XCPS Crosslinked Polystyrene Intelligent Crosslinking Process Parameter Precision Control System

Abstract

This invention belongs to the field of polymer material manufacturing and intelligent control, specifically an intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene. It includes a data acquisition module for constructing a sensing matrix; a digital twin mirror module for reconstructing the internal temperature field based on heat conduction inverse calculation; a kinetic prediction module for predicting temperature rise peaks and reaction inflection points by combining chemical kinetics and neural networks; an intelligent closed-loop control module for executing advanced damping control; and an efficiency evaluation and optimization module for parameter self-learning. This system achieves advanced sensing and precise control of the core state, improving the crosslinking uniformity and dimensional stability of the board.

Description

Technical Field

[0001] This invention belongs to the field of polymer material manufacturing and intelligent control technology, specifically the XCPS cross-linked polystyrene intelligent cross-linking process parameter precise control system. Background Technology

[0002] Cross-linked polystyrene, as a high-performance polymer insulating material, occupies an irreplaceable position in many high-end fields. Its performance depends on the quality of the cross-linking reaction micro-network construction. The formation of a high-quality network is highly dependent on the precise matching of thermodynamic parameters and chemical kinetic processes. Under ideal working conditions, the heat of reaction can be discharged through the mold, but there are significant technical challenges in the actual production of thick sheets.

[0003] In actual industrial production of thick polystyrene sheets, the extremely low thermal conductivity of polystyrene and the violent exothermic cross-linking reaction create an essential physical contradiction. The heat in the core cannot be conducted in time, leading to heat accumulation and causing an adiabatic temperature rise, which causes the core reaction to enter an uncontrollable self-accelerating state.

[0004] Existing control methods rely on monitoring the surface temperature of the mold for feedback adjustment, which has a serious observation lag and cannot intervene in thermal runaway in time. This leads to quality defects in the sheet material, such as bursting, cavities, excessive internal stress, and low mechanical properties. Existing compromise solutions such as slowing down production reduce production efficiency and damage the long-term reliability of materials, making it difficult to achieve intelligent closed-loop optimization of process parameters.

[0005] To address this, the present invention provides an intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene. Summary of the Invention

[0006] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0007] The technical solution adopted by this invention to solve its technical problem is: the intelligent crosslinking process parameter precise control system for XCPS crosslinked polystyrene described in this invention, the system comprising: The multi-source heterogeneous data real-time acquisition module is used to construct a multi-dimensional sensing matrix within the physical field of the XCPS crosslinking reaction, and to capture real-time data on mold wall temperature, initial surface temperature of material, pressure inside the mold, and initiator concentration. The process digital twin virtual mirror module is connected to the multi-source heterogeneous data real-time acquisition module. It is used to map the material surface monitoring parameters to the material internal three-dimensional spatial grid through the heat conduction inverse calculation algorithm based on the multi-dimensional sensing matrix, and restore the temperature gradient field and pressure field distribution inside the material in the time domain. The nonlinear dynamics deep prediction module, connected to the process digital twin virtual mirror module, is used to identify the reaction self-acceleration inflection point and predict the adiabatic temperature rise path and peak value of the material core within a preset time window using a core temperature rise prediction model based on chemical kinetic mapping. The intelligent closed-loop precision control module is connected to the nonlinear dynamics depth prediction module and the external actuator respectively. It is used to perform advanced damping control on the core of the material by adjusting the heat flow boundary of the mold wall and the pressure inside the mold based on the prediction results of the adiabatic temperature rise path and the peak value of the adiabatic temperature rise. The long-term performance evaluation and optimization module is connected to the feedback loop of the nonlinear dynamics depth prediction module. It is used to collect the quality indicators after the sheet is formed and dynamically update the operator coefficients in the core temperature rise prediction model by calculating the deviation between the actual value and the target value.

[0008] Preferably, the multi-source heterogeneous data real-time acquisition module is configured to construct a multi-dimensional sensing matrix within the physical field of the XCPS crosslinking reaction. The acquisition module includes a sensor group deployed on the inner wall of the crosslinking mold, the mold vent, and the material feeding end. Specifically, it includes a high-precision armored thermocouple, an infrared thermal imaging sensor, an in-mold pressure transmitter, and an initiator concentration monitoring unit. The acquisition module captures real-time data on mold wall temperature, initial surface temperature of the material, pressure fluctuations within the mold, and ambient temperature and humidity. The multi-source heterogeneous data real-time acquisition module performs data encapsulation through an industrial communication gateway, converting analog electrical signals into structured data packets. During the data acquisition process, the module performs noise reduction and smoothing processing on the sensor signals and captures the induction period characteristics of the crosslinking reaction initiation stage through a preset sampling frequency, providing physical boundary conditions for subsequent virtual image construction.

[0009] Preferably, the process digital twin virtual mirror module is connected to the multi-source heterogeneous data real-time acquisition module via a logical link. This module is configured to establish a three-dimensional finite element digital model with the same scale as the physical mold within the computing unit. The process digital twin virtual mirror module receives data packets acquired in real time and maps the surface monitoring parameters to the three-dimensional spatial grid inside the material through a thermal conduction inverse calculation algorithm. The module integrates a variable thermal conductivity function for polystyrene material, which dynamically corrects the thermal diffusivity coefficient according to the current degree of crosslinking, temperature, and density of the material. The virtual mirror module maintains a high degree of synchronization with the physical production process in the time domain. By interpolating in real time on the virtual grid nodes, it restores the temperature gradient field and pressure field distribution from the surface to the core inside the material, thereby overcoming the barrier of material thermal resistance to physical detection.

[0010] Preferably, the nonlinear dynamics deep prediction module is the core mechanism unit of the system, and performs data interaction with the process digital twin virtual mirror module. This module is configured to establish a core temperature rise prediction model based on chemical kinetic mapping. The nonlinear dynamics deep prediction module integrates a reaction rate evolution operator based on Arrhenius's law, which correlates the temperature change rate with the crosslinking rate between polystyrene molecular chains. The module adopts a deep recurrent neural network architecture, including an input layer, a feature fusion layer, a time-series prediction layer, and a prediction output layer. The deep prediction model takes historical process curves, initiator activation energy data, and the core temperature rise trend output by the current digital twin model as input. By calculating a loss function minimization strategy, it predicts the peak adiabatic temperature rise in the core region within a preset time window. This prediction module can identify the reaction self-acceleration inflection point caused by positive heat feedback and generate a prediction command before core thermal runaway occurs.

[0011] Preferably, the intelligent closed-loop precision control module is connected to the nonlinear dynamics deep prediction module and the external actuator. This module is configured to dynamically adjust the process control parameters according to the prediction instructions output by the prediction module. The intelligent closed-loop precision control module has predictive control logic deployed inside. When the predicted core temperature rise slope exceeds the safety threshold, the module automatically triggers the cooling system intervention strategy to adjust the circulation flow rate and inlet temperature of the cooling medium on the mold wall. The module can control the output power of the electric heating actuator in real time. By adjusting the heat flow boundary outside the mold, it forcibly intervenes in the temperature field flatness inside the material. The module combines the internal pressure feedback and optimizes the partial pressure of volatile components inside the material by adjusting the opening of the vacuum exhaust valve, thereby synergistically suppressing the quality risk caused by excessive temperature rise in both physical and chemical dimensions.

[0012] Preferably, the long-term performance evaluation and optimization module is connected to the system's feedback loop. This module is configured to perform a closed-loop evaluation of the execution effect after each crosslinking process cycle. The evaluation and optimization module collects the actual density distribution, crosslinking uniformity, and dimensional stability indicators of the formed board, and performs correlation analysis with the parameter trajectory in the control process. By calculating the deviation between the actual value and the target value, the module dynamically updates the neural network weight parameters and the correction coefficients of the Arrhenius mapping operator in the nonlinear dynamics deep prediction module. This self-learning mechanism ensures that the control system can adapt to performance fluctuations under different batches of raw materials, different specifications of molds, and different environmental conditions, and achieves long-term and accurate optimization of process parameters.

[0013] Preferably, the multi-source heterogeneous data real-time acquisition module also includes an initiator decay compensation algorithm. This algorithm corrects the reactant mass baseline in the prediction model in real time based on the monitored initial concentration at the feed end and the reaction running time. This feature ensures that the system can identify kinetic deviations caused by initiator deactivation during long-cycle production.

[0014] Preferably, the process digital twin virtual mirror module adopts an unsteady heat conduction control equation. This equation introduces an internal source term correction during the solution process. The source term represents the instantaneous enthalpy released by the crosslinking reaction. By refining the digital twin mesh to the millimeter level, this module can capture the complex heat transfer boundary between the corner area of ​​the plate and the core area, ensuring the accuracy of the prediction logic in the three-dimensional space dimension.

[0015] Preferably, the nonlinear dynamics deep prediction module introduces physical information perception loss during the training process. This loss function forces the output of the neural network to conform to the law of conservation of energy and the law of conservation of matter, thereby avoiding the logical extrapolation failure that may occur in the pure data-driven model under extreme conditions and improving the reliability of the system in the critical state of thermal runaway.

[0016] Preferably, the intelligent closed-loop precision control module executes the task allocation of the actuator through the model predictive control algorithm. When it receives the cooling prediction command, the algorithm calculates the delay compensation factor and starts the maximum cooling power in advance when the temperature surge reaches the mold wall. This advanced pre-cooling logic effectively offsets the control phase lag caused by the high thermal resistance of polystyrene and realizes the physical suppression of the core temperature rise peak.

[0017] This invention also provides an intelligent control method based on the above system. The method utilizes the above-mentioned multi-source heterogeneous data real-time acquisition module, process digital twin virtual mirror module, nonlinear dynamics deep prediction module, and intelligent closed-loop precise control module to execute the following process: Step 1: Activate multi-source heterogeneous real-time sensing, and collect the initial temperature field, initiator concentration and environmental condition baseline through sensor arrays deployed in various parts of the mold.

[0018] Step 2: Construct a real-time digital twin mapping, inject the collected data as boundary conditions into the finite element physical mirror, and use the heat conduction inverse calculation algorithm to reconstruct the three-dimensional temperature gradient and cross-linking field inside the material in real time.

[0019] Step 3: Perform nonlinear dynamics deduction. Based on the chemical kinetics mapping model and neural network, predict the temperature rise path of the material core within the future time window and identify the abrupt change point of the reaction rate.

[0020] Step 4: Initiate intelligent closed-loop intervention. Based on the prediction results, adjust the cooling system flow rate, electric heating power, and vacuum exhaust pressure in advance to implement advanced damping control for core temperature rise.

[0021] Step 5: Dynamic update and optimization. After each process cycle, analyze the board quality data and correct the prediction operator and mapping function through the backpropagation algorithm to achieve continuous evolution of the system.

[0022] The beneficial effects of this invention are as follows: 1. The XCPS crosslinked polystyrene intelligent crosslinking process parameter precision control system of this invention completely solves the technical defects of observation lag caused by the extremely low thermal conductivity of polystyrene by constructing a process digital twin virtual mirror module. By mapping the measured surface parameters to the internal three-dimensional distribution field, the system can see the physicochemical state of the material core in real time, providing a transformation from black box search to transparent mapping for process control, and improving the determinism of temperature field control inside the material. 2. The XCPS cross-linked polystyrene intelligent cross-linking process parameter precision control system of this invention, through the introduction of a nonlinear dynamics deep prediction module, achieves early warning of adiabatic temperature rise during the cross-linking reaction. By deeply integrating Arrhenius's law with deep learning algorithms, the system can identify the positive feedback trend of the reaction caused by heat accumulation, thereby performing pre-intervention before the material undergoes explosive polymerization or thermal degradation. This mechanism significantly reduces the scrap rate in the production of thick sheets and solves the problems of core explosive polymerization and hollow defects that have plagued the industry for many years. 3. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene described in this invention optimizes the spatial distribution of crosslinking degree of XCPS sheets through a multi-variable collaborative intervention strategy using an intelligent closed-loop precision control module. By introducing coordinated adjustment of pressure and flow rate on top of temperature control, the system can maintain extremely high uniformity of the internal temperature field of the material, thereby effectively suppressing geometric deformation of the molded sheets caused by uneven internal stress, and significantly improving the long-term service performance and dimensional stability of XCPS products. 4. The XCPS cross-linked polystyrene intelligent cross-linking process parameter precision control system of this invention possesses powerful self-adaptive and self-optimizing capabilities. Through continuous iteration of the long-term performance evaluation and optimization module, the system can automatically correct control errors caused by raw material fluctuations or environmental drift. This self-learning characteristic makes the control system exhibit extremely high robustness in the face of different operating conditions, significantly reducing reliance on manual experience and machine adjustment cycles. While ensuring product quality, it significantly improves the overall operating efficiency of the production line. 5. The XCPS crosslinked polystyrene intelligent crosslinking process parameter precision control system of the present invention ensures the absolute safety of the system under high load and extreme reaction conditions through the fusion modeling of physical information perception and data-driven approach. By forcing physical conservation constraints, the system effectively avoids the failure risk of traditional automation solutions in complex nonlinear feedback. This feature makes a significant contribution to solving the problems of long molding cycle, difficult control of reaction heat, low yield and poor quality of large products in the production of thick crosslinked polystyrene. Attached Figure Description

[0023] The invention will now be further described with reference to the accompanying drawings.

[0024] Figure 1 This is a structural block diagram of the intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene in this invention; Figure 2 This is a flowchart of the method for precise control of intelligent crosslinking process parameters of XCPS crosslinked polystyrene in this invention. Detailed Implementation

[0025] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0026] like Figure 1 As shown, the intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene in this embodiment of the invention includes a multi-source heterogeneous data real-time acquisition module configured to construct a multi-dimensional sensing matrix within the physical field of the XCPS crosslinking reaction to achieve comprehensive monitoring of the reaction environment. The acquisition module includes, at the hardware level, a sensor group deployed on the inner wall of the crosslinking mold, the mold exhaust port, and the material feeding end. To ensure accurate data acquisition under high temperature, high pressure and chemically corrosive cross-linking environments, the acquisition module specifically includes a high-precision armored thermocouple, an infrared thermal imaging sensor, an in-mold pressure transmitter and an initiator concentration monitoring unit. The high-precision armored thermocouple is embedded in a pre-set deep hole on the inner side of the mold wall, and the distance between its probe and the surface of the mold cavity is precisely controlled between three and five millimeters, aiming to capture the mold wall temperature that is closest to the instantaneous fluctuation of the material surface. The infrared thermal imaging sensor is installed at the observation window above the mold. It performs high-frequency scanning on the exposed surface of the material to capture the uniformity of the initial temperature distribution of the material surface in real time. The in-mold pressure transmitter is deployed at the geometric center and corner areas of the mold to monitor the pressure fluctuations in the mold caused by gas release and material expansion during the crosslinking reaction. The initiator concentration monitoring unit is located at the outlet of the mixing chamber of the feeding system. It obtains the initial concentration data of the initiator before entering the mold in real time through online spectral analysis technology. The system is also equipped with environmental monitoring sensors to synchronously capture ambient temperature and humidity data, providing environmental background parameters for subsequent process compensation.

[0027] The multi-source heterogeneous data real-time acquisition module performs data encapsulation through an industrial communication gateway. This gateway supports multiple industrial protocol interfaces and can convert the weak analog electrical signals output by the sensors into standardized structured data packets. During the data acquisition process, the built-in signal processor of the module performs noise reduction and smoothing processing on the sensor signal. Specifically, it adopts a strategy that combines Butterworth low-pass filter and Kalman filter algorithm to filter out random noise caused by electromagnetic interference. The acquisition module is configured to capture the induction period characteristics of the crosslinking reaction initiation stage at a preset sampling frequency. The sampling frequency is usually set between 50 Hz and 100 Hz. By accurately identifying the small temperature changes during the induction period, the system can determine the activation start time of the initiator. Furthermore, the multi-source heterogeneous data real-time acquisition module also includes an initiator decay compensation algorithm. This algorithm, based on the monitored initial concentration at the feed end and the cumulative reaction time, combined with the thermal decomposition kinetic curve of the initiator at the current temperature, corrects the reactant mass baseline in the prediction model in real time. This feature ensures that in long-cycle large-scale production, the system can identify and compensate for kinetic calculation deviations caused by premature initiator deactivation, providing physically realistic boundary conditions for subsequent virtual image construction.

[0028] The process digital twin virtual mirror module and the multi-source heterogeneous data real-time acquisition module are connected via a high-speed industrial Ethernet logical link. This module is configured to build a three-dimensional finite element digital model of the same scale as the physical mold within a high-performance computing unit. The process digital twin virtual mirror module receives real-time data packets and maps surface monitoring parameters to a three-dimensional spatial grid inside the material using a thermal conduction inverse calculation algorithm. This module integrates a variable thermal conductivity function for polystyrene. This function not only considers the influence of temperature but also dynamically corrects the thermal diffusivity coefficient based on the material's current degree of crosslinking, density, and phase changes. Since the physicochemical properties of crosslinked polystyrene undergo drastic changes during the reaction process, the variable thermal conductivity function can be calculated through real-time interpolation to ensure that the heat transfer logic in the virtual model is highly aligned with the actual physical process.

[0029] The process digital twin virtual mirror module uses an unsteady heat conduction control equation as the underlying physical driving logic. During the solution process, an internal source term correction is introduced. The source term represents the instantaneous enthalpy released by the crosslinking reaction at different grid nodes and different time steps. To ensure the accuracy of the simulation, the virtual mirror module refines the material's three-dimensional spatial grid to the millimeter level. Especially for thicker plates, the grid density is locally increased in the core region. This module can capture the complex heat transfer boundary conditions between the plate's corner areas and the core region. By solving the energy balance equation in three-dimensional space, it can restore the temperature gradient field and pressure field distribution from the surface to the core of the material in real time. The virtual mirror module maintains a high degree of synchronization with the physical production process in the time domain. By displaying temperature isosurfaces on the virtual grid nodes in real time, operators can overcome the detection barrier caused by the high thermal resistance of polystyrene and achieve transparent observation of the reaction state in the core of the material.

[0030] The nonlinear dynamics deep prediction module, as the core mechanism unit of the system, performs high-frequency data interaction with the process digital twin virtual mirror module. This module is configured to establish a core temperature rise prediction model based on chemical kinetic mapping, aiming to solve the problem of adiabatic temperature rise caused by the violent exothermic reaction of crosslinking. The nonlinear dynamics deep prediction module integrates a reaction rate evolution operator based on the Arrhenius law. This operator performs a deep correlation between the temperature change rate, the crosslinking monomer concentration, and the crosslinking rate between polystyrene molecular chains. In a specific implementation, the module adopts a deep recurrent neural network (DRNN) architecture, which includes an input layer, a feature fusion layer, a temporal prediction layer, and a prediction output layer.

[0031] The input layer of the nonlinear dynamics deep prediction module receives a multi-dimensional parameter sequence, including historical process curves provided by the acquisition module, initiator activation energy data provided by the feed end, and the core temperature rise trend output in real time by the current digital twin model. The feature fusion layer extracts high-order features reflecting the self-acceleration trend of the reaction by performing weighted processing on the above heterogeneous data. The time-series prediction layer uses long short-term memory (LSTM) units to capture the dynamic evolution law in the reaction process. By calculating the loss function minimization strategy, it predicts the peak adiabatic temperature rise of the core region in advance within a preset time window (such as 30 seconds to 3 minutes). This prediction module can keenly identify the reaction self-acceleration inflection point caused by positive heat feedback. That is, when the reaction heat release rate is much greater than the material conduction rate, the system will generate a prediction command before the core thermal runaway occurs.

[0032] Furthermore, to enhance the physical reliability of the predictions, the nonlinear dynamics deep prediction module introduces a Physics-Informed Loss during training. This loss function forces the output of the neural network to conform to the laws of conservation of energy and conservation of matter within each prediction step. This means that the predicted temperature rise must be logically consistent with the enthalpy released by the mass of monomer consumed in the reaction. Through this training mode that combines mechanistic constraints with data-driven approaches, the system effectively avoids the logical extrapolation failures that may occur when the pure data-driven model faces extreme working conditions or abnormal batches of raw materials, greatly improving the reliability of the system's judgment in the critical state of thermal runaway.

[0033] The intelligent closed-loop precision control module is connected to the nonlinear dynamics deep prediction module and the external actuator. This module is configured to dynamically adjust the process control parameters and perform active intervention according to the prediction instructions output by the prediction module. The intelligent closed-loop precision control module is equipped with advanced predictive control logic (MPC), which can plan the action sequence of the actuator in advance according to the predicted temperature rise path. When the predicted core temperature rise slope exceeds the safety threshold, the module automatically triggers the cooling system intervention strategy, adjusting the circulation flow rate and inlet temperature of the cooling medium on the mold wall. Due to the extremely low thermal conductivity of polystyrene, traditional real-time adjustment often produces severe phase lag. Therefore, the intelligent closed-loop precision control module executes the task allocation of the actuator through model predictive control algorithm. By calculating the delay compensation factor, it activates the maximum cooling power in advance when the temperature surge reaches the mold wall. This advanced pre-cooling logic effectively offsets the control phase lag caused by high thermal resistance, achieving physical suppression of the core temperature rise peak.

[0034] The intelligent closed-loop precision control module can control the output power of the electric heating actuator in real time. During the reaction induction period, it ensures that the material quickly reaches the initiation temperature through precise power control. During the period of intense reaction, the heat flow boundary around the mold is adjusted to forcibly intervene in the temperature field flatness inside the material and prevent local overheating. Furthermore, this module, combined with in-mold pressure feedback, optimizes the partial pressure of volatile components inside the material by adjusting the opening of the vacuum exhaust valve. By reducing the in-mold pressure, it can accelerate the gasification and endothermic reaction of low-molecular-weight byproducts generated during the reaction. This synergistically suppresses the quality risks caused by excessively rapid temperature rise from both physical and chemical perspectives. The multi-variable synergistic control strategy ensures that the temperature difference between the core and the surface of the XCPS board is always maintained within the preset process allowable range during the cross-linking process, thereby suppressing the geometric deformation of the board caused by uneven internal stress distribution.

[0035] The long-term performance evaluation and optimization module is connected to the system's feedback loop and is configured to perform closed-loop evaluation and continuous optimization of the performance after each cross-linking process cycle. After each batch of boards is demolded and formed, the evaluation and optimization module collects the actual density distribution, cross-linking uniformity, and dimensional stability indicators of the boards. These indicators are obtained through a laboratory online analysis system or non-contact measurement equipment and are used as real performance feedback input to the system. The evaluation and optimization module performs correlation analysis between these measured values ​​and the parameter trajectories during the control process. It uses a gradient descent-based backpropagation algorithm to dynamically update the neural network weight parameters in the nonlinear dynamics deep prediction module and corrects the frequency factor and activation energy deviation term in the Arrhenius mapping operator.

[0036] This self-learning mechanism endows the system with powerful adaptive capabilities, enabling it to automatically identify and adapt to performance fluctuations under different batches of raw materials (such as changes in monomer purity), different mold specifications (such as differences in mold heat capacity), and different seasonal environmental conditions. Through long-term efficiency accumulation, the long-term efficiency evaluation and optimization module can continuously converge prediction errors, making the control system exhibit extremely high robustness when facing extremely complex nonlinear reaction processes. This not only ensures the quality consistency of high-performance XCPS sheets, but also significantly reduces the machine adjustment cycle when developing new processes or putting new specifications of products into production.

[0037] like Figure 2 As shown, this system constructs a complete technical closed loop through deep logical coupling of multiple modules. In actual production, its operation process is as follows: In step one, the system initiates multi-source heterogeneous real-time sensing. Through high-precision armored thermocouples, infrared sensors, and initiator monitoring units deployed on the inner wall of the mold, the vent, and the feeding end, it collects the initial temperature field, initiator concentration, and the current ambient temperature and humidity baseline. After the acquisition module performs noise reduction processing, it encapsulates the data into a structured data packet and sends it to the twin mirror module.

[0038] In step two, the system constructs a real-time digital twin mapping. The digital twin virtual mirror module injects the received boundary data into a preset three-dimensional finite element mesh. Using the variable thermal conductivity function and the unsteady heat conduction control equation, it calculates and restores the three-dimensional temperature gradient distribution and real-time cross-linking field of the core of the board in real time, breaking through the perception blind spot of physical detection methods inside the insulation material.

[0039] In step three, the system performs nonlinear dynamics deduction. The nonlinear dynamics deep prediction module uses the Arrhenius mapping operator combined with a neural network with physical information perception to perform advanced deduction of the core temperature rise path. The system identifies the abrupt change point of the reaction rate by calculating the loss function and generates a temperature rise prediction command with a timestamp.

[0040] In step four, the system initiates intelligent closed-loop intervention. Based on the temperature rise prediction command, the intelligent closed-loop precision control module starts the high-frequency circulation of the cooling medium in advance before the actual temperature rise surge arrives, and simultaneously adjusts the output power of the electric heating actuator and the vacuum exhaust pressure. Through the combination of advanced pre-cooling and pressure control, the peak temperature rise of the core adiabatic temperature is physically suppressed.

[0041] In step five, the system performs dynamic updates and optimizations. After the process cycle ends, the long-term performance evaluation and optimization module analyzes the key performance indicators (KPIs) such as the density and uniformity of the finished board material, dynamically adjusts the weights and dynamic parameters of the prediction model, and realizes the self-evolution of the system control logic.

[0042] In an example, the intelligent control system of the present invention is used to produce XCPS sheets with a thickness of 100 mm. The system uses digital twin to capture in real time that the core area has an abnormal temperature rise rate six minutes after the reaction starts. The prediction module issues a temperature rise surge warning, and the intelligent control module starts the maximum cooling flow three minutes in advance.

[0043] In comparison, when using a traditional PID constant temperature control system to produce the same specifications of sheet metal, the traditional system, due to observation lag, can only activate the cooling logic after the mold wall temperature has actually risen, at which point the heat in the core of the material has already accumulated significantly.

[0044] The table below shows the comparison data of the two sets of experiments on the core process indicators: The comparative data table above clearly shows that this invention effectively suppresses the core adiabatic temperature rise by constructing predictive control logic, reducing the temperature difference between the core and the surface from 56.7 degrees Celsius to 12.4 degrees Celsius. This extremely high temperature field uniformity directly leads to a precipitous decrease in the explosive runaway rate of large-size finished products and improves the dimensional stability of the board by 82.3%. The experimental data fully demonstrates that this invention, through the coupling of digital twins and dynamic prediction, completely solves the problem of thermal runaway caused by observation lag in XCPS production.

[0045] Furthermore, when production specifications are switched to XCPS products with high density and high flame retardancy, the introduction of flame retardants will change the thermal diffusivity of the material. The process digital twin virtual mirror module quickly adapts to the new thermal conduction boundary by automatically adjusting its built-in variable thermal conductivity function parameters. The nonlinear dynamics deep prediction module, by loading historical fine-tuning parameters for this specification, achieved a temperature rise prediction accuracy of over 97% in the first batch of trial production. The intelligent closed-loop precision control module, by precisely controlling the vacuum exhaust pressure, effectively removes trace amounts of volatiles that may be generated by the flame retardant at high temperatures, ensuring the density and closed-cell rate of the internal structure of the board.

[0046] In the description of the system's physical connection and signal flow, the signals output by the sensor group are aggregated through the field I / O module and transmitted to the central controller via an industrial bus based on the Profinet protocol. The multi-source heterogeneous data acquisition program running inside the central controller performs millisecond-level interrupt scanning to ensure the real-time performance of data acquisition. Subsequently, the encapsulated data packet is uploaded to the edge computing server deployed with a digital twin engine and a deep prediction model via the OPC UA protocol. After the server completes high-density numerical calculations, it sends the control instruction set to the actuators (including proportional-integral cooling water valves, solid-state relay-driven heating tubes, and servo-controlled vacuum valves). The action status signals fed back by the actuators are sent back to the controller via the bus, forming a complete hardware logic loop.

[0047] After each process cycle, the long-term performance evaluation and optimization module automatically triggers the parameter correction process by reading the quality inspection records in the production management system (MES). This optimization process based on backpropagation runs silently in the background and will not interfere with the real-time production process in the foreground. After running continuously for one thousand process cycles, the system's prediction error for the peak adiabatic temperature rise has converged to within ±1.5 degrees Celsius.

[0048] In summary, the intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene provided by this invention constructs a complete technical ecosystem from bottom-level perception and high-dimensional simulation to mechanism prediction and dynamic intervention through deep logical coupling and physical mechanism reconstruction of multiple modules. This invention not only solves the problems of control lag and core burst polymerization caused by the extremely low thermal conductivity of polystyrene materials at the technical level, but also achieves precise deterministic control of complex nonlinear reaction processes at the engineering level. Its physical information perception and prediction capabilities and adaptive learning characteristics provide absolutely reliable technical support for the large-scale, high-quality production of XCPS high-performance insulating materials, and make a significant and substantial contribution to improving the intelligent level and intrinsic safety of my country's special plastics processing industry.

[0049] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene, characterized in that, The system includes: The multi-source heterogeneous data real-time acquisition module is used to construct a multi-dimensional sensing matrix within the physical field of the XCPS crosslinking reaction, and to capture real-time data on mold wall temperature, initial surface temperature of material, pressure inside the mold, and initiator concentration. The process digital twin virtual mirror module is connected to the multi-source heterogeneous data real-time acquisition module. It is used to map the material surface monitoring parameters to the material internal three-dimensional spatial grid through the heat conduction inverse calculation algorithm based on the multi-dimensional sensing matrix, and restore the temperature gradient field and pressure field distribution inside the material in the time domain. The nonlinear dynamics deep prediction module, connected to the process digital twin virtual mirror module, is used to identify the reaction self-acceleration inflection point and predict the adiabatic temperature rise path and peak value of the material core within a preset time window using a core temperature rise prediction model based on chemical kinetic mapping. The intelligent closed-loop precision control module is connected to the nonlinear dynamics depth prediction module and the external actuator respectively. It is used to perform advanced damping control on the core of the material by adjusting the heat flow boundary of the mold wall and the pressure inside the mold based on the prediction results of the adiabatic temperature rise path and the peak value of the adiabatic temperature rise. The long-term performance evaluation and optimization module is connected to the feedback loop of the nonlinear dynamics depth prediction module. It is used to collect the quality indicators after the sheet is formed and dynamically update the operator coefficients in the core temperature rise prediction model by calculating the deviation between the actual value and the target value.

2. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 1, characterized in that, The multi-source heterogeneous data real-time acquisition module includes a sensor group and a signal processing unit; The sensor group includes a high-precision armored thermocouple deployed in a deep hole in the inner wall of the crosslinking mold, an infrared thermal imaging sensor deployed at the mold observation window, an in-mold pressure transmitter deployed at the geometric center and corner areas of the mold, and an initiator concentration monitoring unit deployed at the outlet of the feed hopper. The signal processing unit performs data encapsulation through an industrial communication gateway and performs noise reduction and smoothing processing on the sensor signals. It captures the induction period characteristics of the crosslinking reaction initiation stage by sampling frequency. The noise reduction and smoothing processing adopts a strategy combining Butterworth low-pass filter and Kalman filter algorithm. The signal processing unit also has a built-in initiator decay compensation algorithm, which is configured to: correct the reactant mass base in the core temperature rise prediction model based on the monitored initial concentration at the feed end and the cumulative reaction time, combined with the thermal decomposition kinetic curve of the initiator at the real-time temperature.

3. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 1, characterized in that, The process digital twin virtual mirror module is configured to establish a three-dimensional finite element digital model with the same scale as the physical mold within the computing unit. The three-dimensional finite element digital model performs local mesh refinement in the core region of the material, refining the three-dimensional spatial mesh of the material to the millimeter level. The virtual mirror module keeps synchronized with the physical production process in the time domain. By interpolating in real time on virtual grid nodes, it restores the instantaneous temperature field distribution inside the material from the surface to the core.

4. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 3, characterized in that, The process digital twin virtual mirror module uses unsteady heat conduction control equations as physical driving logic. The unsteady heat conduction control equations are corrected by introducing internal source terms during the solution process. These internal source terms represent the instantaneous enthalpy released by the crosslinking reaction at different grid nodes and different time steps. The process digital twin virtual mirror module integrates a variable thermal conductivity function for polystyrene material. This variable thermal conductivity function dynamically corrects the thermal diffusivity coefficient based on the material's current degree of crosslinking, instantaneous temperature, and density, ensuring that the heat transfer logic in the virtual mirror model aligns with the physical process.

5. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 1, characterized in that, The nonlinear dynamics deep prediction module integrates a reaction rate evolution operator based on the Arrhenius law, which correlates the temperature change rate, crosslinking monomer concentration, and crosslinking rate between polystyrene molecular chains.

6. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 5, characterized in that, The nonlinear dynamics deep prediction module employs a deep recurrent neural network architecture, which includes: The input layer is used to receive parameter sequences, which include historical process curves, initiator activation energy data, and the core temperature rise trend output by the current process digital twin virtual mirror module. The feature fusion layer is used to extract high-order features that reflect the self-accelerating trend of the reaction by performing weighted processing on the heterogeneous data received by the input layer. The time-series prediction layer uses long short-term memory units to capture the dynamic evolution of the reaction process and predicts the peak adiabatic temperature rise of the core region in advance within the future time window by calculating the loss function minimization strategy. The prediction output layer is used to output temperature rise prediction commands.

7. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 6, characterized in that, The nonlinear dynamics deep prediction module introduces a physical information perception loss function during training. This physical information perception loss function forces the output of the neural network to conform to the law of conservation of energy and the law of conservation of matter within each prediction step, and limits the predicted temperature rise value to be logically consistent with the enthalpy value released by the mass of monomer consumed in the reaction.

8. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 7, characterized in that, The intelligent closed-loop precision control module is internally deployed with model predictive control logic, which is configured to plan the action sequence of the actuator in advance based on the temperature rise prediction command: When the predicted core temperature rise slope exceeds the safety threshold, the cooling system intervention strategy is automatically triggered to adjust the circulation flow rate and inlet temperature of the cooling medium on the mold wall. The intelligent closed-loop precision control module calculates the delay compensation factor and activates the maximum cooling power in advance of the temperature surge reaching the mold wall to perform advanced pre-cooling operation to offset the control phase lag caused by the thermal resistance of polystyrene. The intelligent closed-loop precision control module adjusts the heat flow boundary around the mold and intervenes in the temperature field flatness inside the material by controlling the output power of the electric heating actuator.

9. The intelligent crosslinking process parameter precision control system for XCPS crosslinked polystyrene according to claim 8, characterized in that, The intelligent closed-loop precision control module, combined with in-mold pressure feedback, optimizes the partial pressure of volatile components inside the material by adjusting the opening of the vacuum exhaust valve, and accelerates the gasification and heat absorption of low-molecular-weight byproducts generated during the reaction by reducing in-mold pressure, thereby achieving synergistic suppression of core temperature rise in both physical and chemical dimensions.

10. A method for precise control of intelligent crosslinking process parameters of XCPS crosslinked polystyrene, applicable to the intelligent crosslinking process parameter precise control system for XCPS crosslinked polystyrene as described in any one of claims 1-9, characterized in that, The method includes the following steps: Step 1: Initiate multi-source heterogeneous real-time sensing: Through sensor groups deployed on the inner wall of the mold, the vent of the mold, and the material feeding end, the initial temperature field, initiator concentration, and environmental condition benchmark are fully collected, and noise reduction and smoothing processing is performed. Step 2, Constructing a real-time digital twin mapping: The collected data is injected as boundary conditions into the three-dimensional finite element virtual mirror model. Through the variable thermal conductivity function and the unsteady heat conduction control equation, the three-dimensional temperature gradient and cross-linking field inside the material are restored in real time using the heat conduction inverse calculation algorithm. Step 3: Perform nonlinear dynamics deduction: Based on the chemical kinetics mapping model and the neural network with the physical information perception loss function, the temperature rise path of the material core in the future time window is deduced, the reaction rate mutation point is identified, and a temperature rise prediction command with a timestamp is generated. Step 4: Activate intelligent closed-loop intervention: The intelligent closed-loop precision control module executes the model prediction control algorithm according to the temperature rise prediction command, adjusts the cooling system flow rate, electric heating power and vacuum exhaust pressure in advance, and performs advanced damping control to suppress the peak temperature rise of the core adiabatic temperature rise. Step 5, Dynamic Update and Optimization: After the crosslinking process cycle is completed, analyze the actual density distribution, crosslinking degree uniformity and dimensional stability index of the formed board. Use the backpropagation algorithm to correct the neural network weight parameters and the correction coefficient of the Arrhenius mapping operator in the core temperature rise prediction model to realize the self-evolution of the control logic.

Images