A method and system for optimizing the control of the preparation process of multi-layer composite materials based on big data

Abstract

The present invention discloses a method and system for optimizing the process control of preparing multi-layer composite materials based on big data, belonging to the field of process optimization. The method includes: installing a sensor cluster composed of various sensors, collecting the data generated during the operation of a multi-layer co-extrusion and lamination machine, and uploading the data for processing; preprocessing the collected data, mapping it to a unified coordinate system grid after aligning the timestamps, and constructing a sensor historical database; analyzing the data, constructing a dynamic model of key indicators and an optimization control algorithm; according to the real-time data transmitted back by the sensor cluster, calculating a predicted value in combination with the optimization control algorithm, comparing the predicted value with the real-time data of the sensor, and obtaining an optimized operation combination. The present invention greatly improves the production efficiency through automated optimization operations.

Description

Technical Field

[0001] The present invention relates to the field of process optimization, and particularly to a method and system for optimizing the control of the preparation process of multi-layer composite materials based on big data. Background Art

[0002] Multi-layer composite materials are widely used in fields such as packaging and construction. The traditional production process of multi-layer composite materials mainly uses a film laminating machine for production, but there are some problems. First, the mechanical structure of the film laminating machine is relatively fixed, and it is difficult to improve the production efficiency of composite materials by modifying the film laminating machine. Only skilled workers can adjust the various parameters of the film laminating machine according to the produced finished products. However, factors such as temperature changes, pressure distribution, and film thickness evolution in the film laminating machine result in poor product quality and uneven film thickness. The traditional process lacks effective optimization control means and cannot dynamically adjust production parameters according to real-time data, resulting in high energy consumption. To solve the above problems, a method capable of automatically optimizing process control based on the real-time production data of the film laminating machine is needed. Summary of the Invention

[0003] One of the purposes of the present invention is to provide a method for optimizing the control of the preparation process of multi-layer composite materials based on big data, so as to solve the problems of uneven film thickness and high energy consumption in the production process of multi-layer composite materials in the prior art.

[0004] The present invention is realized by the following technical solutions. A method for optimizing the control of the preparation process of multi-layer composite materials includes the following steps: S1. Install a sensor cluster composed of multiple sensors on a multi-layer co-extrusion film laminating machine, collect the data generated during operation, and upload the collected data to a backend processing device for processing; S2. The backend processing device preprocesses the data collected by the sensor cluster, aligns the time stamps of the preprocessed data, maps them to a unified coordinate system grid, integrates and constructs a sensor historical database; S3. Analyze the data in the sensor historical database, construct a key index dynamic model, and based on the key index dynamic model, obtain an optimization control algorithm; S4. According to the real-time data transmitted back by the sensor cluster, calculate the predicted value in combination with the optimization control algorithm, compare the predicted value with the real-time sensor data, dynamically adjust the parameters of the key index dynamic model according to the comparison result, and dynamically adjust the production parameters according to the comparison result to obtain an operation combination, and adjust the production parameters in the process flow based on the operation combination.

[0005] Further, the sensor cluster is constructed through the following sub-steps: S11. Select different types of sensors according to the main factors affecting the forming of multi-layer composite materials, and collect data, including: temperature, pressure, the conveying speed and extrusion speed of the material, the thickness and uniformity of the composite material, the sealing tightness and blistering condition at the edge of the composite material, and the tension of the material during conveying; S12. Use temperature sensors to collect temperature data of the pressure roller, die, and the temperature of the material before entering the pressure roller, pressure sensors to collect data on the pressure distribution and pressure change trend of the pressure roller, speed sensors to collect the conveying speed and extrusion speed of the material, thickness measurement sensors to monitor the thickness of the composite material to ensure the uniformity of the extruded material, infrared sensors to monitor the sealing tightness and blistering condition at the edge, and tension sensors to monitor the tension of the material during conveying; S13. Install the different types of sensors at the corresponding positions of the multi-layer co-extrusion casting machine. The different types of sensors transmit the collected data to the network end in real time through the Internet of Things port. Integrate the data of different types of sensors at the network end to achieve synchronization of sensor data, and send the synchronized data to the backend processing device.

[0006] Further, the preprocessing includes: denoising, complementing, and normalizing the data. Among them, the window sliding average method is used to perform sliding average filtering on the time series data, the wavelet decomposition is used to remove the high-frequency noise of the instantaneous jitter signal of the tension sensor and retain the low-frequency signal, and through outlier detection, the data outside the range of the mean ± 3 times the standard deviation is regarded as an outlier and excluded; after denoising, the data is repaired and complemented, and the linear interpolation method is used to repair the data missing caused by the short-term failure of the sensor or transmission loss; finally, the Z-Score normalization is used to normalize the data.

[0007] Further,

[0008] The dynamic model of key indicators consists of three composite material influencing factor equations with a coupling relationship, including: the temperature index equation used to describe the influence of temperature change on the material, as shown in the following formula:

[0009] , where, is the material density, kg / m 3 ; is the specific heat capacity, J / (kg·K); k is the thermal conductivity, W / (m·K); Q ext is the external heat source, W / m 3 ; v is the flow velocity field of the composite material, m / s; is the gradient of the temperature field; the pressure index equation used to describe the influence of pressure change on the material, as shown in the following formula:

[0010] , where γ’ is the shear rate, which is a scalar measure used to describe the rate of a fluid during shearing; is the viscosity function related to the shear rate, which is used to represent how the viscosity changes according to the shear rate, is the velocity gradient tensor, is the pressure gradient; the index equation used to describe the thickness change of the finished composite material is shown as follows:

[0011] , where h is the thickness of the material, is the Laplace operator, is the product of the material thickness and the flow velocity of the composite material, representing the amount of material passing through a certain cross-sectional area per unit time, is the expansion rate of the material thickness with the flow, representing the expansion or contraction of the material thickness during the flow, Q extrusion is the extruder flow rate, m³ / s; A is the cross-sectional area of the composite material in the transverse direction, m 2 ; in the above three equations, t represents time, represents the rate of change of temperature T with respect to time t, that is, how the temperature changes at different time points; is the rate of change of the velocity field with time; is the rate of change of the material thickness with time; in the above formula, is the partial derivative.

[0012] Furthermore, the viscosity function is represented by the power-law model, and the power-law model is shown as follows:

[0013] , where K is the consistency coefficient, Pa·s; n is the flow index, dimensionless.

[0014] Furthermore, the optimal control algorithm includes an objective function and constraint conditions. Among them, the objective function is:

[0015] , where H is the prediction horizon; is the thickness function at the spatial point (x, y) and time ; is the integration variable, which varies in the range [t, t + H]; h target is the target thickness of the composite material, mm; w is the weight coefficient; is the time-varying control variable; is the control variable; the constraint conditions are set based on the production situation, including physical constraint conditions, which are used to characterize the limiting conditions directly determined by material properties or process requirements; control variable constraint conditions, which are used to characterize the physical limits of the control parameters of the multi-layer co-extrusion casting machine; process constraint conditions, which are used to characterize the limiting conditions related to the performance of the finished composite material.

[0016] Further, step S4 further includes a closed-loop control step. The sensor cluster continuously collects new data. The sensor cluster collects data every Δt time, inputs the latest data into the error function, obtains updated parameters through the gradient descent method, and dynamically calibrates the key indicator dynamic model with the updated parameters.

[0017] Further, the error function is shown as follows:

[0018] , where θ is the model parameter, dimensionless; is the initial value of the model parameter, and N is the total number of samples; is the model prediction value, indicating that the temperature at the point (x i , y i ) and time t depends on the model parameter θ. is the actual measured value, indicating the actual data of the sensor at the point (x i , y i ) and time t; λ is the weight of the regularization term, used to control the influence degree of the regularization term on the total error. By iteratively adjusting the parameter θ, the error is gradually reduced. The parameter update mechanism of J(θ) is expressed as:

[0019] , where θ k+1 is the parameter value after the (k + 1)-th iteration; θ k is the parameter value of the current k-th iteration; α is the control parameter, used to control the step size of each adjustment; is the gradient of the error function at the current parameter.

[0020] On the other hand, the present invention provides an optimization system for controlling the preparation process of a multi-layer composite material based on big data, which includes a multi-layer co-extrusion and lamination machine and a backend processing device. A sensor cluster module is installed on the multi-layer co-extrusion and lamination machine; the backend processing device is used to process the data collected by the sensor cluster and adjust the production parameters of the multi-layer co-extrusion and lamination machine according to the results of data processing. The process control optimization system further includes a data preprocessing module, a data analysis module, and an optimization control module. Among them, the data preprocessing module is configured to preprocess the data collected by the sensor cluster, align the time stamps of the preprocessed data, map them to a unified coordinate system grid, integrate and construct a sensor historical database; the data analysis module is connected to the data preprocessing module and is configured to analyze the data in the sensor historical database, construct a dynamic model of key indicators, and design an optimization control algorithm based on the dynamic model of key indicators; the optimization control module is connected to the data analysis module and the sensor cluster module and is configured to calculate a predicted value according to the real-time data transmitted back by the sensor cluster, combine the optimization control algorithm, compare the predicted value with the real-time sensor data, and dynamically adjust the parameters of the dynamic model of key indicators according to the comparison result.

[0021] Further, the sensor cluster module is configured to install the different types of sensors at corresponding positions on the multi-layer co-extrusion and lamination machine. The different types of sensors transmit the collected data to the network end in real time through the Internet of Things port. At the network end, the data of different types of sensors are integrated to achieve synchronization of sensor data, and the synchronized data is sent to the backend processing device; a temperature sensor is used to collect the temperature data of the pressure roller, the die temperature, and the temperature of the material before entering the pressure roller, a pressure sensor is used to collect the pressure distribution and pressure change trend data of the pressure roller, a speed sensor is used to collect the conveying speed and extrusion speed of the material, a thickness measurement sensor monitors the thickness of the composite material to ensure the uniformity of the extruded material, an infrared sensor monitors the sealing tightness and foaming conditions at the edge, and a tension sensor monitors the tension of the material during the conveying process.

[0022] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0023] 1. By constructing a sensor cluster to collect the operation data of the multi-layer co-extrusion and lamination machine in real time, the present invention can monitor the key parameters in the production process in real time, construct a dynamic model of key indicators by analyzing historical data, design an optimization control algorithm based on this model, and dynamically adjust the production parameters according to the real-time data, thus effectively reducing energy consumption and solving the problem that the prior art cannot dynamically optimize production parameters.

[0024] 2. The present invention greatly improves the production efficiency through automated optimization operations. At the same time, the present invention can not only optimize the production process of the multi-layer co-extrusion lamination machine, but also further optimize other equipment such as the cooling mechanism, preheating mechanism, traction mechanism, and winding mechanism, and can comprehensively improve the production quality and efficiency of multi-layer composite materials. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, form a part of this application, and do not constitute a limitation to the embodiments of the present invention. In the drawings:

[0026] Figure 1 It is a flowchart of the method provided in Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Components of the embodiments of the present invention usually described and illustrated herein can be arranged and designed in various different configurations.

[0028] Embodiment 1

[0029] In this embodiment, a method for optimizing the control of a multi-layer composite material preparation process based on big data is provided. The multi-layer composite material in this embodiment has a three-layer structure, namely a surface composite material, which is obtained by combining a BOPP film and an aluminized film with solvent-free composite A and B adhesives to obtain a surface film (the surface film can also be a single BOPP film). Then, the surface film (composite film) is combined with white non-woven fabric by PP cast lamination to obtain a surface composite material. The middle layer is made of PP as the main material and a coupling agent is added, and is hot-pressed with non-woven fabric, paper or plastic sheet by a plain roller or various patterned rollers to obtain the middle layer. The inner layer film can have various types, which can be a printed BOPP film or a PP aluminized film. At the same time, the inner layer film is not necessarily required, and it can be present or absent. After obtaining the surface composite material, the middle layer, and the inner layer film, a multi-layer lamination composite process is used to form a multi-layer composite material for product outer packaging at one time.

[0030] In this embodiment, by considering the characteristics of the multi-layer co-extrusion film laminating machine used in the multi-layer film laminating composite process, the key parameters to be monitored are determined. Corresponding sensors are installed on the multi-layer co-extrusion film laminating machine to collect sensor data, and an optimized control algorithm is constructed based on the data collected during production. Then, during the subsequent production process, the production process of the multi-layer co-extrusion film laminating machine is regulated according to the real-time data collected by the sensors in combination with the constructed optimized control algorithm, so as to realize the process optimization of the multi-layer composite material. Figure 1 The flowchart of the control optimization method in this embodiment is shown. It can be seen from the figure that this embodiment includes the following steps:

[0031] Step 1: First, construct a sensor cluster composed of multiple sensors, and install this sensor cluster on the multi-layer co-extrusion film laminating machine to collect data when the multi-layer co-extrusion film laminating machine is working, and upload the collected data to the backend processing system.

[0032] Specifically, constructing the sensor cluster includes,

[0033] 1) After analyzing the key data in the multi-layer film laminating composite process, determine the key parameters to be monitored. The parameters affecting material forming mainly include: temperature, pressure, the conveying speed and extrusion speed of the material, the thickness and uniformity of the composite material, the sealing tightness and blistering condition at the edge of the composite material, and the tension of the material during the conveying process.

[0034] 2) According to the above main parameters affecting material forming, select appropriate sensors. Use temperature sensors to collect data on the temperature of the pressure roller, die temperature, and the temperature of the material before entering the pressure roller; use pressure sensors to collect data on the pressure distribution and pressure change trend of the pressure roller; use speed sensors to collect the conveying speed and extrusion speed of the material; use thickness measurement sensors to monitor the thickness of the composite material to ensure the uniformity of the extruded material; use infrared sensors to monitor the sealing tightness and blistering condition at the edge; use tension sensors to monitor the tension of the material during the conveying process.

[0035] 3) After selecting the sensors, install the sensors at key positions of the film laminating machine, such as the pressure roller, die, material inlet, outlet, etc.

[0036] 4) The data collected by the sensors is transmitted to the network end in real time through the Internet of Things (IoT) interface. At the network end, data from multiple sources (temperature sensors, pressure sensors, speed sensors, etc.) is integrated to achieve precise data synchronization, and the synchronized data is sent to the backend processing system.

[0037] Step 2: After the back-end processing system receives the data from the sensor cluster, it first processes the data such as denoising, complementing, and normalizing to ensure the data quality. Then, it aligns the timestamps of the processed data, maps them to a unified coordinate system grid, and integrates and constructs a sensor historical database.

[0038] Specifically, the processing of denoising, complementing, and normalizing the data may include:

[0039] First, use the window sliding average to smooth the short-term fluctuations. Perform sliding average filtering on the time-series data, remove high-frequency noise from the instantaneous jitter signal of the tension sensor through wavelet decomposition, and retain the low-frequency useful signal. Detect and remove outliers. If a data point exceeds the range of the mean ± 3 times the standard deviation, it is regarded as an outlier and removed.

[0040] Then, use the linear interpolation method to repair and complement the denoised data to repair the missing data caused by short-term sensor failures or transmission losses.

[0041] Finally, use Z-Score normalization to normalize the data, so that the sensor data after denoising, complementing, and normalizing can be directly used for model training and optimal control.

[0042] Step 3: Analyze the data in the sensor historical database. By analyzing the data in the historical database, obtain the three key indicators that have the greatest impact on the composite material, namely: temperature change, pressure distribution, and film thickness evolution. Establish a dynamic model based on the coupling relationship of these three key indicators, and design an optimal control algorithm based on the constructed key indicator dynamic model.

[0043] Specifically, in order to accurately describe the production process of the multi-layer co-extrusion film laminating machine, the dynamic change laws of these three key physical quantities, temperature, pressure, and film thickness, need to be expressed by mathematical equations. The core of the key indicator dynamic model is to construct a coupling equation set that can reflect the actual production process by combining the data collected during the production process and the material properties. The coupling equation set may specifically include the following content:

[0044] 1) When the composite material passes through components such as heating rollers and extruders, the temperature will change, and at the same time, the temperature distribution will also directly affect the fluidity, adhesion effect, and final product performance of the material. The heat transfer of the material at different positions and different times after entering the multi-layer co-extrusion film laminating machine can be described by an equation. The following partial differential equation can be used to describe it:

[0045] ,

[0046] Among them, is the material density, kg / m 3 ; is the specific heat capacity, J / (kg·K); k is the thermal conductivity, W / (m·K); Q ext is the external heat source, W / m 3 ; v is the flow velocity field of the composite material, m / s; is the gradient of the temperature field, that is, the rate of change of temperature in space, which is a vector indicating how the temperature changes in each direction.

[0047] It should be noted that in this equation, is the energy storage term, which is used to represent the rate of change of the thermal energy of the composite material per unit volume with time; is the heat conduction term, which is used to represent the heat transferred through heat conduction; is the external heat source term, which is used to represent the heat input from external devices such as heating rollers and extruders; is the convection term, which is used to represent the heat transfer caused by the extrusion and transmission of materials in a multi-layer co-extrusion film laminating machine.

[0048] 2) The composite material will flow during the lamination process using a multi-layer co-extrusion film laminating machine, and the viscoelastic properties of the material will directly affect the pressure distribution and the interfacial bonding strength. The pressure distribution during the bonding process of the material in the multi-layer co-extrusion film laminating machine can be described by an equation. The following partial differential equation can be used to describe it:

[0049] ,

[0050] where, is the shear rate, , which is a scalar measure used to describe the rate during the shear process of the fluid; is the viscosity function related to the shear rate, which is used to represent how the viscosity changes according to the shear rate. This viscosity function can be represented by a power-law model. The specific power-law function can be , where K is the consistency coefficient, Pa·s; n is the flow index. is the flow velocity field of the composite material, which is a vector function used to describe the flow velocity of the material at each point in space over time. For example, in three-dimensional space, it can be expressed as , representing the flow velocities along the x, y, and z directions respectively. is the velocity gradient tensor, which is used to represent the rate of change of velocity in all directions in space and describes how the velocity of the material changes in each direction.

[0051] It should be noted that in this formula is the inertia term, which is used to represent the momentum change brought about by the acceleration of the composite material in the multi-layer co-extrusion film laminating machine; The pressure gradient term is used to represent the lateral spreading of the composite material caused by being extruded by the pressure roller in the multi-layer coextrusion casting machine. The negative sign in front of it is because the negative sign indicates the direction of the pressure's effect on fluid flow, that is, the fluid always flows from the high-pressure area to the low-pressure area. Therefore, the pressure gradient will push the fluid from the high-pressure area to the low-pressure area. The purpose of the negative sign is to ensure that the direction of fluid movement is opposite to the direction of the pressure difference. The viscous force term is used to represent that when the viscosity of the material is high (such as under-heated PP), the flow is slow and a greater pressure is required to drive it.

[0052] 3) When the composite material is laminated using a multi-layer coextrusion casting machine, the thickness of the final finished material is jointly determined by the extrusion volume, material flow rate, and lamination process, which directly affects the appearance and mechanical properties of the final finished product. The equation can be used to predict how the thickness of the material changes with the extrusion speed and pressure. The following partial differential equation can be used to describe it:

[0053] ,

[0054] where h is the thickness of the material; is the rate of change of the material thickness with time, which is used to reflect how the material changes due to factors such as flow, extrusion, and compression during the lamination process. is the Laplace operator, is the product of the material thickness and the flow rate, representing the amount of material passing through a certain cross-sectional area per unit time, is the expansion rate of the material thickness with flow, representing the expansion or contraction of the material thickness during the flow process, reflecting how the flow rate and material thickness jointly affect the flow and distribution of the material. Due to factors such as lamination and flow, the thickness of the material may change spatially. Q extrusion is the extrusion machine flow rate, m³ / s; A is the cross-sectional area of the composite material, m 2 . is the contribution of the material volume provided by the multi-layer coextrusion casting machine per unit time to the thickness.

[0055] It should be noted that in the above three equations t represents time, represents the rate of change of the temperature T with respect to time t , that is, how the temperature changes at different time points; is the rate of change of the velocity field with time; is the rate of change of the material thickness with time; in the above formula, It means partial derivative, which reflects the rate of change of a function along the positive direction of the coordinate axis and is read as "partial". The coupling relationship among the above three equations is temperature-viscosity-flow. An increase in temperature will reduce the material viscosity, thereby affecting the pressure distribution and flow rate. For example, when the heating is insufficient, the material viscosity increases, resulting in the need for greater pressure on the pressure roller to spread evenly. The change in flow rate will directly change the change in film thickness, and the pressure distribution will affect the contact tightness of the materials between different layers, thereby affecting the composite strength.

[0056] The mutual coupling among these three partial differential equations jointly constitutes a dynamic model of key indicators, enabling the quantification of the interaction among temperature, pressure, and thickness. By combining real-time sensor data, these equations can be solved in real time. Through optimization algorithms, based on the results of real-time solution of the dynamic model of key indicators, the future state can be predicted, and the optimal control parameters (such as heating power, extrusion speed) can be found, ultimately achieving high-quality and low-energy consumption production.

[0057] In this embodiment, the purpose of the optimization control algorithm is to dynamically adjust the control parameters (such as heating power, extrusion speed, pressure roller pressure) to achieve the optimal product quality (such as uniform thickness) and the lowest energy consumption under the premise of meeting production constraints.

[0058] The implementation process of the optimization control algorithm highly depends on the prediction ability of the dynamic model of key indicators because the optimization control algorithm needs to dynamically predict the future state based on the results of real-time solution of the model, and then make the best decision in advance.

[0059] Specifically, the construction of the optimization control algorithm includes: 1) Starting from the perspectives of improving product quality and controlling production costs, in order to achieve the purpose of minimizing film thickness deviation and minimizing energy consumption, an objective function is constructed. In this embodiment, the objective function can be expressed by the following formula:

[0060] ,

[0061] where H is the prediction horizon, that is, the time window length for the objective function to predict the future; is the thickness function at the spatial point (x, y) and time ; is the integration variable, which varies in the range [t, t + H]; h target is the target thickness of the composite material, in mm; is the time-varying control variable, that is, the process parameter that needs to be adjusted, which can include the heating power (W) of the multi-layer co-extrusion casting machine, the extrusion speed (rpm), the pressure roller pressure (N / m²), etc. is the control variable.

[0062] Let \(w\) be the weight coefficient, which is used to control the trade-off between energy consumption and quality deviation. It can be adjusted according to the actual production situation. The adjustment strategy is as follows: when it is necessary to preferentially reduce energy consumption during production and allow a certain thickness deviation, \(w\) is set to a larger value; when it is necessary to preferentially ensure thickness uniformity during production and allow higher energy consumption, \(w\) is set to a smaller value.

[0063] It should be noted that and are both control variable functions. Among them, usually represents the control input or operation parameter at a specific time \(t\), and is usually for the control at a certain moment or stage. represents a control variable that changes with time and usually appears in the objective function in integral form, indicating the control input within the entire time interval and is used to show the consideration of the control process within a time interval.

[0064] 2) Set constraint conditions for the objective function based on the actual production situation. In this implementation, the constraint conditions include physical constraint conditions, which are used to characterize the limiting conditions directly determined by material properties or process requirements; control variable constraint conditions, which are used to characterize the physical limits of various control parameters of the multi-layer co-extrusion casting machine; and process constraint conditions, which are used to characterize the limiting conditions related to the performance of the finished composite material.

[0065] For example, in this embodiment, the physical constraint conditions include temperature constraint and pressure constraint. Among them, the temperature constraint is: \(T\) min \(\leq T(x,y,t)\leq T\) max , and the pressure constraint is: \(P\) min \(\leq P(x,y,t)\leq P\) max , the control variable constraint is: \(u\) min \(\leq u(t)\leq u\) max , and the process constraint is: \(h\) min \(\leq h(t)\leq h\) max .

[0066] Among them, \(T(x,y,t)\) is the temperature field, representing the temperature distribution of the material during the production process; \(T\) min is the minimum temperature constraint, indicating that the temperature cannot be lower than this value. A temperature lower than this may cause the material properties not to meet the requirements or instability during the processing. \(T\) max is the maximum temperature constraint, indicating that the temperature cannot exceed this value. A temperature higher than this may cause the material to be overheated, damaged or result in unqualified products. The temperature constraint is to ensure that during the production process, the temperature of the material during heating or cooling is maintained within a reasonable range to guarantee material properties and processing stability.

[0067] P(x, y, t) is the pressure field, representing the pressure distribution of the material during the production process. P min is the minimum pressure constraint, indicating that the pressure of the material during production cannot be lower than this value. Too low pressure may cause uneven material flow or poor lamination effect. P max is the maximum pressure constraint, indicating that the pressure cannot exceed this value. Excessive pressure may cause equipment damage or excessive material deformation. The pressure constraint is to ensure that during the entire production process, the pressure of the material is maintained within an appropriate range to ensure production stability and product quality.

[0068] u(t) is the control variable, representing a certain parameter controlled during the production process. u min is the minimum value constraint of the control variable, indicating that this control parameter cannot be lower than this value. Usually, it is to prevent the equipment from operating at too low efficiency or in an unsafe state. u max is the maximum value constraint of the control variable, indicating that this control parameter cannot be higher than this value. Usually, it is to prevent equipment overload or unstable control process. The control variable constraint is to ensure that during the production process, all control parameters (such as extrusion speed, pressure control, etc.) are within their physical limit ranges to avoid equipment damage or improper operation.

[0069] h(t) is the material thickness, representing the thickness of the composite material during the production process. h min is the minimum thickness constraint, indicating that the thickness of the material cannot be lower than this value. If the thickness is too small, it may result in insufficient product strength or failure to meet the usage requirements. h max is the maximum thickness constraint, indicating that the thickness of the material cannot exceed this value. If the thickness is too large, it may lead to low production efficiency or non-compliant products. The process constraint is used to ensure that during the production process, the thickness of the final product remains within an acceptable range to meet the performance requirements of the product and production capacity.

[0070] Step 4: According to the real-time data transmitted back by the sensor cluster when the multi-layer co-extrusion casting machine is working, calculate the predicted value through an optimized control algorithm, and compare the predicted value with the real-time data of the sensor. The back-end processing system dynamically adjusts the model parameters according to the comparison result.

[0071] At the same time, dynamically adjust parameters such as the heating power and pressure roller pressure of the multi-layer co-extrusion casting machine according to the comparison result, and find an operation combination that can not only ensure uniform film thickness but also save energy. And adjust the production parameters in the process flow based on this operation combination.

[0072] At the same time, the sensor continues to collect new data. If it is found that the actual thickness still does not meet the standard, immediately start a new round of optimization.

[0073] The sensor cluster collects data every Δt time, inputs the latest data into the error function, and obtains updated parameters through the gradient descent method. The new parameter θ new is substituted into the key index dynamic model to generate more accurate predictions. Based on the updated predictions, the optimization algorithm calculates control instructions to complete closed-loop control.

[0074] Since the key index dynamic model is an idealized model, there are many uncertainties in the actual production process. For example, material parameters may fluctuate due to batch or environmental changes; it is difficult to accurately measure factors such as the actual temperature distribution of the heating roller and environmental heat dissipation; factors such as equipment aging and environmental temperature and humidity changes are not modeled. Therefore, through real-time sensor data, the model parameters are dynamically calibrated to make the key index dynamic model more conform to the actual production state.

[0075] In this embodiment, an error function is constructed to dynamically calibrate the model parameters according to real-time sensor data. This error function sets a group of model parameters θ to make the predicted value of the key index dynamic model as close as possible to the sensor measurement value.

[0076] In this embodiment, the error function is shown as the following formula:

[0077] ,

[0078] where θ is the model parameter, which can include k, η, Q ext etc., dimensionless; is the initial value of the model parameter, N is the total number of samples; is the model predicted value, indicating that the temperature at the point (x i , y i ) and time t depends on the model parameter, is the actual measured value, indicating the actual data of the sensor at the point (x i , y i ) and time t; λ is the weight of the regularization term, which is used to control the influence degree of the regularization term on the total error. The model parameters are dynamically calibrated through the above error function to make the prediction result as close as possible to the actual measured value, and at the same time, the regularization term is used to prevent the model from overfitting, making the final result more accurate.

[0079] The parameter update through the above error function specifically includes: by iteratively adjusting the model parameter θ, gradually reducing the error, and the parameter update mechanism of J(θ) can be expressed by the following formula:

[0080] , where θ k+1 is the parameter value after the (k + 1)-th iteration; θ kis the parameter value for the current k-th iteration; α is a control parameter used to control the step size of each adjustment; is the gradient of the error function at the current parameter.

[0081] Embodiment 2

[0082] In this embodiment, an optimization system for controlling the preparation process of multi-layer composite materials based on big data is disclosed. The system includes a multi-layer co-extrusion lamination machine and a backend processing device. A sensor cluster module is installed on the multi-layer co-extrusion lamination machine. The backend processing device is used to process the data collected by the sensor cluster and adjust the production parameters of the multi-layer co-extrusion lamination machine according to the results of data processing.

[0083] The control optimization system further includes a data preprocessing module, a data analysis module, and an optimization control module.

[0084] Among them, the data preprocessing module is configured to preprocess the data collected by the sensor cluster, align the time stamps of the preprocessed data, map them to a unified coordinate system grid, and integrate and construct a sensor historical database.

[0085] Specifically, the data preprocessing module includes: denoising, complementing, and normalizing the data. Among them, the window sliding average method is used to perform sliding average filtering on time series data, wavelet decomposition is used to remove high-frequency noise from the instantaneous jitter signal of the tension sensor and retain the low-frequency signal, and through outlier detection, data outside the range of the mean ± 3 times the standard deviation is regarded as an outlier and excluded; after denoising, the data is repaired and complemented, and the linear interpolation method is used to repair the data missing due to short-term sensor failures or transmission losses; finally, Z-Score normalization is used to normalize the data.

[0086] The data analysis module is connected to the data preprocessing module and is configured to analyze the data in the sensor historical database, construct a dynamic model of key indicators, and design an optimization control algorithm based on the dynamic model of key indicators.

[0087] Specifically, the dynamic model of key indicators consists of three composite material influencing factor equations with a coupling relationship, including:

[0088] The temperature index equation used to describe the influence of temperature change on the material is as follows:

[0089] ,

[0090] Among them, is the material density, kg / m 3 ; is the specific heat capacity, J / (kg·K); k is the thermal conductivity, W / (m·K); Qext is an external heat source, W / m 3 ; v is the flow velocity field of the composite material, m / s; is the gradient of the temperature field, that is, the rate of change of temperature in space, which is a vector indicating the change of temperature in each direction.

[0091] The pressure index equation used to describe the influence of pressure change on the material is shown as follows:

[0092] ,

[0093] where, is the shear rate, , a scalar measure used to describe the rate of the fluid during shear; is the viscosity function related to the shear rate, used to represent how the viscosity changes according to the shear rate. This viscosity function can be represented by the power-law model, and the specific power-law function can be , K is the consistency coefficient, Pa·s; n is the flow index. is the velocity field, which is a vector function used to describe the flow velocity of the material at each point in space over time. For example, in three-dimensional space, it can be represented as , respectively represent the flow velocities in the x, y, and z directions. is the velocity gradient tensor, used to represent the rate of change of velocity in all directions of space, describing how the velocity of the material changes in each direction.

[0094] The index equation used to describe the thickness change of the finished composite material is shown as follows:

[0095] ,

[0096] where, h is the thickness of the material; is the rate of change of the material thickness over time, used to reflect how the material changes due to factors such as flow, extrusion, and compression during the lamination process. is the Laplace operator, is the product of the material thickness and the flow velocity, representing the amount of material passing through a certain cross-sectional area per unit time, is the expansion rate of the material thickness with flow, indicating the expansion or contraction of the material thickness during the flow process, reflecting how the flow velocity and the material thickness jointly affect the flow and distribution of the material. Due to factors such as lamination and flow, the thickness of the material may change in space. Q extrusion is the extruder flow rate, m³ / s; A is the cross-sectional area of the composite material, m 2 . is the contribution of the material volume provided by the multi-layer co-extrusion coating machine per unit time to the thickness. In the above three equations, t represents time, represents the rate of change of temperature T with respect to time t, that is, how the temperature changes at different time points; is the rate of change of the velocity field with time; is the rate of change of the material thickness with time; in the above formula, is the partial derivative.

[0097] The optimization control algorithm includes an objective function and constraint conditions. Among them, the objective function is:

[0098] ,

[0099] where H is the prediction horizon, that is, the length of the time window for the objective function to predict the future; is the thickness function at the spatial point (x, y) and time ; is the integration variable, which varies in the range [t, t + H]; h target is the target thickness of the material, in mm; is the time-varying control variable, that is, the process parameter that needs to be adjusted, which can include the heating power (W) of the multi-layer co-extrusion casting machine, the extrusion speed (rpm), the pressure roller pressure (N / m²), etc. is the control variable.

[0100] The constraint conditions are set based on the production situation, including physical constraint conditions, which are used to characterize the limiting conditions directly determined by material properties or process requirements; control variable constraint conditions, which are used to characterize the physical limits of the control parameters of the multi-layer co-extrusion casting machine; and process constraint conditions, which are used to characterize the limiting conditions related to the performance of the finished composite material.

[0101] The optimization control module is connected to the data analysis module and the sensor cluster module, and is configured to calculate a predicted value according to the real-time data transmitted back by the sensor cluster, combine the optimization control algorithm, compare the predicted value with the real-time sensor data, and dynamically adjust the parameters of the key index dynamic model according to the comparison result.

[0102] Specifically, the optimization control module also includes a closed-loop control. According to the new data continuously collected by the sensor cluster, the latest data is input into the error function, and updated parameters are obtained through the gradient descent method. The key index dynamic model is dynamically calibrated through the updated parameters.

[0103] The error function is shown as follows:

[0104] ,

[0105] where θ is the model parameter, which can include k, η, Q ext etc., dimensionless; is the initial value of the model parameters, and N is the total number of samples; is the model prediction value, indicating that the temperature at the point (x i , y i ) and time t depends on the model parameters. is the actual measured value, representing the actual data of the sensor at the point (x i , y i ) and time t; λ is the weight of the regularization term, used to control the influence degree of the regularization term on the total error. The model parameters are dynamically calibrated through the above error function to make the prediction result as close as possible to the actual measured value, and at the same time, the regularization term is used to prevent the model from overfitting, making the final result more accurate.

[0106] By iteratively adjusting the parameter θ, the error is gradually reduced, and the parameter update mechanism of J(θ) is expressed as:

[0107] ,

[0108] where θ k+1 is the parameter value after the (k + 1)-th iteration; θ k is the parameter value of the current k-th iteration; α is the control parameter, used to control the step size of each adjustment; is the gradient of the error function at the current parameter.

[0109] The specific implementation manners described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is only the specific implementation manners of the present invention and is not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for controlling and optimizing the preparation process of multilayer composite materials based on big data, characterized in that: The process control optimization method comprises: S1. Install a sensor cluster consisting of multiple sensors on the multi-layer co-extrusion laminating machine to collect data generated during operation, and upload the collected data to the back-end processing equipment for processing; S2, the back-end processing equipment pre-processes the data collected by the sensor cluster, aligns the timestamps of the pre-processed data, maps them to a unified coordinate system grid, and integrates and builds a sensor history database; S3. Analyze the data in the sensor history database, construct a key indicator dynamic model, and obtain an optimization control algorithm based on the key indicator dynamic model; S4, according to the real-time data transmitted back by the sensor cluster, combined with the optimization control algorithm, calculate the predicted value, Compare the predicted value with the real-time sensor data, dynamically adjust the parameters of the key indicator dynamic model according to the comparison results, dynamically adjust the production parameters according to the comparison results, obtain the operation combination, and adjust the production parameters in the process flow based on the operation combination; The key indicator dynamic model is composed of three composite material influencing factor equations with coupling relationships, including: The temperature index equation used to describe the effect of temperature changes on materials is shown below: ， in, is the material density, kg / m 3 ; is the specific heat capacity, J / (kg·K); k is the thermal conductivity, W / (m·K); Q ext is the external heat source, W / m 3 ; v is the flow velocity field of the composite material, m / s; is the gradient of the temperature field; The pressure index equation used to describe the effect of pressure changes on materials is shown below: ， Among them, γ' is the shear rate, which is a scalar measure used to describe the rate of fluid during shear process; is the shear rate-dependent viscosity function, which is used to indicate that the viscosity changes according to the shear rate. is the velocity gradient tensor, is the pressure gradient; The equation used to describe the thickness variation of the finished composite material is as follows: ， Where h is the thickness of the material, is the Laplace operator, It is the product of the material thickness and the flow velocity of the composite material, indicating the amount of material passing through a certain cross-sectional area per unit time. Q is the expansion rate of material thickness with flow, indicating the expansion or contraction of material thickness during flow. extrusion is the extruder flow rate, m³ / s; A is the cross-sectional area of ​​the composite material, m 2 ; In the above three equations, t represents time, It indicates the rate of change of temperature T relative to time t, that is, how the temperature changes at different time points; is the rate of change of velocity field with time; is the rate at which the material thickness changes with time; in the above formula, is the partial derivative.

2. The method for controlling and optimizing the preparation process of multi-layer composite materials based on big data according to claim 1, characterized in that: The sensor cluster is constructed by the following sub-steps: S11. According to the main factors affecting the molding of multi-layer composite materials, different types of sensors are selected to collect data. The data include: temperature, pressure, material conveying speed and extrusion speed, thickness and uniformity of the composite material, sealing tightness and blistering at the edge of the composite material, and tension of the material during conveying; S12, using a temperature sensor to collect data on the temperature of the pressure roller, the temperature of the mold, and the temperature of the material before it enters the pressure roller, a pressure sensor to collect data on the pressure distribution and pressure change trend of the pressure roller, a speed sensor to collect data on the conveying speed and extrusion speed of the material, a thickness measurement sensor to monitor the thickness of the composite material to ensure the uniformity of the extruded material, an infrared sensor to monitor the sealing tightness and blistering at the edge, and a tension sensor to monitor the tension of the material during the conveying process; S13, installing the different types of sensors at corresponding positions of the multi-layer co-extrusion laminating machine, and the different types of sensors transmit the collected data to the network end in real time through the Internet of Things port, Integrate data from different types of sensors at the network end, synchronize sensor data, and send the synchronized data to the back-end processing device.

3. The method for controlling and optimizing the preparation process of multi-layer composite materials based on big data according to claim 1, characterized in that: The preprocessing includes: denoising, completing and standardizing the data, wherein: The window sliding average method is used to perform sliding average filtering on the time series data. The instantaneous jitter signal of the tension sensor is decomposed by wavelet to remove high-frequency noise and retain low-frequency signals. Through outlier detection, data that exceeds the mean ±3 times the standard deviation is considered as an outlier and removed. After denoising, the data is repaired and supplemented, and linear interpolation is used to repair data missing due to temporary sensor failure or transmission loss; Finally, the data were standardized using Z-Score standardization.

4. The method for controlling and optimizing the preparation process of multi-layer composite materials based on big data according to claim 1, characterized in that: The viscosity function is represented by a power law model, and the power law model is shown in the following formula: ， Where K is the consistency coefficient, Pa·s; n is the flow index, dimensionless.

5. The method for controlling and optimizing the preparation process of multi-layer composite materials based on big data according to claim 1, characterized in that: The optimization control algorithm includes an objective function and constraints, wherein: The objective function is: ， Among them, H is the prediction time domain; is a spatial point (x, y) and time The thickness function at is the integral variable, which varies in the range [t, t+H]; h target is the target thickness of the composite material, mm; w is the weight coefficient; is a time-varying control variable; is the control variable; Constraints are set based on production conditions, including physical constraints, which are used to characterize constraints directly determined by material properties or process requirements; control variable constraints, which are used to characterize the physical limits of various control parameters of the multi-layer co-extrusion laminating machine; and process constraints, which are used to characterize constraints related to the performance of finished composite materials.

6. The method for controlling and optimizing the preparation process of multi-layer composite materials based on big data according to claim 1, characterized in that: The step S4 also includes a closed-loop control step, in which the sensor cluster continuously collects new data. The sensor cluster collects data every Δt time, inputs the latest data into the error function, obtains updated parameters through the gradient descent method, and dynamically calibrates the key indicator dynamic model through the updated parameters.

7. The method for controlling and optimizing the preparation process of multi-layer composite materials based on big data according to claim 6, characterized in that: The error function is shown below: ， Among them, θ is a model parameter and is dimensionless; is the initial value of the model parameter, N is the total number of samples; is the model prediction value, which means at point (x i ,y i ) and the temperature at time t depends on the model parameter θ, is the actual measured value, indicating that at point (x i ,y i ) and the actual data of the sensor at time t; λ is the weight of the regularization term, which is used to control the influence of the regularization term on the total error. By iteratively adjusting the parameter θ, the error is gradually reduced. The parameter update mechanism of J(θ) is expressed as: ， Among them, θ k+1 is the parameter value after the k+1th iteration; θ k is the parameter value of the current k-th iteration; α is the control parameter used to control the step size of each adjustment; is the gradient of the error function at the current parameter.

Images