A Method and System for Process Control of Evaporation Coating Machine Based on Digital Twin

Abstract

This application discloses a process control method and system for an evaporation coating machine based on digital twins, belonging to the field of vacuum coating and intelligent control of industrial processes. The method includes: real-time acquisition of evaporation source operating parameters and film thickness sensor data; dynamic calculation of effective material margin through data fusion; synchronous updating of the digital twin state to ensure consistency with the material consumption of the physical evaporation source; using the synchronized digital twin, generating a rolling evaporation power adjustment sequence and outputting a first control command when the film thickness deviation exceeds a threshold; executing a process replanning of the remaining film layer to generate a second control command and update the subsequent process formula when the predicted continuous evaporation time of the material is insufficient; and finally issuing and executing the command. This method achieves real-time perception and adaptive compensation of material consumption and process drift through digital twins, effectively solving the problems of low film thickness control accuracy and poor process stability, and significantly improving the yield and intelligence level of high-end coating processes.

Description

Technical Field

[0001] This application relates to the field of vacuum coating and intelligent control of industrial processes, and in particular to a process control method and system for evaporation coating machines based on digital twins. Background Technology

[0002] Evaporation coating technology is widely used in optical thin films, electronic devices, and high-end decorative coatings. Its core process involves heating an evaporation source to vaporize materials and deposit them on a substrate surface to form a film. Film thickness uniformity and film performance directly affect the optical, electrical, and mechanical properties of the final product. In traditional evaporation coating processes, the material consumption state of the evaporation source cannot be directly measured, leading to deviations between the deposition rate and film thickness prediction. Existing control methods typically rely on empirically set power or closed-loop regulation based on fixed models (such as PID control), which has limitations in effectively compensating for changes in material consumption and process parameter drift in real time. This makes it difficult to further improve film thickness control accuracy, and the stability of film performance needs improvement. Furthermore, as product structures become more complex, the control requirements for multi-layer, multi-material precision deposition processes are constantly increasing, posing significant challenges to traditional methods in achieving high yield, high reliability, and intelligent process control. Summary of the Invention

[0003] The purpose of this application is to provide a process control method and system for evaporation coating machines based on digital twins. This addresses technical problems in existing technologies, such as inaccurate deposition rate models due to the unpredictable consumption state of the evaporation source material, accumulated film thickness prediction deviations, insufficient process control precision, and the inability to reprogram process parameters in advance when the material is nearing depletion.

[0004] In view of the above technical problems, this application provides a process control method and system for evaporation coating machine based on digital twin.

[0005] A first aspect of this application provides a process control method for an evaporation coating machine based on digital twins, the method comprising: The operating parameters of the evaporation source and the measurement data of the film thickness sensor are acquired in real time. The operating parameters include at least the evaporation power and the temperature of the evaporation source. The theoretical deposition rate is calculated based on the evaporation power, evaporation source temperature, and the estimated effective material margin at the previous moment. The theoretical deposition rate is fused with the actual deposition rate determined based on film thickness sensor measurement data to obtain the current effective margin of material in the evaporation source; The current effective margin is used as a dynamic input to update the state quantity in the digital twin of the evaporation coating process that represents the material consumption process, so that the material consumption state of the digital twin is consistent with the material consumption state of the physical evaporation source. When the deviation between the cumulative film thickness calculated by the updated digital twin within the first time window and the corresponding target film thickness exceeds the first threshold determined based on the noise measured by the film thickness sensor, the synchronously updated digital twin is used as the prediction model to continuously generate evaporation power adjustment instructions for reducing film thickness prediction deviation in the future finite time domain, which serve as the first control instruction. When the material's sustainable evaporation time, determined based on the current effective margin and the actual deposition rate, is less than the shortest process time required to complete the current film layer, the digital twin, after state synchronization, is used as the simulation environment. By recalculating the deposition time and power allocation of the remaining film layers, the final film system performance meets the target, and a second control command is generated to update the subsequent process parameter sequence. The first control command is sent to the evaporation power supply to adjust the evaporation power, and the second control command is sent to the process controller to execute the new process formula according to the updated process parameter sequence.

[0006] A second aspect of this application provides a process control system for an evaporation coating machine based on digital twins, the system comprising: The data acquisition module is used to acquire the operating parameters of the evaporation source and the measurement data of the film thickness sensor in real time. The operating parameters include at least the evaporation power and the evaporation source temperature. The theoretical deposition rate calculation module is used to calculate the theoretical deposition rate based on the evaporation power, evaporation source temperature and the previous time-lapse estimate of the effective material margin. The effective material margin fusion module is used to fuse the theoretical deposition rate with the actual deposition rate determined based on the film thickness sensor measurement data to obtain the current effective material margin in the evaporation source. The digital twin update module is used to update the state quantity in the digital twin of the evaporation coating process, which is used to characterize the material consumption process, by taking the current effective margin as dynamic input, so that the material consumption state of the digital twin is consistent with the material consumption state of the physical evaporation source. The first control command generation module is used to generate, when the deviation between the cumulative film thickness calculated by the updated digital twin within the first time window and the corresponding target film thickness exceeds the first threshold determined by the noise measured by the film thickness sensor, the synchronously updated digital twin is used as the prediction model to continuously generate evaporation power adjustment commands for reducing the film thickness prediction deviation in the future finite time domain, as the first control command. The second control command generation module is used to generate a second control command for updating the subsequent process parameter sequence when the material sustainable evaporation time determined based on the current effective margin and combined with the actual deposition rate is less than the shortest process time required to complete the current film layer. The module uses the state-synchronized digital twin as the simulation environment and recalculates the deposition time and power allocation of the remaining film layer so that the final film system performance meets the target. The control command execution module is used to send the first control command to the evaporation power supply to adjust the evaporation power, and to send the second control command to the process controller to execute the new process formula according to the updated process parameter sequence.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: The technical solution provided in this application has at least the following technical effects or advantages: It achieves dynamic awareness of material consumption status: By integrating theoretical models and actual measurement data, it estimates the "current effective surplus" in real time, overcoming the industry challenge of not being able to directly measure the inventory of evaporation source materials, and providing a fundamental prerequisite for precise control. It ensures the synchronization and consistency between the process model and the physical process: By dynamically updating the digital twin using the real-time estimated surplus, it always accurately reflects the physical state of the evaporation source, fundamentally solving the problems of prediction inaccuracies and drift compensation failures caused by fixed models. It achieves intelligent feedforward compensation for film thickness deviation: Based on a synchronous and high-fidelity digital twin, it performs rolling time-domain optimization to generate forward-looking power adjustment commands, which can actively and accurately correct film thickness deviations, significantly improving control accuracy and response speed. It endows the system with autonomous decision-making capabilities to cope with resource crises: When it predicts that materials are about to be exhausted, it can perform global replanning of the remaining process based on the digital twin, intelligently generating feasible new formulas, transforming passive alarms into proactive optimization, and greatly improving the yield and reliability of complex processes. A complete "perception-decision-execution" intelligent control closed loop has been constructed: the above effects work synergistically to form an adaptive, predictable, and optimizable intelligent control system, which comprehensively improves the stability, intelligence level, and mass production capability of the evaporation coating process. It solves technical problems in existing technologies such as inaccurate deposition rate models caused by the unpredictable consumption state of the evaporation source material, accumulated film thickness prediction deviations, insufficient process control precision, and the inability to reprogram process parameters in advance when the material is nearing depletion.

[0008] The above description is merely an overview of the technical solution of this application. In order to more clearly explain the technical means of this application, and to enable its implementation in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are described below. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0010] Figure 1 A schematic flowchart of the process control method for an evaporation coating machine based on digital twin provided in the embodiments of this application; Figure 2 A schematic diagram of the structure of the process control system for the evaporation coating machine based on digital twin provided in the embodiments of this application.

[0011] Figure reference numerals: Data acquisition module 10, theoretical deposition rate calculation module 20, material effective margin fusion module 30, digital twin update module 40, film thickness deviation judgment module 50, first control command generation module 60, second control command generation module 70, control command execution module 80. Detailed Implementation

[0012] This application provides a process control method and system for evaporation coating machines based on digital twins, which solves the technical problems in the prior art, such as inaccurate deposition rate models, accumulated film thickness prediction deviations, insufficient process control accuracy, and inability to reprogram process parameters in advance when the material is about to be exhausted, caused by the unpredictable consumption state of the evaporation source material.

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0014] It should be noted that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0015] Example 1, as Figure 1 As shown, this application provides a process control method for an evaporation coating machine based on digital twins, wherein the method includes: The operating parameters of the evaporation source and the measurement data of the film thickness sensor are acquired in real time. The operating parameters include at least the evaporation power and the temperature of the evaporation source. Specifically, in this embodiment, the system acquires the operating parameters of the evaporation source and the measurement data from the film thickness sensor in real time through a data acquisition module and an industrial controller deployed on the evaporation coating machine. The operating parameters include at least evaporation power and evaporation source temperature. The evaporation power is acquired through the standard industrial analog output interface of the evaporation power supply. For example, the evaporation power supply represents its real-time output power as a 4–20 mA current signal. This signal is converted into a digital value by the analog-to-digital converter of the data acquisition module, and then the evaporation power value in watts (W) is calculated based on the current-power correspondence specified by the evaporation power supply at the factory. To reduce the influence of electrical noise, the system can perform moving average filtering on the continuously sampled data. The evaporation source temperature is acquired through a K-type thermocouple placed at a temperature measurement point on the outer wall of the evaporation crucible. The thermoelectric potential generated by the thermocouple is processed by a temperature transmitter for cold junction compensation, linearization, and amplification, and outputs a standard signal (e.g., 4–20 mA), which is acquired by the data acquisition module and converted into a digital temperature value expressed in degrees Celsius (°C). The film thickness sensor measurement data is provided by a quartz crystal oscillating film thickness monitor. Based on the physical law of crystal frequency variation with film thickness, the monitoring instrument calculates the deposition rate and film thickness in real time and sends them to the system via a digital interface at a fixed sampling period. The system parses and verifies the data messages to extract the deposition rate (Å / s) and film thickness (nm) values. To ensure temporal consistency of multi-source data, the system uses the controller's internal clock as a reference and adds a unified timestamp to the evaporation power, evaporation source temperature, and film thickness sensor measurements at each sampling moment. For data with inconsistent sampling periods (such as slower film thickness updates), the system uses a time-aligned interpolation algorithm to generate deposition rate and film thickness data corresponding to the power and temperature data. This results in a set of time-aligned and structured real-time process data packets, stored in a circular buffer, serving as direct input for subsequent calculations of theoretical deposition rates, fusion processing of actual deposition rates, and updates to the digital twin state.

[0016] The theoretical deposition rate is calculated based on the evaporation power, evaporation source temperature, and the estimated effective material margin at the previous moment. Specifically, the theoretical deposition rate R_theoretical(t) is calculated based on the evaporation power P_t, the evaporation source temperature T_t, and the estimated effective material margin M_est(t-1) determined in the previous control cycle at the current sampling time. This calculation is achieved by calling the mathematical model encapsulated in the evaporation source sub-model of the digital twin. The acquisition of the "estimated effective material margin M_est(t-1) at the previous moment" follows the following initialization and state recursion principles: System initialization (t=0): Before the process begins, the evaporation source is loaded with a known mass M_total of coating material. The system sets the initial estimated effective material margin M_est(0) to this loading amount M_total, i.e., M_est(0) = M_total. This loading amount M_total is input by the operator through the human-machine interface. Online recursive update (t≥1): Starting from the first control cycle, M_est(t-1) is derived from the optimal estimate of the "current effective margin" calculated and output by the "fusion processing" step described in the claims in the previous cycle (time t-1). This value is updated at the end of each control cycle and stored in the system's non-volatile memory or status register for recall in the next cycle. State variable definition: M_est(t), as an internal system state variable, represents the system's best estimate of the remaining mass of material available for effective evaporation within the evaporation source, typically in grams (g) or milligrams (mg). In a typical embodiment, the mathematical model for calculating the theoretical deposition rate is expressed as the following function: R_theoretical(t) = η(M_est(t-1)) f(P_t, T_t), where f(P_t, T_t) is the core function characterizing the relationship between deposition rate, evaporation power, and temperature under ideal virgin source conditions. This function is calibrated through process experiments for specific evaporation sources and targets, and can be expressed as, for example, f(P_t, T_t) = k1 P_t^α The efficiency decay function η(M) is defined in the form exp(-k2 / T_t), where k1, k2, and α are constants obtained through the calibration. η(M_est(t-1)) is a correction coefficient (or efficiency decay function) characterizing the effect of the effective material margin on evaporation efficiency, with a value range between (0, 1]. A specific implementation example of the efficiency decay function η(M) is provided below. To ensure model accuracy, the specific form and parameters of η(M) need to be obtained through offline calibration experiments for a specific combination of evaporation source and target material. Two feasible examples are provided below: Example A (lookup table form): Calibration is performed for a certain type of electron beam evaporation source and SiO2 target material. Evaporation continues under constant process conditions until the material is exhausted, and the deposition rate is recorded simultaneously. The percentage of remaining material (M / M_total) is compared with the corresponding normalized efficiency coefficient (η = current rate / initial rate). Construct data pairs and generate lookup tables. For example: when M / M_total=100%, η=1.0; when M / M_total=60%, η=0.95; when M / M_total=20%, η=0.82. During online control, η_t is obtained by looking up the table and interpolating based on the real-time estimated M_est(t). Example B (exponential decay function form): For a certain type of resistive evaporation source and Al target material, a parameterized model is obtained by fitting historical data. For example: η(M) = exp(-γ (M_total - M)), where γ = 0.018 g⁻¹ is the fitting attenuation coefficient. Example C (polynomial fitting form): For a certain type of electron beam evaporation source and MgF2 target, a quadratic polynomial function was obtained by fitting experimental data: η(M) = 1.0 - 0.002 (M_total - M) + 0.00003 (M_total - M)^2, where M is in milligrams and M_total is the initial charge. This form is suitable for processes where the deposition rate changes non-linearly with the remaining material. In online control, M_est(t) is directly substituted into the formula to calculate η_t. Therefore, the steps to calculate R_theoretical(t) include: data preparation, calculating the base rate R_base(t) = f(P_t, T_t), querying or calculating the efficiency coefficient η based on M_est(t-1), and finally synthesizing R_theoretical(t) = η. R_base(t). The specific forms of functions f(P, T) and η(M) and all parameters are determined and preset through offline calibration during the initialization of the digital twin, ensuring the physical consistency of the model.

[0017] The theoretical deposition rate is fused with the actual deposition rate determined based on film thickness sensor measurement data to obtain the current effective margin of material in the evaporation source. Furthermore, the fusion of the theoretical deposition rate with the actual deposition rate determined based on film thickness sensor measurement data is achieved through a Kalman filter algorithm, wherein the current effective margin of material within the evaporation source is set as the system state variable estimated by the Kalman filter algorithm.

[0018] Specifically, after obtaining the theoretical deposition rate R_theoretical(t) predicted by the digital twin and the actual deposition rate R_actual(t) calculated from the film thickness sensor measurement data, the system performs a fusion process to estimate the current effective margin M_est(t) of the material in the evaporation source in the optimal way. 1. Algorithm framework for fusion processing: Kalman filtering. In this embodiment, the fusion process is implemented by the Kalman filtering algorithm. The algorithm defines the current effective margin M_est(t) of the material in the evaporation source as the system state variable to be estimated, uses the theoretical deposition rate R_theoretical(t) as the prediction basis of the state evolution model (system model), and uses the actual deposition rate R_actual(t) as the observation value (measurement model). The two information are fused optimally through iterative calculation. (1) Definition of state space model: State variable (x): x_t = [M_est(t)], that is, the effective margin estimate at the current moment. State evolution equation (system model): x_t = x_{t-1} -(R_theoretical(t)) A Δt) / ρ + w_t Where: x_{t-1} is the state estimate of the previous moment (i.e., M_est(t-1)). R_theoretical(t) is the theoretical deposition rate at the current moment (unit: Å / s or nm / s). A is the preset substrate deposition area (unit: cm²), a known process constant. Δt is the control period (unit: s). ρ is the density of the coating material (unit: g / cm³), a known material constant. w_t is the system process noise, assumed to follow a Gaussian distribution with zero mean and covariance Q. The Q value reflects the degree of confidence in the uncertainty of the theoretical model and is determined through experimental debugging. The physical meaning of this equation: Current margin = Previous moment margin - Theoretically predicted mass of deposited material. Observation equation (measurement model): z_t = (ρ / (A) Δt)) (x_{t-1} - x_t) + v_t = R_actual(t) + v_t, where: z_t is the observed value, i.e., R_actual(t). v_t is the observation noise, which is assumed to follow a Gaussian distribution with zero mean and covariance R. The R value reflects the degree of confidence in the measurement accuracy of the film thickness sensor, which can be determined by sensor calibration data or historical measurement statistics. The physical meaning of this equation is: the observed deposition rate should be consistent with the average deposition rate calculated based on the state change, but there is a measurement error. (2) Kalman filter iteration steps: In each control period t, the following calculations are performed in sequence: State prediction (time update): Prior state estimation: x_t^- = x_{t-1} - (R_theoretical(t)) A Δt) / ρ, prior error covariance: P_t^- = P_{t-1} + Q (P_{t-1} is the posterior estimation error covariance of the previous time step); measurement update: Kalman gain calculation: K_t = P_t^- H^T (H P_t^- H^T + R)^{-1}, (In this one-dimensional system, the observation matrix H = -ρ / (A Δt), but its physical dimensions match those of R in gain calculations, and in practice it can be directly calibrated as the effective gain coefficient. Posterior state estimation (fused output): x_t = x_t^- + K_t (z_t - (ρ / (A Δt)) Simplifying and substituting (x_{t-1} - x_t^-) into z_t = R_actual(t), the core formula is: M_est(t) = M_est(t)^- + K_t The equivalent form of [R_actual(t) - R_theoretical(t)] indicates that the final estimate is based on the model predictions, with Kalman gain weighting adjustments made according to the deviation between the theoretical and actual rates. Update error covariance: P_t = (I - K_t) H) P_t^-. 2. Algorithm initialization and parameter setting, state initialization: As mentioned above, x_0 = M_est(0) = M_total (initial load). Error covariance initialization: P_0 is set to a large value reflecting the initial estimate uncertainty (e.g., (0.1)). M_total)^2). Noise covariance setting: Q (process noise): calibrated based on the prediction accuracy of the theoretical model R_theoretical(t). A higher model confidence level results in a smaller Q value, and vice versa. It can be initially set to (0.01). (R_theoretical(t) A The magnitude is Δt / ρ))^2. R (observation noise): determined according to the accuracy index of the film thickness sensor. For example, if the measurement standard deviation of the sensor rate is σ_R, then R = σ_R^2. 3. Output and Closed Loop After each Kalman filter iteration, the output posterior state estimate x_t is the "current effective margin M_est(t)". ​​This value will: be immediately written into the state memory to update the digital twin (see the next embodiment). be used as M_est(t) necessary for calculating R_theoretical(t+1) in the next cycle (t+1). be used to determine whether to trigger global process replanning (derivative calculations related to M_est(t)). Key Note: Through the above Kalman filter framework, this invention achieves the optimal dynamic estimation of the key unmeasurable state of "material effective margin". It is not a simple comparison of theoretical and actual values, but rather a dynamic trade-off between the predictive ability of the theoretical model and the measurement reliability of the field sensor within a unified probabilistic framework, thereby obtaining a margin estimate that becomes more and more accurate and stable over time. This forms the cornerstone of the entire adaptive control system.

[0019] The current effective margin is used as a dynamic input to update the state quantity in the digital twin of the evaporation coating process that represents the material consumption process, so that the material consumption state of the digital twin is consistent with the material consumption state of the physical evaporation source. Furthermore, the updated digital twin of the evaporation coating process includes: Based on the current effective margin, the efficiency coefficient or decay function parameter related to the material inventory in the evaporation source sub-model of the digital twin is dynamically adjusted.

[0020] Specifically, after obtaining the current effective margin M_est(t) with high confidence through Kalman filtering, the system immediately uses this as a dynamic input to update the state of the digital twin of the evaporation coating process to correct its internal state variables, ensuring that it remains consistent with the actual material consumption state of the physical evaporation source in real time, thereby ensuring the accuracy of subsequent predictions and simulations of the digital twin. This embodiment achieves this by dynamically adjusting the key parameters related to material inventory in the evaporation source sub-model of the digital twin. 1. Based on the parameterized function model update mechanism, in this embodiment, the update is achieved by dynamically adjusting the key parameters related to material inventory in the evaporation source sub-model of the digital twin to "efficiency coefficient or decay function parameter". The specific process is as follows: Parameter mapping: The system maintains a parameter mapping relationship, which defines the correspondence between "effective material margin" and one or more efficiency correction parameters in the evaporation source sub-model. This mapping relationship is predefined when the digital twin is constructed. Parameter calculation / query: Based on the latest estimated value M_est(t), the corresponding new values ​​of the model parameters are calculated or queried according to the mapping relationship. If the model uses the efficiency coefficient η (e.g., R_theoretical = η(M)). If f(P, T) is used, then η is directly updated to η(M_est(t)). That is, the efficiency decay function η(M) is called, and M = M_est(t) is substituted to calculate the latest efficiency coefficient value η_t. If the model uses more complex decay function parameters (for example, the exponential decay function η(M) = exp(-γ)), then... If the attenuation coefficient γ in (M0 - M) is constant, then the function form and parameters (such as γ, M0) remain unchanged. The system calculates the current efficiency coefficient value η_t = η(M_est(t)) by directly substituting M = M_est(t) into this function. Model hot update: The calculated new parameter values ​​are loaded into the running evaporation source sub-model instance in real time through the software interface, replacing the original parameters, thereby completing the synchronization of the material consumption state of the digital twin. State consistency verification (optional): As an internal verification, the system can compare the predicted values ​​of the model for the current process parameters before and after the update. The updated predicted values ​​should be closer to the actual measured values, which is a direct manifestation of "state consistency". 2. Specific embodiment, efficiency coefficient update based on lookup table: As a more specific example, assume that the evaporation source sub-model uses an efficiency coefficient lookup table to represent the η(M) relationship. The lookup table stores a series of (material balance M_i, efficiency coefficient η_i) data pairs. Update action: After obtaining M_est(t), the system performs search and interpolation in the lookup table. Parameter determination: The current efficiency coefficient is calculated using linear interpolation: η_t = η_k + (η_{k+1} - η_k) (M_est(t) - M_k) / (M_{k+1} - M_k), where [M_k, M_{k+1}] is the interval containing M_est(t). Model synchronization: η_t is assigned to the internal variables of the evaporation source sub-model. Subsequently, through the above mechanism, the parameters of the functional relationship η(M) representing the material consumption impact in the digital twin are continuously updated, ensuring that the prediction based on this model is always based on the current optimal estimate of the material state. This realizes the transformation of the digital twin from a static model to a dynamic mirror, forming a key link in the "perception-modeling" closed loop, providing a high-fidelity prediction basis for subsequent intelligent decision-making. 3. Update triggering and cycle: This update step is automatically triggered in each control cycle, immediately after the calculation of "current effective margin M_est(t)", and the update cycle a is strictly synchronized with the main control cycle Δt. 4. Technical effect and closed-loop significance: Through the above update, the digital twin transforms from a static model into a dynamic mirror that can reflect the performance degradation of the physical evaporation source in real time. This step, combined with the previous step (state estimation), completes the closed loop from "perceiving the physical state" to "correcting the virtual model," providing a reliable predictive foundation for all subsequent intelligent decision-making based on digital twins. The interpolation update based on lookup tables and the parameter adjustment based on exponential decay functions described in this section are merely examples of two specific and feasible technical means to achieve "dynamic adjustment of efficiency coefficients or decay function parameters based on the current effective margin," used to fully illustrate and teach how to implement this invention. Universality of parameter acquisition: The data pairs (M_i, η_i) in the lookup table, and the parameters γ, β, M0, etc., in the exponential decay function, can all be obtained through offline process calibration experiments for specific evaporation sources and target materials. The expression "efficiency coefficients or decay function parameters" aims to cover all model correction methods that can characterize the impact of material inventory on evaporation efficiency. In addition to the methods listed in this embodiment, those skilled in the art, under the same inventive concept, can also use other mathematical functions (such as polynomials, piecewise constants) or adaptive learning algorithms (such as online recursive least squares) to establish and update the η(M) relationship. The system design allows for automatic or manual switching of different parameter update strategies based on different process stages or model confidence levels. The core of all the above implementations lies in using the real-time estimated M_est(t) as input to dynamically correct the internal parameters of the digital twin evaporation source sub-model, thereby achieving synchronization between virtual and real states. The specific form and parameters of the lookup table data pairs or decay functions are determined during the digital twin initialization phase through offline process calibration experiments targeting specific evaporation sources and targets.

[0021] The current effective margin is used as a dynamic input to update the state quantity in the digital twin of the evaporation coating process that represents the material consumption process, so that the material consumption state of the digital twin is consistent with the material consumption state of the physical evaporation source. Specifically, after obtaining the current effective margin M_est(t) through Kalman filtering, the system immediately uses this as a dynamic input to perform state synchronization updates on the digital twin of the evaporation coating process. The "updating the state quantity used to characterize the material consumption process in the digital twin of the evaporation coating process" is achieved by dynamically correcting the efficiency parameters of the evaporation source sub-model inside the digital twin. Its core is to update the independent variable in a pre-set, parameterized efficiency decay function model η(M) to M_est(t) and refresh the model output accordingly. The specific implementation steps are as follows: (1) Efficiency parameter calculation: The system calls the function model η(M) established through offline calibration during the initialization of the digital twin. The system uses M_est(t) as input to calculate the current efficiency coefficient value η_t. If η(M) is implemented in the form of a lookup table, η_t is obtained by looking up the table and interpolation. If η(M) is implemented in the form of a continuous parameterized function (e.g., η(M) = exp(-γ) (M0 - M))), then directly calculate η_t =η(M_est(t)). (2) Model hot update: The system loads the calculated η_t into the running evaporation source sub-model instance in real time through the software interface, replacing its original efficiency coefficient parameter. This operation makes all subsequent predictions of the sub-model immediately based on the latest material state. (3) Manifestation and verification of state consistency: After the update, the evaporation source sub-model of the digital twin inherently carries the material consumption state synchronized with the physical evaporation source. Its consistency is directly reflected in: under the same process input, the theoretical deposition rate R_theoretical(t) predicted after the model update will be closer to the actual deposition rate R_actual(t) at the same moment than before the update. Optionally, the system can qualitatively verify the synchronization effect by comparing this deviation. (4) Update timing: This update step is set as a mandatory link of the control cycle. In each control cycle Δt, it is triggered immediately after the calculation of M_est(t) is completed, ensuring the real-time state of the digital twin. Through the steps described above, the digital twin transforms from a static, idealized model into a dynamic mirror image capable of real-time mapping the efficiency degradation of a physical evaporation source due to material consumption. This "perception-correction" closed loop lays a precise model foundation for all subsequent predictions, optimizations, and replanning decisions based on this high-fidelity digital twin.

[0022] When the deviation between the cumulative film thickness calculated by the updated digital twin within the first time window and the corresponding target film thickness exceeds the first threshold determined based on the noise measured by the film thickness sensor; Furthermore, the first threshold includes: Acquire the measurement data sequence of the film thickness sensor within the first time window, and calculate its noise standard deviation; The noise standard deviation is input into a preset monotonically increasing function, and the first threshold is output.

[0023] Specifically, within each first time window, the system dynamically determines the first threshold based on the real-time measurement noise of the film thickness sensor. The specific steps are as follows: In each judgment period (i.e., the first time window), the system performs the following steps to determine the current first threshold Threshold(t): (1) Noise data sequence acquisition: The system extracts all the original film thickness measurements measured by the film thickness sensor within the current first time window (e.g., the most recent N sampling periods) from the real-time database or buffer, forming a measurement data sequence Z = [z(t-N+1), z(t-N+2), ..., z(t)]. The window length N can be preset according to the process characteristics, for example, corresponding to 10 to 30 control periods. (2) Noise standard deviation calculation: The system performs statistical analysis on the data sequence Z and calculates its noise standard deviation σ_z to quantify the intensity of the measurement noise within this time period. During the calculation, the high-frequency noise component in the sequence can be initially separated by first-order difference or high-pass filtering, and then the standard deviation of the component can be calculated. A direct and robust implementation is to calculate the root mean square error (RMSE) of the sequence relative to its local trend line. (3) Adaptive threshold calculation: Input the calculated noise standard deviation σ_z into a preset monotonically increasing function F(·), the output of which is the current first threshold: Threshold(t) = F(σ_z). Example design of function F(·): linear function: Threshold(t) = K σ_z + C. Where K is the sensitivity coefficient (K>0), and C is the basic tolerance constant. This form is simple and direct. Piecewise function: Set multiple σ_z intervals, and use different K values ​​in different intervals to achieve differentiated sensitivity adjustment under different noise levels. Lookup table method: Pre-calibrate a (σ_z, Threshold) lookup table through experiments, and directly obtain the threshold from the table based on σ_z. Function property guarantee: Regardless of the specific form, F(·) must be monotonically increasing, that is, the larger σ_z is, the larger Threshold(t) is, which conforms to the logic of "high noise, wide threshold". Design example of function F(·): Linear function: Threshold(t) = K σ_z + C. Where K is the sensitivity coefficient (K>0), typically ranging from 1.5 to 3.0; C is the basic tolerance constant, typically ranging from 0.5 nm to 2.0 nm. As a preferred example, a linear function can be used with K=2.0 and C=1.0 nm. 3. Application and Update of the Threshold: The calculated Threshold(t) is immediately used for triggering the current time: the absolute value of the deviation |ΔH| between the predicted cumulative film thickness calculated by the digital twin within the same time window and the target cumulative film thickness is compared. If |ΔH|>Threshold(t), it is determined that the deviation exceeds the limit, triggering subsequent rolling optimization control (generating the first control command). This threshold determination process is repeated in each first time window, thereby achieving adaptive updating of the threshold to real-time operating conditions, making the entire control system more resistant to noise interference and more adaptable to the environment. In a specific implementation example, the length N of the first time window (i.e., the number of control steps P) is typically set to 20 to 50, and the control period Δt is 0.5 seconds to 2 seconds; the calculation of the first threshold can preferably adopt the aforementioned linear function form. Technical effect: Through the above method, this invention transforms the triggering criterion of the control system from empirical setting to autonomous decision-making based on the statistical characteristics of objectively measured noise, significantly reducing the probability of false or missed triggering of control commands due to sensor noise, and improving the reliability and accuracy of the intelligent control system under complex operating conditions.

[0024] Using the synchronously updated digital twin as a prediction model, evaporation power adjustment instructions for reducing film thickness prediction deviation are generated in a rolling manner within a finite future time domain, serving as the first control instruction; Furthermore, the step of using the synchronously updated digital twin as a prediction model to continuously generate evaporation power adjustment instructions for reducing film thickness prediction bias within a finite future time domain includes the following steps: Based on the current state of the digital twin after synchronous update of the current effective margin, the film thickness value of the digital twin is predicted in a rolling manner at a series of discrete moments within the first time window in the future, forming a film thickness prediction sequence. The predicted film thickness sequence is compared with the corresponding target film thickness sequence to calculate the film thickness deviation at each discrete moment within the first time window in the future. Based on the film thickness deviation, and in combination with the allowable output power range of the evaporation power supply, the power change rate limit, and the current effective margin's limitation on the deposition rate, the evaporation power adjustment amount corresponding to each discrete moment within the first time window is determined to form an evaporation power adjustment sequence, wherein each evaporation power adjustment amount represents an update amount relative to the current evaporation power setting value; The first evaporation power adjustment value corresponding to the current control cycle in the evaporation power adjustment sequence is used to update the current evaporation power setting value, so as to obtain a new evaporation power setting value for the current control cycle. The new evaporation power setting is used as the first control command and output to the evaporation power supply for execution.

[0025] Specifically, in each control cycle t, the system executes the following steps to generate the first control command: (1) State initialization and film thickness rolling prediction. The system uses the current state of the digital twin, which has been synchronously updated with the current effective margin M_est(t), as the prediction model. The optimization time domain is set to the future P control steps (corresponding to the first time window, for example, P=30). The system drives the digital twin to predict its film thickness value H_pred(t+k|t) at a series of discrete times t+1, t+2,..., t+P in the future time domain, thereby forming the film thickness prediction sequence {H_pred(t+1|t), ..., H_pred(t+P|t)}. (2) Film thickness deviation calculation: The above film thickness prediction sequence {H_pred(t+1|t), ..., H_pred(t+P|t)} is compared point by point with the corresponding target film thickness sequence {H_target(t+1), ..., H_target(t+P)} to calculate the film thickness deviation at each future time: e(t+k|t) = H_target(t+k) - H_pred(t+k|t), for k=1 to P. This deviation sequence is the control target that needs to be corrected. (3) Construction and solution of constrained optimization problem to determine the evaporation power adjustment sequence: The system constructs and solves a constrained optimization problem to determine the optimal evaporation power adjustment amount sequence. Optimization variables: Evaporation power adjustment amount at P future times {ΔP(t|t), ΔP(t+1|t), ..., ΔP(t+P-1|t)}, where each ΔP represents the update amount relative to the current power setpoint (i.e., the evaporation power adjustment sequence). Optimization objective: Minimize the objective function, for example: J=Σ_{k=1}^{P} [w_h e(t+k|t)^2 ] + Σ_{k=0}^{P-1} [ w_p ΔP(t+k|t)^2 ], where w_h and w_p are weighting coefficients. Constraints: a) Allowable output power range of the evaporation power source: P_min ≤ P_current + Σ_{j=0}^{k} ΔP(t+j|t) ≤ P_max, b) Power change rate limit: |ΔP(t+k|t)| ≤ ΔP_max, c) Current effective margin limit on deposition rate: The deposition rate predicted by the digital twin based on the candidate power sequence must not exceed the maximum allowable rate determined by M_est(t) through the efficiency function η(M). The system calls an optimization solver (such as a quadratic programming solver) to solve the problem online, obtaining the evaporation power adjustment sequence {ΔP} that satisfies all constraints. (t|t), ..., ΔP (t+P-1|t)}。(4) Control command generation and output: From the obtained optimal sequence, extract the first evaporation power adjustment ΔP corresponding to the current control cycle. (t|t). Add this to the current evaporation power setpoint P_current(t) to obtain the new evaporation power setpoint: P_new(t) = P_current(t) + ΔP (t|t), where P_new(t) is the first control instruction. The system sends it to the evaporator power supply for execution through the communication interface. (5) Rolling update: In the next cycle, repeat steps (1) to (4) through the above process. This application realizes rolling optimization control based on digital twin prediction. The "decision variable ΔP sequence" is the "evaporator power adjustment sequence".

[0026] When the material's sustainable evaporation time, determined based on the current effective margin and the actual deposition rate, is less than the shortest process time required to complete the current film layer, the digital twin, after state synchronization, is used as the simulation environment. By recalculating the deposition time and power allocation of the remaining film layers, the final film system performance meets the target, and a second control command is generated to update the subsequent process parameter sequence. Furthermore, generating the second control instruction for updating the subsequent process parameter sequence includes the following steps: The material sustainable evaporation time, determined based on the current effective margin and in combination with the actual deposition rate, is set as the upper limit of the total duration of the remaining film deposition process. The digital twin, after state synchronization, is invoked to simulate the remaining film deposition time and evaporation power ratio of multiple candidate schemes, and the final total film thickness prediction value of each candidate scheme is obtained. Under the constraints of the total time limit, the allowable operating range of the evaporation power supply, and the minimum deposition time required for each film layer, a systematic traversal calculation of multiple feasible combinations of the deposition time and evaporation power setting values ​​for the remaining film layers is performed using a discretized search method. By comparing the deviations between the final total film thickness prediction value obtained through the digital twin simulation under each combination and the target total film thickness, the combination sequence with the smallest film thickness deviation is selected as the optimal process parameter sequence. The optimal process parameter sequence is converted into the second control command.

[0027] Specifically, when the sustainable evaporation time of the material, determined based on the current effective margin and the actual deposition rate, is less than the shortest process time required to complete the current film layer, the system enters the process reconfiguration mode. At this time, using the digital twin after state synchronization as the simulation and decision-making core, the following adaptive reprogramming process is executed to generate a second control instruction for updating the subsequent process parameter sequence: (I) Constraint modeling and problem definition: The system first constructs a hard constraint set for the reprogramming problem based on the sustainable evaporation time of the material (T_remain), the allowable operating range of the evaporation power supply ([P_min, P_max]), the minimum deposition time required for each film layer (t_layer_min), and the current effective margin (M_est(t)). The core optimization objective is to minimize the absolute deviation between the final total film thickness (H_total) and the target total film thickness (H_target) while satisfying the above constraints. (II) Accelerated Digital Twin Simulation: To complete the evaluation of a large number of candidate parameter combinations within a limited time, the system adopts the following simulation acceleration measures: Model Reduction: The high-fidelity physical model of the digital twin is reduced in order using a system identification method, retaining key deposition dynamics characteristics to significantly reduce the computation time of a single simulation. Response Surface Approximation: A multi-dimensional response surface model of evaporation power, deposition time, and predicted film thickness is pre-constructed or built online for rapid initial screening and approximate prediction of a large number of candidate combinations. Parallel Computation: Multi-core processors or computing clusters are used to perform parallel simulations of different candidate process parameter sequences. Caching Strategy: Results of the same or similar sub-process sequences that have been simulated are cached to avoid redundant calculations. (III) Implementation of Gradient Guided Beam Search Algorithm: To efficiently search the high-dimensional parameter space, this embodiment adopts the gradient guided beam search algorithm. The specific steps are as follows: Discretization and Candidate Generation: The deposition time of each remaining film layer is discretized with a step size ΔT (e.g., 1–5 s); the evaporation power is discretized with a step size ΔP (e.g., 0.5–2 W) to generate initial candidate nodes. Cost Function Design: The cost function used to evaluate path quality in the bundle search is defined as: Cost = α·|H_target - H_pred| + β·Power_Var, where H_pred is the total predicted film thickness from the digital twin simulation, Power_Var is the variance of the evaporation power variation in the process sequence, and α and β are adjustable weighting coefficients used to balance film thickness accuracy and process stability. Gradient-Guided Pruning and Expansion: When expanding nodes at each layer, not only is the current cumulative cost of the path considered, but also the local gradient information (∂H_pred / ∂P and ∂H_pred / ∂T) provided by the digital twin is utilized. The system prioritizes retaining and expanding nodes whose gradient directions match the current film thickness deviation correction requirements (i.e., can effectively reduce deviations) and have large gradient magnitudes. The bundle width B is set to be dynamically adjustable, with a large initial value to maintain search diversity, and then gradually narrowed according to the distribution of costs of the explored paths to focus computational resources.Simulation and Evaluation: For each complete parameter sequence retained in the bundle, a digital twin (or an accelerated simulation model) is called to perform simulation to obtain accurate H_pred and Power_Var, and its final replacement value is calculated. (IV) Gradient Calculation and Approximation Method: The local gradients ∂H_pred / ∂P and ∂H_pred / ∂T can be calculated in real time using the finite difference method: ∂H_pred / ∂P ≈ [H_pred(P+δP) - H_pred(P)] / δP and ∂H_pred / ∂T ≈ [H_pred(T+δT) - H_pred(T)] / δT, where the perturbation step size δP and δT are determined by the system according to the model sensitivity and noise level to achieve a balance between gradient effectiveness and computational stability. (V) Optimal Sequence Decision and Instruction Generation: After the search is completed, the parameter sequence with the smallest replacement value is selected from the final B retained paths as the optimal process parameter sequence. The system converts the sequence into a standardized second control instruction, whose data structure is represented as: {t_start, Layer_ID, P_set(t), T_deposit}. This instruction can be directly loaded into the execution queue of the process controller, thereby realizing real-time and accurate reconstruction of the subsequent deposition process. (VI) Technical effects: Through the above embodiments, this method realizes intelligent online process reconstruction when the evaporation source material is nearing depletion, enabling the prediction of the digital twin to more accurately track the actual process state, and can efficiently search for the optimal parameter combination that balances film thickness accuracy and process stability.

[0028] The first control command is sent to the evaporation power supply to adjust the evaporation power, and the second control command is sent to the process controller to execute the new process formula according to the updated process parameter sequence.

[0029] Specifically, in this embodiment, the first control command and the second control command are respectively sent to the corresponding execution units to achieve real-time closed-loop control of the evaporation coating process. When the first control command is sent, the system sends the first control command (an evaporation power adjustment calculated based on the film thickness prediction deviation of the digital twin) generated through rolling optimization to the evaporation power supply via the industrial control interface. After receiving the command, the evaporation power supply adjusts the output power P_set(t) of the evaporation source in real time according to the set power update amplitude and rate limit to correct the current film thickness deposition rate deviation. To ensure execution safety and power accuracy, the system can perform power smoothing processing before the command is sent, for example, by using low-pass filtering or a rate limiting function to control instantaneous power changes within the allowable range of the equipment. When the second control command is sent, the system converts the optimal process parameter sequence obtained based on the material's sustainable evaporation time, current effective margin, and digital twin simulation optimization into a standardized second control command. This command is sent to the process controller via the process controller interface. The controller automatically executes the new process formulation based on parameters such as the deposition time of each film layer, the evaporation power setpoint, and the substrate temperature in the command sequence to achieve accurate deposition of subsequent film layers. The system performs monitoring and feedback. During the issuance and execution of commands, it continuously collects evaporation power, film thickness sensor data, and digital twin prediction status, forming a real-time feedback closed loop. The first control command corrects instantaneous deviations in real time, while the second control command ensures the overall process executes according to the optimal parameter sequence. These two commands complement each other, ensuring that the actual deposited film thickness is highly consistent with the target film thickness, while simultaneously guaranteeing controllable material consumption and process stability. Through the above embodiments, the system achieves closed-loop adaptive control of the evaporation coating process: the first control command quickly corrects short-term deviations, and the second control command optimizes the remaining film deposition process. Combining digital twin simulation and real-time sensor data, it achieves an optimal balance between film thickness accuracy, material consumption management, and power adjustment speed, thereby significantly improving process consistency, reliability, and output quality.

[0030] Example 2, based on the same inventive concept as the digital twin-based evaporation coating machine process control method in the foregoing examples, such as... Figure 2 As shown, this application provides a process control system for an evaporation coating machine based on digital twins. The system and method embodiments in this application are based on the same inventive concept. The system includes: The data acquisition module 10 is used to acquire the operating parameters of the evaporation source and the measurement data of the film thickness sensor in real time. The operating parameters include at least the evaporation power and the evaporation source temperature. The theoretical deposition rate calculation module 20 is used to calculate the theoretical deposition rate based on the evaporation power, evaporation source temperature and the estimated effective material margin at the previous moment. The material effective margin fusion module 30 is used to fuse the theoretical deposition rate with the actual deposition rate determined based on the film thickness sensor measurement data to obtain the current effective margin of the material in the evaporation source. The digital twin update module 40 is used to update the state quantity in the digital twin of the evaporation coating process, which is used to characterize the material consumption process, by taking the current effective margin as a dynamic input, so that the material consumption state of the digital twin is consistent with the material consumption state of the physical evaporation source. The first control command generation module 50 is used to generate, when the deviation between the cumulative film thickness calculated by the updated digital twin within the first time window and the corresponding target film thickness exceeds the first threshold determined by the noise measured by the film thickness sensor, the synchronously updated digital twin is used as the prediction model to continuously generate evaporation power adjustment commands for reducing the film thickness prediction deviation in the future finite time domain, as the first control command. The second control instruction generation module 60 is used to generate a second control instruction for updating the subsequent process parameter sequence when the material sustainable evaporation time determined based on the current effective margin and combined with the actual deposition rate is less than the shortest process time required to complete the current film layer. The second control instruction generation module 60 uses the state-synchronized digital twin as the simulation environment and recalculates the deposition time and power allocation of the remaining film layer so that the final film system performance meets the target. The control instruction execution module 70 is used to send the first control instruction to the evaporation power supply to adjust the evaporation power, and to send the second control instruction to the process controller to execute the new process formula according to the updated process parameter sequence.

[0031] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0032] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0033] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A process control method for an evaporation coating machine based on digital twins, characterized in that, include: The operating parameters of the evaporation source and the measurement data of the film thickness sensor are acquired in real time. The operating parameters include at least the evaporation power and the temperature of the evaporation source. The theoretical deposition rate is calculated based on the evaporation power, evaporation source temperature, and the estimated effective material margin at the previous moment. The theoretical deposition rate is fused with the actual deposition rate determined based on film thickness sensor measurement data to obtain the current effective margin of material in the evaporation source; The current effective margin is used as a dynamic input to update the state quantity in the digital twin of the evaporation coating process that represents the material consumption process, so that the material consumption state of the digital twin is consistent with the material consumption state of the physical evaporation source. When the deviation between the cumulative film thickness calculated by the updated digital twin within the first time window and the corresponding target film thickness exceeds the first threshold determined based on the noise measured by the film thickness sensor, the synchronously updated digital twin is used as the prediction model to continuously generate evaporation power adjustment instructions for reducing film thickness prediction deviation within a limited time domain in the future, which serve as the first control instruction. When the material's sustainable evaporation time, determined based on the current effective margin and the actual deposition rate, is less than the shortest process time required to complete the current film layer, the digital twin, after state synchronization, is used as the simulation environment. By recalculating the deposition time and power allocation of the remaining film layers, the final film system performance meets the target, and a second control command is generated to update the subsequent process parameter sequence. The first control command is sent to the evaporation power supply to adjust the evaporation power, and the second control command is sent to the process controller to execute the new process formula according to the updated process parameter sequence.

2. The process control method for an evaporation coating machine based on digital twins according to claim 1, characterized in that, The theoretical deposition rate is fused with the actual deposition rate determined based on film thickness sensor measurement data, which is achieved through a Kalman filter algorithm. The current effective margin of material in the evaporation source is set as the system state variable estimated by the Kalman filter algorithm.

3. The process control method for an evaporation coating machine based on digital twins according to claim 1, characterized in that, The updated digital twin of the evaporation coating process includes: Based on the current effective margin, the efficiency coefficient or decay function parameter related to the material inventory in the evaporation source sub-model of the digital twin is dynamically adjusted.

4. The process control method for an evaporation coating machine based on digital twins according to claim 1, characterized in that, The first threshold includes: Acquire the measurement data sequence of the film thickness sensor within the first time window, and calculate its noise standard deviation; The noise standard deviation is input into a preset monotonically increasing function, and the first threshold is output.

5. The process control method for an evaporation coating machine based on digital twins according to claim 1, characterized in that, The process of using the synchronously updated digital twin as a prediction model to continuously generate evaporation power adjustment instructions for reducing film thickness prediction bias within a finite future time domain includes the following steps: Based on the current state of the digital twin after synchronous update of the current effective margin, the film thickness value of the digital twin is predicted in a rolling manner at a series of discrete moments within the first time window in the future, forming a film thickness prediction sequence. The predicted film thickness sequence is compared with the corresponding target film thickness sequence to calculate the film thickness deviation at each discrete moment within the first time window in the future. Based on the film thickness deviation, and in combination with the allowable output power range of the evaporation power supply, the power change rate limit, and the current effective margin's limitation on the deposition rate, the evaporation power adjustment amount corresponding to each discrete moment within the first time window is determined to form an evaporation power adjustment sequence, wherein each evaporation power adjustment amount represents an update amount relative to the current evaporation power setting value; The first evaporation power adjustment value corresponding to the current control cycle in the evaporation power adjustment sequence is used to update the current evaporation power setting value, so as to obtain a new evaporation power setting value for the current control cycle. The new evaporation power setting is used as the first control command and output to the evaporation power supply for execution.

6. The process control method for an evaporation coating machine based on digital twins according to claim 1, characterized in that, The generation of the second control instruction for updating the subsequent process parameter sequence includes the following steps: The material sustainable evaporation time, determined based on the current effective margin and in combination with the actual deposition rate, is set as the upper limit of the total duration of the remaining film deposition process. The digital twin, after state synchronization, is invoked to simulate the remaining film deposition time and evaporation power ratio of multiple candidate schemes, and the final total film thickness prediction value of each candidate scheme is obtained. Under the constraints of the total time limit, the allowable operating range of the evaporation power supply, and the minimum deposition time required for each film layer, a systematic traversal calculation of multiple feasible combinations of the deposition time and evaporation power setting values ​​for the remaining film layers is performed using a discretized search method. By comparing the deviations between the final total film thickness prediction value obtained through the digital twin simulation under each combination and the target total film thickness, the combination sequence with the smallest film thickness deviation is selected as the optimal process parameter sequence. The optimal process parameter sequence is converted into the second control command.

7. A process control system for an evaporation coating machine based on digital twins, characterized in that, The system is used to implement the digital twin-based evaporation coating machine process control method according to any one of claims 1 to 6, and the system comprises: The data acquisition module is used to acquire the operating parameters of the evaporation source and the measurement data of the film thickness sensor in real time. The operating parameters include at least the evaporation power and the evaporation source temperature. The theoretical deposition rate calculation module is used to calculate the theoretical deposition rate based on the evaporation power, evaporation source temperature and the previous time-lapse estimate of the effective material margin. The effective material margin fusion module is used to fuse the theoretical deposition rate with the actual deposition rate determined based on the film thickness sensor measurement data to obtain the current effective material margin in the evaporation source. The digital twin update module is used to update the state quantity in the digital twin of the evaporation coating process, which is used to characterize the material consumption process, by taking the current effective margin as dynamic input, so that the material consumption state of the digital twin is consistent with the material consumption state of the physical evaporation source. The first control command generation module is used to generate, when the deviation between the cumulative film thickness calculated by the updated digital twin within the first time window and the corresponding target film thickness exceeds the first threshold determined by the noise measured by the film thickness sensor, the synchronously updated digital twin is used as the prediction model to generate, in a rolling manner, evaporation power adjustment commands for reducing the film thickness prediction deviation within a limited time domain in the future, as the first control command. The second control command generation module is used to generate a second control command for updating the subsequent process parameter sequence when the material sustainable evaporation time determined based on the current effective margin and combined with the actual deposition rate is less than the shortest process time required to complete the current film layer. The module uses the state-synchronized digital twin as the simulation environment and recalculates the deposition time and power allocation of the remaining film layer so that the final film system performance meets the target. The control command execution module is used to send the first control command to the evaporation power supply to adjust the evaporation power, and to send the second control command to the process controller to execute the new process formula according to the updated process parameter sequence.

Images