Optical thin film coating temperature control method and system fused with infrared temperature measurement

Abstract

The invention relates to the technical field of vacuum coating machines, and particularly discloses an optical thin film coating temperature control method and system fused with infrared temperature measurement. The method comprises the following steps: selecting a temperature monitoring strategy according to target film system process requirements and substrate geometrical characteristics, calling a substrate quality mapping model, and measuring an angle set and an angle-region mapping table; traversing the measurement angle set to collect temperature in the coating process to form a time sequence temperature sequence; calculating a strategic local state vector based on the current strategy, and obtaining a comprehensive quality index through a quality transfer model; and performing system health degree diagnosis based on the time sequence temperature sequence and the comprehensive quality index, switching a normal control mode or a fault-tolerant control mode according to a diagnosis result, and executing a corresponding control instruction. According to the invention, the spanning from single-point temperature measurement to global quality perception and from temperature tracking to quality optimization control is realized, and the film layer quality consistency and the production process robustness are effectively improved.

Description

Technical Field

[0001] This application relates to the field of vacuum coating machine technology, specifically to a method and system for controlling the temperature of optical thin film coating that integrates infrared thermometry. Background Technology

[0002] In the vacuum deposition process of high-precision optical thin films, the thermal environment on the substrate surface has a significant impact on the adhesion, stress state, microstructure, and optical properties of the film. Uneven temperature distribution or dynamic fluctuations can lead to problems such as stress concentration, abnormal crystallization, and refractive index deviation in the film, which in turn can cause spectral characteristic shifts or film cracking, affecting the reliability and batch consistency of the device. Monitoring and controlling the substrate temperature has always been a key technical aspect of optical coating process control.

[0003] Currently, infrared thermography is widely used in industry for non-contact temperature measurement. This method does not require direct contact with the substrate, is suitable for vacuum environments, and has advantages such as fast response and high safety. However, due to the limitations of the coating cavity structure and the fixed measurement angle, infrared thermography can usually only obtain the temperature value of a local area on the substrate surface at a specific moment. Under the process conditions of continuous substrate rotation, this discrete, single-point measurement method is difficult to comprehensively reflect the temperature field distribution characteristics of the entire substrate surface and lacks effective spatial resolution. The mainstream control strategy still relies on the traditional PID algorithm, the core objective of which is to make the single-point temperature value track the preset temperature curve. The limitations of this method are: firstly, it is essentially a "reactive" control, and its response to dynamic thermal disturbances introduced by changes in deposition rate and plasma-assisted heating is lagging. Summary of the Invention

[0004] It solves the core technical problems of one-sided temperature sensing and lack of intelligent fault tolerance in existing technologies, and realizes the leap from single-point temperature measurement to full-domain quality perception, and from temperature tracking to quality optimization control, thereby achieving the technical effect of improving the consistency of film quality and the robustness of production process.

[0005] In view of the above problems, this application provides an optical thin film coating temperature control method integrating infrared thermometry, the method comprising: Obtain the process requirements of the target film system and the geometric characteristics of the target substrate, select the corresponding temperature monitoring strategy, and call the substrate quality mapping model, measurement angle set and angle-region mapping table related to the strategy from the process database; During the coating process, the measurement angle set is traversed, and an infrared thermometer is triggered at each angle to collect the temperature and record the process parameters simultaneously. Based on the angle-region mapping table, the collected temperature value is identified in real time as the characteristic temperature of the corresponding annular monitoring area, thus obtaining the time-series temperature sequence and process parameter sequence set of the current cycle. Based on the determined temperature monitoring strategy, the corresponding strategic local state vector is calculated according to the time-series temperature sequence of the current period, and the global perturbation estimation is performed to obtain the perturbation estimate. Based on the strategic local state vector and the perturbation estimate, the comprehensive quality index is obtained by using the pre-stored substrate quality mapping model. Based on the current period's time-series temperature sequence and the obtained comprehensive quality index, a system health diagnosis is performed: the signal-to-noise ratio of the time-series temperature sequence is analyzed, and it is determined whether the comprehensive quality index is below a preset threshold. If any of the above indicators is less than the preset threshold, the reliability of the temperature measurement system is determined to have decreased, a system degradation warning is generated, and the fault-tolerant control mode is triggered. If both of the above indicators are greater than or equal to the preset threshold, the system will remain in normal control mode. The decision-based control model generates and executes corresponding control commands. If it is in normal control mode, the trend of change is calculated based on the comprehensive quality index of the current cycle and the previous cycle, and the corresponding control command is generated and executed according to the trend. In fault-tolerant control mode, the heater power and deposition source parameters are locked at the current operating values, and basic safeguard control commands to maintain the continuous operation of the process are generated and executed.

[0006] This application also provides an optical thin film coating temperature control system that integrates infrared thermometry, the system comprising: The strategy configuration and model calling module is used to obtain the process requirements of the target film system and the geometric characteristics of the target substrate, select the corresponding temperature monitoring strategy, and call the substrate quality mapping model, measurement angle set and angle-region mapping table related to the strategy from the process database. The multi-angle synchronous temperature measurement and data acquisition module is used to traverse the set of measurement angles during the coating process, trigger the infrared thermometer to collect the temperature at each angle and record the process parameters synchronously, and identify the collected temperature value as the characterization temperature of the corresponding annular monitoring area in real time according to the angle-region mapping table, so as to obtain the time-series temperature sequence and process parameter sequence set of the current cycle. The comprehensive quality index calculation module is used to calculate the corresponding strategy-based local state vector based on the current period's time-series temperature sequence, perform global perturbation estimation to obtain perturbation estimate, and obtain the comprehensive quality index based on the strategy-based local state vector and perturbation estimate using a pre-stored substrate quality mapping model. The system health diagnosis module is used to diagnose system health based on the current period's time-series temperature sequence and the obtained comprehensive quality index: it analyzes the signal-to-noise ratio of the time-series temperature sequence and determines whether the comprehensive quality index is below a preset threshold. If any of the above indicators is less than the preset threshold, the reliability of the temperature measurement system is determined to have decreased, a system degradation warning is generated, and the fault-tolerant control mode is triggered. If both of the above indicators are greater than or equal to the preset threshold, the system will remain in normal control mode. The control command generation and execution module generates and executes corresponding control commands based on the control mode determined by the system health diagnosis module. If it is in normal control mode, the trend of change is calculated based on the comprehensive quality index of the current cycle and the previous cycle, and the corresponding control command is generated and executed according to the trend. In fault-tolerant control mode, the heater power and deposition source parameters are locked at the current operating values, and basic safeguard control commands to maintain the continuous operation of the process are generated and executed.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages: 1) Based on a single infrared thermometer, this approach achieves a leap from single-point temperature measurement to global quality control, significantly improving the consistency of membrane quality. By constructing a "virtual dual-zone state pair" and calculating the "comprehensive quality index Q," this solution transforms fragmented temperature data into a quantitative evaluation value directly linked to the final membrane quality. 2) An integrated intelligent closed loop of "sensing-diagnosis-control" was constructed, significantly enhancing the robustness and fault tolerance of the production system. This solution achieves real-time monitoring of the health status of the infrared temperature measurement system by analyzing the signal-to-noise ratio of the time-series temperature data online and performing joint diagnosis with comprehensive quality indicators; 3) Through strategic collaborative optimization control, thermal field disturbances under complex operating conditions are effectively suppressed. To address different process requirements such as center stability and edge uniformity, this solution employs differentiated monitoring strategies and models, enabling precise quantification of the thermal coupling effects between different regions. Based on this, the generated control commands can collaboratively adjust heater power and deposition source parameters, while simultaneously optimizing temperature levels and spatial distribution. This effectively suppresses complex thermal disturbances caused by deposition self-heating and airflow changes, ensuring film quality under harsh process conditions.

[0008] In summary, this application systematically solves the pain point of temperature control in high-precision optical coating by using information fusion, intelligent diagnosis and strategic control as technical approaches, and ultimately achieves comprehensive technical effects such as improving product yield, enhancing process adaptability and ensuring production safety. Attached Figure Description

[0009] Figure 1 This is a flowchart illustrating a method for controlling the temperature of optical thin film coating that integrates infrared thermometry, as provided in an embodiment of this application.

[0010] Figure 2 This is a schematic diagram of an optical thin film coating temperature control system that integrates infrared thermometry, provided as an embodiment of this application.

[0011] Figure labeling: Strategy configuration and model calling module 11, multi-angle synchronous temperature measurement and data acquisition module 12, comprehensive quality index calculation module 13, system health diagnosis module 14, control command generation and execution module 15. Detailed Implementation

[0012] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it 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, the following are specific embodiments of this application.

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] This application provides an embodiment of an optical thin film coating temperature control method that integrates infrared thermometry, such as... Figure 1 As shown, the method includes: application to a vacuum evaporation or ion beam sputtering coating apparatus equipped with a rotating substrate holder. The substrate is mounted on the rotating substrate holder and rotates periodically with it; an infrared thermometer is located outside the vacuum chamber, and its optical path is aligned with the substrate surface through an infrared transmission window on the vacuum chamber wall, achieving non-contact multi-point temperature monitoring during the substrate rotation.

[0015] S1: Obtain the process requirements of the target film system and the geometric characteristics of the target substrate, select the corresponding temperature monitoring strategy, and call the substrate quality mapping model, measurement angle set and angle-region mapping table related to the strategy from the process database.

[0016] Specifically, before operating the vacuum coating equipment, the process formula is called up: the operator inputs the target film system for this production task (e.g., ...) through the human-machine interface. The system considers three key elements: antireflective coating, substrate material (such as fused silica, silicon wafer, BK7 glass, etc.), and substrate geometry (including size, thickness, whether it is planar or curved, and geometric shape). Based on these three elements, a process identification code is generated and used as an index to query the pre-stored process database in the control system. From the pre-stored process database, the system retrieves the corresponding three core elements: substrate quality mapping model, measurement angle set, and angle-region mapping table. According to the selected temperature monitoring strategy and substrate size, the radial installation position parameters of the infrared thermometer are automatically determined. Then, the mechanical positioning mechanism is driven to precisely move it to the target position, ensuring that the temperature measurement spot completely covers and falls within the substrate surface monitoring area specified by the strategy. The measurement angle set is a discrete set of angle positions distributed within the substrate rotation period [0°, 360°) used to trigger sampling by the infrared thermometer. Its design follows the principle of global coverage and regional emphasis; specifically, for the central annular monitoring area, the measurement angle should focus on capturing rotational symmetry. Angle sets typically consist of M angles uniformly distributed within [0°, 360°), such as {0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°}. This setup can indiscriminately sense temperature fluctuations in the central region in all directions, effectively assessing its stability. For the peripheral annular monitoring area, the measured angles should, while ensuring global coverage, focus on capturing process disturbances in specific directions. Angle sets can use N angles, with most angles quasi-uniformly distributed within [0°, 360°) to ensure global coverage; simultaneously, local refinement is performed within a sector near key process directions (such as the tangential direction pointed to by the deposition source). Example: If the deposition source is fixed at 90°, a complete angle set can be designed as {0°, 45°, 75°, 80°, 85°, 90°, 95°, 100°, 105°, 135°, 180°, 225°, 270°, 315°}. Among them, angles of 0°, 45°, and 135° ensure global coverage, while the densified angles in the range of 75° to 105° enable high-resolution monitoring of the areas with the strongest deposition flux.

[0017] Further temperature monitoring strategies include: The temperature monitoring strategy divides the substrate surface into two concentric annular monitoring areas according to the process requirements of the film system, and assigns a unique spatial semantic label to each annular area. Decision logic of temperature monitoring strategy: The IF process requires a focus on monitoring the stability of the film deposition in the central region of the substrate. The THEN strategy is defined as central stability monitoring, and its execution steps are as follows: install an infrared thermometer at a radial position relative to the center of rotation of the substrate so that its spot trajectory falls into the central annular region, and assign a semantic label of central temperature to this region. The IF process requires a focus on monitoring the uniformity of film deposition in the edge region of the substrate. The THEN strategy is defined as edge uniformity monitoring, and its execution steps are: to install an infrared thermometer at a radial position relative to the rotation center of the substrate so that its spot trajectory falls into the edge annular region, and to assign an edge temperature semantic label to this region. The system is divided into two concentric ring-shaped monitoring areas, including: The central annular region is a circular area on the substrate surface with the geometric center as the center and a radius ranging from 0% to 25% of the radius of its circumscribed circle, including the endpoints; The edge annular region is an annular region on the substrate surface with the geometric center as the center and a radius ranging from 75% to 100% of the radius of its circumscribed circle, including the endpoints.

[0018] Specifically, the temperature monitoring strategy relies on the analysis of the target film system's process requirements and the acquisition of the target substrate's geometric features. These geometric features include, but are not limited to, the substrate's geometry, thickness, and circumscribed circle radius. This embodiment primarily focuses on circular substrates, whose geometric center coincides with the rotation center of the rotating substrate holder. The principle for dividing the annular monitoring area is based on different process concerns, dividing the circular substrate surface into two concentric annular monitoring areas, and assigning each area a unique spatial semantic label. The specific divisions are as follows: Central annular area: defined as a circular area (including endpoints) with the substrate's geometric center as the center and a radius of 0% to 25% of its circumscribed circle radius. Edge annular area: defined as an annular area (including endpoints) with the substrate's geometric center as the center and a radius of 75% to 100% of its circumscribed circle radius. The technical basis and preferred scheme for area division are as follows: The determination of the above division range is based on the following technical considerations: Determination of the central area range (0%-25%): This range aims to capture the most critical "point source-like" thermal behavior. The area within a 25% radius typically reflects the stability of the central heat source most directly and is least affected by edge effects. Verified through both production practice and expert methods, 15% is an optimal radius ratio. This ratio allows for the most sensitive detection of central temperature fluctuations while providing a suitably sized projection target area for the infrared thermometer's spot. Determining the edge region range (75%-100%): This range aims to capture the "boundary" effect, where uniformity issues are most likely to occur. The approximately 25% ring around the outer edge of the substrate is the area with the largest temperature gradient and the most likely film thickness inhomogeneity. A 75% starting point ensures the monitoring area fully covers this high-gradient region while avoiding overlap with the central stable region. Verified through both production practice and expert methods, 85% is an optimal starting radius ratio. This ratio accurately covers the key area where uniformity issues occur, providing the most representative data sample for uniformity control based on edge temperature data. Based on the above regional division, the strategy decision-making and execution logic is as follows: The IF process requires monitoring the stability of the film deposition in the central region of the substrate (e.g., center thickness uniformity), and the THEN strategy is defined as center stability monitoring. The IF process requires monitoring the uniformity of the film deposition in the edge regions of the substrate (e.g., edge thickness uniformity), and the THEN strategy is defined as edge uniformity monitoring. Execution steps: Based on the selected strategy, the target monitoring area (central area or edge area) is determined. Combining this with the preferred size proportions of the area (e.g., 15% of the radius for the central area, and 15% of the outer edge for the edge area, i.e., 85%-100%), the specific physical extent of this area on the substrate is calculated. Using a mechanically automated device, the infrared thermometer is precisely moved and positioned radially such that the trajectory formed by its light spot as the substrate rotates falls entirely within the determined preferred physical extent. This precise positioning based on preferred dimensions ensures that the temperature signal can most effectively characterize the temperature characteristics of the target area, laying a solid foundation for subsequent high-precision control.

[0019] Furthermore, the angle-region mapping table includes: To obtain the requirements for the target membrane system, determine the current temperature monitoring strategy, and calibrate the angle-region mapping table under this strategy offline. The calibration process is as follows: The current temperature monitoring strategy defines a ring-shaped monitoring area on the surface of a standard substrate. Based on the geometry of the standard substrate to be calibrated, a radial mounting position is associated, and the infrared thermometer is mounted at this associated radial mounting position. In the vacuum coating equipment, install and calibrate the rotating substrate holder and its rotary encoder to ensure that the substrate rotation center is aligned with the geometric center; A standard substrate with known geometric features is mounted on a calibrated rotating substrate holder. The zero-degree reference position of the rotary encoder is defined, and the initial projection position of the infrared thermometer spot on the substrate surface at this reference position is determined as the reference coordinate point. Based on the geometric relationship between the reference coordinate point and the center of rotation of the substrate, the theoretical trajectory of the light spot formed on the surface of the substrate when the standard substrate rotates within the measurement angle set is calculated. According to the current temperature monitoring strategy, the annular area corresponding to the theoretical trajectory is marked with the spatial language label corresponding to the current strategy, and the correspondence between the measurement angle and the spatial semantic label is established. Summarize all the correspondences under the current strategy, generate an angle-region mapping table, associate the mapping table with the geometric feature information of the standard substrate, and store it in the process database.

[0020] Specifically, the angle-region mapping table pre-stored in the process database is constructed through an offline calibration process. The specific steps are as follows: (1) System calibration preparation: In the vacuum coating equipment, install and calibrate the rotating substrate holder and its high-precision rotary encoder to ensure that the rotation center is aligned with the geometric center of the substrate; at the same time, fix the infrared thermometer so that its light path passes through the infrared window and is aligned with the substrate. Here, the substrate rotation center refers to the projection point of the mechanical rotation axis of the rotating substrate holder on the substrate plane, that is, the physical center point O around which the substrate rotates. rotSubstrate geometric center: refers to the center of symmetry of the substrate's own geometric shape (such as the center of a circular substrate), denoted as O. geo The rotation center and geometric center of the substrate must be aligned. Eccentricity will cause the substrate to wobble during rotation, resulting in the temperature measurement spot trajectory not being an ideal circle. The projection positions of different batches at the same angle will be inconsistent, the temperature field reconstruction will be distorted, and the control accuracy will be affected. To achieve center alignment, the geometric center of the substrate is made to coincide with the rotation center after installation by using a precision fixture or automatic centering chuck on the rotating substrate holder. (2) Determining the reference position: Install a standard substrate with known geometric features on the rotating substrate holder, define the zero-degree reference position of the rotary encoder, and at this position, accurately determine the projection position of the infrared thermometer spot on the surface of the standard substrate by auxiliary optical means (such as laser pointer) or mechanical positioning device, and record this position as the reference coordinate point. Here, the geometric features must at least include the geometric shape and size of the substrate. The geometric features directly determine the radial boundary division of the central area, transition area and edge area of ​​the substrate surface. (3) Trajectory calculation and region association: Drive the rotating substrate holder to rotate it step by step according to the measurement angle set. At each angle θ i Below, the theoretical projected coordinates of the infrared spot are calculated using the geometric relationship between the reference coordinate point and the center of rotation, i.e., the coordinate transformation relationship under rigid body rotation. For example, the initial projected coordinates (x0, y0) of the infrared thermometer spot at the zero-degree reference position are transformed to the current angle θ. i Theoretical projected coordinates (x) i y i The transformation relationship is as follows: ; Where, θ i The rotation angle is relative to the zero-degree reference. Its positive direction is defined according to the device coordinate system. When the actual rotation direction is opposite to this positive direction, the system automatically adjusts θ. i A negative sign is used to ensure accurate coordinate calculation. The calculation of theoretical projected coordinates is based on a rigid body rotational kinematics model, applicable to various substrates rotating around a geometric center. For each measured angle θ... i Its corresponding theoretical projection coordinates (x i y i The value θ represents the physical spatial location of the infrared thermometer spot on the substrate surface. Based on predefined substrate region division rules, this spatial location is matched with the spatial extent of each region to determine its substrate region affiliation. For example, for a circular substrate, its affiliation can be determined by calculating the radial distance from this point to the substrate's rotation center and comparing it with the region boundary defined based on the substrate radius. Thus, a system is established based on the measurement angle θ. i The correspondence to the substrate area. (4) Mapping table generation and storage: collect all measured angles θ iFor the corresponding region, a final angle-region mapping table is generated and stored in the process database along with the geometric feature information of the corresponding standard substrate for use in the online coating process.

[0021] S2: During the coating process, the measurement angle set is traversed. At each angle, the infrared thermometer is triggered to collect the temperature and record the process parameters simultaneously. Based on the angle-region mapping table, the collected temperature value is identified in real time as the characteristic temperature of the corresponding annular monitoring area, thus obtaining the time-series temperature sequence and process parameter sequence set of the current cycle.

[0022] Specifically, after the coating process starts, the system synchronously acquires data and forms sequences. The specific steps are: (1) Real-time monitoring and triggering: The system continuously reads the angle signal θ(t) emitted by the high-precision rotary encoder installed on the rotating substrate holder. (2) Synchronous sampling: When θ(t) is compared with any angle θ in the set of measured angles loaded in S1... i During matching, the system simultaneously performs two actions: Triggering temperature acquisition: Sending a trigger signal to the infrared thermometer outside the vacuum chamber to obtain the surface temperature T(t) at the current time t. At this time, according to the angle-region mapping table established by S1, the temperature value is recorded as the temperature of the corresponding region. Recording process parameters: Simultaneously recording all relevant process parameters at the current time t, including evaporation source power, substrate rotation speed, ion beam current, working gas pressure, gas flow, and real-time film thickness measured by the crystal oscillator film thickness gauge, etc. (3) Sequence construction: Repeating steps (1) and (2) until all angles in the measurement angle set have been traversed and sampled within a complete rotation cycle. At this time, the system aligns and packages all timestamp data collected in this cycle to form the time-series temperature sequence of the current cycle and the process parameter sequence set of the current cycle.

[0023] S3: Based on the determined temperature monitoring strategy, calculate the corresponding strategyd local state vector according to the time-series temperature sequence of the current period, perform global perturbation estimation to obtain the perturbation estimate, and obtain the comprehensive quality index by using the pre-stored substrate quality mapping model based on the strategyd local state vector and the perturbation estimate.

[0024] Specifically, when the selected strategy is center stability monitoring, a strategy-based local state vector is calculated. Based on S1, the angle-region mapping table of the relevant strategy is called to map all temperature data points in the current periodic temperature sequence to center temperature semantic labels. The standard deviation of these center region temperature data is calculated; this standard deviation is the strategy-based local state vector under this strategy, used to quantify the temperature fluctuations in the core region. Global perturbation estimation and thermal perturbation propagation are performed. The calculated standard deviation of the center region temperature is input into a pre-stored thermal perturbation propagation model. This model describes the quantitative relationship of the propagation of fluctuations from the center region to the edge region, outputting an equivalent edge temperature perturbation value in °C. This perturbation estimate is used to quantify the potential impact of the current fluctuations in the center region on the overall process uniformity. When the selected strategy is edge uniformity monitoring, a strategy-based local state vector is calculated. Based on S1, the angle-region mapping table of the relevant strategy is called to map all temperature data points in the current periodic time-series temperature series to edge temperature semantic labels. The range of these edge region temperature data (the difference between the maximum and minimum values) is calculated. This range is the strategy-based local state vector under this strategy, used to directly characterize the temperature uniformity of the edge region. Global perturbation estimation and thermal perturbation source tracing are performed. The edge region temperature range calculated in the previous steps is input into a pre-stored thermal perturbation source tracing model. This model can infer the possible sources of edge non-uniformity and output an equivalent central region temperature fluctuation intensity, in °C. This perturbation estimate is used to quantify the fluctuation intensity of the central heat source most likely to cause the current edge non-uniformity problem.

[0025] Furthermore, based on the current period's time-series temperature sequence, the corresponding strategy-based local state vector is calculated, and a global perturbation estimate is performed to obtain the perturbation estimate, including: When the strategy is to monitor the center stability, the local state vector is the temperature standard deviation of the center region. The temperature standard deviation of the center region and the feature values ​​of the extracted process parameter sequence set are input into the pre-stored thermal disturbance transfer model to obtain the equivalent edge temperature disturbance value. This value is used to quantify the degree of thermal impact of the center fluctuation on the edge region. When the strategy is edge uniformity monitoring, the local state vector is the temperature range of the edge region. The feature values ​​of the edge region temperature range and the extracted process parameter sequence set are input into the pre-stored thermal disturbance tracing model to obtain the equivalent central region temperature fluctuation intensity. This value is used to quantify the central heat source fluctuation intensity most likely to cause edge non-uniformity. The calculation of the temperature standard deviation in the central region and the temperature range in the peripheral regions includes: Based on a defined temperature strategy, an angle-region mapping table related to the strategy is obtained, and each value in the time-series temperature sequence of the current period is mapped to a relevant spatial language label based on the table. The corresponding semantic label is the center temperature. The temperature standard deviation of the current period's time-series temperature sequence is calculated to obtain the temperature standard deviation of the center region. The corresponding semantic label is edge temperature. The temperature range of the current period's time-series temperature sequence is calculated to obtain the temperature range of the edge region.

[0026] Specifically, the construction and training method of the thermal perturbation transfer model, under the central stability monitoring strategy, requires the construction of a thermal perturbation transfer model to achieve a quantitative assessment of the thermal impact of temperature fluctuations in the central region on the edge region. This model employs a nonlinear data-driven method, and the specific construction process is as follows: First, in the offline calibration stage, multiple sets of controlled experiments are conducted based on the geometric characteristics of the target film system and substrate. During the experiments, thermal perturbations of different intensities and frequencies are introduced by actively adjusting the power of the central heater. Simultaneously, a high-precision infrared thermal imager is used to synchronously collect temperature distribution data of the central and edge regions of the substrate surface, and a complete set of process parameter sequences is recorded. The process parameter sequence set includes: real-time heater power, deposition source beam intensity, evaporation rate, working gas flow rate in the vacuum chamber, substrate rotation speed, and chamber background pressure. For each set of experimental data, the standard deviation of the central region temperature is extracted as the local state input, and the range of the edge region temperature is calculated as the model's expected output label, i.e., the equivalent edge temperature perturbation value. This value is not a direct measurement of the edge temperature, but rather a comprehensive quantitative indicator representing the degree of edge temperature non-uniformity caused by the transfer of thermal fluctuations from the central region to the edge through physical mechanisms such as heat conduction, radiation, and airflow convection under current process conditions. The unit is degrees Celsius, and a higher value indicates poorer edge thermal stability. Simultaneously, key features are extracted from the process parameter sequence: the mean and standard deviation of heater power (reflecting the overall power level and fluctuations of the heater); the rate of change of the deposition source beam (indicating the stability of the deposition source and its potential impact on the temperature field); the integral value of the evaporation rate (representing the mass of accumulated deposits, indirectly affecting temperature distribution); the fluctuation amplitude of gas flow rate (reflecting the modulation effect of airflow on the temperature field); and the stability index of rotation speed (i.e., rotational speed deviation rate) (assessing whether the substrate rotation is smooth, affecting temperature uniformity). These features serve as auxiliary inputs to characterize the modulation effect of the current process environment on the thermal disturbance transmission path. Subsequently, the standard deviation of the central region temperature and the above process features are combined to form a joint input vector, with the equivalent edge temperature disturbance value as the output. The model is then trained using nonlinear regression algorithms such as support vector regression or random forest. The model's hyperparameters are optimized through cross-validation to ensure its generalization ability under different operating conditions. After the trained thermal disturbance transfer model is validated, it is bound to the corresponding film system and substrate information and stored in the process database for use in online control, enabling real-time prediction of edge thermal effects.

[0027] The construction and training method of the thermal perturbation source tracing model: Under the edge uniformity monitoring strategy, in order to identify the intensity of the core heat source perturbation that causes edge temperature non-uniformity, a thermal perturbation source tracing model needs to be constructed. This model adopts a nonlinear modeling method, aiming to infer the central heat source cause of edge non-uniformity phenomena. The specific construction process is as follows: In the offline stage, for specific film systems and substrate configurations, multiple sets of comparative experiments are designed, applying different types of artificial perturbations to construct a training dataset with clear causal relationships. Multi-source perturbation data refers to the types of thermal field perturbations applied intentionally from different sources in different experimental groups. Specifically, these include: Central heat source perturbation group: actively adjusting only the power of the central heater to introduce sinusoidal or step fluctuations while maintaining other parameters stable; Edge heat source perturbation group: adjusting only the power of the edge heating zone or changing the edge gas spray intensity; Deposition source perturbation group: adjusting the electron beam deflection angle or the position of the evaporation source crucible to cause deposition flux eccentricity; Airflow perturbation group: changing the inlet distribution or gas flow pattern to induce uneven convection within the cavity; Composite perturbation group: simultaneously introducing two or more perturbations to simulate complex operating conditions. In each experiment, infrared thermography sequences of the edge region and a complete set of process parameter sequences are collected synchronously. The process parameter sequence set includes: heater power in each zone (real-time values ​​of the central and edge regions), deposition source voltage and current (reflecting beam stability), evaporation rate (instantaneous and cumulative values ​​obtained from a quartz crystal monitor), working gas pressure and gas flow rate (reflecting cavity environmental stability), and substrate rotation speed (used to calculate rotational deviation rate and periodic fluctuations). For each set of experimental data, the temperature range of the edge region is first calculated as the local state input. Based on the known perturbation type and intensity in the experimental design, the actual temperature fluctuation intensity (characterized by standard deviation) of the central heating region at the corresponding time is extracted as the model output label, namely, the "equivalent central region temperature fluctuation intensity". This value is not the real-time measured central temperature, but a "equivalent central thermal perturbation source intensity" obtained through model inversion that can explain the current edge non-uniformity. The unit is degrees Celsius. The higher the value, the more likely the edge non-uniformity is dominated by central thermal field fluctuations. At the same time, features are extracted from the process parameter sequence, including: the ratio of central to edge heating power (reflecting the thermal field distribution tendency); the rate of change of deposition source voltage and current (indicating deposition flux stability); the dynamic coupling degree of gas flow rate and evaporation rate (extracted through cross-correlation analysis); and the periodic fluctuation frequency and amplitude of substrate rotation speed (reflecting mechanical stability). These features, together with the edge temperature range, constitute the model's input feature vector. "Multi-source perturbation data" is fully incorporated into the model's training dataset, meaning that each set of experimental data (including input feature vectors and output labels) is added to the training set as an independent sample.Since each set of data corresponds to a specific type of disturbance source, the model can learn during training that: when the input features include "large fluctuations in central heating power, stable deposition source, and uniform airflow," edge non-uniformity is more likely caused by the central heat source, resulting in a higher output of "equivalent central region temperature fluctuation intensity"; conversely, when the input features include "large changes in edge heating power and stability at the center," the model should output a lower value, indicating that edge non-uniformity is not caused by the center. Through this multi-category, multi-condition comparative learning, the model not only learns the mapping relationship between input and output but also enhances its ability to identify different disturbance modes, avoiding the misattribution of edge disturbances to the central heat source. Training employs non-linear algorithms such as gradient boosting regression trees or neural networks, with hyperparameters optimized through cross-validation. After the trained thermal disturbance tracing model is tested on an independent validation set, it is associated with the corresponding process configuration and stored in the process database for use in online control, enabling rapid tracing and quantitative evaluation of the root causes of edge non-uniformity.

[0028] Furthermore, using a pre-stored substrate quality mapping model, a comprehensive quality index is obtained, including: Based on the determined temperature monitoring strategy, a virtual two-zone state pair (Vmsr, Vinf) consisting of the corresponding strategy-based local state vector and the model-inferred perturbation estimate is constructed: If a central stability monitoring strategy is adopted, the temperature standard deviation of the central region is used as Vmsr, and the equivalent edge temperature disturbance value is used as Vinf. If the edge uniformity monitoring strategy is adopted, the temperature range of the edge region is taken as Vmsr, and the temperature fluctuation intensity of the central region is taken as Vinf; a) Calculate the imbalance D of the virtual dual-zone state pair using the following formula: D=(Vmsr×Vinf) / (Vmsr²+Vinf²+δ)+λ·(Vmsr / (Vinf+ε)+Vinf / (Vmsr+ε)); In the formula, the first term is the normalized coupling strength term, the second term is the relative deviation penalty term, δ is the normalized stabilizer to prevent Vmsr and Vinf from being 0 at the same time, with the dimension [T²], ε is the ratio stabilizer to prevent the ratio calculation from being divided by zero when Vmsr or Vinf is 0, with the dimension [T], and λ is the deviation penalty coefficient, T = ℃ or K; Establish the mapping relationship from the degree of imbalance D to the comprehensive quality index Q, and its calculation formula is as follows: Q = exp(-D / Dref); Where Dref is the reference imbalance degree, and the comprehensive quality index Q is a dimensionless quantity between 0 and 1, and its value is positively correlated with the film quality. The deviation penalty coefficient and reference imbalance are defined as configurable parameters associated with the target membrane system type and stored in the process database.

[0029] Specifically, the calculation method of comprehensive quality indicators To further achieve a comprehensive and strategic assessment of the coating process quality, this invention introduces the concept of a "virtual dual-zone state pair" and constructs a comprehensive quality index based on it. The core of this method lies in combining locally observable temperature states with model-inferred global perturbation estimates to form a virtual state pair that reflects the degree of imbalance in the "local-global" coupling, thereby quantifying the film quality risk. The construction and function of the virtual dual-zone state pair are crucial because the temperature distribution on the substrate surface directly affects the stress, adhesion, and refractive index uniformity of the film during the coating process. However, limited by the observational limitations of single-point infrared thermometry, the system can only acquire temperature information for local areas. Therefore, this invention proposes the construction of a "virtual dual-zone state pair," the structure of which is as follows: When employing a center stability monitoring strategy, the standard deviation of the temperature in the center region is used as the primary state variable (denoted as the first state variable) to characterize the stability of the central thermal field; the equivalent edge temperature disturbance value output by the thermal disturbance transmission model is used as the inferred state variable (denoted as the second state variable) to characterize the potential impact of center fluctuations on the edge region. Together, they constitute a virtual dual-zone state pair. When employing an edge uniformity monitoring strategy, the temperature range in the edge region is used as the primary state variable to characterize the uniformity level of the edge temperature; the intensity of the equivalent central region temperature fluctuation output by the thermal disturbance tracing model is used as the inferred state variable to characterize whether the edge non-uniformity is caused by a central heat source disturbance. Together, they constitute a virtual dual-zone state pair. The purpose of constructing a virtual dual-zone state pair is to overcome the spatial limitations of single-point temperature measurement and, through an "observation + inference" approach, virtually reconstruct the thermal state correlation between the two key regions of the substrate surface—the center and the edge. This state pair not only includes measured information but also incorporates the influence of the process environment on thermal disturbance transmission or tracing, thus providing a more comprehensive input for subsequent quality assessment. The calculation of the imbalance degree of virtual dual-zone state pairs and its physical meaning: To quantify the degree of "imbalance" of virtual dual-zone state pairs, the imbalance degree D is defined, and its calculation formula is as follows: Imbalance = (Master state quantity × Inferred state quantity) / (Master state quantity² + Inferred state quantity² + δ) + Deviation penalty coefficient × (Master state quantity / (Inferred state quantity + ε) + Inferred state quantity / (Master state quantity + ε)) The first term, "(Main State Variable × Inferred State Variable) / (Main State Variable² + Inferred State Variable² + δ)," is called the normalized coupling strength term, used to measure the degree of coordination between the two state variables. When both increase or decrease in the same direction, this term approaches its maximum value of 0.5, indicating that the system is in a certain "consistent" disturbance state; when one of them is zero, this term approaches zero, indicating that the disturbance exists in isolation. The second term, "Deviation Penalty Coefficient × (Main State Variable / (Inferred State Variable + ε) + Inferred State Variable / (Main State Variable + ε))," is called the relative deviation penalty term, used to penalize situations where the two state variables are severely mismatched. For example, if the main state variable is large while the inferred state variable is small, or vice versa, it indicates a contradiction between observation and model inference, and the system may have unmodeled disturbances or sensor anomalies. δ is a normalized stabilizer, with the dimension of temperature squared (°C² or K²), used to prevent the denominator from being zero when both the main state variable and the inferred state variable approach zero. Its value ranges from 0.01 to 1, with a preferred value of 0.1. ε is a ratio stabilizer, with the dimension of temperature (°C or K), used to prevent division by zero errors when calculating ratios. Its value ranges from 0.1 to 5, with a preferred value of 1.0. The deviation penalty coefficient λ is a dimensionless parameter used to adjust the weight of the second term in the total imbalance. Its value ranges from 0.1 to 2.0. When using a center stability monitoring strategy, more attention is paid to whether center fluctuations cause edge problems; therefore, λ is preferably 0.5 to 1.0. When using an edge uniformity monitoring strategy, a stricter penalty is needed for deviations between observations and source tracing results; therefore, λ is preferably 1.0 to 2.0. When using a center stability monitoring strategy, the system mainly focuses on whether center thermal fluctuations cause edge thermal disturbances, allowing for a certain degree of model inference bias; therefore, a lower penalty weight should be set. Under this strategy, the preferred value for the deviation penalty coefficient λ is 0.8. When employing an edge uniformity monitoring strategy, the system must rigorously identify the root cause of edge non-uniformity. If a significant edge temperature difference is observed but the model's source tracing does not point to a central disturbance, it likely indicates the presence of unmodeled local edge disturbances (such as window contamination, airflow deflection, substrate edge occlusion, etc.), which must be given high weight and warned accordingly. Under this strategy, the preferred value for the deviation penalty coefficient λ is 1.5. This formula is called the "imbalance degree of the virtual dual-region state pair" because it comprehensively evaluates the "consistency" and "coordination" between the states of the two virtual regions. For the center stability monitoring strategy, if the center temperature fluctuates greatly (large master state variable) and the edge perturbations inferred by the model are also large (large inferred state variable), the system is in an "interpretable" imbalance state with a moderate degree of imbalance. If the center fluctuates greatly but the edge perturbations are small, the model may have failed or there may be other interferences. The deviation penalty term will significantly increase the degree of imbalance, indicating a system anomaly. For the edge uniformity monitoring strategy, if the edge non-uniformity is severe (large master state variable) and the source tracing points to strong center perturbations (large inferred state variable), the degree of imbalance reflects the true risk. If the edge is non-uniform but the source tracing shows that the center is stable, the deviation penalty term will significantly increase the degree of imbalance, indicating that there may be local edge problems (such as occlusion, airflow disturbances, etc.). The establishment and significance of the comprehensive quality index: After obtaining the imbalance degree D, a mapping relationship from D to the comprehensive quality index Q is established. The calculation formula is: Comprehensive Quality Index = exp(-Imbalance Degree / Reference Imbalance Degree). The significance of this formula is that it transforms the nonlinear imbalance degree into a dimensionless index between 0 and 1. The higher the value, the more stable the thermal state of the coating process and the more guaranteed the film quality. When the imbalance degree approaches zero, the comprehensive quality index approaches 1, indicating that the system is in an ideal state; when the imbalance degree increases, the comprehensive quality index decays exponentially, reflecting a rapid increase in quality risk. It is called a "comprehensive quality index" because it integrates local temperature measurement, model inference, the influence of process parameters, and strategic logic. It is a comprehensive and quantitative assessment of the thermal stability and quality risk of the coating process, rather than a direct reflection of a single physical quantity. The reference imbalance degree is the key parameter of this mapping relationship, used to set the "time scale" of quality degradation. Its physical meaning is: when the imbalance degree reaches the reference imbalance degree, the comprehensive quality index decays to about 0.37 (i.e., 1 / e), indicating that the quality has significantly deteriorated. The value of the reference imbalance is related to the process sensitivity of the target film system: for film systems with high tolerance, such as ordinary antireflection films, the reference imbalance can be set to 1.0 to 2.0, with an optimal value of 1.5 for the reference imbalance Dref; for film systems that are extremely sensitive to temperature uniformity, such as high-reflectivity mirrors and narrow-band filters, the reference imbalance should be set to 0.5 to 1.0, with an optimal value of 0.7 for the reference imbalance Dref, to trigger quality warnings earlier. The reference imbalance, along with the deviation penalty coefficient λ, serves as a configurable parameter associated with the target film system type. During the system debugging phase, it is calibrated using historical data or set using expert experience and stored in the process database for use by different processes. In summary, this invention, by constructing virtual dual-zone state pairs, calculating the imbalance, and mapping it to a comprehensive quality index, achieves a strategic, intelligent, and configurable evaluation of the coating process quality, providing a reliable basis for subsequent control mode decisions.

[0030] S4: Based on the current period's time-series temperature sequence and the obtained comprehensive quality index, perform system health diagnosis: analyze the signal-to-noise ratio of the time-series temperature sequence, and simultaneously determine whether the comprehensive quality index is below a preset threshold. If any of the above indicators < the preset threshold, it is determined that the reliability of the temperature measurement system has decreased, a system degradation warning is generated, and the fault tolerance control mode is triggered; If both of the above indicators ≥ the preset threshold, the system remains in the normal control mode.

[0031] Specifically, this step is used to evaluate the health status of the infrared temperature measurement system itself, and the specific implementation is as follows: Calculate the rotation fundamental frequency: Obtain the set rotation speed R (unit: RPM) of the rotating substrate holder during the coating process, and calculate the fundamental frequency f0 of the substrate rotation accordingly. f0 = R / 60 (unit: Hz), and this component reflects the periodic temperature signal caused by the substrate rotation and strictly corresponding to the monitoring area position. Spectrum transformation and feature extraction: Perform a fast Fourier transform on the time-series temperature sequence of the current period to obtain its spectrum. Extract two key amplitudes from this spectrum: a) Signal amplitude Asignal: It is the amplitude of the spectrum at the substrate rotation fundamental frequency f0; b) Noise amplitude Anoise: It is the average amplitude of the spectrum within the preset noise frequency band, and the preset noise frequency band is defined as the union of [0.1f0, 0.9f0] and [[1.1f0, 2.5f0]. This frequency band range is designed to exclude the fundamental frequency and its main harmonics, so as to capture the non-periodic interference caused by electrical noise, random fluctuations of background radiation, etc. Based on the signal amplitude and the noise amplitude, calculate the signal-to-noise ratio: SNR = 10×log 10 (Asignal / Anoise); Signal health determination: Compare the calculated SNR with the preset signal-to-noise ratio threshold (SNR threshold ) for comparison. If the signal-to-noise ratio SNR < the preset threshold, it is determined that the health status of the infrared temperature measurement system is abnormal, and the signal reliability is insufficient to support high-quality control. Based on the process state and model effectiveness diagnosis of the comprehensive quality index, this step is used to evaluate the potential of the current process to produce a qualified film layer and the effectiveness of the quality assessment model, and the specific implementation is as follows: Directly read the calculated comprehensive quality index Q. This index is a dimensionless quantity between 0 and 1, and its numerical value is positively correlated with the film layer quality. Compare Q with the preset quality index threshold (Qthreshold). This threshold is a risk boundary value determined based on a large amount of historical successful process data statistics. If Q < Qthreshold, it indicates that the current process state has a high risk of producing unqualified film layers, and / or the substrate quality mapping model used to calculate Q may have become inaccurate under the current working conditions.

[0032] System health comprehensive diagnosis and control mode decision-making, based on the above two independent diagnosis results, perform system-level state fusion judgment and decision-making: Diagnostic condition A (system degradation): If SNR < SNR thresholdOr Q < Q_threshold, perform Action A: The system determines that the reliability of the temperature measurement system has decreased. At this time, the system automatically generates and reports a system degradation warning, indicating the source of the abnormality ("signal quality deterioration" or "quality index exceeding the limit"). At the same time, the control core immediately triggers the fault-tolerant control mode and enters the safe operation state. Diagnostic condition B (system health): If SNR ≥ SNR threshold and Q ≥ Qthreshold, perform Action B: The system determines that it is in a healthy state, maintains the normal control mode, and continues to perform the optimization control based on the quality trend as defined in claim 7. This step realizes the online monitoring of the hardware state and signal quality of the infrared temperature measurement system itself by introducing the calculation of the signal-to-noise ratio of the spectrum analysis based on the rotation fundamental frequency; through the joint diagnosis with the comprehensive index Q reflecting the final film quality, a dual health firewall covering "signal acquisition reliability" and "process output quality risk" is constructed. This design can accurately distinguish whether it is the normal fluctuation of the process itself or the abnormality caused by the degradation of the measurement system performance (such as window contamination, sensor offset) or model misalignment. Therefore, when the system reliability decreases, a conservative and stable fault-tolerant control strategy is triggered in a timely manner, effectively preventing the batch product scrapping caused by executing control actions under the guidance of incorrect data, and significantly improving the robustness and production safety of the entire coating system.

[0033] Furthermore, based on the time-series temperature sequence of the current cycle and the obtained comprehensive quality index, system health diagnosis is carried out, including: Obtain the set rotation speed R of the rotating substrate holder during the coating process, and calculate the substrate rotation fundamental frequency f0, f0 = R / 60; Perform a fast Fourier transform on the time-series temperature sequence of the current cycle to obtain its spectrum, and extract two key amplitudes from the spectrum: a) Signal amplitude Asignal: It is the amplitude of the spectrum at the substrate rotation fundamental frequency f0; b) Noise amplitude Anoise: It is the average amplitude of the spectrum within the preset noise frequency band, and the preset noise frequency band is defined as the union of [0.1f0, 0.9f0] and [[1.1f0, 2.5f0]; Based on the signal amplitude and the noise amplitude, calculate the signal-to-noise ratio: SNR = 10 × log 10 (Asignal / Anoise); If the signal-to-noise ratio SNR < the preset threshold, it indicates that the health of the infrared temperature measurement system is abnormal; For the obtained comprehensive quality index Q, if the comprehensive quality index Q < the preset threshold, it indicates that there is a high risk of producing unqualified film layers.

[0034] Specifically, the implementation of the system health diagnosis method aims to further improve the reliability and robustness of the temperature control system. This invention introduces a system health diagnosis mechanism during the coating process, combining the signal-to-noise ratio of the infrared thermometer signal with comprehensive quality indicators to achieve a dual assessment of the temperature measurement system's status and process risks. The specific implementation process is as follows: First, the set rotational speed R (unit: revolutions per minute) of the substrate holder during the coating process is obtained. Based on this, the substrate rotation fundamental frequency is calculated. This fundamental frequency represents the theoretical frequency at which any point on the substrate periodically passes through the infrared thermometer's spot area during rotation, serving as the physical benchmark for judging whether the temperature measurement signal is normal. Subsequently, the time-series temperature sequence acquired within the current control cycle is processed. This time-series temperature sequence is obtained by sampling the infrared thermometer within one or more complete rotation cycles of the substrate, reflecting the temperature change process of the measured area on the substrate surface over time. A Fast Fourier Transform (FFT) is performed on this sequence to transform it from the time domain to the frequency domain, obtaining the corresponding temperature signal spectrum. This spectrum is a distribution graph with frequency as the horizontal axis and the amplitude of each frequency component as the vertical axis, characterizing the periodic components and their intensity contained in the temperature change. Under normal operating conditions, due to the uniform rotation of the substrate, the temperature signal periodically enters the temperature measurement field of view from the same spatial position. The temperature sequence should exhibit periodic fluctuations strongly correlated with the fundamental frequency of rotation, thus a significant peak should appear at f0 in the spectrum. If mechanical vibration, airflow disturbance, or temperature measurement interference exists, abnormal energy distribution will appear at other frequencies. The signal-to-noise ratio (SNR) is calculated in decibels (dB), quantifying the prominence of the effective rotating signal relative to background noise. The basis for setting the preset SNR threshold is not empirical guesswork, but rather derived from extensive historical data and fault case calibration. Statistics show that when the infrared window is clean, the temperature measurement optical path is aligned, and the rotation is stable, the SNR is generally higher than 20 dB; however, when window contamination, optical path misalignment, substrate shaking, or temperature measurement malfunction occurs, the SNR drops significantly, typically below 15 dB. To balance sensitivity and false alarm rate, the preset SNR threshold is set to 18 dB. This value is optimized through receiver operating characteristic (ROC) curve analysis and verified in long-term operation across multiple devices and various membrane systems, demonstrating an accuracy rate exceeding 95% in identifying temperature measurement system anomalies. This threshold can be stored as a default configuration in the system and can be fine-tuned based on equipment model or process level. The basis for setting the preset threshold for the comprehensive quality index Q is as follows: Q is a dimensionless quantity between 0 and 1, with higher values ​​indicating better process quality. The determination of its preset threshold is directly related to the quality acceptance standards of the target membrane system. Specifically: for membrane systems with high tolerance, such as ordinary antireflective films and protective films, a certain degree of temperature fluctuation is allowed. Extensive historical data and expert analysis have shown that when Q > 0.6, the final film's optical performance and adhesion both meet factory standards. Therefore, the preset threshold for Q in this type of process is set to 0.6.For high-precision film systems such as high-reflectivity mirrors and narrow-band filters, temperature uniformity is extremely important. Data analysis shows that only when Q > 0.85 can the spectral center wavelength drift be controlled within ±0.5 nm. Therefore, the preset Q threshold for this type of process is set to 0.85. These thresholds were determined through statistical regression analysis of hundreds of batches of historical coating data and established a strong correlation with the final product testing results (such as spectrophotometer measurements and laser damage tests), providing sufficient data support and engineering applicability. In summary, when the real-time calculated signal-to-noise ratio is lower than the preset threshold, the infrared temperature measurement system is deemed abnormally healthy, possibly indicating sensor failure or optical path interference. When the comprehensive quality index is lower than the preset threshold, the process is deemed to have a high risk, potentially leading to unqualified film performance. The combined criteria of these two indicators can effectively distinguish between "system failure" and "process instability," providing a reliable decision-making basis for subsequently triggering fault-tolerant control modes.

[0035] S5: Based on the extracted current equivalent temperature and current temperature gradient, a basic control command is generated. At the same time, the equivalent temperature and temperature gradient at future time moments are used as feedforward signals to dynamically compensate the basic control command, resulting in an optimized comprehensive control command. This comprehensive control command is executed to achieve coordinated optimization control of the substrate temperature level and spatial distribution. The basic control command includes: generating a power control command for the substrate temperature actuator based on the deviation between the current equivalent temperature and the temperature control target curve; and generating a parameter adjustment command for the deposition source system based on the current temperature gradient.

[0036] Specifically, under normal control mode, the system executes a multi-level optimization strategy based on quality indices. First, the system accurately captures key features characterizing the uniformity of temperature spatial distribution, as well as equivalent temperature gradient information reflecting the thermal coupling strength between different regions, by calculating strategic local state vectors (such as the temperature standard deviation of the central region and the temperature range of the edge region) and global disturbance estimates (such as equivalent edge temperature disturbances and equivalent central fluctuation intensity). Subsequently, these features are integrated into the calculation of the comprehensive quality index Q. The core of this index—the imbalance degree D—is designed to penalize both situations where "temperature level and uniformity issues coexist" and "both are severely imbalanced," thus internalizing the goal of collaborative optimization into a single objective of maximizing Q. Finally, the system generates control commands based on the trend of Q (ΔQ): it primarily regulates the overall temperature level by fine-tuning the heater power and influences the heat load distribution and primarily regulates the spatial temperature gradient by adjusting the deposition source parameters. This hierarchical response strategy achieves fine, gentle, and orderly collaborative optimization of the process, effectively avoiding over-adjustment and oscillations. Meanwhile, the fault-tolerant control mode serves as a reliable guarantee for the collaborative optimization process, embodying a "fail-safe" design principle. When the system diagnoses a decrease in temperature measurement reliability, it immediately locks the heater power and deposition source parameters to their safe operating values ​​before triggering. This "freezes" the control loop, shielding the actuators from interference from abnormal signals. Thus, even when local system functions degrade, it maximizes the maintenance of the established stable balance between temperature levels and uniformity, ensuring process continuity and product consistency. In summary, this control method, through the synergy and seamless switching between "optimization" and "stability maintenance" modes, not only significantly improves the resilience and reliability of the entire production system but also achieves advanced collaborative optimization control of substrate temperature levels and spatial distribution uniformity through a sophisticated closed-loop system of perception, decision-making, and execution.

[0037] Furthermore, normal control modes include: Calculate the difference between the current period's comprehensive quality index Q(t) and the previous period's comprehensive quality index Q(t-1): ΔQ = Q(t) - Q(t-1); Based on the value of ΔQ, execute the corresponding preset control strategy: If ΔQ≥0, the quality status is determined to be stable or tending to be optimized, and the heater power and deposition source parameters are maintained at the current operating values. If ΔQ < 0, the quality condition is determined to be deteriorating, and the following adjustment steps are executed in sequence: First response: Increase the heater power by a preset fixed step size δP based on the current operating value, while keeping the deposition source parameters unchanged; If ΔQ<0 in the next cycle, then the second response is to increase the deposition source parameters by a preset fixed step size δD based on the operating values ​​of the cycle before the first degradation, and at the same time restore the heater power to the operating values ​​of the cycle before the first degradation. If ΔQ < 0 in the next cycle, it is judged as a comprehensive anomaly and a high-level process alarm is generated. The preset fixed step size is the minimum unit change that is set in advance during the system debugging phase and is used to fine-tune the actuator.

[0038] Furthermore, fault-tolerant control modes include: When the fault-tolerant control mode is triggered, execute immediately: Store the heater power operating value of the previous sampling period before the trigger time as the safe power setting value; Store the running values ​​of the sedimentation source parameters from the sampling period preceding the trigger time as safety parameter settings; The heater power guarantee command is always equal to the aforementioned safe power setting value; The sedimentation source parameter safeguard command is always equal to the safety parameter setting value; During the fault-tolerant control mode, the system shields the influence of the collected time-series temperature sequence on the control commands and continuously applies two safeguard command generation rules until the mode is canceled by manual intervention or a command from a higher-level system.

[0039] Specifically, in vacuum optical coating systems, substrate temperature is not a directly controllable independent variable, but rather a "controlled process variable" determined by multiple factors, including heater power, deposition energy, chamber pressure, substrate rotation speed, and material thermal conductivity. Unlike common constant-temperature chambers, the thermal equilibrium during coating is extremely complex: deposition particles (such as metal atoms evaporated by an electron beam) carry kinetic energy and impact the substrate, generating a significant self-heating effect; different film materials have different thermal absorptivity and emissivity, leading to dynamic temperature changes under the same heating power; substrate rotation causes its surface to be periodically exposed to heat sources and cold zones, forming a dynamic thermal field. Therefore, there is no fixed "temperature setpoint" that can guarantee the quality stability of the entire coating process. Attempting to "pull" the temperature to a certain setpoint using traditional PID control is prone to causing oscillations or response lags in multilayer films, variable materials, and variable speed processes. The role of infrared thermometry is "fusion." The core of "fusion infrared thermometry" in this invention is not to achieve "constant temperature control," but to utilize infrared thermometry as a "process state sensor" to achieve the following functions: real-time sensing of the actual dynamic temperature of the substrate surface, rather than relying on "open-loop estimation" of the heater's set power; extracting local and global information about the substrate's thermal state through the statistical characteristics (such as standard deviation and range) and spectral characteristics of the temperature sequence; and constructing a comprehensive quality index Q by combining process parameters, elevating the physical quantity of "temperature" to a quantitative assessment of film quality risk. Therefore, infrared thermometry is the sensing and evaluation layer of the entire control closed loop, rather than a simple feedback signal. Normal Control Mode: "Feedforward-Feedback Coordinated Control" Based on Quality Trends. In normal control mode, the system does not directly control the temperature, but dynamically adjusts the heater power and deposition source parameters according to the changing trend (ΔQ) of the comprehensive quality index Q. The physical significance of this design is as follows: When ΔQ < 0, it indicates that the quality is deteriorating, possibly caused by excessively low temperature (poor film adhesion) or excessively high temperature (high stress, coarse grains); the first response is to increase the heater power to raise the substrate temperature and improve film crystallization and adhesion; if the deterioration continues, it indicates that simply raising the temperature is ineffective, possibly due to insufficient or excessive deposition flux, so the second response adjusts the deposition source parameters to restore thermal balance; "restoring the heater power to the level before the first deterioration" is to isolate variables and avoid control chaos caused by simultaneous changes in multiple parameters. This strategy is essentially a quality optimization control based on infrared temperature feedback, a direct manifestation of "integrated infrared temperature measurement". Fault-tolerant control mode: Protects process continuity under temperature measurement failure. When the signal-to-noise ratio is too low or the Q value is too low, the system determines that the temperature measurement signal is unreliable or the process has deviated significantly. At this time, it enters fault-tolerant mode, freezes the heater and deposition source parameters to the most recent "healthy" operating values, and shields abnormal temperature measurement data.The brilliance of this design lies in: avoiding unnecessary shutdowns to prevent costly coating interruptions and material waste; maintaining the best known process conditions, essentially "flying with a defect," giving operators time to address issues; and using "safety instructions" based on historical infrared temperature data to determine "safe points," highlighting the crucial role of temperature information in defining decision boundaries. The control is not about "temperature," but rather "the quality of the film as characterized by temperature." The control objective of this invention, from beginning to end, is to ensure the quality stability of the optical thin film. Infrared temperature measurement is the core sensing method for achieving this objective, while heater power and deposition source parameters are operable execution variables. The system achieves a leap from "temperature monitoring" to "quality control" through a closed loop of "temperature measurement → feature extraction → quality assessment → trend judgment → parameter adjustment."

[0040] Based on the aforementioned embodiment of the optical thin film coating temperature control method incorporating infrared thermometry, and using the same inventive concept, this application also provides an optical thin film coating temperature control system incorporating infrared thermometry. Please refer to the appendix. Figure 2 The system includes: Strategy configuration and model calling module 11 is used to obtain the process requirements of the target film system and the geometric features of the target substrate, select the corresponding temperature monitoring strategy, and call the substrate quality mapping model, measurement angle set and angle-region mapping table related to the strategy from the process database. Multi-angle synchronous temperature measurement and data acquisition module 12 is used to traverse the measurement angle set during the coating process, trigger the infrared thermometer to collect the temperature at each angle and record the process parameters synchronously, and mark the collected temperature value as the characteristic temperature of the corresponding ring monitoring area in real time according to the angle-region mapping table, so as to obtain the time-series temperature sequence and process parameter sequence set of the current cycle. The comprehensive quality index calculation module 13 is used to calculate the corresponding strategic local state vector based on the current period's time-series temperature sequence, perform global disturbance estimation to obtain the disturbance estimate, and obtain the comprehensive quality index based on the strategic local state vector and the disturbance estimate using the pre-stored substrate quality mapping model. System health diagnosis module 14 is used to perform system health diagnosis based on the current period's time-series temperature sequence and the obtained comprehensive quality index: analyzing the signal-to-noise ratio of the time-series temperature sequence and determining whether the comprehensive quality index is lower than a preset threshold. If any of the above indicators is less than the preset threshold, the reliability of the temperature measurement system is determined to have decreased, a system degradation warning is generated, and the fault-tolerant control mode is triggered. If both of the above indicators are greater than or equal to the preset threshold, the system will remain in normal control mode. Control command generation and execution module 15 is used to generate and execute corresponding control commands based on the control mode determined by system health diagnosis module 14. If it is in normal control mode, the trend of change is calculated based on the comprehensive quality index of the current cycle and the previous cycle, and the corresponding control command is generated and executed according to the trend. In fault-tolerant control mode, the heater power and deposition source parameters are locked at the current operating values, and basic safeguard control commands to maintain the continuous operation of the process are generated and executed.

[0041] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for controlling the temperature of optical thin film coating by integrating infrared thermometry, applied to a vacuum coating equipment including a rotating substrate holder, wherein the substrate is placed on the rotating substrate holder, and an infrared thermometer is installed outside the vacuum coating equipment, with its optical path aligned with the substrate through an infrared viewing window, characterized in that... The method includes the following steps: Obtain the process requirements of the target film system and the geometric characteristics of the target substrate, select the corresponding temperature monitoring strategy, and call the substrate quality mapping model, measurement angle set and angle-region mapping table related to the strategy from the process database; During the coating process, the measurement angle set is traversed, and an infrared thermometer is triggered at each angle to collect the temperature and record the process parameters simultaneously. Based on the angle-region mapping table, the collected temperature value is identified in real time as the characteristic temperature of the corresponding annular monitoring area, thus obtaining the time-series temperature sequence and process parameter sequence set of the current cycle. Based on the determined temperature monitoring strategy, the corresponding strategic local state vector is calculated according to the time-series temperature sequence of the current period, and the global perturbation estimation is performed to obtain the perturbation estimate. Based on the strategic local state vector and the perturbation estimate, the comprehensive quality index is obtained by using the pre-stored substrate quality mapping model. Based on the current period's time-series temperature sequence and the obtained comprehensive quality index, a system health diagnosis is performed: the signal-to-noise ratio of the time-series temperature sequence is analyzed, and it is determined whether the comprehensive quality index is below a preset threshold. If any of the above indicators is less than the preset threshold, the reliability of the temperature measurement system is determined to have decreased, a system degradation warning is generated, and the fault-tolerant control mode is triggered. If both of the above indicators are greater than or equal to the preset threshold, the system will remain in normal control mode. The decision-based control model generates and executes corresponding control commands. If it is in normal control mode, the trend of change is calculated based on the comprehensive quality index of the current cycle and the previous cycle, and the corresponding control command is generated and executed according to the trend. In fault-tolerant control mode, the heater power and deposition source parameters are locked at the current operating values, and basic safeguard control commands to maintain the continuous operation of the process are generated and executed.

2. The optical thin film coating temperature control method integrating infrared thermometry as described in claim 1, characterized in that, Temperature monitoring strategies include: The temperature monitoring strategy divides the substrate surface into two concentric annular monitoring areas according to the process requirements of the film system, and assigns a unique spatial semantic label to each annular area. Decision logic of temperature monitoring strategy: The IF process requires a focus on monitoring the stability of the film deposition in the central region of the substrate. The THEN strategy is defined as central stability monitoring, and its execution steps are as follows: install an infrared thermometer at a radial position relative to the center of rotation of the substrate so that its spot trajectory falls into the central annular region, and assign a semantic label of central temperature to this region. The IF process requires a focus on monitoring the uniformity of film deposition in the edge region of the substrate. The THEN strategy is defined as edge uniformity monitoring, and its execution steps are: to install an infrared thermometer at a radial position relative to the rotation center of the substrate so that its spot trajectory falls into the edge annular region, and to assign an edge temperature semantic label to this region. The system is divided into two concentric ring-shaped monitoring areas, including: The central annular region is a circular area on the substrate surface with the geometric center as the center and a radius ranging from 0% to 25% of the radius of its circumscribed circle, including the endpoints; The edge annular region is an annular region on the substrate surface with the geometric center as the center and a radius ranging from 75% to 100% of the radius of its circumscribed circle, including the endpoints.

3. The optical thin film coating temperature control method integrating infrared thermometry as described in claim 2, characterized in that, Angle-region mapping table, including: To obtain the requirements for the target membrane system, determine the current temperature monitoring strategy, and calibrate the angle-region mapping table under this strategy offline. The calibration process is as follows: The current temperature monitoring strategy defines a ring-shaped monitoring area on the surface of a standard substrate. Based on the geometry of the standard substrate to be calibrated, a radial mounting position is associated, and the infrared thermometer is mounted at this associated radial mounting position. In the vacuum coating equipment, install and calibrate the rotating substrate holder and its rotary encoder to ensure that the substrate rotation center is aligned with the geometric center; A standard substrate with known geometric features is mounted on a calibrated rotating substrate holder. The zero-degree reference position of the rotary encoder is defined, and the initial projection position of the infrared thermometer spot on the substrate surface at this reference position is determined as the reference coordinate point. Based on the geometric relationship between the reference coordinate point and the center of rotation of the substrate, the theoretical trajectory of the light spot formed on the surface of the substrate when the standard substrate rotates within the measurement angle set is calculated. According to the current temperature monitoring strategy, the annular area corresponding to the theoretical trajectory is marked with the spatial language label corresponding to the current strategy, and the correspondence between the measurement angle and the spatial semantic label is established. Summarize all the correspondences under the current strategy, generate an angle-region mapping table, associate the mapping table with the geometric feature information of the standard substrate, and store it in the process database.

4. The optical thin film coating temperature control method integrating infrared thermometry as described in claim 3, characterized in that, Based on the current period's time-series temperature sequence, the corresponding strategy-based local state vector is calculated, and a global perturbation estimate is obtained, including: When the strategy is to monitor the center stability, the local state vector is the temperature standard deviation of the center region. The temperature standard deviation of the center region and the feature values ​​of the extracted process parameter sequence set are input into the pre-stored thermal disturbance transfer model to obtain the equivalent edge temperature disturbance value. This value is used to quantify the degree of thermal impact of the center fluctuation on the edge region. When the strategy is edge uniformity monitoring, the local state vector is the temperature range of the edge region. The feature values ​​of the edge region temperature range and the extracted process parameter sequence set are input into the pre-stored thermal disturbance tracing model to obtain the equivalent central region temperature fluctuation intensity. This value is used to quantify the central heat source fluctuation intensity most likely to cause edge non-uniformity. The calculation of the temperature standard deviation in the central region and the temperature range in the peripheral regions includes: Based on a defined temperature strategy, an angle-region mapping table related to the strategy is obtained, and each value in the time-series temperature sequence of the current period is mapped to a relevant spatial language label based on the table. The corresponding semantic label is the center temperature. The temperature standard deviation of the current period's time-series temperature sequence is calculated to obtain the temperature standard deviation of the center region. The corresponding semantic label is edge temperature. The temperature range of the current period's time-series temperature sequence is calculated to obtain the temperature range of the edge region.

5. The optical thin film coating temperature control method integrating infrared thermometry as described in claim 4, characterized in that, Using a pre-stored substrate quality mapping model, a comprehensive quality index is obtained, including: Based on the determined temperature monitoring strategy, a virtual two-zone state pair (Vmsr, Vinf) consisting of the corresponding strategy-based local state vector and the model-inferred perturbation estimate is constructed: If a central stability monitoring strategy is adopted, the temperature standard deviation of the central region is used as Vmsr, and the equivalent edge temperature disturbance value is used as Vinf. If the edge uniformity monitoring strategy is adopted, the temperature range of the edge region is taken as Vmsr, and the temperature fluctuation intensity of the central region is taken as Vinf; a) Calculate the imbalance D of the virtual dual-zone state pair using the following formula: D=(Vmsr×Vinf) / (Vmsr²+Vinf²+δ)+λ·(Vmsr / (Vinf+ε)+Vinf / (Vmsr+ε)); In the formula, the first term is the normalized coupling strength term, the second term is the relative deviation penalty term, δ is the normalized stabilizer to prevent Vmsr and Vinf from being 0 at the same time, with the dimension [T²], ε is the ratio stabilizer to prevent the ratio calculation from being divided by zero when Vmsr or Vinf is 0, with the dimension [T], and λ is the deviation penalty coefficient, T = ℃ or K; Establish the mapping relationship from the degree of imbalance D to the comprehensive quality index Q, and its calculation formula is as follows: Q = exp(-D / Dref); Where Dref is the reference imbalance degree, and the comprehensive quality index Q is a dimensionless quantity between 0 and 1, and its value is positively correlated with the film quality. The deviation penalty coefficient and reference imbalance are defined as configurable parameters associated with the target membrane system type and stored in the process database.

6. The optical thin film coating temperature control method integrating infrared thermometry as described in claim 5, characterized in that, Based on the current period's time-series temperature sequence and the obtained comprehensive quality index, a system health diagnosis is performed, including: Obtain the set rotation speed R of the substrate holder during the coating process, and calculate the fundamental frequency f0 of the substrate rotation based on it, f0 = R / 60; Perform a Fast Fourier Transform on the current period's temperature sequence to obtain its spectrum, and extract two key amplitudes from the spectrum: a) Signal amplitude Asignal: The amplitude of the spectrum at the substrate rotation fundamental frequency f0; b) Noise Amplitude: The average amplitude of the spectrum within a preset noise band, which is defined as the union of [0.1f0, 0.9f0] and [[1.1f0, 2.5f0]; Based on the signal amplitude and noise amplitude, the signal-to-noise ratio (SNR) is calculated as follows: SNR = 10 × log 10 (Asignal / Anoise); If the signal-to-noise ratio (SNR) is less than the preset threshold, it indicates that the infrared temperature measurement system is in abnormal condition. If the comprehensive quality index Q is less than the preset threshold, it indicates a high risk of producing substandard membrane layers.

7. The optical thin film coating temperature control method integrating infrared thermometry as described in claim 1, characterized in that, Normal control modes include: Calculate the difference between the current period's comprehensive quality index Q(t) and the previous period's comprehensive quality index Q(t-1): ΔQ = Q(t) - Q(t-1); Based on the value of ΔQ, execute the corresponding preset control strategy: If ΔQ≥0, the quality status is determined to be stable or tending to be optimized, and the heater power and deposition source parameters are maintained at the current operating values. If ΔQ < 0, the quality condition is determined to be deteriorating, and the following adjustment steps are executed in sequence: First response: Increase the heater power by a preset fixed step size δP based on the current operating value, while keeping the deposition source parameters unchanged; If ΔQ<0 in the next cycle, then the second response is to increase the deposition source parameters by a preset fixed step size δD based on the operating values ​​of the cycle before the first degradation, and at the same time restore the heater power to the operating values ​​of the cycle before the first degradation. If ΔQ < 0 in the next cycle, it is judged as a comprehensive anomaly and a high-level process alarm is generated. The preset fixed step size is the minimum unit change that is set in advance during the system debugging phase and is used to fine-tune the actuator.

8. The optical thin film coating temperature control method integrating infrared thermometry as described in claim 1, characterized in that, Fault-tolerant control modes include: When the fault-tolerant control mode is triggered, execute immediately: Store the heater power operating value of the previous sampling period before the trigger time as the safe power setting value; Store the running values ​​of the sedimentation source parameters from the sampling period preceding the trigger time as safety parameter settings; The heater power guarantee command is always equal to the aforementioned safe power setting value; The sedimentation source parameter safeguard command is always equal to the safety parameter setting value; During the fault-tolerant control mode, the system shields the influence of the collected time-series temperature sequence on the control commands and continuously applies two safeguard command generation rules until the mode is canceled by manual intervention or a command from a higher-level system.

9. An optical thin-film coating temperature control system integrating infrared thermometry, the system comprising: The strategy configuration and model calling module is used to obtain the process requirements of the target film system and the geometric characteristics of the target substrate, select the corresponding temperature monitoring strategy, and call the substrate quality mapping model, measurement angle set and angle-region mapping table related to the strategy from the process database. The multi-angle synchronous temperature measurement and data acquisition module is used to traverse the set of measurement angles during the coating process, trigger the infrared thermometer to collect the temperature at each angle and record the process parameters synchronously, and identify the collected temperature value as the characterization temperature of the corresponding annular monitoring area in real time according to the angle-region mapping table, so as to obtain the time-series temperature sequence and process parameter sequence set of the current cycle. The comprehensive quality index calculation module is used to calculate the corresponding strategy-based local state vector based on the current period's time-series temperature sequence, perform global perturbation estimation to obtain perturbation estimate, and obtain the comprehensive quality index based on the strategy-based local state vector and perturbation estimate using a pre-stored substrate quality mapping model. The system health diagnosis module is used to diagnose system health based on the current period's time-series temperature sequence and the obtained comprehensive quality index: it analyzes the signal-to-noise ratio of the time-series temperature sequence and determines whether the comprehensive quality index is below a preset threshold. If any of the above indicators is less than the preset threshold, the reliability of the temperature measurement system is determined to have decreased, a system degradation warning is generated, and the fault-tolerant control mode is triggered. If both of the above indicators are greater than or equal to the preset threshold, the system will remain in normal control mode. The control command generation and execution module generates and executes corresponding control commands based on the control mode determined by the system health diagnosis module 14. If it is in normal control mode, the trend of change is calculated based on the comprehensive quality index of the current cycle and the previous cycle, and the corresponding control command is generated and executed according to the trend. In fault-tolerant control mode, the heater power and deposition source parameters are locked at the current operating values, and basic safeguard control commands to maintain the continuous operation of the process are generated and executed.

Images