A method and system for controlling the production of synchronous biaxially stretched films based on process optimization

Abstract

The application relates to the technical field of thin film production control, and discloses a synchronous bidirectional stretching thin film production control method and system based on process optimization. The method comprises the following steps: collecting BOPA thin film production line process parameters in real time and performing subspace identification, and establishing a state space prediction model; based on the model, performing multi-objective weighted combination on quality indexes and energy consumption to form a performance optimization function; setting a process constraint boundary, and constructing a quadratic programming constraint set; applying a gradient projection algorithm to perform rolling optimization, calculating an optimal control sequence, and executing a first control instruction. The application solves the quality fluctuation problem caused by the insufficient model self-adaptive capability in the BOPA thin film production process.

Description

Technical Field

[0001] This application relates to the field of thin film production control technology, and in particular to a method and system for controlling the production of synchronous biaxially stretched thin films based on process optimization. Background Technology

[0002] Bis-axially oriented nylon (BOPA) film is widely used in food packaging, medical devices, and electronic products due to its excellent mechanical properties, gas barrier properties, and transparency. Traditional BOPA film production control mainly relies on fixed parameter settings and manual adjustments based on experience. PID control methods are used for simple closed-loop control of stretching temperature, speed, and tension, or open-loop control strategies based on statistical models are used to preset process parameters. With the development of automation technology, some production lines have begun to adopt single-objective optimization model predictive control methods, using simplified linear models to optimize and control single quality indicators such as thickness uniformity.

[0003] However, traditional fixed-parameter control and PID control cannot cope with the complex coupling relationships between process parameters, making it difficult to simultaneously guarantee thickness uniformity, width stability, and optical performance. Secondly, existing model predictive control is mostly based on offline modeling and lacks online adaptive capability, making it unable to adapt to fluctuations in raw material properties and changes in equipment status. Thirdly, single-objective optimization control ignores energy consumption factors and control stability, leading to equipment fluctuations and energy waste caused by frequent adjustments. Fourthly, existing control methods do not accurately handle constraints on key process parameters such as stretching ratio and temperature gradient, posing a risk of violating process safety boundaries.

[0004] Meanwhile, subspace identification methods suffer from problems such as difficulty in determining the model order and untimely parameter updates; the weight setting of multi-objective optimization functions lacks scientific basis, making it difficult to balance quality indicators and energy consumption indicators; and the constraint boundary handling methods are simple and crude, often using hard truncation to handle constraint violations, resulting in discontinuous control. Summary of the Invention

[0005] This application provides a synchronous biaxially oriented film production control method and system based on process optimization, which is used to solve the quality fluctuation problem caused by insufficient model adaptive capability in the BOPA film production process.

[0006] In a first aspect, this application provides a process optimization-based synchronous biaxially oriented film production control method. The method includes: real-time acquisition and subspace identification processing of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature, and tension data from the BOPA film production line to obtain a state-space prediction model; multi-objective weighted combination processing of film thickness deviation, width deviation, optical performance indicators, and energy consumption parameters based on the state-space prediction model to obtain a performance optimization function; constraint boundary setting processing based on the performance optimization function to obtain a quadratic programming constraint set; inputting the quadratic programming constraint set into a gradient projection algorithm for rolling time-domain optimization to obtain the optimal control sequence, and applying the first control vector to the longitudinal and transverse stretching actuators.

[0007] Optionally, the real-time acquisition and subspace identification processing of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature, and tension data of the BOPA film production line to obtain a state-space prediction model includes:

[0008] The melt temperature sensor, longitudinal tensile speed encoder, transverse tensile speed encoder, heating zone temperature sensor and tension sensor are sampled 10 times per second to obtain an input-output data sequence containing 5000 moments.

[0009] The singular value decomposition of the Hankel matrix is ​​performed based on the input and output data sequence to obtain the system order parameters and the dimension-reduced observability matrix.

[0010] Based on the observability matrix, the extended observability matrix is ​​decomposed into orthogonal triangular matrices to obtain the system state transition matrix, input matrix, output matrix, and disturbance matrix.

[0011] The system state transition matrix, input matrix, output matrix, and disturbance matrix are input into the discretization equation to construct a state-space model, resulting in a discrete-time state-space model containing state equations and output equations.

[0012] Based on the discrete-time state-space model, the historical prediction error is processed by recursive least squares calculation to obtain the model parameter update matrix and gain vector.

[0013] Based on the model parameter update matrix and gain vector, the system parameters of the discrete-time state-space model are updated online to obtain a state-space prediction model of the dynamic characteristics of the thin film stretching process with parameter adaptive capability.

[0014] Optionally, the step of performing multi-objective weighted combination processing on the film thickness deviation, width deviation, optical performance indicators, and energy consumption parameters according to the state-space prediction model to obtain the performance optimization function includes:

[0015] The system state for the next 30 sampling periods is iteratively calculated based on the state-space prediction model to obtain a prediction output sequence that includes the predicted thickness distribution, predicted width distribution, and predicted optical performance of the thin film.

[0016] Based on the predicted output sequence, the difference calculation process is performed on the product standard specification value to obtain the deviation sequence between the predicted thickness and the standard thickness, the deviation sequence between the predicted width and the standard width, and the deviation sequence between the predicted optical performance and the standard optical performance.

[0017] The thickness deviation sequence, width deviation sequence, and optical performance deviation sequence are processed by element-wise squaring to obtain the thickness deviation squared sequence, width deviation squared sequence, and optical performance deviation squared sequence.

[0018] The weighted summation of the thickness deviation square sequence, width deviation square sequence, and optical performance deviation square sequence is performed to obtain the weighted summation of product quality deviations in the prediction time domain.

[0019] Based on the control input of the state-space prediction model, differential calculation processing is performed on the control quantities at adjacent time points to obtain a control increment sequence containing longitudinal stretching speed change, lateral stretching speed change, and temperature regulation change.

[0020] The weighted sum of product quality deviations, the sum of squares of control increment sequences, and energy consumption parameters are linearly combined with weighted coefficients to obtain a performance optimization function that comprehensively considers quality weight, stability weight, and energy consumption weight.

[0021] Optionally, the constraint boundary setting process based on the performance optimization function to obtain the quadratic programming constraint set includes:

[0022] The ratio of longitudinal stretching speed to transverse stretching speed is calculated based on the performance optimization function to obtain the stretching ratio sequence at the current sampling time.

[0023] The numerical sequence of stretch ratios is compared with the preset upper limit of 4.5 and lower limit of 2.5 to obtain a set of stretch ratio inequality constraints.

[0024] Based on the control variables of the performance optimization function, the temperature control quantity of each heating zone is limited in numerical range to obtain a set of temperature constraint conditions with an upper limit of 160 degrees Celsius and a lower limit of 70 degrees Celsius for the temperature control quantity.

[0025] Based on the set of temperature constraints, the temperature difference between adjacent heating zones is calculated to obtain a temperature gradient constraint that the temperature gradient does not exceed 10 degrees Celsius per meter.

[0026] By comparing the tension control variables in the performance optimization function with the rated tension range of the equipment, a set of tension constraints with an upper limit of 500 Newtons and a lower limit of 50 Newtons is obtained.

[0027] Based on the set of stretch ratio inequality constraints, temperature constraints, temperature gradient constraints, and tension constraints, a matrix combination process is performed to obtain a set of quadratic programming constraints that includes inequality constraint matrices and constraint boundary vectors.

[0028] Optionally, the step of inputting the quadratic programming constraint set into the gradient projection algorithm for rolling time-domain optimization to obtain the optimal control sequence, and applying the first control vector to the longitudinal and transverse stretching actuators, includes:

[0029] By combining the set of quadratic programming constraints with the performance optimization function, a complete quadratic programming problem is obtained, which includes the objective function, the inequality constraint matrix, and the constraint boundary vector.

[0030] The gradient of the initial control sequence is calculated based on the complete quadratic programming problem to obtain the gradient vector of the objective function with respect to the control variables.

[0031] Based on the gradient vector, the current control sequence is searched for negative gradient direction to obtain the control sequence update direction under unconstrained conditions.

[0032] The control sequence update direction is projected onto the constrained feasible region for projection operation to obtain a feasible control sequence update direction that satisfies all constraints.

[0033] The current control sequence is iteratively updated according to the feasible control sequence update direction to obtain the optimal control sequence for the next 30 periods after convergence.

[0034] The first control vector in the optimal control sequence is decomposed into longitudinal stretching speed command, transverse stretching speed command, and temperature adjustment command, which are then applied to the longitudinal stretching servo motor, the transverse stretching chain clamp mechanism, and the heating system actuator, respectively.

[0035] Optionally, the step of projecting the control sequence update direction onto the constrained feasible region for projection operation to obtain a feasible control sequence update direction that satisfies all constraints includes:

[0036] The current control sequence is linearly combined according to the control sequence update direction to obtain an unprojected temporary control sequence;

[0037] The temporary control sequence is compared item by item with the inequality constraints in the quadratic programming constraint set to obtain the control component index set that violates the constraints.

[0038] Based on the set of control component indices that violate the constraints, the corresponding control components are subjected to boundary truncation to obtain corrected control components that satisfy the boundary constraints.

[0039] Based on the difference between the corrected control component and the original control component, a reverse compensation calculation is performed to obtain the projection adjustment vector on the constraint boundary.

[0040] The projection adjustment vector is applied to the control sequence update direction for vector correction processing to obtain feasible direction components on the constraint boundary tangent plane.

[0041] Based on the feasible direction components, all constraints are verified to obtain feasible control sequence update directions that strictly satisfy all constraints.

[0042] Optionally, the step of applying the projection adjustment vector to the control sequence update direction for vector correction processing to obtain feasible direction components on the constraint boundary tangent plane includes:

[0043] The orthogonalization calculation is performed on the constraint boundary normal vector corresponding to the projection adjustment vector to obtain the set of unit normal vectors of the constraint boundary.

[0044] The inner product operation is performed between the control sequence update direction and the unit normal vector set to obtain the projection coefficients of the update direction on the normals of each constraint boundary.

[0045] Based on the projection coefficients, the set of unit normal vectors is linearly combined to obtain the normal components that need to be removed from the update direction.

[0046] The normal component is subtracted from the control sequence update direction by a vector subtraction operation to obtain the tangential component retained in the tangential space of the constraint boundary.

[0047] The satisfaction of the constraint conditions is verified a second time based on the tangential components to obtain the feasible direction components on the tangential plane of the constraint boundary after verification.

[0048] Based on the feasible direction components, the step size parameter is adaptively adjusted to obtain a feasible control sequence update direction that guarantees convergence.

[0049] Secondly, this application provides a process-optimized synchronous biaxially oriented film production control system, the process-optimized synchronous biaxially oriented film production control system comprising:

[0050] The identification module is used to collect and process subspace identification data of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature and tension in real time from the BOPA film production line to obtain a state space prediction model.

[0051] The combination module is used to perform multi-objective weighted combination processing on the film thickness deviation, width deviation, optical performance index and energy consumption parameter according to the state space prediction model to obtain the performance optimization function;

[0052] The processing module is used to perform constraint boundary setting processing based on the performance optimization function to obtain a set of quadratic programming constraints;

[0053] The solution module is used to input the set of quadratic programming constraints into the gradient projection algorithm for rolling time-domain optimization to obtain the optimal control sequence, and to apply the first control vector to the longitudinal and transverse stretching actuators.

[0054] Thirdly, a process-optimized synchronous biaxially oriented film production control device is provided, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor invokes the instructions in the memory to cause the process-optimized synchronous biaxially oriented film production control device to execute the aforementioned process-optimized synchronous biaxially oriented film production control method.

[0055] Fourthly, a computer-readable storage medium is provided, wherein instructions are stored therein, which, when executed on a computer, cause the computer to perform the above-described process-optimized synchronous biaxially stretched film production control method.

[0056] The technical solution provided in this application constructs an adaptive state-space prediction model by real-time acquisition and subspace identification processing of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature, and tension data from a BOPA film production line. This effectively solves the problem that traditional fixed-parameter models cannot cope with fluctuations in process parameters, enabling the control system to adapt to changes in raw material performance and equipment status fluctuations in real time, significantly improving the stability of film quality. Subspace identification algorithms, as an advanced system identification method, fully utilize their dimensionality reduction and rapid modeling characteristics in the specific application field of BOPA films. This allows the model to capture the dynamic characteristics of complex nonlinear systems while maintaining low computational complexity, making it suitable for online real-time applications. Based on the state-space prediction model, multi-objective weighted combination processing is performed on film thickness deviation, width deviation, optical performance indicators, and energy consumption parameters to construct a performance optimization function that comprehensively considers quality, stability, and energy consumption. This solves the problem of traditional single-objective optimization neglecting energy consumption factors and control stability, achieving coordinated optimization of product quality, equipment life, and energy consumption. The application of this multi-objective optimization algorithm in the BOPA film production field enables the system to minimize energy consumption and equipment wear while ensuring product quality, improving the economic efficiency and sustainability of production. Based on the performance optimization function, a quadratic programming constraint set was constructed, including key process parameters such as stretching ratio, temperature, temperature gradient, and tension. This solves the problem of inaccurate handling of process parameter constraints in traditional control methods. By using explicit mathematical expressions, the process safety boundary is transformed into a form that the optimization algorithm can handle, effectively preventing the risk of violating the process safety boundary. The quadratic programming constraint set is then input into the gradient projection algorithm for rolling time-domain optimization, yielding the optimal control sequence that strictly satisfies all process constraints. This solves the problems of simplistic and crude constraints and discontinuous control in traditional constraint handling methods.

[0057] In the specific application of BOPA film production, the gradient projection algorithm ensures that the control quantity always meets process constraints through projection operations, while maintaining the stability of optimization, significantly improving control accuracy and system stability. The rolling time-domain optimization strategy enables the system to continuously adjust the control sequence based on the latest feedback, enhancing its resistance to disturbances. By applying the first control vector to the longitudinal and transverse stretching actuators, precise closed-loop control of the BOPA film production process is achieved, comprehensively improving product quality, production efficiency, and energy utilization. The application of artificial intelligence algorithms in this solution, particularly subspace identification and gradient projection optimization algorithms, fully considers the special needs of BOPA film production, specifically addressing key issues such as modeling accuracy, multi-objective trade-offs, and constraint handling. This demonstrates the technological innovation and value enhancement brought about by the deep integration of algorithmic features and application domains. Attached Figure Description

[0058] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a schematic diagram of an embodiment of the synchronous biaxially stretched film production control method based on process optimization in this application.

[0060] Figure 2 This is a schematic diagram of an embodiment of the synchronous biaxially stretched film production control system based on process optimization in this application.

[0061] Figure 3 This is a schematic block diagram of the synchronous biaxially stretched film production control equipment based on process optimization in this embodiment of the invention. Detailed Implementation

[0062] This application provides a method and system for controlling the production of synchronous biaxially oriented thin films based on process optimization. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0063] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the synchronous biaxially stretched film production control method based on process optimization in this application includes:

[0064] Step S101: Real-time acquisition and subspace identification processing of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature and tension data of BOPA film production line to obtain state space prediction model.

[0065] Step S102: Based on the state-space prediction model, perform multi-objective weighted combination processing on the film thickness deviation, width deviation, optical performance index and energy consumption parameter to obtain the performance optimization function;

[0066] Step S103: Perform constraint boundary setting processing based on the performance optimization function to obtain the quadratic programming constraint set;

[0067] Step S104: Input the set of quadratic programming constraints into the gradient projection algorithm for rolling time-domain optimization to obtain the optimal control sequence, and apply the first control vector to the longitudinal and transverse stretching actuators.

[0068] It is understood that the executing entity of this application can be a synchronous biaxially oriented film production control system based on process optimization, or it can be a terminal or a server; the specific implementation is not limited here. This application's embodiments use a server as an example for illustration.

[0069] Specifically, a state-space prediction model is obtained by real-time acquisition and subspace identification processing of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature, and tension data from the BOPA film production line. This process employs high-frequency sampling of 10 times per second to collect input-output data sequences at 5000 time points. These data come from melt temperature sensors, longitudinal stretching speed encoders, transverse stretching speed encoders, heating zone temperature sensors, and tension sensors. Subspace identification processing is a system identification method that extracts system order parameters and a dimension-reduced observability matrix by performing singular value decomposition on the Hankel matrix. The Hankel matrix is ​​a special structure matrix constructed from input-output data sequences, and singular value decomposition effectively reduces system complexity. Based on the obtained observability matrix, the extended observability matrix is ​​further orthogonally triangularly decomposed to obtain the system state transition matrix, input matrix, output matrix, and disturbance matrix. These matrices are then input into the discretized equations to construct a discrete-time state-space model containing state equations and output equations. Subsequently, the historical prediction error is processed by recursive least squares calculation to obtain the model parameter update matrix and gain vector, realizing the online update of the model parameters, thereby obtaining a state-space prediction model of the dynamic characteristics of the thin film stretching process with parameter adaptive capability.

[0070] Based on the state-space prediction model, a multi-objective weighted combination processing of thin film thickness deviation, width deviation, optical performance indicators, and energy consumption parameters is performed to obtain a performance optimization function. This process first uses the state-space prediction model to iteratively calculate the system state for the next 30 sampling periods, obtaining a predicted output sequence containing the predicted thin film thickness distribution, predicted width distribution, and predicted optical performance. Then, the difference between the predicted output sequence and the product standard specification value is calculated, resulting in three deviation sequences: the deviation sequence between predicted thickness and standard thickness, the deviation sequence between predicted width and standard width, and the deviation sequence between predicted optical performance and standard optical performance. Next, element-wise squaring is performed on these three deviation sequences to obtain the corresponding square sequences. These square sequences are then weighted and summed to obtain the weighted sum of product quality deviations in the prediction time domain. Simultaneously, the control inputs at adjacent time points are differentially calculated to obtain a control increment sequence, including changes in longitudinal stretching speed, changes in transverse stretching speed, and changes in temperature regulation. Finally, the weighted sum of product quality deviations, the sum of squares of control increment sequences, and energy consumption parameters are linearly combined with weighted coefficients to form a performance optimization function that comprehensively considers quality weight, stability weight, and energy consumption weight.

[0071] Constraint boundary setting is performed based on the performance optimization function to obtain a set of quadratic programming constraints. Specifically, firstly, the ratio of longitudinal tension speed to transverse tension speed is calculated to obtain the tension ratio numerical sequence at the current sampling time, and this sequence is compared with the preset upper limit of 4.5 and lower limit of 2.5 to form a set of tension ratio inequality constraints. Simultaneously, the temperature control values ​​for each heating zone are limited to a range of 160 degrees Celsius for the upper limit and 70 degrees Celsius for the lower limit, forming a set of temperature constraints. Further, the gradient of the temperature difference between adjacent heating zones is calculated to ensure that the temperature gradient does not exceed 10 degrees Celsius per meter. In addition, the tension control variable is compared with the rated tension range of the equipment, setting a set of tension constraints with an upper limit of 500 Newtons and a lower limit of 50 Newtons. Finally, these constraints are matrix-combined to obtain a set of quadratic programming constraints containing inequality constraint matrices and constraint boundary vectors. The set of quadratic programming constraints is then input into a gradient projection algorithm for rolling time-domain optimization to obtain the optimal control sequence, and the first control vector is applied to the longitudinal and transverse tension actuators. The gradient projection algorithm is a method for solving constrained optimization problems, capable of searching for the optimal solution while ensuring that the constraints are satisfied. First, the set of quadratic programming constraints is combined with the performance optimization function to form a quadratic programming problem. Then, the gradient vector of the objective function with respect to the control variables is calculated, and the search proceeds along the negative gradient direction to obtain the control sequence update direction under unconstrained conditions. Subsequently, the control sequence update direction is projected onto the constrained feasible region, ensuring that all constraints are satisfied. This projection process includes linearly combining the current control sequence to obtain a temporary control sequence, identifying control components that violate constraints and trunculating their boundaries, calculating the projection adjustment vector on the constraint boundaries through back-compensation, and finally ensuring that the control sequence strictly satisfies all constraints. After iterative updates, the convergent optimal control sequence for the next 30 cycles is obtained, and the first control vector is decomposed into longitudinal stretching speed commands, lateral stretching speed commands, and temperature adjustment commands, which are applied to the longitudinal stretching servo motor, the lateral stretching chain clamp mechanism, and the heating system actuator, respectively.

[0072] In one specific embodiment, the process of performing step S101 may specifically include the following steps:

[0073] The melt temperature sensor, longitudinal tensile speed encoder, transverse tensile speed encoder, heating zone temperature sensor and tension sensor are sampled 10 times per second to obtain an input-output data sequence containing 5000 moments.

[0074] Singular value decomposition is performed on the Hankel matrix based on the input and output data sequences to obtain the system order parameters and the dimension-reduced observability matrix;

[0075] Based on the observability matrix, the extended observability matrix is ​​decomposed into orthogonal triangular matrices to obtain the system state transition matrix, input matrix, output matrix, and disturbance matrix.

[0076] The system state transition matrix, input matrix, output matrix, and disturbance matrix are input into the discretization equation to construct a state-space model, resulting in a discrete-time state-space model containing state equations and output equations.

[0077] Based on the discrete-time state-space model, the historical prediction error is processed by recursive least squares calculation to obtain the model parameter update matrix and gain vector;

[0078] The system parameters of the discrete-time state-space model are updated online based on the model parameter update matrix and gain vector, resulting in a state-space prediction model of the dynamic characteristics of the thin film stretching process with parameter adaptive capability.

[0079] Specifically, data sampling is performed 10 times per second on the melt temperature sensor, longitudinal stretching speed encoder, transverse stretching speed encoder, heating zone temperature sensor, and tension sensor. First, the data acquisition module performs high-frequency sampling on the five sensors on the BOPA film production line at a sampling frequency of 10Hz, meaning 10 data points are collected from each sensor per second. This sampling continues for approximately 8 minutes and 20 seconds, resulting in an input-output data sequence containing 5000 time points. The input data includes melt temperature, longitudinal stretching speed, transverse stretching speed, and heating zone temperature, while the output data is the tension value. When performing singular value decomposition (SVD) on the Hankel matrix based on the input-output data sequence, the collected input-output data sequence is reconstructed into a Hankel matrix form. A Hankel matrix is ​​a matrix with a special structure where all elements on each antidiagonal are identical. In BOPA film production control, the input and output sequences are arranged in chronological order to construct a Hankel matrix. Then, singular value decomposition is performed on this matrix, resulting in three matrices: a left singular matrix, a singular value matrix, and a right singular matrix. By analyzing the magnitude of singular values, the order parameter of the system, i.e., the system complexity, is determined. Generally, the vectors corresponding to the first few larger singular values ​​are selected to perform dimensionality reduction on the original high-dimensional data, resulting in the dimensionality-reduced observability matrix. The observability matrix represents the mapping relationship between the system state and the output.

[0080] When performing orthogonal triangular decomposition on the extended observability matrix based on the observability matrix, the extended observability matrix is ​​first constructed. This matrix contains the original observability matrix and additional information about the input and output data. Orthogonal triangular decomposition, or QR decomposition, is then performed on the extended observability matrix, decomposing it into the product of an orthogonal matrix and an upper triangular matrix. Algebraic operations are then performed on the decomposition results to extract the system state transition matrix, input matrix, output matrix, and disturbance matrix. The state transition matrix describes the time-varying nature of the system's internal state; the input matrix represents the influence of control inputs on the system state; the output matrix represents how the system state maps to the observable output; and the disturbance matrix represents the direct impact of external disturbances on the system output. When inputting the system state transition matrix, input matrix, output matrix, and disturbance matrix into the discretized equations for state-space model construction, a discrete-time state-space expression is used, including state equations and output equations. The state equations describe how the system state at the next time step is determined by the current state and control inputs, and the output equations describe how the system output is determined by the current state and control inputs. A discrete-time state-space model is constructed using these two equations, which can describe the dynamic relationships between variables during the stretching process of BOPA films.

[0081] When processing historical prediction errors using recursive least squares calculations based on a discrete-time state-space model, the system's future output is first predicted using the established state-space model. Then, the predicted values ​​are compared with actual measurements to calculate the prediction error. For these prediction errors, a recursive least squares algorithm is applied for parameter estimation. This algorithm continuously adjusts model parameters as new data arrives, making the model predictions closer to the actual system behavior. The recursive least squares algorithm obtains the model parameter update matrix and gain vector by minimizing the sum of squared errors with a forgetting factor. The parameter update matrix is ​​the covariance matrix of the parameter estimates, and the gain vector determines the degree of influence of new observation data on the parameter estimates. When updating the system parameters of the discrete-time state-space model online based on the model parameter update matrix and gain vector, the parameter correction is calculated using the recursive least squares method with each new input-output data pair, and the system matrix in the state-space model is updated. This online parameter update mechanism enables the model to adapt to process fluctuations and equipment state changes, resulting in a state-space prediction model with adaptive parameter capabilities for the dynamic characteristics of the thin film stretching process.

[0082] In one specific embodiment, the process of performing step S102 may specifically include the following steps:

[0083] The system state for the next 30 sampling periods is iteratively calculated based on the state-space prediction model to obtain a prediction output sequence that includes the predicted thickness distribution, predicted width distribution, and predicted optical performance of the thin film.

[0084] The difference between the predicted output sequence and the product standard specification value is calculated to obtain the deviation sequence between the predicted thickness and the standard thickness, the deviation sequence between the predicted width and the standard width, and the deviation sequence between the predicted optical performance and the standard optical performance.

[0085] The thickness deviation sequence, width deviation sequence, and optical performance deviation sequence are processed by element-wise squaring to obtain the thickness deviation squared sequence, width deviation squared sequence, and optical performance deviation squared sequence.

[0086] The weighted sum of the product quality deviations in the prediction time domain is obtained by performing a weighted summation on the squared sequences of thickness deviation, width deviation, and optical performance deviation.

[0087] Based on the state-space prediction model, the control inputs of adjacent time-time control quantities are differentially calculated to obtain a control increment sequence containing longitudinal stretching speed changes, lateral stretching speed changes, and temperature regulation changes.

[0088] By performing a linear combination of the weighted sum of product quality deviations, the sum of squares of the control increment sequence, and the energy consumption parameter with weighted coefficients, a performance optimization function that comprehensively considers quality weight, stability weight, and energy consumption weight is obtained.

[0089] Specifically, using the established state-space model, starting from the current moment, the system state and output for the next 30 sampling periods are calculated sequentially. The current system state and control input are substituted into the state equation to calculate the system state for the next moment. Then, the calculated system state is substituted into the output equation to calculate the corresponding system output. This process is repeated 30 times to obtain the system state sequence and output sequence for the next 30 sampling periods. The system output is a predicted output sequence containing the predicted thickness distribution, predicted width distribution, and predicted optical performance of the thin film. These output parameters directly reflect the key quality indicators of the BOPA thin film. When performing difference calculations on the product standard specification values ​​based on the predicted output sequences, the film thickness distribution, width distribution, and optical performance obtained in the previous step are compared with the product standard specification values. The standard specification values ​​are target values ​​pre-set according to product quality requirements, such as a standard thickness of 25 micrometers, a standard width of 1500 millimeters, and standard optical performance indicators such as specific transmittance or complex refractive index values. By calculating the differences between the predicted values ​​and the standard values, the deviation sequences of predicted thickness and standard thickness, predicted width and standard width, and predicted optical performance and standard optical performance are obtained respectively. These deviation sequences reflect the extent to which product quality may deviate from the target value within the next 30 sampling periods.

[0090] When performing element-wise squaring on the thickness deviation sequence, width deviation sequence, and optical performance deviation sequence, each element in the three deviation sequences obtained in the previous step is squared individually. The purpose of this process is to unify positive and negative deviations into non-negative values ​​and to impose a higher penalty on larger deviations, making the optimization process more focused on reducing large deviations. Element-wise squaring involves squaring each value in the deviation sequence individually to obtain the corresponding sequence of squared values: the thickness deviation squared sequence, the width deviation squared sequence, and the optical performance deviation squared sequence.

[0091] When performing a weighted summation of the squared sequences of thickness deviation, width deviation, and optical performance deviation, weight coefficients are assigned to each sequence to reflect the relative importance of each quality indicator. For example, if thickness uniformity is more critical, the weight coefficient for the squared sequence of thickness deviation will be set higher. The weighted summation of the three squared sequences, along with the weighted squared deviations at all times, yields the weighted total product quality deviation over the prediction time domain. This total reflects the degree to which the overall product quality deviates from the target over the next 30 sampling periods.

[0092] When performing differential calculations on control inputs from state-space prediction models for adjacent time-series control quantities, the difference between two adjacent control quantities over the next 30 sampling periods is calculated. The control quantities include longitudinal stretching speed, lateral stretching speed, and the heating zone temperature setpoint. By calculating the difference between adjacent control quantities, a control increment sequence is obtained, containing changes in longitudinal stretching speed, lateral stretching speed, and temperature regulation. These changes reflect the smoothness of the control operation; larger changes indicate more drastic control, potentially leading to increased mechanical wear and energy consumption. When performing a weighted linear combination of the weighted sum of product quality deviations, the sum of squares of the control increment sequence, and energy consumption parameters, the sum of squares of the control increment sequence is first calculated by squaring each element and summing the results. Then, appropriate weighting coefficients are selected to linearly combine the weighted sum of product quality deviations, the sum of squares of the control increment sequence, and energy consumption parameters (such as total electrical energy consumption) to form a performance optimization function. This optimization function comprehensively considers three aspects: product quality (represented by quality deviation), control smoothness (represented by the sum of squares of control increments), and energy consumption (represented by energy consumption parameters). By adjusting the weight coefficients of each part, the relationship between the three objectives can be balanced, forming a performance optimization function that comprehensively considers quality weight, stability weight, and energy consumption weight.

[0093] In one specific embodiment, the process of executing step S103 may specifically include the following steps:

[0094] The ratio of longitudinal stretching speed to transverse stretching speed is calculated based on the performance optimization function to obtain the stretching ratio sequence at the current sampling time.

[0095] The numerical sequence of stretch ratios is compared with the preset upper limit of 4.5 and lower limit of 2.5 to obtain the set of stretch ratio inequality constraints.

[0096] Based on the control variables of the performance optimization function, the numerical range of the temperature control quantity of each heating zone is limited to obtain a set of temperature constraints with an upper limit of 160 degrees Celsius and a lower limit of 70 degrees Celsius for the temperature control quantity.

[0097] Based on the set of temperature constraints, the temperature difference between adjacent heating zones is calculated to obtain a temperature gradient constraint that the temperature gradient does not exceed 10 degrees Celsius per meter.

[0098] By comparing the tension control variables in the performance optimization function with the rated tension range of the equipment, a set of tension constraints with an upper limit of 500 Newtons and a lower limit of 50 Newtons is obtained.

[0099] By performing matrix-based combination processing on the set of stretch ratio inequality constraints, temperature constraints, temperature gradient constraints, and tension constraints, a quadratic programming constraint set containing inequality constraint matrices and constraint boundary vectors is obtained.

[0100] Specifically, the ratio of longitudinal stretching speed to transverse stretching speed is calculated based on the performance optimization function. The longitudinal stretching speed is divided by the original film speed, and the transverse stretching speed is divided by the original film width to obtain the longitudinal stretching ratio and transverse stretching ratio, respectively. Multiplying these two ratios yields the total stretching ratio sequence at the current sampling time. The stretching ratio is the ratio of the film size after stretching to the size before stretching, directly affecting the film's mechanical and optical properties. When comparing the stretching ratio sequence with preset upper and lower limits of 4.5 and 2.5, the upper and lower limits of the stretching ratio are expressed as mathematical inequalities, forming a set of stretching ratio inequality constraints. This constraint ensures that the film will not break due to overstretching or fail to meet performance standards due to understretching. Based on the control variables of the performance optimization function, the temperature control variables for each heating zone are limited to an upper limit of 160 degrees Celsius and a lower limit of 70 degrees Celsius, forming a set of temperature constraints. Temperature control is a critical parameter in BOPA film production; excessively high temperatures will cause the film to melt and deform, while excessively low temperatures will prevent effective stretching. The temperature gradient between adjacent heating zones is calculated based on a set of temperature constraints. The temperature gradient is obtained by dividing the temperature difference between adjacent heating zones by the distance between them, and this gradient is limited to no more than 10 degrees Celsius per meter, thus forming the temperature gradient constraint. This temperature gradient constraint ensures uniform heating of the film on the production line, preventing stress concentration and deformation caused by rapid local temperature changes.

[0101] The tension control variable in the performance optimization function is compared with the rated tension range of the equipment, setting an upper limit of 500 Newtons and a lower limit of 50 Newtons to form a set of tension constraints. Tension control is a crucial parameter for ensuring film flatness and thickness uniformity; excessive tension can lead to film tearing, while insufficient tension can cause film relaxation and wrinkling. Based on the set of constraints related to stretch ratio inequalities, temperature constraints, temperature gradient constraints, and tension constraints, a matrix-based combination process is performed, converting all constraints into standard linear inequalities. These constraints are then combined into an inequality constraint matrix and constraint boundary vectors, forming a set of quadratic programming constraints.

[0102] In one specific embodiment, the process of executing step S104 may specifically include the following steps:

[0103] By combining the set of quadratic programming constraints with the performance optimization function, a complete quadratic programming problem is obtained, which includes the objective function, the inequality constraint matrix, and the constraint boundary vector.

[0104] The gradient of the initial control sequence is calculated based on the complete quadratic programming problem to obtain the gradient vector of the objective function with respect to the control variables.

[0105] Based on the gradient vector, the negative gradient direction search process is performed on the current control sequence to obtain the control sequence update direction under unconstrained conditions.

[0106] Project the control sequence update direction onto the constrained feasible region and perform projection operations to obtain a feasible control sequence update direction that satisfies all constraints.

[0107] Specifically, the current control sequence is linearly combined according to the control sequence update direction to obtain an unprojected temporary control sequence; the temporary control sequence is compared item by item with the inequality constraints in the quadratic programming constraint set to obtain a set of control component indices that violate the constraints; based on the set of control component indices that violate the constraints, the corresponding control components are truncated at the boundary to obtain corrected control components that satisfy the boundary constraints.

[0108] The projection adjustment vector on the constraint boundary is obtained by performing reverse compensation calculation based on the difference between the corrected control component and the original control component.

[0109] The projection adjustment vector is applied to the control sequence update direction for vector correction processing to obtain feasible direction components on the constraint boundary tangent plane. This includes: performing orthogonalization calculations on the constraint boundary normal vectors corresponding to the projection adjustment vector to obtain the set of unit normal vectors of the constraint boundary; performing an inner product operation on the control sequence update direction and the set of unit normal vectors to obtain the projection coefficients of the update direction on the normals of each constraint boundary; performing a linear combination operation on the set of unit normal vectors based on the projection coefficients to obtain the normal components that need to be removed from the update direction; performing a vector subtraction operation on the control sequence update direction minus the normal components to obtain the tangential components retained in the constraint boundary tangent space; performing a secondary verification process on the satisfaction of the constraint conditions based on the tangential components to obtain the verified feasible direction components on the constraint boundary tangent plane; and adaptively adjusting the step size parameter based on the feasible direction components to obtain a feasible control sequence update direction that guarantees convergence.

[0110] Based on the feasible direction components, all constraints are satisfied to obtain feasible control sequence update directions that strictly satisfy all constraints.

[0111] The current control sequence is iteratively updated according to the feasible control sequence update direction to obtain the optimal control sequence for the next 30 periods after convergence.

[0112] The first control vector in the optimal control sequence is decomposed into longitudinal stretching speed command, lateral stretching speed command, and temperature adjustment command, which are then applied to the longitudinal stretching servo motor, the lateral stretching chain clamp mechanism, and the heating system actuator, respectively.

[0113] Specifically, combining the inequality constraint matrix and constraint boundary vector with the performance optimization function constitutes a standard constrained optimization problem. This quadratic programming problem comprises three parts: the objective function (i.e., the performance optimization function), constraints (including stretching ratio, temperature, temperature gradient, and tension constraints), and optimization variables (control sequence). When performing gradient calculation on the initial control sequence according to the complete quadratic programming problem, the derivative of the objective function at the current control point is calculated, i.e., the rate of change of the objective function with respect to each control variable is solved, yielding the gradient vector representing the direction of the fastest descent. Gradient calculation is the core step of the optimization algorithm; through numerical differentiation methods, the partial derivatives of the performance optimization function with respect to the control variables are calculated to form the gradient vector.

[0114] When performing a negative gradient direction search on the current control sequence based on the gradient vector, the algorithm moves along the opposite direction of the gradient because the gradient points to the direction of the fastest growth of the function, and the optimization objective is to minimize the objective function. Therefore, the search needs to proceed along the negative gradient direction. This step yields the control sequence update direction under unconstrained conditions, which is the direction of the optimal search under ideal conditions. When projecting the control sequence update direction onto the constrained feasible region, the algorithm checks whether moving along the negative gradient direction violates the constraints. If it does, the update direction needs to be corrected to a direction within the constrained feasible region. This projection process is crucial to the gradient projection algorithm, ensuring that the optimization process always satisfies the technological constraints.

[0115] The current control sequence is linearly combined based on the control sequence update direction, i.e., the current control value is added to the update direction by a certain step size, resulting in an unprojected temporary control sequence. Then, the temporary control sequence is compared one by one with the inequality constraints in the quadratic programming constraint set to check if each constraint is violated, identifying the control component indices that violate the constraints and forming an index set. Based on the control component index set that violates the constraints, the corresponding control components are truncated at the boundary, i.e., control values ​​exceeding the constraint range are corrected to constraint boundary values, resulting in corrected control components that satisfy the boundary constraints. Backward compensation calculation is performed based on the difference between the corrected control components and the original control components to obtain the projection adjustment vector on the constraint boundary. This vector represents the adjustment required to satisfy the constraints in the original direction. The projection adjustment vector is applied to the control sequence update direction for vector correction processing, resulting in feasible direction components on the constraint boundary tangent plane. This process includes: first, orthogonalizing the constraint boundary normal vectors corresponding to the projection adjustment vector to obtain the set of unit normal vectors of the constraint boundary; then, performing an inner product operation between the control sequence update direction and the set of unit normal vectors to calculate the projection coefficients of the update direction on the normals of each constraint boundary; performing a linear combination of the set of unit normal vectors based on the projection coefficients to obtain the normal components that need to be removed from the update direction; subtracting the normal components from the control sequence update direction to obtain the tangential components retained in the tangential space of the constraint boundary; performing a second verification on the tangential components to ensure that they satisfy all constraint conditions; and finally, adjusting the step size parameter based on the feasible direction components to ensure the convergence of the optimization algorithm.

[0116] Based on the feasible direction components, the satisfaction of all constraints is verified again to obtain the feasible control sequence update direction that strictly satisfies all constraints. The current control sequence is iteratively updated according to this update direction, repeatedly performing gradient calculation, negative gradient search, and constraint projection until convergence is achieved, yielding the optimal control sequence for the next 30 cycles. Finally, the first control vector in the optimal control sequence is decomposed into longitudinal stretching speed command, lateral stretching speed command, and temperature adjustment command, which are then transmitted to the corresponding actuators.

[0117] The above describes the synchronous biaxially oriented film production control method based on process optimization in the embodiments of this application. The following describes the synchronous biaxially oriented film production control system based on process optimization in the embodiments of this application. Please refer to [link to relevant documentation]. Figure 2 One embodiment of the synchronous biaxially stretched film production control system based on process optimization in this application includes:

[0118] The identification module 201 is used to collect and process subspace identification data of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature and tension in real time from the BOPA film production line to obtain a state space prediction model.

[0119] Combination module 202 is used to perform multi-objective weighted combination processing on thin film thickness deviation, width deviation, optical performance index and energy consumption parameter according to the state space prediction model to obtain performance optimization function;

[0120] Processing module 203 is used to perform constraint boundary setting processing based on the performance optimization function to obtain a set of quadratic programming constraints;

[0121] The solution module 204 is used to input the set of quadratic programming constraints into the gradient projection algorithm for rolling time-domain optimization to obtain the optimal control sequence, and to apply the first control vector to the longitudinal and transverse stretching actuators.

[0122] above Figure 2 The synchronous biaxially oriented film production control system based on process optimization in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The synchronous biaxially oriented film production control equipment based on process optimization in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0123] Figure 3 This is a schematic diagram of a process-optimized synchronous biaxially oriented film production control device 300 provided in an embodiment of the present invention. This process-optimized synchronous biaxially oriented film production control device 300 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the process-optimized synchronous biaxially oriented film production control device 300. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the process-optimized synchronous biaxially oriented film production control device 300 to implement the steps of the above-described process-optimized synchronous biaxially oriented film production control method.

[0124] The process-optimized synchronous biaxially oriented film production control equipment 300 may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The illustrated structure of the process-optimized synchronous biaxially oriented film production control equipment does not constitute a limitation on the process-optimized synchronous biaxially oriented film production control equipment provided by the present invention. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0125] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the process-optimized synchronous biaxially stretched film production control method.

[0126] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0127] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a process-optimized synchronous biaxially stretched thin film production control device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0128] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A process optimization based simultaneous biaxial stretching film production control method, characterized by, The method comprises: Real-time acquisition and subspace identification processing of melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature and tension data of the BOPA film production line to obtain a state space prediction model, including: 10 times per second data sampling processing of the melt temperature sensor, longitudinal stretching speed encoder, transverse stretching speed encoder, heating zone temperature sensor and tension sensor to obtain an input-output data sequence containing 5000 time points; singular value decomposition processing of the Hankel matrix according to the input-output data sequence to obtain system order parameters and an observable matrix after dimension reduction; orthogonal triangular decomposition processing of an extended observable matrix based on the observable matrix to obtain a system state transition matrix, an input matrix, an output matrix and a disturbance matrix; state space model construction processing of the system state transition matrix, the input matrix, the output matrix and the disturbance matrix input into a discretization equation to obtain a discrete-time state space model containing state equations and output equations; recursive least squares calculation processing of historical prediction errors according to the discrete-time state space model to obtain a model parameter update matrix and a gain vector; online update processing of system parameters of the discrete-time state space model based on the model parameter update matrix and the gain vector to obtain a state space prediction model of the dynamic characteristics of the film stretching process with parameter self-adaptive capability; Multi-objective weighted combination processing of film thickness deviation, width deviation, optical performance indicators and energy consumption parameters according to the state space prediction model to obtain a performance optimization function; Constraint boundary setting processing based on the performance optimization function to obtain a quadratic programming constraint set; Rolling horizon optimization solving processing of the quadratic programming constraint set input into a gradient projection algorithm to obtain an optimal control sequence, and the first control vector is applied to the longitudinal and transverse stretching actuators, including: combination processing of the quadratic programming constraint set and the performance optimization function to obtain a complete quadratic programming problem containing a target function, an inequality constraint matrix and a constraint boundary vector; gradient calculation processing of an initial control sequence according to the complete quadratic programming problem to obtain a gradient vector of the target function with respect to the control variable; negative gradient direction search processing of a current control sequence based on the gradient vector to obtain a control sequence update direction under unconstrained conditions; projection operation processing of the control sequence update direction projected to a constraint feasible region to obtain a feasible control sequence update direction meeting all constraint conditions; iterative update processing of the current control sequence according to the feasible control sequence update direction to obtain an optimal control sequence of the next 30 periods after convergence; decomposition of the first control vector in the optimal control sequence into longitudinal stretching speed instructions, transverse stretching speed instructions and temperature adjustment instructions, which are respectively applied to the longitudinal stretching servo motor, the transverse stretching chain clamp mechanism and the heating system actuator.

2. The process optimization based simultaneous biaxial stretching film production control method according to claim 1, characterized by, The multi-objective weighted combination processing of the film thickness deviation, the width deviation, the optical performance index and the energy consumption parameter according to the state space prediction model obtains a performance optimization function, including: The iterative calculation processing of the system state in the future 30 sampling periods according to the state space prediction model obtains a prediction output sequence containing the film prediction thickness distribution, the prediction width distribution and the prediction optical performance; The difference calculation processing of the product standard specification value based on the prediction output sequence obtains the deviation sequence of the prediction thickness and the standard thickness, the deviation sequence of the prediction width and the standard width and the deviation sequence of the prediction optical performance and the standard optical performance; The element-by-element square operation processing of the thickness deviation sequence, the width deviation sequence and the optical performance deviation sequence obtains the thickness deviation square sequence, the width deviation square sequence and the optical performance deviation square sequence; The weighted summation processing of the thickness deviation square sequence, the width deviation square sequence and the optical performance deviation square sequence obtains the product quality deviation weighted total sum in the prediction time domain; The difference calculation processing of the adjacent time control amount based on the control input of the state space prediction model obtains a control increment sequence containing the longitudinal stretching speed change amount, the transverse stretching speed change amount and the temperature adjustment change amount; The linear combination processing of the product quality deviation weighted total sum, the square sum of the control increment sequence and the energy consumption parameter with a weight coefficient obtains a performance optimization function considering the quality weight, the stability weight and the energy consumption weight.

3. The process optimization based simultaneous biaxial stretching film production control method according to claim 1, characterized by, The constraint boundary setting processing based on the performance optimization function obtains a quadratic programming constraint set, including: The ratio calculation processing of the longitudinal stretching speed and the transverse stretching speed according to the performance optimization function obtains a stretching ratio numerical sequence at the current sampling time; The comparison processing of the stretching ratio numerical sequence and the preset stretching ratio upper limit value 4.5 and the lower limit value 2.5 obtains a stretching ratio inequality constraint condition set; The numerical range limiting processing of the heating area temperature control amount based on the control variable of the performance optimization function obtains a temperature control amount upper limit 160 degrees Celsius and a lower limit 70 degrees Celsius of the temperature control amount of the temperature control amount; The gradient calculation processing of the temperature difference value between adjacent heating areas according to the temperature constraint condition set obtains a temperature gradient constraint condition that the temperature gradient does not exceed 10 degrees Celsius per meter; The comparison processing of the tension control variable in the performance optimization function and the equipment rated tension range obtains a tension constraint condition set that the tension upper limit is 500 Newton and the lower limit is 50 Newton; The matrix combination processing based on the stretching ratio inequality constraint condition set, the temperature constraint condition set, the temperature gradient constraint condition and the tension constraint condition set obtains a quadratic programming constraint set containing an inequality constraint matrix and a constraint boundary vector.

4. The process optimization based simultaneous biaxial stretching film production control method according to claim 1, characterized by, The projection operation processing of the control sequence update direction projection to the constraint feasible region obtains a feasible control sequence update direction meeting all the constraint conditions, including: The linear combination processing of the current control sequence according to the control sequence update direction obtains a temporary control sequence without projection. The temporary control sequence is compared with inequality constraints in the quadratic programming constraint set one by one to obtain a constraint-violating control component index set; A boundary truncation process is performed on the corresponding control component based on the constraint-violating control component index set to obtain a modified control component satisfying the boundary constraint; A reverse compensation calculation process is performed according to the difference between the modified control component and the original control component to obtain a projection adjustment vector on the constraint boundary; The projection adjustment vector is applied to the control sequence update direction for vector modification to obtain a feasible direction component on the constraint boundary tangent plane; Based on the feasible direction component, a satisfaction verification process is performed on all constraint conditions to obtain a feasible control sequence update direction that strictly satisfies all constraint conditions.

5. The process optimization based simultaneous biaxial stretching film production control method according to claim 4, characterized by, The projection adjustment vector is applied to the control sequence update direction for vector modification to obtain a feasible direction component on the constraint boundary tangent plane, comprising: An orthogonalization calculation process is performed on the constraint boundary normal vector corresponding to the projection adjustment vector to obtain a unit normal vector set of the constraint boundary; An inner product operation process is performed on the control sequence update direction and the unit normal vector set to obtain a projection coefficient of the update direction on each constraint boundary normal; A linear combination process is performed on the unit normal vector set based on the projection coefficient to obtain a normal component that needs to be removed from the update direction; A vector subtraction operation process is performed on the control sequence update direction by subtracting the normal component to obtain a tangent component that remains in the constraint boundary tangent space; A secondary verification process is performed on the satisfaction of the constraint condition according to the tangent component to obtain a feasible direction component on the constraint boundary tangent plane that has been verified; An adaptive adjustment process is performed on the step length parameter based on the feasible direction component to obtain a feasible control sequence update direction that guarantees convergence.

6. A process optimization based simultaneous biaxial stretching film production control system characterized in that, The process optimization-based synchronous biaxial stretching film production control system for implementing the process optimization-based synchronous biaxial stretching film production control method according to any one of claims 1-5, comprising: An identification module configured to collect and perform subspace identification on melt temperature, longitudinal stretching speed, transverse stretching speed, heating zone temperature, and tension data of the BOPA film production line to obtain a state space prediction model; A combination module configured to perform multi-objective weighted combination on film thickness deviation, width deviation, optical performance index, and energy consumption parameter based on the state space prediction model to obtain a performance optimization function; A processing module configured to perform constraint boundary setting based on the performance optimization function to obtain a quadratic programming constraint set; A solving module configured to input the quadratic programming constraint set into a gradient projection algorithm to perform rolling horizon optimization solving to obtain an optimal control sequence, and apply the first control vector to the longitudinal and transverse stretching actuators.

7. A process optimization based simultaneous biaxial stretching film production control apparatus, characterized by, A memory and a processor, the memory storing a computer program capable of running on the processor, and the processor implements the process optimization-based synchronous biaxial stretching film production control method according to any one of claims 1-5 when executing the computer program. A memory and a processor, the memory storing a computer program capable of running on the processor, and the processor implements the process optimization-based synchronous biaxial stretching film production control method according to any one of claims 1-5 when executing the computer program.

8. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program, when executed by a processor, causes the processor to perform the process-optimization-based simultaneous biaxial stretching film production control method of any one of claims 1 to 5.

Images