group's submatrix and the global deviation sequence output in step one. Multiplying the results yields a column vector. The L2 norm of this column vector, i.e., the square root of the sum of the squares of its elements, is taken as the contribution score for that functional group. The contribution score is recorded as Where g is the functional group number, and in this embodiment g∈{A,B,C,D}, the physical meaning of the contribution score is: the projection length of the parameter space of the functional group onto the current deviation vector, which reflects the potential explanatory power of the functional group for the deviation.

[0081] The contribution scores of all functional groups are summed to obtain the total contribution score. The contribution score of each functional group is then divided by the total contribution score to obtain the normalized contribution score of that functional group, ensuring that the total contribution score of all functional groups equals 1. A minimum contribution threshold is then set. Below the minimum contribution threshold Functional groups are marked as frozen and do not participate in the inversion; the remaining functional groups are arranged in descending order of normalized contribution to form an inversion priority sequence; among them, the minimum contribution threshold value is... Its initial value was set at 0.05. The calibration basis was that the deviation explained by functional groups with a normalization contribution of less than 5% was less than one-twentieth of the global deviation, and the inversion benefit was disproportionate to the overfitting risk.

[0082] Set the initial residual sequence to be equal to the global bias sequence from step one: Starting from the functional group with the highest contribution, perform regularized least squares solution level by level according to the inversion priority sequence; for the th in the sequence Level functional group, denoted by its submatrix. With the current residual sequence Using the input as input, solve for the parameter correction amount of this functional group. Where i is the level index in the inversion priority sequence, and i=1,2,…,M, M is the total number of active functional groups after contribution filtering, corresponding to the order of each functional group currently participating in the inversion in the inversion priority sequence; the parameter correction is calculated by multiplying the transpose of the functional group submatrix by itself to obtain a square matrix, and then adding a smoothing adjustment coefficient. With the identity matrix The product of these values ​​forms the regularization coefficient matrix. Multiplying the inverse of this regularization coefficient matrix by the product of the transpose of the submatrix and the current residual yields a vector, which represents the parameter correction for this level. The corresponding calculation formula is as follows:

[0083] ;in express The transpose of the matrix, To and Identity matrices of the same size For the first Smoothing adjustment coefficient for functional groups at all levels; The rule for determining the value is as follows: the baseline value of the smoothing adjustment coefficient is denoted as parameter. Its initial value is set to 0.01. The calibration basis for this value is: performing condition number analysis on the compression sensitivity matrix of a typical 5-layer optical film product in this embodiment, and the functional submatrices of each... The condition number of a square matrix is ​​between 100 and 1000. Setting it to 0.01 reduces the condition number after regularization to between 10 and 100, which meets the engineering requirements for the stability of numerical solutions. equal Divide by the normalized contribution of that functional group; the higher the contribution, the better. The smaller the value, the larger the allowable correction range, and the lower the contribution. The larger the value, the more it suppresses over-correction;

[0084] The first Correction quantity obtained by level solution Substituting into its submatrix, the amount of bias explained by this level of correction is calculated and subtracted from the current residuals to obtain the updated residual sequence; the corresponding calculation formula is expressed as:

[0085] ;in To deduct the first The residual sequence after the explanation of the functional group level. For the first The residual sequence before the first stage of inversion; the updated residual is the remaining bias that has not yet been explained, which will be used as the input for the next stage of inversion;

[0086] After each level of inversion is completed, a convergence check is performed: the ratio of the L2 norm of the current residual sequence to the L2 norm of the initial global bias sequence is calculated, and this ratio is defined as the residual convergence ratio;

[0087] The preset convergence threshold is denoted as parameter. If the residual convergence ratio is still higher than If there are still unprocessed functional groups in the sequence, the next level of inversion will continue;

[0088] Preset convergence threshold Its initial value is set to 0.10. The calibration basis of this value is: when the residual convergence ratio is less than 10%, it indicates that more than 90% of the deviation energy has been explained by each level of inversion. The remaining unexplained deviation magnitude is lower than the product optical tolerance limit. Continuing the inversion will not produce a correction amount with process significance.

[0089] After all cascaded inversions are completed, the adjustments to all functional groups that have undergone inversions are performed according to the parameters. The original column positions are concatenated and assembled into a joint parameter correction vector. The correction amount for the corresponding parameter bits of the freezing functional group that did not participate in the inversion was set to zero; Substitute into the complete compression sensitivity matrix Calculate prediction bias Check whether the absolute values ​​of each component of the prediction bias exceed the back-substitution safety threshold; the back-substitution safety threshold is denoted as parameter. If the prediction bias exceeds... The component indicates that the batch needs to be manually verified, and the specific compliance standard is set according to actual needs; the initial value of the back-substitution safety threshold is set to 1.5 times the maximum allowable deviation of the optical indicators bound to the current product model. The calibration basis of this value is: to add a 50% safety margin on the basis of the optical tolerance limit to cover the superimposed effects of the linear approximation error of the sensitivity matrix and the production line execution deviation.

[0090] For example, the normalized contribution values ​​of the four functional groups in a 5-layer product after contribution calculation are as follows: Functional group A is 0.52, functional group B is 0.31, functional group C is 0.12, and functional group D is 0.05; the contribution value of functional group D is equal to... The function was frozen and excluded; the inversion priority sequence was A→B→C; after solving for functional group A at the first level, the residual convergence ratio decreased from 1.00 to 0.41; after solving for functional group B at the second level, it decreased to 0.15; after solving for functional group C at the third level, it decreased to 0.07, which is lower than... The cascade terminates; the three-level corrections are concatenated and assembled into a joint parameter correction vector. ;

[0091] This step outputs three results: the first is the contribution score and inversion priority sequence of each functional group; the second is the trajectory of the convergence ratio of residuals at each level; and the third is the joint parameter correction vector. The convergence status indicator is attached and passed to step four.

[0092] Please see Figure 5As shown, step four: Based on the intrinsic sensitivity values ​​of each parameter in the compression sensitivity matrix and the optical tolerance limits bound to the product, calculate the process tolerance adjustment range for each parameter in the joint parameter correction vector. Convert the corrected physical parameter values ​​into electromechanical control parameters through the production line equipment process model, encapsulate them into equipment command messages, and send them to the corresponding workstation controller for execution. Specifically, this includes:

[0093] Read the joint parameter correction vector from step three. and its convergence status identifier; read the compression sensitivity matrix from step two. Retrieve the optical tolerance limit associated with the current product model from the product database. This value represents the maximum allowable deviation of optical indicators according to product quality standards. In this embodiment... The value is set at 0.5 percentage points, and this value is determined by the product's factory inspection standards.

[0094] joint parameter correction vector For each parameter component in the matrix, calculate the allowable adjustment range of that parameter while satisfying the overall optical tolerance requirements; the method for calculating the adjustment range is as follows: take the compression sensitivity matrix. The sensitivity value of the main diagonal of the column corresponding to this parameter is defined as the intrinsic sensitivity of the parameter at the wavelength position of its most directly related optical index, and is a direct product of the perturbation differential calculation in step two. Dividing the optical tolerance limit by the absolute value of this intrinsic sensitivity yields the maximum allowable deviation of the parameter from the correction center value. The process tolerance adjustment range of the parameter is equal to the correction center value plus or minus this maximum deviation. The corresponding calculation formula is expressed as:

[0095] ;in The components of the joint parameter correction vector are the correction center values. The vector composed of the main diagonal elements of the compressibility sensitivity matrix represents the intrinsic sensitivity of each parameter. This is the optical tolerance limit; the physical meaning of this formula is that the higher the sensitivity of a parameter, the narrower its allowable adjustment margin, while the lower the sensitivity of a parameter, the more generous its allowable adjustment margin; the resulting interval includes both the correction center value and the upper and lower limits of process feasibility.

[0096] It should be noted that when an element on the main diagonal of the compression sensitivity matrix is ​​masked to zero in step two, it means that the parameter has been determined to have no effective effect on any optical index, and its tolerance range no longer has physical meaning; for such parameters, the system keeps its correction amount at zero and marks it as not participating in adjustment in the process parameter recommendation table;

[0097] The corrected physical parameter values, i.e., the current process parameter settings, plus the correction center value, are converted into corresponding electromechanical control parameters through the production line equipment process model. The production line equipment process model is a pre-established set of parameter mapping relationships, stored in the equipment control database, containing conversion functions from physical parameters of each station to equipment control quantities. In this embodiment, the main mapping relationships involved include: the film thickness correction amount is mapped to the metering pump speed adjustment amount through the coating process model, and its conversion coefficient is determined by the coating die gap, coating liquid viscosity, and linear velocity and calibrated during the equipment debugging stage; the material concentration correction amount is mapped to the magnetron power adjustment amount through the sputtering process model, and its conversion coefficient is determined by the target utilization rate and gas flow rate calibration.

[0098] The converted electromechanical control parameter values ​​and their corresponding tolerance ranges are grouped by workstation number and encapsulated into standardized equipment instruction messages. The data structure of the instruction message includes the following fields: workstation number, control parameter identifier, control quantity setting value corresponding to the correction center value, upper and lower limits of the control quantity corresponding to the tolerance range, timestamp, and batch number. Before the instruction is issued, the system performs equipment limit verification on the correction center value of each parameter: the corrected physical parameter value is compared with the parameter adjustment upper and lower limits pre-stored for that equipment workstation. If the corrected value exceeds the parameter adjustment upper limit or falls below the parameter adjustment lower limit, the system clamps the correction center value to the corresponding limit value and marks the equipment limit clamping triggered in the recommendation table. After the parameter triggers the equipment limit clamping, the clamping difference is automatically included in the residual, and a fast secondary inversion is initiated to ensure that the total deviation still meets the optical tolerance requirements.

[0099] The encapsulated instruction message is sent to the controller at the corresponding workstation for execution via the production line bus communication protocol;

[0100] When the convergence status output in step three indicates that manual review is required, the system only generates a recommended process parameter table for process engineers to review, and does not automatically issue equipment instruction messages. The issuance will be manually triggered after the engineer confirms the message.

[0101] For example, the joint parameter correction vector of a 5-layer product It contains 8 parameter components; the first parameter is the thickness correction of the first layer +2.3 nm, and its sensitivity is -1.25 percentage points per nanometer (this value is obtained from the perturbation differential calculation in step two). The values ​​of the corresponding elements on the main diagonal (after being preserved by masking) and the half-width of the tolerance interval are... Equal to 0.4 nanometers, with a tolerance range of [+1.9 nanometers, +2.7 nanometers]; this thickness correction is converted into an increase of 12 revolutions per minute in metering pump speed through the coating process model; the 5th layer is a functional protective layer belonging to the freezing functional group D, with a correction of zero, marked as not participating in adjustment; the recommended items of the 8 parameters are arranged in sequence to form a process parameter recommendation table, and the electromechanical control parameters are grouped by workstation and packaged into 3 instruction messages, which are sent to the controllers of the coating workstation, sputtering workstation and curing workstation respectively.

[0102] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0103] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0104] Those skilled in the art will recognize that the modules and algorithm modules of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0105] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0106] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0107] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for recommending parameters of multilayer optical films by back-calculating transmittance and haze indices, characterized in that, Includes the following steps: Collect the measured transmittance and haze values ​​of current multilayer optical film products, subtract them from the corresponding standard transmittance and haze value sequences, and then stitch them together to obtain the global deviation sequence. Based on the current product model, retrieve the membrane layer functional division table, construct an initial sensitivity matrix based on the perturbation differential response of each membrane layer's adjustable parameters, perform physical threshold mask determination and functional linkage mask determination on the initial sensitivity matrix to obtain the threshold mask matrix and functional mask matrix, perform element-wise logical AND operation on the threshold mask matrix and functional mask matrix and multiply them element-wise by the initial sensitivity matrix to obtain the compressed sensitivity matrix; Extract each functional submatrix from the compression sensitivity matrix according to the grouping of the membrane functional division table, calculate the contribution score of each functional group to the global deviation sequence, normalize and sort to form an inversion priority sequence, perform regularized least squares solution step by step according to the inversion priority sequence to obtain the parameter correction amount at each level, and after each level of solution, deduct the explained deviation from the current residual and pass it to the next level until the residual converges or all functional groups are processed, and concatenate the parameter correction amounts at each level into a joint parameter correction vector; Based on the compression sensitivity matrix and the product-bound optical tolerance limit, the process tolerance adjustment range is calculated for each parameter in the joint parameter correction vector, and then converted into an equipment instruction message and sent to the corresponding workstation controller for execution.

2. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 1, characterized in that, Collect the measured transmittance and haze values ​​of current multilayer optical film products, including: The production line uses an online spectrometer to scan the current product at fixed wavelength sampling intervals within the visible light band, obtaining several wavelength sampling points. For each wavelength sampling point, the arithmetic mean of the number of consecutive sampling integrations is taken as the final reading for that wavelength point. When the standard deviation between the consecutive sampling values ​​of a certain wavelength sampling point exceeds the preset sampling stability threshold, the number of samplings is increased to twice the number of consecutive sampling integrations, and the mean is taken again. The measured transmittance values ​​of all wavelength sampling points are arranged in ascending order of wavelength to form a transmittance measured value sequence, and the measured haze values ​​are arranged in the same order to form a haze measured value sequence.

3. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 2, characterized in that, The global deviation sequence is obtained by subtracting from the corresponding transmittance standard value sequence and concatenating them, and then including: The transmittance deviation sequence is obtained by subtracting the measured transmittance value sequence from the transmittance standard value sequence element by element. The haze deviation sequence is obtained by subtracting the measured haze value sequence from the haze standard value sequence element by element. The transmittance deviation sequence is arranged in order of wavelength from low to high as the upper half of the column vector, and the haze deviation sequence is arranged in order of the same wavelength as the lower half of the column vector, forming the global deviation sequence.

4. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 1, characterized in that, An initial sensitivity matrix is ​​constructed based on the perturbation differential response of each film layer's tunable parameters, including: Using the wavelength sampling point as the row reference, perturbation differential calculation is performed on each adjustable parameter one by one. For each adjustable parameter, a positive perturbation increment is applied to the parameter from its nominal value, and the nominal values ​​of the other parameters are substituted into the optical thin film transfer matrix calculation model. The transmittance sequence and haze sequence under the perturbation state are calculated point by point at each wavelength. Then, all nominal values ​​are substituted into the same model to calculate the transmittance sequence and haze sequence under the nominal state. The perturbation state sequence of each adjustable parameter is subtracted element by element from the nominal state sequence and divided by the positive perturbation increment to obtain the sensitivity elements of the adjustable parameter at each wavelength index position. The elements are arranged in row and column order to form the initial sensitivity matrix.

5. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 4, characterized in that, Physical threshold masking and functional linkage masking are performed on the initial sensitivity matrix to obtain the threshold mask matrix and the functional mask matrix, including: The functional division table of film layers is retrieved from the process database. This table divides all film layers of the current product into four functional groups: Functional group A is the transmission-dominant layer, characterized by film layers that mainly regulate light transmittance through interference effects; Functional group B is the scattering-dominant layer, characterized by film layers that mainly affect haze; Functional group C is the interface coupling layer, characterized by transitional film layers that couple with the optical behavior of adjacent layers between different functional layers; and Functional group D is the functional protection layer, characterized by film layers that have a very weak impact on optical indicators. The physical threshold mask determination is as follows: for each element position of the initial sensitivity matrix, check whether the absolute value of the sensitivity is less than the minimum response threshold. If the absolute value of the sensitivity is less than the minimum response threshold, mark the corresponding position in the threshold mask matrix as 0; otherwise, mark the corresponding position in the threshold mask matrix as 1. The functional linkage mask determination is performed on each element of the initial sensitivity matrix based on the functional assignment of each membrane layer and the physical positional relationship between layers in the membrane layer functional division table. Specifically: for the parameters of functional group C, i.e. the interface coupling layer, only when the threshold mask of the upper and lower adjacent layers of the interface coupling layer is 1 in the corresponding index row, the corresponding position in the functional mask matrix is ​​marked as 1; otherwise, the corresponding position in the functional mask matrix is ​​marked as 0. For the parameters of functional group B, i.e. the scattering-dominant layer, the corresponding position in the transmittance index row is marked as 1 only if the ratio of the design thickness of the scattering-dominant layer to the corresponding wavelength exceeds the scattering effectiveness threshold; otherwise, the corresponding position in the functional mask matrix is ​​marked as 0. For other positions, the corresponding position in the functional mask matrix is ​​assigned a value according to whether the corresponding film layer in the film layer functional division table is marked as the effective influence layer of the corresponding index.

6. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 5, characterized in that, Calculate and normalize the contribution scores of each functional group to the global deviation sequence to obtain the normalized contribution score, including: From the compressed sensitivity matrix, extract the column blocks corresponding to each of the four functional groups defined in the membrane functional division table to obtain four functional sub-matrices. Multiply the transpose of each functional sub-matrix by the global deviation sequence to obtain the corresponding column vector. Take the L2 norm of the column vector as the contribution score of the functional group. The contribution score reflects the degree of matching between the effective sensitivity retained by the functional group in the compressed sensitivity matrix after physical threshold masking and functional linkage masking and the current deviation. For functional groups whose non-zero elements in the functional sub-matrix are largely masked by the element-wise logical AND operation result of the threshold mask matrix and the functional mask matrix, the contribution score is reduced accordingly. Sum the contribution scores of all functional groups to obtain the total contribution score. Divide the contribution score of each functional group by the total contribution score to obtain the normalized contribution of each functional group. Functional groups whose normalized contribution is below the minimum contribution threshold are marked as frozen and not allowed to participate in subsequent inversion; the remaining functional groups are arranged in descending order of normalized contribution to form an inversion priority sequence.

7. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 6, characterized in that, The parameter corrections for each level are obtained by performing regularized least squares solution step by step according to the inversion priority sequence, including: Let the initial residual sequence be equal to the global bias sequence; starting from the functional group with the highest contribution according to the inversion priority sequence, solve the parameter correction amount of the functional group at level i with the current residual sequence as input; where i is the level index in the inversion priority sequence, and i=1,2,…,M, M is the total number of active functional groups after contribution filtering. The parameter correction is calculated by multiplying the transpose of the functional group submatrix by itself to obtain a square matrix, then adding the product of the smoothing adjustment coefficient of that level and the identity matrix to form a regularization coefficient matrix. The regularization coefficient matrix is ​​then inverted and multiplied by the product of the transpose of the functional group submatrix and the current residual sequence. The smoothing adjustment coefficient is equal to the smoothing adjustment coefficient baseline value divided by the normalized contribution of the functional group.

8. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 7, characterized in that, After each level of solution, the explained bias is subtracted from the current residual and passed on to the next level, until the residual converges or all functional groups have finished processing, including: Multiply the parameter correction obtained from the i-th level solution with the functional group submatrix to obtain the explanation of the bias for that level correction. Subtract the explanation from the current residual sequence to obtain the updated residual sequence. Calculate the ratio of the L2 norm of the updated residual sequence to the L2 norm of the initial global bias sequence to obtain the residual convergence ratio. When the residual convergence ratio is lower than the preset convergence threshold, the inversion is terminated and the remaining functional groups no longer participate in the inversion; otherwise, the updated residual sequence is used as the input for the next level of inversion and the next level of inversion is continued until all functional groups have been processed.

9. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 1, characterized in that, Based on the compression sensitivity matrix and product-bound optical tolerance limits, the process tolerance adjustment range is calculated for each parameter in the joint parameter correction vector, and converted into equipment instruction messages which are then sent to the corresponding workstation controllers for execution, including: For each parameter component in the joint parameter correction vector, the diagonal sensitivity value of the corresponding column in the compressed sensitivity matrix is ​​taken as the intrinsic sensitivity of the parameter. The maximum allowable deviation of the parameter from the correction center value is obtained by dividing the optical tolerance limit by the absolute value of the intrinsic sensitivity. The diagonal sensitivity is the sensitivity value corresponding to the most sensitive wavelength point of the parameter. The process tolerance adjustment range is equal to the correction center value plus or minus the maximum amplitude; the corrected physical parameter values ​​are converted into corresponding electromechanical control parameters through the production line equipment process model, and grouped and encapsulated into equipment instruction messages according to the workstation number and sent to the controller of the corresponding workstation for execution; among them, the production line equipment process model is a set of pre-established parameter mapping relationships stored in the equipment control database, which contains the conversion function from the physical parameters of each workstation to the equipment control quantity.

10. The method for recommending parameters of multilayer optical films based on the reverse calculation of transmittance and haze indices according to claim 9, characterized in that, Before converting the corrected physical parameter values ​​into corresponding electromechanical control parameters using the production line equipment process model, and grouping and encapsulating them into equipment instruction messages according to workstation numbers and sending them to the controllers of the corresponding workstations for execution, the process also includes: For each parameter's correction center value, perform equipment limit verification: compare the corrected physical parameter value with the parameter adjustment upper limit and lower limit pre-stored at the workstation. If the corrected value exceeds the parameter adjustment upper limit or falls below the parameter adjustment lower limit, clamp the correction center value to the corresponding limit value and mark the equipment limit clamping triggered in the process parameter recommendation table. After the parameter triggers the equipment limit clamping, automatically include the clamping difference in the residual and start the secondary inversion to ensure that the total deviation still meets the optical tolerance limit requirements.

Images