Method and system for accurately regulating and controlling coating thickness of optical lens and intelligently processing optical lens

Abstract

The invention discloses an optical lens coating thickness accurate regulation and control and intelligent processing method and system, and belongs to the technical field of automatic process control, and the method comprises the steps: building a technological process digital twin model based on an optical lens three-dimensional model and a target film system structure, carrying out the virtual coating simulation, and generating an ideal thickness growth curve; reflection spectrum data of a plurality of spectrum reflection probes preset in the coating cavity are collected for synchronous inversion calculation, a real-time thickness distribution diagram is generated, and the real-time thickness distribution diagram is compared with an ideal thickness growth curve to generate a film thickness deviation feature vector; and a technological process digital twinborn model is input to carry out compensation strategy deduction to generate a dynamic technological compensation instruction, and a control signal is sent to a coating equipment execution mechanism to regulate and control coating physical parameters. According to the method, the technological process digital twinborn model is combined with closed-loop control of multi-point real-time monitoring, and dynamic compensation and coordinated regulation and control of space distribution and time rate of film thickness growth can be realized, so that the precision and uniformity of film thickness control are improved.

Description

Technical Field

[0001] This invention relates to the field of automated process control technology, and in particular to a method and system for precise control and intelligent processing of optical lens coating thickness. Background Technology

[0002] Optical thin films are a core technology that alters the light wave transmission characteristics by depositing one or more layers of specific materials on the surface of optical elements. They are widely used in laser systems, optical communications, display technology, and precision instruments to achieve functions such as anti-reflection, reflection, and beam splitting. The performance of optical thin films, such as reflectivity and transmittance, directly depends on the physical thickness and refractive index of each layer in the film system. Therefore, precise control of the film thickness during the coating process is a crucial step in ensuring that optical elements meet their design specifications.

[0003] In existing optical coating processes, film thickness control mainly relies on quartz crystal oscillation monitoring or single-point optical monitoring. Quartz crystal oscillation monitoring indirectly estimates the quality of the film deposited on its surface by measuring changes in the resonant frequency of the quartz crystal, and then converts this into thickness. This method is low-cost and simple to operate, but its monitoring point is usually far from the actual optical lens workpiece and is sensitive to temperature changes, leading to a systematic deviation between the measured value and the actual film thickness on the workpiece. Single-point optical monitoring involves placing a monitoring plate inside the coating chamber and monitoring the film thickness by measuring changes in reflectivity or transmittance at a specific wavelength in real time. Coating is stopped when the optical signal reaches a preset extreme point.

[0004] However, the indirect measurement method of quartz crystal oscillation monitoring makes it difficult to accurately reflect the true condition of the workpiece surface, especially when depositing alloy or compound films, where differences in response to different components lead to a decrease in thickness control accuracy. While single-point optical monitoring can directly measure optical thickness, its monitoring range is limited to a single point. For large-diameter or complex curvature aspherical lenses, due to the non-uniform distribution of the evaporation source particle cloud and the geometric effects caused by the rotation of the workpiece disk, the deposition rate varies significantly at different locations. Single-point monitoring cannot reflect the film thickness distribution across the entire lens surface, often resulting in inconsistent film thickness between the center and edge regions, i.e., poor uniformity, which seriously affects the overall performance and yield of optical components. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for precise control and intelligent processing of optical lens coating thickness. By employing a digital twin model of the process combined with a closed-loop control strategy involving multi-point real-time monitoring, it achieves dynamic compensation and coordinated control of the spatial distribution and temporal rate of film thickness growth, thereby improving the accuracy and uniformity of coating thickness control.

[0006] The above objectives can be achieved through the following approach:

[0007] A method for precise control and intelligent processing of optical lens coating thickness includes: establishing a digital twin model of the process based on a three-dimensional model of the optical lens and the target film structure; performing virtual coating simulation using the digital twin model to generate an ideal thickness growth curve; collecting reflection spectral data from multiple preset spectral reflection probes within the coating cavity and simultaneously inverting the reflection spectral data to generate a real-time thickness distribution map; comparing the real-time thickness distribution map with the ideal thickness growth curve to identify spatial distribution differences and temporal growth rate deviations in film thickness growth, generating a film thickness deviation feature vector; inputting the film thickness deviation feature vector into the digital twin model to perform compensation strategy deduction, selecting a compensation scheme that can simultaneously correct spatial distribution differences and temporal growth rate deviations, and generating a dynamic process compensation command; and sending control signals to the coating equipment actuator according to the dynamic process compensation command to adjust the coating physical parameters.

[0008] Optionally, generating the ideal thickness growth curve includes: analyzing the three-dimensional geometric data of the optical lens from the three-dimensional model of the optical lens to obtain surface feature parameters; obtaining the evaporation physical properties of the coating material based on the target film structure; fusing the surface feature parameters with the evaporation physical properties of the coating material to construct a spatial particle cloud distribution model of the evaporation source; combining the spatial particle cloud distribution model and the motion trajectory parameters of the workpiece disk to construct a digital twin model of the process, and generating the ideal thickness growth curve based on the digital twin model of the process.

[0009] Optionally, generating the film thickness deviation feature vector includes: analyzing the real-time thickness distribution map in the spatial dimension to extract the spatial morphological component that characterizes the film thickness uniformity; performing gradient calculation on multiple monitoring areas in the real-time thickness distribution map in the time dimension to obtain a multi-point growth rate vector; and coupling the spatial morphological component with the multi-point growth rate vector to generate the film thickness deviation feature vector.

[0010] Optionally, after generating the film thickness deviation feature vector, the method further includes: using the multi-point growth rate vector to perform trend prediction and generate a future thickness deviation trend; when the future thickness deviation trend meets the preset intervention conditions, generating a forward-looking intervention signal; and appending the forward-looking intervention signal to the film thickness deviation feature vector to obtain an updated film thickness deviation feature vector.

[0011] Optionally, generating dynamic process compensation instructions includes: simulating and adjusting physical parameters affecting the spatial distribution of film thickness in the digital twin model of the process based on the spatial morphology component in the updated film thickness deviation feature vector, generating a spatial uniformity compensation scheme set; simulating and adjusting physical parameters affecting the overall growth rate of film thickness in the digital twin model of the process based on the multi-point growth rate vector in the film thickness deviation feature vector, generating a global growth rate compensation scheme set; performing cross-validation evaluation on the spatial uniformity compensation scheme set and the global growth rate compensation scheme set, selecting the combination that has the least negative impact on the growth rate while correcting the spatial morphology, and generating dynamic process compensation instructions.

[0012] Optionally, the method further includes: after the coating process is completed, obtaining the final measured film thickness data of the optical lens; performing correlation analysis between the final measured film thickness data and the dynamic process compensation command to generate a model error correction factor; and using the model error correction factor to iteratively update the process digital twin model to generate an updated process digital twin model.

[0013] Optionally, the generated model error correction factor includes: calculating the residual between the final measured film thickness data and the thickness predicted by the digital twin model of the process at the end of the process to obtain a predicted residual distribution map; tracing the dynamic process compensation instructions corresponding to the deviation points that meet the preset conditions in the predicted residual distribution map to obtain inefficient compensation instruction features; quantifying the influence weight of the inefficient compensation instruction features on the physical parameters to generate the model error correction factor.

[0014] Optionally, the method further includes: evaluating the convergence of the predicted residual distribution map generated by the updated process digital twin model, generating a model confidence score; and dynamically adjusting the intervention conditions based on the model confidence score.

[0015] Optionally, sending a control signal to the coating equipment actuator according to the dynamic process compensation instruction includes: decomposing the dynamic process compensation instruction into a time-division control sequence for multiple independent coating physical parameters; performing smooth transition processing on each control point in the time-division control sequence to generate a continuous control curve; and sending the continuous control curve to the coating equipment actuator.

[0016] Based on the same inventive concept, this invention also provides a precise control and intelligent processing system for optical lens coating thickness. The system includes: a digital twin model generation module, used to establish a digital twin model of the process based on a three-dimensional model of the optical lens and the target film structure, and to perform virtual coating simulation using the digital twin model to generate an ideal thickness growth curve; a real-time monitoring module, used to collect reflection spectral data from multiple preset spectral reflection probes within the coating cavity, and to perform synchronous inversion calculations on the reflection spectral data to generate a real-time thickness distribution map; a deviation analysis module, used to compare the real-time thickness distribution map with the ideal thickness growth curve, identify spatial distribution differences and temporal growth rate deviations in film thickness growth, and generate a film thickness deviation feature vector; an intelligent decision-making module, used to input the film thickness deviation feature vector into the process digital twin model for compensation strategy deduction, select a compensation scheme that can simultaneously correct spatial distribution differences and temporal growth rate deviations, and generate dynamic process compensation instructions; and a control execution module, used to send control signals to the coating equipment execution mechanism according to the dynamic process compensation instructions to adjust the coating physical parameters.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. This invention pre-generates an ideal thickness growth curve by constructing a digital twin model of the process, and generates a real-time thickness distribution map by combining real-time monitoring data from multi-point spectral probes. This enables a comprehensive quantitative comparison of the coating process from an ideal baseline to the actual state. This closed-loop feedback control method based on the digital twin model can decompose the complex film thickness control target into an ideal trajectory that can be accurately tracked. This allows the control system to no longer rely solely on endpoint judgment, but to guide and correct the entire growth process, fundamentally improving the accuracy and predictability of film thickness control.

[0019] 2. This invention introduces a forward-looking control strategy, which predicts the future trend of thickness deviation by analyzing the real-time growth rate vector and generates a forward-looking intervention signal before the deviation exceeds the acceptable range. This predictive control mechanism enables the system to shift from passively responding to current errors to actively preventing future errors, intervening in advance and adjusting process parameters, effectively suppressing the cumulative effect of process disturbances in the time dimension, significantly enhancing the dynamic response capability of the control system and the stability of the process, and avoiding the occurrence of major deviations.

[0020] 3. This invention possesses adaptive learning and self-optimization capabilities. After each coating process, the system analyzes the residual between the final measured data and the model prediction, identifies and quantifies the difference between the model and physical reality, and generates a model error correction factor to iteratively update the digital twin model of the process. This closed-loop learning mechanism enables the digital twin model to continuously approximate the real physical process, continuously improving its prediction and simulation accuracy. At the same time, the system dynamically adjusts the initiative of the intervention strategy according to the change in model confidence, realizing the adaptive evolution of the control algorithm, and making the system exhibit higher robustness and accuracy in long-term operation.

[0021] 4. The intelligent decision-making module of this invention can decouple and collaboratively compensate for spatial distribution differences and temporal growth rate deviations. The system uses a digital twin model to quickly deduce and cross-validate different compensation scheme combinations, and selects the optimal scheme that minimizes the negative impact on the overall growth rate while correcting the spatial uniformity of film thickness. This multi-objective collaborative optimization decision-making method solves the problem of neglecting one aspect when adjusting a single parameter in traditional control methods, and ensures the maximization of the comprehensive benefits of compensation actions, thereby enabling the simultaneous realization of complex coating processes with high uniformity and high precision.

[0022] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating a method for precise control and intelligent processing of optical lens coating thickness according to an embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the structure of an optical lens coating thickness precision control and intelligent processing system according to an embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram illustrating the compensation control function of an embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram comparing the uniformity of film thickness spatial distribution in an embodiment of the present invention.

[0028] Figure 5 This is a schematic diagram illustrating the model's self-learning iteration effect in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0030] Reference Figure 1 One embodiment of the present invention proposes a method for precise control and intelligent processing of optical lens coating thickness. By adopting a digital twin model of the process combined with a closed-loop control strategy of multi-point real-time monitoring, dynamic compensation and coordinated control of the spatial distribution and time rate of film thickness growth can be achieved, thereby improving the accuracy and uniformity of coating thickness control.

[0031] The method described in this embodiment specifically includes:

[0032] A digital twin model of the process was established based on the three-dimensional model of the optical lens and the target film structure. The process digital twin model was then used to perform virtual coating simulation and generate an ideal thickness growth curve.

[0033] The reflection spectrum data of multiple preset spectral reflection probes in the coating cavity are collected, and the reflection spectrum data are synchronously inverted and calculated to generate a real-time thickness distribution map.

[0034] Specifically, spectral reflectance probes are first installed at multiple key locations within the coating chamber. These probes are strategically positioned to cover different areas of the optical lens or its representative monitoring sheet, thereby capturing the spatial distribution of the film thickness. After coating begins, each spectral reflectance probe continuously illuminates the substrate surface directly below it with a broadband beam of light, simultaneously collecting the light reflected back from the growing thin film. This reflected light is then sent to a spectrometer for dispersion, yielding the reflectance spectral data for each probe location at the current moment. This data is essentially a curve describing the variation of reflected light intensity with wavelength.

[0035] Next, the system performs synchronous inversion calculations on the reflectance spectral data acquired by all probes. This is a core process for deriving physical parameters from optical phenomena. The goal of this calculation is to accurately determine the current film thickness at each probe location. The principle is based on a known thin-film optical theory model, which can accurately calculate the theoretically expected reflectance spectrum based on a given film thickness, the material's optical constants (refractive index and extinction coefficient), and the substrate's optical properties. The inversion calculation is the reverse operation, finding the film thickness value that minimizes the difference between the theoretical and measured spectra by fitting and comparing the measured reflectance spectral data with the theoretical model. This difference is typically quantified using a least-squares evaluation function.

[0036] E=∑[R meas (λ i )-R theo (λ i ,d)] 2 ,

[0037] Where E is the error function that needs to be minimized. R meas( λ i ) is at a specific wavelength λ i The reflectance value is obtained from actual measurements using a spectral reflectance probe. R theo (λ i d) is the theoretical reflectivity calculated based on an optical model, which is the wavelength λ. i The thickness of the film layer, denoted by d, is a function of the thickness of the film layer to be determined, where the optical constants of the film layer material are input as known parameters. By running a numerical optimization algorithm, the system can quickly solve for the d that minimizes the value of E, which is the real-time film layer thickness at the probe monitoring point. Since the data from all probes are processed synchronously, the system can obtain a set of discrete thickness data points in a short time, each data point corresponding to a spatial coordinate within the cavity. Finally, using spatial interpolation algorithms, such as radial basis function interpolation or polynomial surface fitting, these discrete thickness data points are fitted into a continuous two-dimensional surface. This surface is a real-time thickness distribution map that can intuitively reflect the film thickness distribution state of the entire optical lens surface.

[0038] By comparing the real-time thickness distribution map with the ideal thickness growth curve, the spatial distribution differences and time growth rate deviations of film thickness growth are identified, and a film thickness deviation feature vector is generated.

[0039] The film thickness deviation feature vector is input into the digital twin model of the process to perform compensation strategy deduction, and a compensation scheme that can simultaneously correct spatial distribution differences and time growth rate deviations is selected to generate dynamic process compensation instructions.

[0040] According to the dynamic process compensation command, a control signal is sent to the actuator of the coating equipment to adjust the physical parameters of the coating.

[0041] Specifically, the process begins by establishing a digital twin model that accurately reflects the physical process. Based on this model, virtual coating simulation is performed, and an ideal thickness growth curve representing the optimal process path is pre-generated as a benchmark. During the actual coating process, the system acquires reflectance spectral data from multiple spectral reflectance probes in real time and performs synchronous inversion calculations to obtain a real-time thickness distribution map characterizing the current state. This real-time state is then precisely compared with the ideal benchmark to quantify and identify spatial distribution differences and temporal growth rate deviations in film thickness growth, encapsulating them into a structured film thickness deviation feature vector. This deviation feature vector is fed back to the digital twin model of the process. Leveraging its powerful simulation capabilities, the model quickly seeks the optimal compensation scheme that can simultaneously correct both deviations and generates dynamic process compensation commands. Finally, these commands are sent to the actuators of the coating equipment. By adjusting the coating physical parameters, the optimal decision is transformed into precise correction actions, thus forming a continuously iterative, self-correcting closed-loop control process.

[0042] Optionally, generating the ideal thickness growth curve includes:

[0043] Analyze the three-dimensional geometric data of the optical lens from its three-dimensional model to obtain surface feature parameters;

[0044] The evaporation physical properties of the coating material are obtained based on the target film structure.

[0045] By integrating the surface feature parameters with the evaporation physics properties of the coating material, a spatial particle cloud distribution model of the evaporation source is constructed.

[0046] By combining the spatial particle cloud distribution model and the motion trajectory parameters of the workpiece disk, a digital twin model of the process is constructed, and an ideal thickness growth curve is generated based on the digital twin model of the process.

[0047] Specifically, by reading the 3D model file of the optical lens, the core surface feature parameters characterizing its geometry are analyzed and extracted. These parameters precisely describe the radius of curvature, aspheric coefficient, effective aperture, and precise coordinates of each point in 3D space. Simultaneously, based on the target film structure in the design requirements, the specific coating material to be used is determined, and its evaporation physics properties in a vacuum environment are obtained from a material database or through experimental calibration. These properties primarily include its saturated vapor pressure at a specific operating temperature, the relationship between evaporation rate and power, and the density of the film after deposition. Subsequently, the above two pieces of information are integrated—combining the evaporation physics properties of the coating material with the geometry of the evaporation source—to establish a spatial particle cloud distribution model describing the diffusion behavior of evaporating particles in a vacuum. This model can predict the density and directional distribution of the particle stream emitted from the evaporation source within the cavity. Finally, this spatial particle cloud distribution model is coupled with the motion trajectory parameters of the workpiece disk, such as the workpiece disk's rotational angular velocity, revolution angular velocity, and the fixed position of the optical lens on the workpiece disk, thereby constructing a complete digital twin model of the process. This model can accurately simulate the particle flux received at any point on the surface of an optical lens during the entire coating process under the combined motion of the workpiece disk. By running a virtual simulation of this digital twin model of the process, the theoretically optimal path for the film thickness accumulation over time at different locations on the optical lens surface can be calculated, ultimately generating a cluster of ideal thickness growth curves representing the ideal growth state of the film thickness in both spatial and temporal dimensions. The theoretical deposition rate at any point can be characterized by the following relationship:

[0048]

[0049] Among them, R p ρ represents the instantaneous deposition rate at point p on the optical lens surface, and M represents the mass evaporation rate per unit time of the evaporation source, which is determined by the evaporation physics properties of the coating material and the input process parameters. m The material density of the deposited thin film is also determined by the evaporation physics properties. It is a comprehensive spatial factor function that integrates the particle exit angle θ and the incident angle. The influence of the distance r from the evaporation source to point p on the deposition rate was also considered. These angles and distances were calculated by combining the spatial particle cloud distribution model, the surface feature parameters of the optical lens, and the motion trajectory parameters of the workpiece disk.

[0050] Optionally, the generation of the film thickness deviation feature vector includes:

[0051] The real-time thickness distribution map is analyzed in the spatial dimension to extract the spatial morphology components that characterize the uniformity of film thickness.

[0052] Gradient calculations are performed on multiple monitoring areas in the real-time thickness distribution map over time to obtain a multi-point growth rate vector.

[0053] The spatial morphological component is coupled with the multi-point growth rate vector to generate a film thickness deviation feature vector.

[0054] Specifically, firstly, to address the spatial non-uniformity of film thickness distribution, a spatial dimension analysis is performed on the real-time thickness distribution map obtained through spectral inversion. This analysis compares the real-time thickness data map with the ideal thickness distribution at the same time, resulting in a deviation distribution map. Then, a set of orthogonal basis functions, such as Zernike polynomials suitable for circular optical lenses, are used to fit and decompose this deviation distribution map. The series of coefficients obtained after decomposition constitute the spatial morphology components characterizing the film thickness uniformity. Each coefficient in this component corresponds to a specific spatial distribution morphology, such as overall thickness or thinness, thickness difference between the center and the edge, or a wedge-shaped linear thickness change, thus graphically representing the complex spatial error as a set of structured numerical values. Secondly, to capture the dynamic change trend of film growth, the system performs gradient calculations on multiple preset monitoring areas in the real-time thickness distribution map in the time dimension. Specifically, in several selected key monitoring areas, the real-time values ​​of the film thickness are continuously tracked, and the rate of change of thickness over time, i.e., the instantaneous growth rate, is calculated as follows:

[0055]

[0056] Among them, v j Let d be the instantaneous growth rate of the j-th monitoring area. j Let t be the real-time thickness value of the region and t be the time. By collecting the growth rate data from all monitored areas, a multi-point growth rate vector is constructed. Finally, the spatial morphology component describing the static spatial error pattern is coupled with the multi-point growth rate vector reflecting the dynamic growth trend. Technically, this coupling operation typically involves concatenating the two vectors to form a single vector with a higher dimension and more comprehensive information. This final generated vector is the film thickness deviation feature vector.

[0057] Optionally, after generating the film thickness deviation feature vector, the process also includes:

[0058] The multi-point growth rate vector is used to predict the future thickness deviation trend.

[0059] When the future thickness deviation trend meets the preset intervention conditions, a forward-looking intervention signal is generated;

[0060] The prospective intervention signal is appended to the film thickness deviation feature vector to obtain the updated film thickness deviation feature vector.

[0061] Specifically, after generating the film thickness deviation feature vector representing the current state, this method introduces the prediction and evaluation of future process trends to endow the system with forward-looking control capabilities. First, trend prediction is performed using the obtained real-time multi-point growth rate vector. Specifically, for each monitored area, an extrapolation calculation is performed based on its current thickness deviation and growth rate deviation to generate a future thickness deviation trend. This prediction process can be described by the following relationship:

[0062]

[0063] in, This represents the predicted future thickness deviation value in the j-th monitoring area. It is the difference between the real-time thickness and the ideal thickness of the region at the current moment, that is, the current actual thickness deviation, which is obtained by directly comparing the real-time thickness distribution map with the ideal thickness growth curve. It is the real-time growth rate of the j-th monitoring area extracted from the multi-point growth rate vector. The theoretical ideal growth rate is obtained by differentiating the ideal thickness growth curve at the current time point. Δt is a preset prediction time window, representing the forward-looking time length of the system. Through this calculation, the system can predict the extent to which the thickness deviation in each monitored area will develop after time Δt. Subsequently, the system compares this prediction result, representing the future thickness deviation trend, with a preset intervention condition. This intervention condition is usually one or more deviation thresholds. Once the predicted deviation value exceeds this threshold, it means that if no measures are taken, the process result will soon exceed the allowable tolerance range. When the future thickness deviation trend meets this preset intervention condition, the system immediately generates a forward-looking intervention signal. This signal serves as a clear indicator, indicating that compensation adjustment needs to be initiated in advance. Finally, this forward-looking intervention signal with early warning properties is appended to the original film thickness deviation feature vector, forming an updated film thickness deviation feature vector with richer information dimensions, including the current state and future warnings, for use by subsequent decision-making modules.

[0064] Optionally, the generation of dynamic process compensation instructions includes:

[0065] Based on the spatial morphology component in the updated film thickness deviation feature vector, the physical parameters affecting the spatial distribution of film thickness are simulated and adjusted in the digital twin model of the process, generating a set of spatial uniformity compensation schemes.

[0066] Based on the multi-point growth rate vector in the film thickness deviation feature vector, the physical parameters affecting the overall film thickness growth rate are simulated and adjusted in the digital twin model of the process, and a global growth rate compensation scheme set is generated.

[0067] The spatial uniformity compensation scheme set and the global growth rate compensation scheme set are cross-validated and evaluated. The combination that has the least negative impact on the growth rate while correcting the spatial morphology is selected, and a dynamic process compensation instruction is generated.

[0068] Specifically, upon receiving an updated film thickness deviation feature vector containing the current state and future warnings, the system initiates a multi-objective collaborative optimization decision-making process to generate dynamic process compensation instructions. First, a spatial morphology component specifically describing the uneven spatial distribution of film thickness is extracted from the feature vector. For the deviation morphology revealed by this component, such as a thinner center and thicker edges or a wedge-shaped distribution, the system conducts directional virtual simulation experiments in the digital twin model of the process. In the simulation, a series of physical parameters that primarily affect the spatial distribution of film thickness are systematically simulated and adjusted, such as changing the geometry or position of the correction baffle, or fine-tuning the ratio of the workpiece disk's rotation to revolution speed. By rapidly extrapolating the compensation effects under different parameter combinations, the system generates a set of spatial uniformity compensation schemes containing multiple feasible adjustment options. Simultaneously, a multi-point growth rate vector is extracted from the film thickness deviation feature vector. This vector reflects the deviation between the actual growth rate and the ideal rate of the coating layer in different regions. To address this deviation, the system again utilizes a digital twin model of the process to simulate and adjust physical parameters affecting the overall film thickness growth rate, such as the heating power of the evaporation source or the beam intensity of the electron gun. This process also generates a global growth rate compensation scheme set containing multiple rate adjustment options. Finally, the two scheme sets are cross-validated and evaluated. This process does not simply add the two schemes independently, but rather pairs each scheme in the spatial uniformity compensation scheme set with each scheme in the global growth rate compensation scheme set, and uses the digital twin model of the process to evaluate the overall effect achievable by each combination. The evaluation goal is to find an optimal combination that corrects the spatial morphology components to the greatest extent while minimizing the negative perturbation to the overall growth rate, and vice versa. This optimization process can be quantified by a comprehensive evaluation function:

[0069] J min =min(w s *E spatial +w r *E rate ),

[0070] Among them, J min This represents the minimum overall cost sought. E spatial E represents the residual spatial non-uniformity predicted by a digital twin model of the process after applying a certain compensation scheme. rate This represents the deviation between the predicted average growth rate and the ideal rate. sWith w r The preset weighting factors represent the control system's emphasis on spatial uniformity error and growth rate error, respectively. These two weighting factors are dimensionless parameters pre-set according to specific process requirements. By calculating the comprehensive cost value of all possible combinations, the combination scheme that minimizes the comprehensive cost value is selected and transformed into a specific sequence of executable parameters for the equipment, ultimately generating dynamic process compensation instructions.

[0071] Optionally, the method further includes:

[0072] After the coating process is completed, the final measured film thickness data of the optical lens is obtained;

[0073] The final measured film thickness data is correlated with the dynamic process compensation command to generate a model error correction factor.

[0074] The process digital twin model is iteratively updated using the model error correction factor to generate an updated process digital twin model.

[0075] Specifically, after the entire coating process is completed, to achieve self-evolution and long-term accuracy of the model, this method introduces a closed-loop learning and updating mechanism. First, a final offline precision inspection is performed on the coated optical lens, for example, using an ellipsometer or spectrophotometer, to obtain the final measured film thickness data covering the entire lens surface. This data represents the final physical manifestation of the process results. Next, the system precisely compares this final measured film thickness data with the final thickness distribution map predicted by the digital twin model of the process at the end of the coating process, thereby identifying the deviation between the model prediction and physical reality. A key step is to perform a temporal and spatial correlation analysis of this deviation with a series of dynamic process compensation commands issued by the system throughout the coating process. This analysis aims to trace and identify which specific compensation commands failed to achieve the expected results, i.e., inefficient or ineffective compensation behaviors. By quantifying the gap between the expected and actual results of a particular compensation command, the system can generate a targeted model error correction factor. This correction process can be conceptually described as follows:

[0076]

[0077] in, It is the i-th key parameter to be updated in the digital twin model of the process, such as a coefficient in the particle cloud distribution model of the evaporation source or the response coefficient of the material evaporation rate and power. This is the original value of the parameter. F(ΔD,C) i ) is a correction function whose inputs are the model prediction bias ΔD and the specific dynamic process compensation command C associated with that bias. iThe function's output value reflects the input to the original parameters. The appropriate adjustment ratio should be determined. Finally, using this set of calculated model error correction factors, the relevant physical models or empirical coefficients within the digital twin model of the process are adjusted and calibrated, thereby generating a more accurate and realistically updated digital twin model of the process, providing a more reliable simulation basis for the next coating task.

[0078] Optionally, the generative model error correction factor includes:

[0079] Calculate the residual between the final measured film thickness data and the thickness predicted by the digital twin model of the process at the end of the process to obtain the predicted residual distribution map;

[0080] By tracing the dynamic process compensation instructions corresponding to the deviation points that meet the preset conditions in the predicted residual distribution map, the characteristics of inefficient compensation instructions are obtained.

[0081] The influence weight of the inefficient compensation command features on physical parameters is quantified to generate a model error correction factor.

[0082] Specifically, to generate a model error correction factor capable of accurately calibrating the model, this method performs a meticulous diagnostic and quantification process. First, the point-by-point difference between the final measured film thickness data and the thickness predicted by the digital twin model of the process at the end of the process is calculated, generating a visualized prediction residual distribution map. This map accurately reveals the magnitude and direction of the model prediction deviation at various locations on the optical lens surface. Subsequently, the prediction residual distribution map is analyzed to identify deviation points where the deviation value exceeds a preset tolerance range. For each significant deviation point, its historical data throughout the coating process is traced, and the final error at that point is correlated with a series of historical dynamic process compensation commands acting on that point or its neighboring area. Through this tracing, inefficient compensation commands that fail to effectively reduce the deviation or may even worsen it can be identified. Finally, for each inefficient compensation command feature, the difference between the influence weights of its associated physical parameters in the model and in the real world is quantified, thereby generating a model error correction factor. The calculation of this factor can be expressed by the following relationship:

[0083]

[0084] Where, γ k This is a model error correction factor for the influence relationship of the k-th physical parameter in the model; it is a dimensionless proportionality coefficient. ΔO actualThis represents the amount of observable process result change in the real physical world caused by the inefficient compensation command. This value is calculated by back-calculating the residual value of the corresponding region in the final predicted residual distribution map, combined with the state before and after compensation. This represents the change in process outcome predicted by the k-th physical parameter model in the digital twin model of the process when executing the same inefficient compensation command. This value can be obtained by performing an independent micro-simulation within the digital twin model. This factor γ k It directly quantifies the imbalance between the model's predictive power and the actual physical process response.

[0085] Optionally, the method further includes:

[0086] Evaluate the convergence of the predicted residual distribution map generated by the updated process digital twin model and generate a model confidence score;

[0087] The intervention conditions are dynamically adjusted based on the model confidence score.

[0088] Specifically, to achieve self-evolution of the digital twin model of the process and adaptive optimization of the control strategy, this method introduces a higher-order evaluation and feedback loop after model update. This loop first evaluates the performance convergence of the updated process digital twin model. Specifically, after completing another coating batch, the newly acquired final measured film thickness data is compared with the final thickness predicted by the updated model to generate the latest prediction residual distribution map. By calculating the global error statistic of this distribution map, such as the root mean square value of the residuals across the entire optical lens surface, and comparing it with the error statistics of previous batches, it is determined whether the model prediction accuracy shows a stable decreasing convergence trend. Based on the evaluation results of this convergence trend, a quantitative model confidence score is generated. This score reflects the reliability of the current model; if the model prediction error continues to decrease, the model confidence score will increase accordingly, and vice versa. Subsequently, the system dynamically adjusts the intervention conditions used to trigger proactive intervention based on this model confidence score. The intervention conditions are essentially the deviation thresholds at which the system decides when to intervene in the real-time process. The adjustment process follows a core principle: the model's confidence level determines the aggressiveness of the control strategy. When the model confidence score is high, it indicates that the digital twin model accurately simulates the physical process. The system then tightens intervention conditions, lowering the deviation threshold that triggers compensation actions. This allows the control system to intervene early even for minor future deviation trends, aiming for higher coating accuracy. Conversely, when the model confidence score is low, it means the model currently has significant uncertainty. The system then relaxes intervention conditions, raises the deviation threshold, and adopts a more conservative control strategy, responding only to more significant deviation trends. This avoids erroneous adjustments due to inaccurate model predictions, thus ensuring the stability of the process. This dynamic adjustment process can be expressed by the following formula:

[0089] T new =T max -S model *(T max -T min ),

[0090] Among them, T new This is the newly generated threshold for the intervention condition after adjustment. max T is the preset, most lenient intervention threshold, representing the conservative control boundary of the system at the lowest confidence level. min It is the preset, most stringent intervention threshold, representing the system's control objective of pursuing ultimate precision at the highest confidence level. model The confidence score between 0 and 1 is generated in the aforementioned steps. This score is obtained by analyzing the rate of error reduction of the prediction residual distribution map across multiple consecutive production batches.

[0091] Optionally, sending a control signal to the coating equipment actuator according to the dynamic process compensation command includes:

[0092] The dynamic process compensation command is decomposed into a time-division control sequence for multiple independent coating physical parameters;

[0093] Perform smooth transition processing on each control point in the time-sharing control sequence to generate a continuous control curve;

[0094] The continuous control curve is sent to the actuator of the coating equipment.

[0095] Specifically, after the intelligent decision-making module generates a dynamic process compensation instruction that achieves the optimal compensation effect, this method transforms the high-level instruction into direct hardware control to ensure that the instruction can be executed accurately and stably by the physical equipment. This process first decomposes the comprehensive dynamic process compensation instruction. A dynamic process compensation instruction typically includes composite requirements for multiple process objectives, such as simultaneously adjusting the overall film growth rate and improving its spatial distribution uniformity. Therefore, it needs to be parsed into specific control objectives for multiple independent coating physical parameters. These physical parameters include, but are not limited to, the heating power of the evaporation source, the electron gun beam intensity, the revolution and rotation speed of the workpiece disk, and the geometric position of the correction baffle. After decomposition, a time-sharing control sequence is generated for each controlled physical parameter. This sequence consists of discrete control points composed of a series of timestamps and target setpoints. However, directly sending these discrete step control points to the equipment would cause severe oscillations in the system state and mechanical shocks to the physical actuators, thereby compromising process stability. Therefore, a smooth transition process is then performed on the control points in each time-sharing control sequence. This process employs interpolation algorithms, such as cubic spline interpolation, to connect discrete control points, generating a mathematically continuous control curve with continuous derivatives. This curve ensures a smooth and gradual transition from one setpoint to the next, avoiding any instantaneous abrupt changes. Finally, this continuous control curve, generated for each physical parameter, is sent to the actuator of the coating equipment. Technically, the control system samples this continuous control curve at an extremely high frequency, converting it into a dense stream of digital setpoints, which is then transmitted in real-time to the corresponding equipment controllers, such as power controllers and motor drivers, thereby driving the physical actuators to perform smooth and precise adjustment actions.

[0096] Based on the same inventive concept, such as Figure 2 As shown, the present invention also provides a system for precise control and intelligent processing of optical lens coating thickness, the system comprising:

[0097] The digital twin model generation module is used to establish a digital twin model of the process based on the three-dimensional model of the optical lens and the target film structure, and to use the digital twin model of the process to perform virtual coating simulation and generate an ideal thickness growth curve.

[0098] The real-time monitoring module is used to collect the reflection spectrum data of multiple preset spectral reflection probes in the coating cavity, and to perform synchronous inversion calculation on the reflection spectrum data to generate a real-time thickness distribution map.

[0099] The deviation analysis module is used to compare the real-time thickness distribution map with the ideal thickness growth curve, identify the spatial distribution differences and time growth rate deviations of film thickness growth, and generate a film thickness deviation feature vector.

[0100] The intelligent decision-making module is used to input the film thickness deviation feature vector into the digital twin model of the process to deduce the compensation strategy, select a compensation scheme that can simultaneously correct spatial distribution differences and time growth rate deviations, and generate dynamic process compensation instructions.

[0101] The control execution module is used to send control signals to the execution mechanism of the coating equipment according to the dynamic process compensation instructions, and to adjust the physical parameters of the coating.

[0102] To verify the feasibility of this invention in practice, it was applied to a coating workshop in a precision optical component manufacturing facility. This workshop needed to coat a batch of high-value, large-aperture (Ф150mm) aspherical optical lenses with multilayer antireflective coatings (AR coatings). Traditional processes mainly rely on quartz crystal oscillation monitoring, i.e., crystal control. When dealing with complex curved surfaces, the slight fluctuations in the evaporation source state and the "umbrella effect" often lead to poor film thickness uniformity, low batch-to-batch repeatability, and ultimately, a low yield of the final product.

[0103] In this embodiment, the three-dimensional CAD model of the aspherical lens is first read using the digital twin model generation module. Combined with the material physical properties of the target antireflective coating, such as a four-layer film system composed of alternating Ta2O5 and SiO2, a complete digital twin model of the process is constructed. This model, through virtual coating simulation, generates ideal thickness growth curves for each layer at all locations on the lens surface.

[0104] In the actual coating process, eight spectral reflectance probes installed inside the coating chamber are distributed at different radii, such as the center, 0.5R, and 0.8R. These probes, through a real-time monitoring module, synchronously collect the reflectance spectrum data of the growing thin film. The system performs real-time inversion calculations on this data, generating a real-time thickness distribution map covering the entire lens surface once per second.

[0105] To verify the beneficial effects of this invention, a batch of 50 aspherical lenses was selected for coating experiments, and compared with a control group of 50 lenses using the traditional crystal control method. The experiment lasted for 5 consecutive production batches, focusing on evaluating the uniformity of the film thickness, the final thickness accuracy, and the yield of the finished product.

[0106] During the deposition of the second layer of high-refractive-index material Ta2O5, after approximately 300 seconds, the deviation analysis module detected an anomaly, such as... Figure 3 As shown. By comparing the real-time thickness distribution map with the ideal thickness growth curve, it was found that the growth rate in the central region of the lens was about 4% slower than that in the edge region, and the overall growth rate was 2% lower than the ideal value. The system immediately generated a film thickness deviation feature vector containing spatial morphological components (thin at the center, thick at the edge) and a multi-point growth rate vector (the rate at each point is lower).

[0107] The deviation feature vector is fed into the intelligent decision-making module. The module uses a digital twin model of the process for rapid simulation, evaluating multiple compensation strategies in less than 0.5 seconds. Ultimately, the system selects an optimal combination: fine-tuning the position of the correction baffle used to adjust film thickness uniformity by 1.5 mm, while simultaneously increasing the electron gun beam intensity by 3%. This solution is encapsulated as a dynamic process compensation command.

[0108] After receiving the instruction, the control execution module decomposes it into a time-division control sequence for the baffle motor and electron gun power supply, and after smoothing, generates a continuous control curve, which is then sent to the equipment actuator. During the subsequent coating process, the spatial uniformity of the film thickness and the growth rate are rapidly corrected and stabilized near the ideal values.

[0109] After the entire coating process was completed, the final film thickness of both the experimental and control groups of lenses in this batch was measured using a high-precision ellipsometer. The measured data from the experimental group was used for the model's self-learning and iteration. The system compared the final measured film thickness with the final thickness predicted by the model and calculated the prediction residual distribution map. By tracing the historical compensation instructions corresponding to areas with large residuals, the system identified a slight deviation in the model's prediction of the relationship between electron gun beam intensity and Ta2O5 evaporation rate. Based on this, the system generated a model error correction factor and iteratively updated the digital twin model. In subsequent batches, the model's prediction accuracy and control performance were further improved.

[0110] Based on the data comparison, the intelligent processing system of the present invention shows significant advantages in film thickness control accuracy, uniformity and production yield.

[0111] Table 1 Comparison of film thickness uniformity (taking the third SiO2 layer as an example)

[0112]

[0113] Table 2 Comparison of Film Thickness Accuracy and Yield

[0114]

[0115]

[0116] Table 3. Model Self-Learning Iteration Effect Table

[0117]

[0118] As can be seen from the data in Tables 1 to 3 above, the film thickness control effect of the experimental group was greatly improved after applying the present invention. Table 1 shows that the film thickness uniformity (measured by standard deviation) of the experimental group was far superior to that of the control group, such as... Figure 4 As shown, with model self-learning, uniformity further improved, and the standard deviation decreased from the initial 0.8 nm to 0.5 nm. Table 2 shows that the final film thickness accuracy of the experimental group was extremely high, with the average deviation controlled within ±0.6%, far less than the ±2.5% of the control group, directly leading to a significant improvement in yield, from approximately 80% in the control group to nearly 100%. Table 3 clearly demonstrates the self-learning and evolutionary capabilities of the system of this invention, such as... Figure 5 As shown, the model prediction error (RMS) continuously decreases with the increase in production batches, indicating that the digital twin model is becoming increasingly closer to the real physical process. Simultaneously, the system dynamically tightens the threshold conditions for triggering intervention based on the improvement in model confidence, making the control strategy more precise and proactive.

[0119] In summary, the embodiments of the present invention fully demonstrate that the method and system, by constructing a digital twin model of the process, performing real-time monitoring and deviation analysis, intelligent decision compensation, and model self-learning iteration, can effectively solve the problems of inaccurate thickness control and poor uniformity in optical lens coating, greatly improve the coating quality and production yield of complex optical components, and have significant technical advantages and application value.

[0120] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0121] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for precise control and intelligent processing of optical lens coating thickness, characterized in that, The method includes: A digital twin model of the process was established based on the three-dimensional model of the optical lens and the target film structure. The process digital twin model was then used to perform virtual coating simulation and generate an ideal thickness growth curve. The reflection spectrum data of multiple preset spectral reflection probes in the coating cavity are collected, and the reflection spectrum data are synchronously inverted and calculated to generate a real-time thickness distribution map. By comparing the real-time thickness distribution map with the ideal thickness growth curve, the spatial distribution differences and time growth rate deviations of film thickness growth are identified, and a film thickness deviation feature vector is generated. The film thickness deviation feature vector is input into the digital twin model of the process to perform compensation strategy deduction, and a compensation scheme that can simultaneously correct spatial distribution differences and time growth rate deviations is selected to generate dynamic process compensation instructions. According to the dynamic process compensation command, a control signal is sent to the actuator of the coating equipment to adjust the physical parameters of the coating.

2. The method for precise control and intelligent processing of optical lens coating thickness according to claim 1, characterized in that, The process of generating the ideal thickness growth curve includes: Analyze the three-dimensional geometric data of the optical lens from its three-dimensional model to obtain surface feature parameters; The evaporation physical properties of the coating material are obtained based on the target film structure. By integrating the surface feature parameters with the evaporation physics properties of the coating material, a spatial particle cloud distribution model of the evaporation source is constructed. By combining the spatial particle cloud distribution model and the motion trajectory parameters of the workpiece disk, a digital twin model of the process is constructed, and an ideal thickness growth curve is generated based on the digital twin model of the process.

3. The method for precise control and intelligent processing of optical lens coating thickness according to claim 1, characterized in that, The generated film thickness deviation feature vector includes: The real-time thickness distribution map is analyzed in the spatial dimension to extract the spatial morphology components that characterize the uniformity of film thickness. Gradient calculations are performed on multiple monitoring areas in the real-time thickness distribution map over time to obtain a multi-point growth rate vector. The spatial morphological component is coupled with the multi-point growth rate vector to generate a film thickness deviation feature vector.

4. The method for precise control and intelligent processing of optical lens coating thickness according to claim 3, characterized in that, After generating the film thickness deviation feature vector, the following steps are also included: The multi-point growth rate vector is used to predict the future thickness deviation trend. When the future thickness deviation trend meets the preset intervention conditions, a forward-looking intervention signal is generated; The prospective intervention signal is appended to the film thickness deviation feature vector to obtain the updated film thickness deviation feature vector.

5. The method for precise control and intelligent processing of optical lens coating thickness according to claim 4, characterized in that, The generated dynamic process compensation instructions include: Based on the spatial morphology component in the updated film thickness deviation feature vector, the physical parameters affecting the spatial distribution of film thickness are simulated and adjusted in the digital twin model of the process, generating a set of spatial uniformity compensation schemes. Based on the multi-point growth rate vector in the film thickness deviation feature vector, the physical parameters affecting the overall film thickness growth rate are simulated and adjusted in the digital twin model of the process, and a global growth rate compensation scheme set is generated. The spatial uniformity compensation scheme set and the global growth rate compensation scheme set are cross-validated and evaluated. The combination that has the least negative impact on the growth rate while correcting the spatial morphology is selected, and a dynamic process compensation instruction is generated.

6. The method for precise control and intelligent processing of optical lens coating thickness according to claim 4, characterized in that, The method further includes: After the coating process is completed, the final measured film thickness data of the optical lens is obtained; The final measured film thickness data is correlated with the dynamic process compensation command to generate a model error correction factor. The process digital twin model is iteratively updated using the model error correction factor to generate an updated process digital twin model.

7. The method for precise control and intelligent processing of optical lens coating thickness according to claim 6, characterized in that, The generated model error correction factor includes: Calculate the residual between the final measured film thickness data and the thickness predicted by the digital twin model of the process at the end of the process to obtain the predicted residual distribution map; By tracing the dynamic process compensation instructions corresponding to the deviation points that meet the preset conditions in the predicted residual distribution map, the characteristics of inefficient compensation instructions are obtained. The influence weight of the inefficient compensation command features on physical parameters is quantified to generate a model error correction factor.

8. The method for precise control and intelligent processing of optical lens coating thickness according to claim 7, characterized in that, The method further includes: Evaluate the convergence of the predicted residual distribution map generated by the updated process digital twin model and generate a model confidence score; The intervention conditions are dynamically adjusted based on the model confidence score.

9. The method for precise control and intelligent processing of optical lens coating thickness according to claim 1, characterized in that, Sending control signals to the coating equipment actuator according to the dynamic process compensation command includes: The dynamic process compensation command is decomposed into a time-division control sequence for multiple independent coating physical parameters; Perform smooth transition processing on each control point in the time-sharing control sequence to generate a continuous control curve; The continuous control curve is sent to the actuator of the coating equipment.

10. A system for precise control and intelligent processing of optical lens coating thickness, applied to the method for precise control and intelligent processing of optical lens coating thickness as described in any one of claims 1-9, characterized in that, The system includes: The digital twin model generation module is used to establish a digital twin model of the process based on the three-dimensional model of the optical lens and the target film structure, and to use the digital twin model of the process to perform virtual coating simulation and generate an ideal thickness growth curve. The real-time monitoring module is used to collect the reflection spectrum data of multiple preset spectral reflection probes in the coating cavity, and to perform synchronous inversion calculation on the reflection spectrum data to generate a real-time thickness distribution map. The deviation analysis module is used to compare the real-time thickness distribution map with the ideal thickness growth curve, identify the spatial distribution differences and time growth rate deviations of film thickness growth, and generate a film thickness deviation feature vector. The intelligent decision-making module is used to input the film thickness deviation feature vector into the digital twin model of the process to deduce the compensation strategy, select a compensation scheme that can simultaneously correct spatial distribution differences and time growth rate deviations, and generate dynamic process compensation instructions. The control execution module is used to send control signals to the coating equipment actuator according to the dynamic process compensation instructions, and to adjust the coating physical parameters.

Images