A film production line surface temperature homogenization temperature control system

Abstract

The present application relates to the field of high polymer material processing and industrial automation control technology, in particular to a film production line surface temperature homogenization temperature control system, comprising: a data acquisition module for obtaining real-time state data in the film production process; a parameter inversion module for obtaining the thermal physical property parameters representing the film material characteristics and the comprehensive convective heat transfer coefficient parameters representing the environmental interference; an entropy monitoring module for calculating the thermal coupling entropy index; a game control module for constructing a state space model, solving and generating multi-temperature zone heating power control instructions; a closed-loop correction module for calculating a correction increment to update the actuator health factor, and feeding back the updated actuator health factor to the entropy monitoring module; the present application solves the mismatch problem caused by the traditional fixed parameter model failing to adapt to the raw material rheology and environmental interference, significantly improving the temperature prediction and control accuracy under high-speed transmission and complex working conditions.

Description

Technical Field

[0001] This invention relates to the field of polymer material processing and industrial automation control technology, specifically a temperature control system for surface temperature uniformity in a thin film production line. Background Technology

[0002] Polymer film stretching production lines are industrial production systems that use mechanical stretching devices to extend heated polymer melts or sheets at specific temperatures to improve the physical properties of films. During high-speed transport, to ensure film forming quality, multi-temperature zone heating devices are typically used to precisely control the film surface temperature, aiming to maintain uniform temperature distribution throughout the line. However, actual production environments are complex and variable, with raw material rheological properties often undergoing abrupt changes, and environmental interference factors such as wind fields are difficult to avoid. Existing technologies typically use fixed-parameter control models, which cannot accurately reflect the thermal properties of the film material over time and the environmental heat transfer coefficient, leading to model prediction mismatches. Furthermore, strong thermal coupling effects exist between adjacent heating zones, often resulting in conflicting control objectives. Additionally, heating actuators experience hardware aging and efficiency degradation after long-term operation. If parameter changes cannot be detected in real time and the interplay between temperature zones cannot be coordinated, uneven film heating, thermal stress concentration, and even film breakage can easily occur, posing significant production quality and safety risks. Therefore, a solution is urgently needed to address the problems existing in current technologies. Summary of the Invention

[0003] To solve the above-mentioned technical problems, the present invention provides a temperature control system for surface temperature uniformity in a thin film production line. Specifically, the technical solution of the present invention includes:

[0004] The data acquisition module is used to acquire real-time status data during the thin film production process. The real-time status data includes the spatiotemporal temperature distribution data of the thin film, real-time linear velocity data, heater input power data, and ambient temperature data.

[0005] The parameter inversion module is used to invert the thermal property parameters characterizing the properties of thin film materials and the comprehensive convective heat transfer coefficient parameters characterizing environmental disturbances based on spatiotemporal temperature distribution data and real-time linear velocity data using a preset physical information neural network model.

[0006] The entropy monitoring module is used to calculate the thermal coupling entropy index based on heater input power data, spatiotemporal temperature distribution data and preset actuator health factors. The thermal coupling entropy index is a value used to quantify the degree of conflict between control targets between adjacent heating temperature zones.

[0007] The game-theoretic control module is used to construct a state-space model based on thermal property parameters and comprehensive convective heat transfer coefficient parameters, establish the objective function of model predictive control, dynamically adjust the weight of the smoothing term in the objective function according to the thermal coupling entropy exponent, and solve to generate multi-temperature zone heating power control commands.

[0008] The closed-loop correction module is used to calculate the correction increment to update the actuator health factor based on the deviation between the predicted response of the state-space model and the spatiotemporal temperature distribution data, and then feeds the updated actuator health factor back to the entropy monitoring module.

[0009] Optionally, the parameter inversion module, based on spatiotemporal temperature distribution data and real-time linear velocity data, uses a preset physical information neural network model to invert the thermal property parameters and comprehensive convective heat transfer coefficient parameters of the thin film, specifically by performing the following steps:

[0010] A neural network with embedded fluid dynamics and thermodynamics equations is constructed, and the residuals of the physical equations are calculated using automatic differentiation techniques.

[0011] The physical equation residuals are constructed based on the transient one-dimensional energy conservation equation of the moving medium. The calculation of the residuals involves the sum of squares of the following terms: the energy storage change of the thin film over time, the convection term with displacement, the heat conduction term, the heating input term introduced by the heater input power, and the environmental heat dissipation term between the thin film surface and the environment.

[0012] By training a neural network to minimize the residuals of the physical equations, the outputs time-varying thermophysical parameters and comprehensive convective heat transfer coefficient parameters are obtained.

[0013] Specifically, the thermophysical parameters include the specific heat capacity of the thin film and the density of the thin film.

[0014] Optionally, the entropy monitoring module calculates the thermal coupling entropy exponent by performing the following steps:

[0015] Obtain the heater input power data of two adjacent temperature zones, divide them by the product of the rated power and the corresponding actuator health factor, calculate the absolute value of the difference between the two, and obtain the load rate difference.

[0016] Obtain the spatiotemporal temperature distribution data of two adjacent temperature zones, calculate the absolute value of the difference between their temperature values, and obtain the absolute value of the temperature difference;

[0017] Multiply the absolute value of the temperature difference by the preset thermocoupler coefficient, and perform an exponential function operation on the product to obtain the temperature difference influence factor;

[0018] Calculate the product of the load rate difference and the temperature difference influence factor, and then sum and average the products of all adjacent temperature zones to obtain the thermal coupling entropy index.

[0019] Optionally, the game-theoretic control module constructs a state-space model based on thermal property parameters and comprehensive convective heat transfer coefficient parameters, and specifically executes the following steps:

[0020] The inverted thin film specific heat capacity parameters, thin film density parameters, and comprehensive convective heat transfer coefficient parameters are used as time-varying physical coefficients and substituted into the preset one-dimensional partial differential equation of heat conduction.

[0021] The one-dimensional heat conduction partial differential equation is discretized in space and time to generate a linear state-space equation containing the state transition matrix and the input matrix.

[0022] The input matrix includes an actuator health factor, which is used to adjust the gain of the control input to characterize the actual output efficiency of the actuator.

[0023] Optionally, the game-theoretic control module dynamically adjusts the weights of the objective function for model prediction and control by combining the thermal coupling entropy exponent, specifically by performing the following steps:

[0024] Construct a quadratic programming cost function, which contains at least three terms: a tracking error term representing the temperature tracking accuracy, a control increment smoothing term representing the change in control action in the time dimension, and a spatial second derivative term representing the difference in output between adjacent actuators in the spatial dimension.

[0025] A variable weighting coefficient is assigned to the second derivative term in the spatial domain, and a positive correlation mapping relationship is established between this weighting coefficient and the thermal coupling entropy exponent.

[0026] In each control cycle, the weight coefficients are updated based on the real-time calculated thermal coupling entropy exponent, and the control sequence that minimizes the quadratic programming cost function is solved. The first element of the sequence is extracted as the multi-temperature zone heating power control command.

[0027] Optionally, a positive correlation mapping relationship can be established between the weighting coefficients and the thermal coupling entropy exponent, specifically including:

[0028] Preset a thermal coupling entropy threshold and a basic weight value;

[0029] The mapping rules are set using the logic of the S-shaped function: when the real-time calculated thermal coupling entropy exponent is less than the thermal coupling entropy threshold, the weight coefficient remains near the basic weight value; when the thermal coupling entropy exponent exceeds the thermal coupling entropy threshold, the weight coefficient increases non-linearly and rapidly with the increase of the entropy value, so as to forcibly penalize the output difference of adjacent temperature zones in the cost function.

[0030] Optionally, the closed-loop correction module corrects the actuator health factor by performing the following steps:

[0031] The predicted temperature rise rate at the current moment is calculated using a state-space model;

[0032] The measured temperature rise rate at the current moment is calculated using the collected spatiotemporal temperature distribution data;

[0033] Calculate the ratio of the measured temperature rise rate to the predicted temperature rise rate, and calculate the deviation between this ratio and the reference value;

[0034] The cumulative deviation over time is integrated, and the result of the integration is multiplied by a preset adaptive update gain to obtain the correction increment.

[0035] Add the correction increment to the executor health factor from the previous moment to obtain the updated executor health factor for the current moment.

[0036] Optionally, modifying the actuator health factor also includes the following safety verification steps:

[0037] Determine whether the updated actuator health factor is lower than the preset lower threshold.

[0038] If the value is below the lower threshold, the corresponding heater is determined to have a hardware aging failure, and a hardware replacement alarm signal is generated.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1. This invention uses a parameter inversion module to invert the thermal properties of thin films and the environmental heat transfer coefficient in real time using a physical information neural network; it solves the mismatch problem caused by the inability of traditional fixed parameter models to adapt to raw material rheology and environmental interference, ensures the real-time consistency between the control model parameters and the actual physical process of the production line, and significantly improves the temperature prediction and control accuracy under high-speed transmission and complex working conditions.

[0041] 2. This invention introduces an entropy monitoring and game-theoretic control mechanism, which uses the thermal coupling entropy index to dynamically quantify the degree of control conflict between adjacent temperature zones. The system adaptively adjusts the weight of the smoothing term in the model predictive control according to the index. In low-risk situations, it prioritizes ensuring temperature tracking accuracy, while in high-risk situations, it forces the smoothing of the output differences between adjacent actuators, effectively resolving the thermal stress concentration caused by strong coupling between multiple temperature zones and reducing the risk of membrane failure.

[0042] 3. This invention establishes a closed-loop correction mechanism based on actuator health factors. By comparing the deviation between the measured temperature rise rate and the predicted temperature rise rate, the actual output efficiency of the actuator is identified and updated online. This mechanism can automatically compensate for model prediction errors caused by heater aging, maintain the high-performance operation of the system throughout its entire life cycle, and trigger hardware fault alarms in a timely manner when the health factor is below the threshold.

[0043] 4. This invention constructs a state-space model that integrates time-varying physical parameters and equipment health status, realizing dual adaptation of control strategies to fluctuations in the production environment and equipment hardware status; through physical equation residual constraints and multi-objective optimization solutions, the system can achieve coordinated optimization of heating power in multiple temperature zones while ensuring the stability of numerical calculations, thus ensuring the uniformity of polymer film molding quality and production continuity. Attached Figure Description

[0044] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0045] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. Example 1:

[0047] Please see Figure 1 A temperature control system for surface temperature uniformity in a thin film production line, comprising:

[0048] The data acquisition module is used to acquire real-time status data during the thin film production process. The real-time status data includes the spatiotemporal temperature distribution data of the thin film, real-time linear velocity data, heater input power data, and ambient temperature data.

[0049] The parameter inversion module is used to invert the thermal property parameters characterizing the properties of thin film materials and the comprehensive convective heat transfer coefficient parameters characterizing environmental disturbances based on spatiotemporal temperature distribution data and real-time linear velocity data using a preset physical information neural network model.

[0050] The entropy monitoring module is used to calculate the thermal coupling entropy index based on heater input power data, spatiotemporal temperature distribution data and preset actuator health factors. The thermal coupling entropy index is a value used to quantify the degree of conflict between control targets between adjacent heating temperature zones.

[0051] The game-theoretic control module is used to construct a state-space model based on thermal property parameters and comprehensive convective heat transfer coefficient parameters, establish the objective function of model predictive control, dynamically adjust the weight of the smoothing term in the objective function according to the thermal coupling entropy exponent, and solve to generate multi-temperature zone heating power control commands.

[0052] The closed-loop correction module is used to calculate the correction increment to update the actuator health factor based on the deviation between the predicted response of the state-space model and the spatiotemporal temperature distribution data, and then feeds the updated actuator health factor back to the entropy monitoring module.

[0053] This embodiment provides a temperature control system for surface temperature uniformity in a thin film production line. The system is mainly used in polymer film stretching production lines and aims to solve the problem of uneven temperature control caused by sudden changes in the rheological properties of raw materials and environmental interference during high-speed transmission.

[0054] The system includes a data acquisition module, a parameter inversion module, an entropy monitoring module, a game-theoretic control module, and a closed-loop correction module. During system operation, the data acquisition module acquires real-time status data during the thin film production process. This real-time status data specifically includes the spatiotemporal temperature distribution data of the thin film during transmission, acquired through a high-frequency infrared thermal imager array. Real-time linear velocity data acquired through the encoder of the drive motor Heater input power data obtained through feedback from the power controller And ambient temperature data collected by environmental sensors The parameter inversion module, based on the aforementioned spatiotemporal temperature distribution data and real-time linear velocity data, uses a pre-constructed physical information neural network model to perform calculations, inverting the thermal property parameters characterizing the thin film material properties and the comprehensive convective heat transfer coefficient parameters characterizing environmental disturbances. At the same time, the entropy monitoring module, based on the heater input power data, spatiotemporal temperature distribution data, and preset actuator health factors, calculates the thermal coupling entropy index, which is used to quantify the degree of control target conflict between adjacent heating temperature zones.

[0055] The game-theoretic control module constructs a state-space model based on the inverted thermophysical parameters and the comprehensive convective heat transfer coefficient parameters, and establishes an objective function for model predictive control. It dynamically adjusts the weight of the smoothing term in the objective function according to the thermal coupling entropy exponent, thereby solving and generating multi-temperature zone heating power control commands. The closed-loop correction module calculates the correction increment based on the deviation between the predicted response of the state-space model and the actual collected spatiotemporal temperature distribution data to update the actuator health factor, and feeds the updated actuator health factor back to the entropy monitoring module to form closed-loop control. Example 2:

[0056] The parameter inversion module, based on spatiotemporal temperature distribution data and real-time linear velocity data, uses a preset physical information neural network model to invert the thermal property parameters and comprehensive convective heat transfer coefficient parameters of the thin film. Specifically, it performs the following steps:

[0057] A neural network with embedded fluid dynamics and thermodynamics equations is constructed, and the residuals of the physical equations are calculated using automatic differentiation techniques.

[0058] The physical equation residuals are constructed based on the transient one-dimensional energy conservation equation of the moving medium. The calculation of the residuals involves the sum of squares of the following terms: the energy storage change of the thin film over time, the convection term with displacement, the heat conduction term, the heating input term introduced by the heater input power, and the environmental heat dissipation term between the thin film surface and the environment.

[0059] By training a neural network to minimize the residuals of the physical equations, the outputs time-varying thermophysical parameters and comprehensive convective heat transfer coefficient parameters are obtained.

[0060] Specifically, the thermophysical parameters include the specific heat capacity of the thin film and the density of the thin film.

[0061] To ensure consistency between the inversion results and physical facts, the neural network employs a multi-task learning architecture, with a training loss function number of... It not only includes the residuals of the above physical equations It also explicitly includes data matching residuals. Specifically, neural networks use spatiotemporal coordinates Input: Predicted temperature Data matching residuals are defined as follows: ,in The data acquisition module obtains the measured spatiotemporal temperature distribution data. The final total loss function is... ,in To balance physical constraints with adaptive weights for data fitting;

[0062] This embodiment further illustrates the specific operating mechanism of the parameter inversion module; the core of this module lies in solving the mismatch problem caused by fixed parameters in traditional control models;

[0063] Specifically, the parameter inversion module constructs a neural network embedding fluid dynamics and thermodynamics equations; this network uses automatic differentiation techniques to calculate the residuals of the physical equations, which serve as the loss function for network training; the residuals of the physical equations... The transient one-dimensional energy conservation equation based on the moving medium is mathematically expressed as follows:

[0064] ;

[0065] in: The total number of training sample points within the sampling period is derived from the total number of data points in the data acquisition module within the sampling period; This is a parameter for film density, expressed in kilograms per cubic meter. Its initial value comes from the raw material property table and is fine-tuned and inverted by the network; This is the specific heat capacity parameter for thin films, expressed in joules per kilogram (Kelvin). These are the key thermophysical parameters to be inverted; Thin film temperature, in Kelvin. ; Time, in seconds ; Displacement along the direction of film movement, in meters. ; The real-time linear velocity of the thin film is expressed in meters per second. ; Thermal conductivity, measured in watts per meter Kelvin. In this embodiment, is set as a function of temperature; The heating input term introduced for the heater's input power, i.e., the corresponding data collected. The unit is watts. ; The volume of the heating element is expressed in cubic meters. It is determined by the geometry of the production line; The parameters for the comprehensive convective heat transfer coefficient are expressed in watts per square kelvin. , are the parameters that need to be inverted to characterize the environmental wind field disturbance; The heat dissipation surface area is expressed in square meters. ; The ambient temperature;

[0066] By training a neural network to minimize the residuals of the above physical equations, the parameter inversion module can output the thin film specific heat capacity parameter as it changes over time in real time. Thin film density parameters and comprehensive convective heat transfer coefficient parameters This allows for the accurate capture of material property fluctuations during the production process. Example 3:

[0067] The entropy monitoring module calculates the thermal coupling entropy index by performing the following steps:

[0068] Obtain the heater input power data of two adjacent temperature zones, divide them by the product of the rated power and the corresponding actuator health factor, calculate the absolute value of the difference between the two, and obtain the load rate difference.

[0069] Obtain the spatiotemporal temperature distribution data of two adjacent temperature zones, calculate the absolute value of the difference between their temperature values, and obtain the absolute value of the temperature difference;

[0070] Multiply the absolute value of the temperature difference by the preset thermocoupler coefficient, and perform an exponential function operation on the product to obtain the temperature difference influence factor;

[0071] Calculate the product of the load rate difference and the temperature difference influence factor, and then sum and average the products of all adjacent temperature zones to obtain the thermal coupling entropy index.

[0072] This embodiment details the calculation method of the thermal coupling entropy index in the entropy monitoring module; thermal coupling entropy index It is a key indicator used to assess the risk of internal control conflicts in a system, and its calculation formula is as follows:

[0073] ;

[0074] in, Indicates the total number of heating temperature zones; and They represent the first The and the first Real-time input power of heaters in each temperature zone; The rated maximum power of the heating actuator is taken as a system inherent constant. and The actuator health factor corresponds to the temperature range, and its value range is as follows: The feedback comes from the closed-loop correction module; the first part of the formula This represents the load rate difference, used to quantify the inconsistency in output load between adjacent actuators;

[0075] In the formula The thermal coupling coefficient is expressed in inverted Kelvin. In this embodiment, The values ​​are derived from logistic regression analysis of historical membrane breakage failure data from the production line. The specific calculation method is as follows: constructing a logistic regression model. ,in This indicates a membrane breakage failure. The maximum temperature difference between adjacent temperature zones in the historical fault sample; This is the intercept term of the regression model; after solving the regression coefficients through maximum likelihood estimation, the weighting coefficients corresponding to the temperature difference characteristics are extracted. And set the thermal coupling coefficient to This coefficient directly reflects the marginal effect of temperature difference on the logarithmic probability of membrane rupture, thus giving it physical meaning; it is used to adjust the sensitivity of temperature difference to entropy increase. This factor is the temperature difference influence factor, which is determined when the temperature difference between adjacent temperature zones... As the factor increases, it grows exponentially, thus amplifying the weight of the load factor difference on system risk; the calculated... It is sent to the game control module for subsequent control decisions. Example 4:

[0076] The game-theoretic control module constructs a state-space model based on thermal property parameters and comprehensive convective heat transfer coefficient parameters, and specifically executes the following steps:

[0077] The inverted thin film specific heat capacity parameters, thin film density parameters, and comprehensive convective heat transfer coefficient parameters are used as time-varying physical coefficients and substituted into the preset one-dimensional partial differential equation of heat conduction.

[0078] The one-dimensional heat conduction partial differential equation is discretized in space and time to generate a linear state-space equation containing the state transition matrix and the input matrix.

[0079] The input matrix includes an actuator health factor, which is used to adjust the gain of the control input to characterize the actual output efficiency of the actuator.

[0080] The game-theoretic control module dynamically adjusts the weights of the objective function for model prediction and control by combining the thermal coupling entropy exponent, specifically executing the following steps:

[0081] Construct a quadratic programming cost function, which contains at least three terms: a tracking error term representing the temperature tracking accuracy, a control increment smoothing term representing the change in control action in the time dimension, and a spatial second derivative term representing the difference in output between adjacent actuators in the spatial dimension.

[0082] A variable weighting coefficient is assigned to the second derivative term in the spatial domain, and a positive correlation mapping relationship is established between this weighting coefficient and the thermal coupling entropy exponent.

[0083] In each control cycle, the weight coefficients are updated based on the real-time calculated thermal coupling entropy exponent, and the control sequence that minimizes the quadratic programming cost function is solved. The first element of the sequence is extracted as the multi-temperature zone heating power control command.

[0084] Establish a positive correlation between the weighting coefficients and the thermal coupling entropy exponent, specifically including:

[0085] Preset a thermal coupling entropy threshold and a basic weight value;

[0086] The mapping rules are set using the logic of the S-shaped function: when the real-time calculated thermal coupling entropy exponent is less than the thermal coupling entropy threshold, the weight coefficient remains near the basic weight value; when the thermal coupling entropy exponent exceeds the thermal coupling entropy threshold, the weight coefficient increases non-linearly and rapidly with the increase of the entropy value, so as to forcibly penalize the output difference of adjacent temperature zones in the cost function.

[0087] This embodiment details the operating logic of the game control module; this module is based on the inverted thin film specific heat capacity parameters. Thin film density parameters and comprehensive convective heat transfer coefficient parameters Constructing a state-space model; specifically, this involves substituting these time-varying physical coefficients into a one-dimensional partial differential equation for heat conduction, and discretizing it using the forward time difference and central spatial difference scheme (FTCS); where the state transition matrix... It is a tridiagonal matrix, and its main diagonal elements are The second diagonal, i.e., the adjacent node elements are ;in The sampling period is This represents the spatial sampling interval of the external thermal imager. To ensure the numerical stability of the discretized state-space model, the system automatically verifies the CFL condition during initialization, which must be satisfied. If the current setting is and This can lead to computational divergence. The system will automatically perform coarse-grained merging of the spatial grid to increase its size. Or reduce the sampling period of the controller. This continues until the aforementioned stability inequality is satisfied. Furthermore, in this embodiment... The axis specifically refers to the machine direction of thin-film transport, adjacent nodes. and The heating temperature zones are arranged longitudinally along the production line; and the equation is discretized in space and time to generate the following linear state-space equation.

[0088] ;

[0089] in, The state transition matrix is ​​the input matrix. From the nominal input matrix The result is obtained by multiplying the diagonal matrix of the actuator health factor, i.e. ;in This represents the operation of constructing a diagonal matrix using vector elements, thereby characterizing the actual heating efficiency decay of actuators in each temperature zone due to aging at the model level, and accurately characterizing the actual output efficiency decrease of actuators due to aging and other reasons.

[0090] Based on this, the game control module constructs a quadratic programming cost function that includes risk game terms. :

[0091] ;

[0092] in, For prediction in the time domain; The tracking error term characterizes the temperature tracking accuracy, with a weight of . ; To control the incremental smoothing term, which represents the change in control action over time, the weight is... ; The second derivative term in the spatial domain represents the difference in output between adjacent actuators in the spatial dimension, with a weight of . The specific calculation formula is as follows: The smoothness of the heating power distribution in three adjacent temperature zones is quantified by using a second-order central difference operator to prevent spatial power abrupt changes from causing uneven thermal stress in the thin film.

[0093] To achieve adaptive game control, this embodiment establishes weighting coefficients. With thermal coupling entropy index The positive correlation mapping relationship is specifically achieved using the following S-shaped function logic:

[0094] ;

[0095] in, The preset thermal coupling entropy threshold is set based on the statistically determined critical entropy values ​​before system instability in historical production data. Basic weight value; and The closed-loop correction modules all use preset positive adjustment coefficients; when the thermal coupling entropy exponent is calculated in real time... Less than hour, Keeping the temperature tracking accuracy near the baseline weight value, the control system prioritizes ensuring temperature tracking accuracy; when Exceed hour, The entropy increases rapidly and non-linearly, forcibly penalizing output differences between adjacent temperature zones, forcing the system into survival mode to prioritize production continuity; in each control cycle, the module adjusts the parameters based on real-time data. renew Solve the above cost function and issue the first element of the optimal control sequence as an instruction. Example 5:

[0096] The closed-loop correction module corrects the actuator health factors by performing the following steps:

[0097] The predicted temperature rise rate at the current moment is calculated using a state-space model;

[0098] The measured temperature rise rate at the current moment is calculated using the collected spatiotemporal temperature distribution data;

[0099] Calculate the ratio of the measured temperature rise rate to the predicted temperature rise rate, and calculate the deviation between this ratio and the reference value;

[0100] The cumulative deviation over time is integrated, and the result of the integration is multiplied by a preset adaptive update gain to obtain the correction increment.

[0101] Add the correction increment to the actuator health factor from the previous moment to obtain the updated actuator health factor for the current moment;

[0102] The correction of actuator health factors also includes the following safety verification steps:

[0103] Determine whether the updated actuator health factor is lower than the preset lower threshold.

[0104] If the value is below the lower threshold, the corresponding heater is determined to have a hardware aging failure, and a hardware replacement alarm signal is generated.

[0105] This embodiment describes the closed-loop correction module's effect on actuator health factors. The online identification and update process aims to eliminate model prediction bias caused by equipment aging.

[0106] Specifically, the module uses a state-space model to calculate the predicted temperature rise rate at the current moment. The measured temperature rise rate at the current moment is calculated using the collected spatiotemporal temperature distribution data. The health factor is updated using the following integral correction formula:

[0107] ;

[0108] in, The preset adaptive update gain is in countdown seconds. , used to control the correction speed; To prevent small constants with a denominator of zero, for example If the measured rate is consistently lower than the predicted rate, the integral term accumulates to a negative value, leading to... The decrease reflects a reduction in actuator efficiency;

[0109] After updating and obtaining the executor health factor at the current moment, the system will perform a security verification step: determine the updated... Is it below the preset lower threshold? In this embodiment, The value is set according to the hardware fatigue limit of the heater, for example, 0.4; if The system determines that the corresponding heater has suffered an irreversible hardware aging failure and immediately generates a hardware replacement alarm signal to prompt maintenance personnel to handle the situation.

[0110] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A thin film production line surface temperature homogenizing temperature control system, characterized by, include: The data acquisition module is used to acquire real-time status data during the thin film production process. The real-time status data includes the spatiotemporal temperature distribution data of the thin film, real-time linear velocity data, heater input power data, and ambient temperature data. The parameter inversion module is used to invert the thermal property parameters characterizing the properties of thin film materials and the comprehensive convective heat transfer coefficient parameters characterizing environmental disturbances based on spatiotemporal temperature distribution data and real-time linear velocity data using a preset physical information neural network model. The entropy monitoring module is used to calculate the thermal coupling entropy index based on heater input power data, spatiotemporal temperature distribution data and preset actuator health factors. The thermal coupling entropy index is a value used to quantify the degree of conflict between control targets between adjacent heating temperature zones. The game-theoretic control module is used to construct a state-space model based on thermal property parameters and comprehensive convective heat transfer coefficient parameters, establish the objective function of model predictive control, dynamically adjust the weight of the smoothing term in the objective function according to the thermal coupling entropy exponent, and solve to generate multi-temperature zone heating power control commands. The closed-loop correction module is used to calculate the correction increment to update the actuator health factor based on the deviation between the predicted response based on the state-space model and the spatiotemporal temperature distribution data, and to feed the updated actuator health factor back to the entropy monitoring module. Thermally coupled entropy index The calculation formula is as follows: ； in, Indicates the total number of heating temperature zones; and They represent the first The and the first Real-time input power of heaters in each temperature zone; The rated maximum power of the heating actuator is taken as a system inherent constant. and The actuator health factor corresponds to the temperature range, and its value range is as follows: The feedback comes from the closed-loop correction module; the first part of the formula This represents the load rate difference, used to quantify the inconsistency in output load between adjacent actuators; The thermal coupling coefficient is expressed in inverted Kelvin. ; The temperature difference is an influencing factor, when the temperature difference between adjacent temperature zones... As the factor increases, it grows exponentially. The health factor is updated using the following integral correction formula: The closed-loop correction module corrects the actuator health factor by performing the following steps: Calculate the predicted temperature rise rate at the current moment using a state-space model. ; Calculate the measured temperature rise rate at the current moment using the collected spatiotemporal temperature distribution data. ; Calculate the ratio of the measured temperature rise rate to the predicted temperature rise rate, and calculate the deviation between this ratio and the reference value; The cumulative deviation over time is integrated, and the result of the integration is multiplied by a preset adaptive update gain to obtain the correction increment. Add the correction increment to the actuator health factor from the previous time step to obtain the updated actuator health factor for the current time step. ; ； in, The preset adaptive update gain is in countdown seconds. , used to control the correction speed; To prevent small constants with a denominator of zero, for example If the measured rate is consistently lower than the predicted rate, the integral term accumulates to a negative value, leading to... The decrease reflects a reduction in actuator efficiency.

2. The temperature control system for surface temperature uniformity in a thin film production line according to claim 1, characterized in that, The parameter inversion module, based on spatiotemporal temperature distribution data and real-time linear velocity data, uses a preset physical information neural network model to invert the thermal property parameters and comprehensive convective heat transfer coefficient parameters of the thin film. Specifically, it performs the following steps: A neural network with embedded fluid dynamics and thermodynamics equations is constructed, and the residuals of the physical equations are calculated using automatic differentiation techniques. The physical equation residuals are constructed based on the transient one-dimensional energy conservation equation of the moving medium. The calculation of the residuals involves the sum of squares of the following terms: the energy storage change of the thin film over time, the convection term with displacement, the heat conduction term, the heating input term introduced by the heater input power, and the environmental heat dissipation term between the thin film surface and the environment. By training a neural network to minimize the residuals of the physical equations, the outputs time-varying thermophysical parameters and comprehensive convective heat transfer coefficient parameters are obtained. Specifically, the thermophysical parameters include the specific heat capacity of the thin film and the density of the thin film.

3. The temperature control system for surface temperature uniformity in a thin film production line according to claim 1, characterized in that, The game-theoretic control module constructs a state-space model based on thermal property parameters and comprehensive convective heat transfer coefficient parameters, and specifically executes the following steps: The inverted thin film specific heat capacity parameters, thin film density parameters, and comprehensive convective heat transfer coefficient parameters are used as time-varying physical coefficients and substituted into the preset one-dimensional partial differential equation of heat conduction. The one-dimensional heat conduction partial differential equation is discretized in space and time to generate a linear state-space equation containing the state transition matrix and the input matrix. The input matrix includes an actuator health factor, which is used to adjust the gain of the control input to characterize the actual output efficiency of the actuator.

4. The surface temperature uniformity control system for a thin film production line according to claim 1, characterized in that, The game-theoretic control module dynamically adjusts the weights of the objective function for model prediction and control by combining the thermal coupling entropy exponent, specifically executing the following steps: Construct a quadratic programming cost function, which contains at least three terms: a tracking error term representing the temperature tracking accuracy, a control increment smoothing term representing the change in control action in the time dimension, and a spatial second derivative term representing the difference in output between adjacent actuators in the spatial dimension. A variable weighting coefficient is assigned to the second derivative term in the spatial domain, and a positive correlation mapping relationship is established between this weighting coefficient and the thermal coupling entropy exponent. In each control cycle, the weight coefficients are updated based on the real-time calculated thermal coupling entropy exponent, and the control sequence that minimizes the quadratic programming cost function is solved. The first element of the sequence is extracted as the multi-temperature zone heating power control command.

5. The surface temperature uniformity control system for a thin film production line according to claim 4, characterized in that, Establish a positive correlation between the weighting coefficients and the thermal coupling entropy exponent, specifically including: Preset a thermal coupling entropy threshold and a basic weight value; The mapping rules are set using the logic of the S-shaped function: when the real-time calculated thermal coupling entropy exponent is less than the thermal coupling entropy threshold, the weight coefficient remains near the basic weight value; when the thermal coupling entropy exponent exceeds the thermal coupling entropy threshold, the weight coefficient increases non-linearly and rapidly with the increase of the entropy value, so as to forcibly penalize the output difference of adjacent temperature zones in the cost function.

6. The surface temperature uniformity control system for a thin film production line according to claim 5, characterized in that, The modified actuator health factor also includes the following security verification steps: Determine whether the updated actuator health factor is lower than the preset lower threshold. If the value is below the lower threshold, the corresponding heater is determined to have a hardware aging failure, and a hardware replacement alarm signal is generated.

Images