Segmented temperature control closed type rubber step mixing intelligent control method

Abstract

This invention discloses a segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing, relating to the field of rubber processing and mixing technology. The invention constructs a segmented closed-loop mixing digital twin with a cross-stage micro-parameter transmission mechanism. Differentiated coupling strategies are built for the three stages of plasticizing, filler dispersion, and homogenization final mixing. Combined with a segmented dynamic weight optimization algorithm with stage coupling correction coefficients, virtual pre-simulation and global optimization of process parameters are achieved. Ultrasonic scattering detection components with embedded alumina ceramic acoustic lining collect rubber compound condition data online. After time-synchronized calibration, this data is compared with the twin prediction data, and a comprehensive deviation coefficient is calculated. When the deviation exceeds a threshold, the sub-model parameters are corrected before an adjustment command is output, realizing virtual pre-simulation optimization and intelligent closed-loop control of the entire step-by-step mixing process.

Description

Technical Field

[0001] This invention relates to the field of rubber processing and mixing technology, specifically to a segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing. Background Technology

[0002] Closed-loop mixing is a core process in rubber product manufacturing. The effectiveness of plasticization, filler dispersion, and additive homogenization during mixing directly determines the mechanical properties and batch-to-batch consistency of the finished rubber compound. As the quality requirements for high-end rubber products continue to rise, stepwise mixing processes, with their segmented feeding and staged temperature and shear control, are widely used in the production of high-reinforcement, high-performance rubber formulations. The development of digital twin technology provides a new technical path for the virtual simulation and intelligent control of the mixing process.

[0003] The current step-by-step mixing process parameters are highly dependent on operator experience for adjustment, and the development of new formulations requires repeated physical trials, resulting in long development cycles and significant raw material losses. Existing mixing simulation technologies mostly involve independent calculations of single physical fields, failing to establish differentiated coupled models for the dominant physical processes in each stage of plasticizing, dispersion, and final mixing. Furthermore, there is a lack of transfer and connection of microscopic parameters between stages, limiting the accuracy of simulation predictions. Simultaneously, the thick-walled metal structure of the mixing chamber makes online detection of filler dispersion difficult. Current control methods often rely on a single temperature parameter, which cannot match the dynamic adjustment of the rubber compound's microscopic state, easily leading to problems such as over-mixing causing molecular chain degradation or under-mixing resulting in insufficient dispersion, and significant batch-to-batch performance deviations in rubber compounds.

[0004] In summary, existing technologies cannot simultaneously meet the requirements of efficient development of stepwise mixing processes and precise closed-loop control of the production process. Therefore, it is particularly important to develop an intelligent control method that combines high-precision segmented digital twin simulation with online status detection to solve the technical problems of insufficient simulation accuracy, low R&D efficiency, and poor control accuracy in existing technologies. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing. This method constructs a segmented closed-loop mixing digital twin with a cross-stage micro-parameter transfer mechanism. Differentiated coupling strategies are built for the three stages of plasticizing, filler dispersion, and homogenization final mixing. Combined with a segmented dynamic weight optimization algorithm with stage coupling correction coefficients, the method completes virtual pre-simulation and global optimization of process parameters. Ultrasonic scattering detection components with embedded alumina ceramic acoustic lining collect rubber compound condition data online. After time-synchronized calibration, this data is compared with the twin prediction data, and a comprehensive deviation coefficient is calculated. When the deviation exceeds a threshold, the sub-model parameters are corrected before an adjustment command is output, achieving virtual pre-simulation optimization and intelligent closed-loop control of the entire step-by-step mixing process.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a segmented temperature-controlled closed-loop intelligent control method for stepwise rubber mixing, the method comprising the following components:

[0007] Step 1: Construct a segmented closed-loop digital twin of the entire mixing process, integrate a multi-physics field coupling simulation engine, build corresponding coupling sub-models for the three mixing stages of plasticization, filler dispersion, and homogenization final mixing, establish a cross-stage micro-parameter transfer mechanism between sub-models, input the internal mixer equipment structural parameters and rubber formulation component parameters, and complete the initialization and calibration of the twin model.

[0008] Step 2: Map the segmented temperature control curves, feeding timing nodes, rotor speed sequence, and top bolt pressure parameters of each stage of the step-by-step mixing process to the corresponding sub-models of the digital twin, and restore the boundary conditions of the complete mixing process of the three stages.

[0009] Step 3: Start the multiphysics coupling simulation engine and run each coupling sub-model in the phase sequence. In the plasticizing stage, pre-simulate the temperature field distribution and Mooney viscosity change trend of the rubber compound. In the filler dispersion stage and the homogenization and final refining stage, pre-simulate the filler dispersion evolution process simultaneously.

[0010] Step 4: Using the mixing quality index and energy consumption index as multiple optimization objectives, call the piecewise dynamic weight optimization algorithm with stage coupling correction to iteratively optimize the input process parameters and output the globally optimal combination of process parameters.

[0011] Step 5: During the actual production process, the mixing condition data is collected in real time through the on-site sensing unit. The measured data after synchronous calibration is compared point by point with the real-time prediction data of the twin model, and the comprehensive deviation coefficient is calculated.

[0012] Step 6: When the comprehensive deviation coefficient exceeds the preset threshold, the twin model first corrects the simulation input parameters of the corresponding sub-model, and then calculates and generates the adjusted process instructions based on the corrected sub-model, which are then sent to the internal mixer control system for execution to dynamically compensate for fluctuations in actual working conditions.

[0013] Furthermore, in step one, the cavity, rotor, top bolt, and jacket structure of the physical internal mixer are reproduced at a 1:1 geometric scale. The entered equipment structural parameters include the number of rotor edges, effective volume of the mixing chamber, jacket heat exchange area, and maximum stroke of the top bolt. The entered formula component parameters include the rheological coefficient, thermal conductivity, density, and specific heat capacity of each rubber compound and compounding agent. The three coupled sub-models adopt differentiated coupling strategies. The plasticizing segment model adopts rheological-heat transfer bidirectional coupling, in which temperature field data is substituted into the rheological equation to update the viscosity at each time step, and viscosity field data is substituted into the heat transfer equation to update the shear heat source term. The filler dispersion segment model adopts rotor dynamics-rheological bidirectional coupling. The homogenization and final refining stage model adopts a two-way coupling of heat transfer and reaction kinetics. The cross-stage microscopic parameter transfer mechanism is as follows: the molecular chain deentanglement degree output by the plasticizing stage model is used as the initial rheological parameter of the filler dispersion stage model, and the bound rubber content output by the filler dispersion stage model is used as the initial reaction kinetic parameter of the homogenization and final refining stage model. The model initialization calibration is based on the Cross rheological model, calling no less than 12 sets of historical production data of the same formula, and using the least squares method to fit and correct the viscosity-temperature coefficient and shear heat generation coefficient. The objective function is to minimize the average relative error of rubber temperature and Mooney viscosity, so that the average relative error of static prediction of the twin model is controlled within 3%.

[0014] Furthermore, in step two, process parameters are sequentially entered into the corresponding sub-models according to the time sequence of the plasticizing section, the filler dispersion section, and the homogenization and final refining section. Each parameter includes the temperature control setting value, the rotor rated speed, the working pressure of the top plug, and the preset duration of the stage. The filling dispersion section additionally enters the feeding sequence nodes of the reinforcing filler and activator, and the homogenization and final refining section additionally enters the feeding sequence nodes of the vulcanization system. During the parameter mapping process, boundary verification is automatically performed. The verification rules include a temperature control parameter range of 60℃-160℃, a rotor speed range of 10rpm-80rpm, a top plug pressure range of 0.2MPa-1.2MPa, and a stage duration range of 1min-15min. If any parameter exceeds the rated range, the mapping is intercepted and an abnormal prompt is output. After all parameters pass the verification, the virtual working condition initialization of each sub-model is completed.

[0015] Furthermore, in step three, the multiphysics coupling simulation adopts an adaptive variable step size mechanism. The step size for the plasticizing stage is set to 1 second, the step size for the filler dispersion stage is set to 0.5 seconds, and the step size for the homogenization and final refining stage is set to 1 second. After each time step is completed, the physical field data of all nodes in the corresponding sub-model are updated synchronously. Every 20 seconds, the plasticizing stage outputs a set of global temperature field average and average viscosity data. Every 10 seconds, the filler dispersion stage outputs a set of filler particle size distribution statistics and shear rate field cloud map. Every 15 seconds, the homogenization and final refining stage outputs a set of average viscosity data and cavity local temperature extreme value data. During the simulation, abnormal nodes that exceed the set temperature range or have a dispersion growth rate of zero are automatically identified, and their corresponding spatial coordinates and time nodes are marked. After the entire process simulation is completed, parameter evolution curves for each stage and a summary table of abnormal nodes are generated.

[0016] Furthermore, in step four, the piecewise dynamic weight optimization algorithm with stage coupling correction uses a stage-adapted fitness function combined with stage matching correction coefficients for iterative optimization, as shown in the formula: ,in For overall adaptability throughout the entire process, , , The adaptability of each stage—plasticizing section, filler dispersion section, and homogenization and final refining section—is assessed. , , The global weights for each stage are 0.28, 0.42, and 0.30, respectively. The stage parameter matching correction coefficient ranges from 0.92 to 1.08 and is dynamically adjusted based on the matching deviation of micro-parameters between adjacent stages. The single-stage fitness function of the plasticizing stage only includes two indicators: viscosity and energy consumption. The formula is as follows: ,in The Mooney viscosity range is simulated for the plasticizing stage. This represents the maximum permissible Mooney viscosity range in the plasticizing stage. The value is set to 8 Mooney units, determined based on the statistical upper limit of the decrease in Mooney viscosity of natural rubber within its typical plasticizing temperature range. This refers to the energy consumption per unit of rubber compound mixing in the plasticizing section. This is the benchmark energy consumption for the plasticizing section with the same formula. The value of 0.12 kWh / kg is determined by calculating the product of the no-load power of the internal mixer and the average current in a typical plasticizing section. , The values ​​are 0.68 and 0.32 respectively; the fitness function of each segment in the packing dispersion section and the homogenization and final refining section includes three indicators: dispersion degree, viscosity, and energy consumption. The corresponding weights for the packing dispersion section are 0.47, 0.31, and 0.22, and the corresponding weights for the homogenization and final refining section are 0.25, 0.30, and 0.45; iterative optimization is used to... Maximize as the convergence objective, when the convergence is achieved in 11 consecutive iterations When the relative change is less than 0.7%, the optimization is considered complete, and the corresponding combination of process parameters is output.

[0017] Furthermore, in step five, the field sensing unit uniformly arranges three sets of temperature sensors, two sets of pressure sensors, one set of rheology sensors, and two sets of ultrasonic scattering detection components along the circumference of the mixing chamber inner wall. The ultrasonic scattering detection components have countersunk holes in the side wall of the mixing chamber, and are fitted with an 8mm thick alumina ceramic acoustic liner. The ultrasonic probe is attached to the outside of the liner, with the detection surface flush with the inner wall of the chamber. The sampling frequency of the temperature sensors is set to 1Hz, and the sampling frequencies of the pressure, rheology, and ultrasonic scattering sensors are uniformly set to 2Hz. The collected operating data is transmitted to the twin system via an industrial Ethernet bus, with data transmission delay controlled within 100ms. All collected data are uniformly timestamped at the millisecond level for synchronous calibration, eliminating the response time difference between different sensors and ensuring a one-to-one correspondence between the twin prediction data and the measured data time nodes. After each batch of production is completed, all collected data is automatically archived and stored, forming a production dataset for subsequent model iteration calibration.

[0018] Furthermore, in step five, the packing dispersion is obtained by joint inversion of the ultrasonic backscattering detection component and the spindle power data. The specific calculation logic is as follows: extract two feature values, the main frequency attenuation rate and the proportion of harmonic energy of the ultrasonic backscattering signal, combine them with the real-time spindle power parameters, and substitute them into the pre-calibrated ternary quadratic fitting curve to calculate the real-time packing dispersion; the fitting curve is pre-calibrated using offline detection data of the same formulation standard sample, with no less than 15 calibration points, and the relative error of the dispersion calculation is controlled within 6%; after acquiring continuous data at a sampling frequency of 2Hz, each feature value is averaged in a 5s sliding window, and a dispersion data point is output every 5s for comparison and calculation with twin prediction data.

[0019] Furthermore, in step five, the formula for calculating the comprehensive deviation coefficient is as follows: ,in This is the comprehensive deviation coefficient. The difference between the measured temperature and the twin-predicted temperature. The preset temperature deviation threshold is set to 2℃. This represents the difference between the measured dispersion and the twin-predicted dispersion. The preset dispersion deviation threshold is set to 5%. The difference between the measured Mooney viscosity and the twin-predicted Mooney viscosity is given. The preset viscosity deviation threshold is set to 3 Mooney units (MU). , , The weighting coefficients for each deviation item are 0.38, 0.37, and 0.25, respectively; when When the value is greater than 1, process parameter adjustment is triggered. If the value is less than or equal to 1, the current process parameters will be maintained.

[0020] Furthermore, in step six, when the temperature deviation triggers process adjustment, the heat transfer coefficient and heat generation coefficient parameters of the corresponding sub-model are first corrected. Then, an adjustment command is generated based on the corrected sub-model. Priority is given to adjusting the jacket refrigerant flow rate, with the flow rate adjustment step size set to 5% of the rated flow rate, and the interval between single flow rate adjustments not less than 25 seconds. If the temperature deviation still does not fall back to the threshold range after two consecutive flow rate adjustments, the rotor speed is adjusted synchronously, with the speed adjustment step size set to 2 rpm, and the total speed adjustment amplitude not exceeding ±15% of the rated speed of the corresponding stage. The speed adjustment synchronously calculates the change in shear heat generation rate and matches the refrigerant flow rate increment at a ratio of 1.15 times the shear heat generation increment. After the parameter adjustment is completed, the temperature parameters are continuously monitored for 45 seconds to confirm that they are stable before resuming normal simulation and production synchronous operation.

[0021] Furthermore, in step six, when the dispersion deviation triggers process adjustment, the shear rheological parameters of the corresponding sub-model are first corrected, and then an adjustment command is output. Priority is given to extending the runtime of the corresponding mixing stage, with a duration adjustment step size of 10 seconds. The maximum extension duration for a single adjustment does not exceed 20% of the original set duration for that stage. During the extended mixing process, the temperature control temperature, rotor speed, and top bolt pressure parameters of that stage are maintained unchanged. Dispersion data is collected every 10 seconds for verification. Once the dispersion falls back to the threshold range, the extension is immediately stopped and the process proceeds to the next mixing stage. If the dispersion still fails to meet the standard after reaching the maximum extension duration, a feeding anomaly warning signal is triggered and uploaded to the internal mixer control. The system's human-machine interface; when Mooney viscosity deviation triggers process adjustment, it first corrects the reaction kinetic parameters of the final refining stage model, then outputs adjustment commands to fine-tune the temperature control parameters for the homogenization final refining stage. The temperature control adjustment step size is set to 2℃, and the total amplitude of a single adjustment does not exceed ±5℃. The temperature control adjustment is synchronized with the rotor speed; when the temperature rises, the speed is reduced by 2 rpm, and when the temperature falls, the speed is increased by 2 rpm, so that the shear strength deviation within the stage does not exceed 8%. After each parameter adjustment, it waits 20 seconds to collect viscosity data to verify the adjustment effect until the viscosity deviation falls back to the threshold range. After the adjustment is completed, it continuously monitors the viscosity data throughout the homogenization final refining stage until the end of the stage.

[0022] Compared with existing technologies, this segmented temperature-controlled closed-loop intelligent control method for stepwise rubber mixing has the following advantages:

[0023] I. This invention significantly improves process development efficiency and mixing quality stability. Through a phased differentiated coupling digital twin model and a cross-phase micro-parameter transfer mechanism, it accurately reproduces the multi-physics coupling process of each stage of step-by-step mixing. Combined with a dynamic weight optimization algorithm with phase coupling correction, it automatically completes the optimization of process parameters, greatly reducing the number of physical trial mixings, shortening the new formulation development cycle, and effectively reducing the fluctuation range of Mooney viscosity and filler dispersion between batches.

[0024] II. This invention achieves precise closed-loop control of the production process and reduces operating energy consumption. An ultrasonic scattering detection component with an embedded alumina ceramic acoustic liner enables reliable online detection of filler dispersion. Combined with a comprehensive deviation coefficient judgment mechanism and a control logic that first corrects the model and then outputs instructions, it dynamically compensates for fluctuations in operating conditions, avoiding problems such as molecular chain degradation from over-mixing and insufficient dispersion from under-mixing. Temperature control and shear parameters at each stage are matched as needed, effectively reducing the energy consumption per unit of rubber compound during mixing.

[0025] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0027] Figure 1 This is the main flow chart of a segmented temperature-controlled closed-loop intelligent control method for stepwise rubber mixing.

[0028] Figure 2 A flowchart of segmented digital twin multiphysics coupling and parameter optimization for a segmented temperature-controlled closed-loop rubber step-by-step mixing intelligent control method.

[0029] Figure 3 This is a flowchart illustrating the online detection and closed-loop intelligent control process of a segmented temperature-controlled closed-loop rubber step-by-step mixing intelligent control method. Detailed Implementation

[0030] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0031] In the mass production scenario of natural rubber-based carbon black reinforced heavy-duty tire tread compound using a 65L meshing closed internal mixer, the segmented temperature-controlled closed-loop rubber step-by-step mixing intelligent control method of this invention is applied to carry out new formula process development and daily production closed-loop quality control.

[0032] The main equipment used in this embodiment is a 65L synchronous meshing internal mixer, equipped with a four-sided shear rotor, an effective mixing chamber volume of 65L, a total jacket heat exchange area of ​​0.82㎡, a maximum working stroke of the top bolt of 120mm, a rated rotor speed range of 0-80rpm, and a jacket coolant of circulating cooling water with a flow rate adjustment range of 0-8m³ / h. The applied formulation is a basic formulation for heavy-duty tire tread rubber, comprising 100 parts of natural rubber SCR5, 50 parts of carbon black N330, 4 parts of zinc oxide, 2 parts of stearic acid, 1.2 parts of sulfenamide accelerator NS, 1.8 parts of sulfur, and 6 parts of environmentally friendly aromatic oil.

[0033] I. Digital Twin Construction and Model Initialization and Calibration

[0034] A virtual three-dimensional model is constructed according to the 1:1 geometric scale of the physical internal mixer, which fully restores the geometric structure of the mixing chamber, rotor ridges, top plug, and jacket flow channel. Equipment structural parameters such as the number of rotor ridges, effective volume of the mixing chamber, jacket heat exchange area, and maximum stroke of the top plug are entered. At the same time, the rheological properties and rheological parameters of each component in the formula, such as rheological coefficient, thermal conductivity, density, and specific heat capacity, are imported.

[0035] For the three mixing stages—plasticization, filler dispersion, and homogenization final refining—corresponding coupled sub-models were built, employing differentiated coupling strategies. The plasticization sub-model uses a rheology-heat transfer bidirectional coupling. After each time step calculation, the global temperature field data is substituted into the rheological equation to update the viscosity values ​​at each node. Then, the updated viscosity field data is substituted into the heat transfer equation to correct the intensity of the heat source term generated by shearing, achieving real-time interaction between rheological behavior and the heat transfer process. The filler dispersion sub-model uses rotor dynamics-rheology bidirectional coupling. The apex shear rate and flow field distribution are obtained through rotor dynamics calculations, combined with the rheological model to calculate the rubber compound viscosity field. The viscosity field is then used to infer the fluid resistance experienced by the rotor, correcting the load parameters calculated by the dynamics. The homogenization final refining sub-model uses heat transfer-reaction kinetics bidirectional coupling. Temperature field data is substituted into the vulcanization reaction kinetic equation to calculate the crosslinking precursor formation rate. Then, the exothermic reaction term is superimposed onto the heat source term in the heat transfer equation, achieving coupling between the reaction process and the temperature field.

[0036] A cross-stage microscopic parameter transfer mechanism is established among the three sub-models. The molecular chain deentanglement degree calculated in the plasticization segment is used. Defined as ,in For energy storage modulus, The loss modulus is directly output from the viscoelastic model in the plasticizing section simulation. Its value ranges from 0 to 1, with a higher value indicating a higher degree of plasticization. The value range is calibrated based on experimental data showing that the degree of plasticization of natural rubber increases from approximately 0.2 to approximately 0.7 within the range of 100-130℃. It serves as the initial rheological parameter for the filler dispersion section model, specifically used to correct the zero-shear viscosity in the Cross rheological model. The correction method is as follows The physical meaning of this correction factor is as follows: Molecular chain deentanglement reduces the initial flow resistance of the compound; for every 0.1 increase in deentanglement degree, the zero-shear viscosity decreases by approximately 6%. The bound rubber content calculated from the filler dispersion section... This refers to the volume fraction of the rubber phase adsorbed on the filler surface. In this embodiment, the toluene swelling method is used for offline calibration. ,Establish The relationship between the ultrasonic attenuation coefficient A and the ultrasonic attenuation coefficient A is as follows: ,in The measured ultrasonic attenuation coefficient of pure natural rubber at a frequency of 2MHz is 0.85dB / mm. The measured attenuation coefficient of the compound at the same frequency is used as the initial reaction kinetic parameter of the homogenization final stage model to correct the vulcanization reaction rate constant. The correction method is as follows The physical basis for this modified relationship is that the binding layer reduces the mobility of the rubber molecular chains, thereby improving the contact efficiency between the accelerator and the rubber.

[0037] The model initialization and calibration were based on a Cross rheological model to describe the shear-thinning behavior of the rubber compound. Twelve sets of historical production data from the same formulation were used, and the least squares method was employed to fit the corrected viscosity-temperature coefficient and the shear-generated heat power conversion coefficient. The objective function for fitting was to minimize the average relative error between the rubber compound temperature and Mooney viscosity. After calibration, the twin model showed an average relative error of 2.7% for static prediction of rubber compound temperature and 2.9% for static prediction of Mooney viscosity, both controlled within 3%, meeting the simulation accuracy requirements.

[0038] II. Process Parameter Mapping and Virtual Operating Condition Boundary Loading

[0039] According to the production sequence of the plasticizing section, filler dispersion section, and homogenization and final refining section, the initial process parameters are entered into the corresponding sub-models in sequence. For the plasticizing section, the temperature control setting is entered as 115℃, the rotor rated speed as 40rpm, the upper jack working pressure as 0.6MPa, and the stage preset duration as 2min. For the filler dispersion section, the temperature control setting is entered as 105℃, the rotor rated speed as 55rpm, the upper jack working pressure as 0.8MPa, and the stage preset duration as 3min. At the same time, the feeding sequence node for reinforcing filler carbon black and activators zinc oxide, stearic acid, and aromatic oil is entered as 2min. For the homogenization and final refining section, the temperature control setting is entered as 85℃, the rotor rated speed as 30rpm, and the stage preset duration as 2.5min. At the same time, the feeding sequence node for vulcanization system accelerator and sulfur is entered as 5min.

[0040] During parameter mapping, boundary checks are automatically performed. The check rules include the allowable range of temperature control parameters (60℃-160℃), rotor speed (10rpm-80rpm), top bolt pressure (0.2MPa-1.2MPa), and stage duration (1min-15min). In this embodiment, all parameters entered are within the rated range. After the check passes, the virtual operating condition initialization of each sub-model is completed, and the model enters the ready-to-run state.

[0041] III. Multiphysics Coupled Simulation and Full-Process Virtual Pre-simulation

[0042] A multiphysics coupled simulation engine was launched, and a simulation platform based on the finite volume method was used for numerical solution. An adaptive variable step size mechanism was employed to run the simulation calculations. In the plasticizing stage, the viscosity of the rubber compound changed gradually, so a step size of 1 second was set to balance computational efficiency. In the filler dispersion stage, the parameters of the filler agglomerate breakup process changed drastically, so a step size of 0.5 seconds was set to ensure simulation accuracy. In the homogenization and final refining stage, the vulcanization reaction rate was stable, so a step size of 1 second was set to control the computational load. After each time step of calculation was completed, the physical field data of all nodes within the corresponding sub-model were updated synchronously.

[0043] During the simulation, characteristic data is output in stages. In the plasticizing stage, a set of global temperature field average and average viscosity data is output every 20 seconds to record the evolution of the raw rubber's plasticization degree. In the filler dispersion stage, a set of filler particle size distribution statistics and shear rate field cloud map is output every 10 seconds to track the dynamic process of agglomerate breakage and dispersion. In the homogenization and final refining stage, a set of average viscosity data and local temperature extreme values ​​of the cavity are output every 15 seconds to monitor temperature fluctuations and viscosity change trends after the addition of the vulcanizing agent. The simulation process automatically identifies abnormal nodes where the temperature exceeds the set range or the dispersion growth rate is zero, marking the corresponding spatial coordinates and time nodes. After the entire process simulation is completed, parameter evolution curves for each stage and a summary table of abnormal nodes are generated.

[0044] The preliminary results under the initial process parameters showed that the final carbon black dispersion of the rubber compound was 87.3%, the Mooney viscosity range was 4.2 units, and the energy consumption for mixing the unit rubber compound was 0.28 kWh / kg. The dispersion and viscosity fluctuations did not meet the quality index requirements of the tread rubber, and parameter optimization was required.

[0045] IV. Piecewise Dynamic Weight Algorithm Optimization with Stage Coupling Correction

[0046] With mixing quality and energy consumption as multiple optimization objectives, a segmented dynamic weighted optimization algorithm with stage coupling correction is used for iterative optimization. The overall fitness expression for the entire process is: overall fitness equals the sum of the product of the stage matching correction coefficient and the fitness of each stage segment plus its corresponding global weight. The global weights for each stage are 0.28, 0.42, and 0.30, respectively, determined based on the degree of influence of each stage on the final product quality. The filler dispersion stage has the greatest impact on product performance and has the highest weight. The stage matching correction coefficient ranges from 0.92 to 1.08, calculated based on the relative deviation between the micro-parameters output from the previous stage and the initial set value of the current stage. For every 1% increase in deviation, the coefficient decreases by 0.01, with a coefficient of 1.0 when the deviation is 0. This is used to correct the impact of stage connection errors on the global fitness.

[0047] The fitness function for a single stage is configured with indicators according to the stage. In the plasticizing stage, no filler is added, and the fitness function only includes viscosity and energy consumption, with corresponding weights of 0.68 and 0.32. The core objective of the plasticizing stage is to achieve uniform plasticization, therefore viscosity has a higher weight. The fitness functions for the filler dispersion stage and the homogenization and final refining stage include dispersion, viscosity, and energy consumption. The corresponding weights for the filler dispersion stage are 0.47, 0.31, and 0.22, with dispersion as the core control objective. The corresponding weights for the homogenization and final refining stage are 0.25, 0.30, and 0.45, with temperature and energy consumption control as the core, while also ensuring scorch safety.

[0048] Iterative optimization aims to maximize the overall fitness of the entire process. Optimization is considered complete when the relative change in overall fitness after 11 consecutive iterations is less than 0.7%. In this embodiment, the convergence condition was met after 141 iterations, outputting the optimal combination of process parameters: plasticizing section temperature control 112℃, rotor speed 45rpm, stage duration 1.8min; filler dispersion section temperature control 103℃, rotor speed 50rpm, feeding node 1.8min, stage duration 3.2min; homogenization final refining section temperature control 84℃, rotor speed 35rpm, feeding node 5.0min, stage duration 2.5min. Preliminary verification shows that under the optimal process, the carbon black dispersion reaches 91.5%, the Mooney viscosity range is 2.8 units, and the unit mixing energy consumption is 0.257kWh / kg, all indicators meeting the requirements.

[0049] V. On-site sensor deployment and synchronous acquisition of production data

[0050] Before actual production, sensing units are evenly arranged circumferentially on the inner wall of the mixing chamber. These units include three sets of platinum resistance temperature sensors, two sets of piezoelectric pressure sensors, one set of online rheological sensors, and two sets of ultrasonic scattering detection components. The ultrasonic scattering detection components have countersunk holes in the side wall of the mixing chamber, into which an 8mm thick alumina ceramic acoustic liner is embedded. The liner thickness is determined based on the principle of acoustic impedance matching; its thickness is an integer multiple of half the wavelength of ultrasound in alumina ceramic, which reduces the acoustic impedance mismatch between the steel wall and the rubber compound, improving ultrasonic transmission efficiency. A 5MHz ultrasonic probe is mounted on the outside of the liner, with the detection surface flush with the inner wall of the chamber to avoid interfering with the flow of the rubber compound.

[0051] The temperature sensor sampling frequency is set to 1Hz, while the pressure, rheology, and ultrasonic scattering sensors are all uniformly set to 2Hz. The acquired operating data is transmitted to the twin system via an industrial Ethernet bus, with data transmission latency controlled within 100ms. All acquired data is synchronized with millisecond-level timestamps to eliminate response time differences between different sensors, ensuring a one-to-one correspondence between the twin's predicted data and the measured data. After each batch of production is completed, all acquired data is automatically archived and stored, forming a production dataset for subsequent model iteration and calibration.

[0052] VI. Dispersion Inversion, Deviation Calculation and Closed-Loop Control

[0053] During production, the packing dispersion is obtained by joint inversion using ultrasonic backscattering detection components and spindle power data. Two characteristic values ​​are extracted from the ultrasonic backscattering signal: the main frequency attenuation rate and the proportion of harmonic energy. These are combined with real-time spindle power normalization parameters and substituted into a pre-calibrated ternary quadratic fitting curve to calculate the real-time dispersion. The fitting curve is pre-calibrated using offline detection data from 15 sets of standard samples with the same formulation. The relative error of the dispersion calculation is controlled within 6%. At a sampling frequency of 2Hz, each characteristic value is averaged over a 5-second sliding window, and a dispersion data point is output every 5 seconds.

[0054] The measured data after synchronous calibration are compared point-by-point with the real-time predicted data of the twin model to calculate the comprehensive deviation coefficient. The deviation coefficient is obtained by weighted summation of the normalized values ​​of the deviations of temperature, dispersion, and viscosity, with weights of 0.38, 0.37, and 0.25 respectively. The weights are determined based on the degree of influence of each indicator on the final vulcanization performance of the rubber compound. Temperature is directly related to scorch safety and has the highest weight. The temperature deviation threshold is set at 2℃, determined based on the scorch temperature safety window of the tread rubber formulation. The vulcanization initiation temperature of this formulation is approximately 95℃, and a 2℃ deviation ensures that the temperature in the final mixing stage remains within a safe range. The dispersion deviation threshold is set at 5%, corresponding to the ASTM D7723 dispersion grade 6 qualification requirement, ensuring that the product always remains within the qualification range of grade 6 or above. The viscosity deviation threshold is set at 3 Mooney units to match the allowable fluctuation range of the feed viscosity in the subsequent extrusion process. When the deviation coefficient is greater than 1, process parameter adjustments are triggered; when it is less than or equal to 1, the current parameters are maintained.

[0055] Table 1 below provides example data for calculating the comprehensive deviation coefficient K value at a typical production moment:

[0056] temperature 106.3℃ 105.0℃ 1.3℃ 2℃ 0.65 Dispersion 88.7% 91.5% 2.8% 5% 0.56 Mooney viscosity 64.2MU 63.0MU 1.2MU 3MU 0.40

[0057] At this moment, K = 0.38 × 0.65 + 0.37 × 0.56 + 0.25 × 0.40 = 0.247 + 0.207 + 0.100 = 0.554, which is less than 1, so the current parameters are maintained.

[0058] When a deviation triggers adjustment, the simulation input parameters of the corresponding sub-model are corrected first, and then adjustment commands are output based on the corrected model. When a temperature deviation triggers adjustment, the heat transfer coefficient and heat generation coefficient of the sub-model are corrected first, and then the jacket refrigerant flow rate is adjusted first. The flow rate adjustment step is 5% of the rated flow rate, and the interval between single adjustments is not less than 25 seconds. The interval length is determined based on the thermal inertia characteristics of the jacket temperature control to avoid temperature oscillations caused by frequent adjustments. If the temperature deviation does not decrease after two consecutive flow rate adjustments, the rotor speed is adjusted synchronously. The speed adjustment step is 2 rpm, and the total amplitude does not exceed ±15% of the rated speed at the corresponding stage. The speed adjustment synchronously calculates the change in shear heat generation rate, matches the refrigerant flow rate increment at a ratio of 1.15 times the shear heat generation increment, considers the heat loss coefficient of jacket heat exchange at a ratio of 1.15 times, and reserves a 15% heat exchange margin to avoid heat generation exceeding the heat exchange capacity. After the parameter adjustment is completed, continuous monitoring is performed for 45 seconds to confirm temperature stability before resuming normal operation.

[0059] When dispersibility deviation triggers adjustment, first correct the shear rheological parameters of the sub-model, then extend the runtime of the corresponding stage by 10 seconds, with a maximum extension not exceeding 20% ​​of the original runtime to avoid excessive mixing leading to molecular chain degradation. During the extension, keep other parameters constant and verify the dispersibility every 10 seconds. Once the target is met, immediately proceed to the next stage; if the target is not met even after reaching the maximum runtime, trigger a feeding anomaly warning and upload it to the human-machine interface. When Mooney viscosity deviation triggers adjustment, first correct the reaction kinetic parameters of the final refining stage sub-model, then fine-tune the temperature control parameters of the final refining stage by 2°C, with a total range not exceeding ±5°C. Temperature control adjustment is synchronized with rotation speed; if the temperature increases, decrease the rotation speed by 2 rpm; if the temperature decreases, increase the rotation speed by 2 rpm, ensuring that the shear strength deviation within the stage does not exceed 8%. After each adjustment, wait 20 seconds to verify the effect until the deviation falls back to the threshold range.

[0060] This embodiment achieves high-precision simulation and reproduction of the entire step-by-step mixing process through a segmented digital twin model and a cross-stage micro-parameter transfer mechanism. Combined with a dynamic weight optimization algorithm with stage coupling correction, it can automatically output the optimal process parameters for the adapted formulation. This embodiment has been verified to reduce the number of physical trial mixing times for new formulation process development from 8 to 2, a 75% reduction, and shorten the development cycle by approximately 55%, compared to the traditional trial mixing method. Table 2 below presents the statistical results of rubber compound quality from 30 consecutive production batches:

[0061] Carbon black dispersion (%) 91.2 1.35 2.8 ≥88 Mooney viscosity (MU) 63.0 0.98 2.1 60-66 Energy consumption per unit (kWh / kg) 0.254 0.008 0.023 ≤0.28

[0062] Validation after 30 consecutive batches showed that the average carbon black dispersion of the rubber compound reached 91.2%, with batch-to-batch fluctuations controlled within ±2.8%. The average Mooney viscosity was 63.0 MU, with fluctuations controlled within ±2.1 Mooney units. Batch consistency was significantly improved, and the energy consumption per unit of rubber compound was reduced by approximately 9.2% compared to traditional processes. The entire method is highly feasible and can effectively improve the process development efficiency and production quality stability of step-by-step mixing.

[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A segmented temperature-controlled, closed-loop intelligent control method for step-by-step rubber mixing, characterized in that, The method comprises the following components: Step 1: Construct a segmented closed-loop digital twin of the entire mixing process, integrate a multi-physics field coupling simulation engine, build corresponding coupling sub-models for the three mixing stages of plasticization, filler dispersion, and homogenization final mixing, establish a cross-stage micro-parameter transfer mechanism between sub-models, input the internal mixer equipment structural parameters and rubber formulation component parameters, and complete the initialization and calibration of the twin model. Step 2: Map the segmented temperature control curves, feeding timing nodes, rotor speed sequence, and top bolt pressure parameters of each stage of the step-by-step mixing process to the corresponding sub-models of the digital twin, and restore the boundary conditions of the complete mixing process of the three stages. Step 3: Start the multiphysics coupling simulation engine and run each coupling sub-model in the phase sequence. In the plasticizing stage, pre-simulate the temperature field distribution and Mooney viscosity change trend of the rubber compound. In the filler dispersion stage and the homogenization and final refining stage, pre-simulate the filler dispersion evolution process simultaneously. Step 4: Using the mixing quality index and energy consumption index as multiple optimization objectives, call the piecewise dynamic weight optimization algorithm with stage coupling correction to iteratively optimize the input process parameters and output the globally optimal combination of process parameters. Step 5: During the actual production process, the mixing condition data is collected in real time through the on-site sensing unit. The measured data after synchronous calibration is compared point by point with the real-time prediction data of the twin model, and the comprehensive deviation coefficient is calculated. Step 6: When the comprehensive deviation coefficient exceeds the preset threshold, the twin model first corrects the simulation input parameters of the corresponding sub-model, and then calculates and generates the adjusted process instructions based on the corrected sub-model, which are then sent to the internal mixer control system for execution to dynamically compensate for fluctuations in actual working conditions.

2. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step one, the cavity, rotor, top bolt, and jacket structure of the physical internal mixer are reproduced at a 1:1 geometric scale. The recorded equipment structural parameters include the number of rotor edges, effective volume of the mixing chamber, jacket heat exchange area, and maximum stroke of the top bolt. The recorded formulation component parameters include the rheological coefficient, thermal conductivity, density, and specific heat capacity of each rubber compound and compounding agent. The three coupled sub-models adopt differentiated coupling strategies. The plasticizing segment model adopts rheology-heat transfer bidirectional coupling. At each time step, temperature field data is substituted into the rheological equation to update the viscosity, and viscosity field data is substituted into the heat transfer equation to update the shear heat source term. The filler dispersion segment model adopts rotor dynamics-rheology bidirectional coupling, and the homogenization final mixing segment model adopts heat transfer-reaction kinetics bidirectional coupling. The cross-stage microscopic parameter transfer mechanism is as follows: the molecular chain deentanglement degree output by the plasticizing segment model is used as the initial rheological parameter of the filler dispersion segment model, and the bound rubber content output by the filler dispersion segment model is used as the initial reaction kinetic parameter of the homogenization final mixing segment model. The model initialization and calibration are based on the Cross rheological model, which calls no less than 12 sets of historical production data of the same formula. The least squares method is used to fit and correct the viscosity-temperature coefficient and the shear heat generation coefficient. The objective function is to minimize the average relative error between the rubber temperature and the Mooney viscosity, so that the average relative error of the static prediction of the twin model is controlled within 3%.

3. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step two, process parameters are sequentially entered into the corresponding sub-models according to the time sequence of the plasticizing section, the filler dispersion section, and the homogenization and final refining section. Each parameter includes the temperature control setpoint, the rotor rated speed, the working pressure of the top plug, and the preset duration of the stage. The filling dispersion section additionally enters the feeding sequence nodes of the reinforcing filler and activator, and the homogenization and final refining section additionally enters the feeding sequence nodes of the vulcanization system. During the parameter mapping process, boundary verification is automatically performed. The verification rules include a temperature control parameter range of 60℃-160℃, a rotor speed range of 10rpm-80rpm, a top plug pressure range of 0.2MPa-1.2MPa, and a stage duration range of 1min-15min. If any parameter exceeds the rated range, the mapping is intercepted and an abnormal prompt is output. After all parameters pass the verification, the virtual working condition initialization of each sub-model is completed.

4. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step three, the multiphysics coupling simulation adopts an adaptive variable step size mechanism. The step size for the plasticizing stage is set to 1 second, the step size for the filler dispersion stage is set to 0.5 seconds, and the step size for the homogenization and final refining stage is set to 1 second. After each time step is completed, the physical field data of all nodes in the corresponding sub-model are updated synchronously. Every 20 seconds, the plasticizing stage outputs a set of global temperature field average and average viscosity data. Every 10 seconds, the filler dispersion stage outputs a set of filler particle size distribution statistics and shear rate field cloud map. Every 15 seconds, the homogenization and final refining stage outputs a set of average viscosity data and local temperature extreme value data of the cavity. During the simulation, abnormal nodes that exceed the set temperature range or have a dispersion growth rate of zero are automatically identified, and their corresponding spatial coordinates and time nodes are marked. After the entire simulation is completed, parameter evolution curves for each stage and a summary table of abnormal nodes are generated.

5. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step four, the piecewise dynamic weight optimization algorithm with stage coupling correction uses a stage-adapted fitness function combined with stage matching correction coefficients for iterative optimization, as shown in the formula: ,in For overall adaptability throughout the entire process, , , The adaptability of each stage—plasticizing section, filler dispersion section, and homogenization and final refining section—is assessed. , , This represents the global weight for each stage.

6. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step five, the field sensing unit uniformly arranges three sets of temperature sensors, two sets of pressure sensors, one set of rheology sensors, and two sets of ultrasonic scattering detection components along the circumference of the mixing chamber inner wall. The ultrasonic scattering detection components have countersunk holes in the side wall of the mixing chamber, and are fitted with an 8mm thick alumina ceramic acoustic liner. The ultrasonic probe is attached to the outside of the liner, with the detection surface flush with the inner wall of the chamber. The sampling frequency of the temperature sensors is set to 1Hz, and the sampling frequencies of the pressure, rheology, and ultrasonic scattering sensors are uniformly set to 2Hz. The collected operating data is transmitted to the twin system via an industrial Ethernet bus, with data transmission delay controlled within 100ms. All collected data are uniformly timestamped at the millisecond level for synchronous calibration, ensuring a one-to-one correspondence between the twin prediction data and the measured data time points. After each batch of production is completed, all collected data is automatically archived and stored, forming a production dataset for subsequent model iteration calibration.

7. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step five, the packing dispersion is obtained by joint inversion of the ultrasonic backscattering detection component and the spindle power data. The specific calculation logic is as follows: extract two characteristic values, namely the main frequency attenuation rate and the proportion of harmonic energy, of the ultrasonic backscattering signal, combine them with the real-time spindle power parameters, and substitute them into the pre-calibrated ternary quadratic fitting curve to calculate the real-time packing dispersion; the fitting curve is pre-calibrated using offline detection data of the same formulation standard sample, with no less than 15 calibration points, and the relative error of the dispersion calculation is controlled within 6%; the dispersion data is updated every 5 seconds and used for comparison calculation with twin prediction data.

8. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step five, the formula for calculating the comprehensive deviation coefficient is as follows: ,in This is the comprehensive deviation coefficient. The difference between the measured temperature and the twin-predicted temperature. To preset the temperature deviation threshold, This represents the difference between the measured dispersion and the twin-predicted dispersion. To preset the dispersion deviation threshold, The difference between the measured Mooney viscosity and the twin-predicted Mooney viscosity is given. To preset the viscosity deviation threshold, , , These are the weighting coefficients corresponding to each deviation term.

9. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step six, when the temperature deviation triggers process adjustment, the heat transfer coefficient and heat generation coefficient parameters of the corresponding sub-model are first corrected. Then, an adjustment command is generated based on the corrected sub-model. The jacket refrigerant flow rate is adjusted first, with the flow rate adjustment step size set to 5% of the rated flow rate. The interval between single flow rate adjustments is not less than 25 seconds. If the temperature deviation still does not fall back to the threshold range after two consecutive flow rate adjustments, the rotor speed is adjusted synchronously, with the speed adjustment step size set to 2 rpm. The total speed adjustment amplitude does not exceed ±15% of the rated speed of the corresponding stage. The speed adjustment is synchronized with the calculation of the change in shear heat generation rate, and the refrigerant flow rate increment is matched at a ratio of 1.15 times the shear heat generation increment.

10. The segmented temperature-controlled closed-loop intelligent control method for step-by-step rubber mixing according to claim 1, characterized in that, In step six, when the dispersion deviation triggers process adjustment, the shear rheological parameters of the corresponding sub-model are first corrected, and then an adjustment command is output, prioritizing the extension of the running time of the corresponding mixing stage. During the extended mixing process, the temperature control temperature, rotor speed, and top bolt pressure parameters of this stage are kept constant. Dispersion data is collected every 10 seconds for verification. Once the dispersion falls back to the threshold range, the extension is immediately stopped and the process proceeds to the next mixing stage. If the dispersion still does not meet the standard after reaching the maximum extension time, an abnormal feeding warning signal is triggered and uploaded to the human-machine interface of the internal mixer control system. When the Mooney viscosity deviation triggers process adjustment, the reaction kinetic parameters of the final refining stage sub-model are first corrected, and then an adjustment command is output to fine-tune the temperature control parameters for the homogenization final refining stage. The temperature control adjustment is synchronized with the rotor speed. When the temperature rises, the speed is reduced by 2 rpm, and when the temperature falls, the speed is increased by 2 rpm, so that the shear strength deviation within the stage does not exceed 8%. After each parameter adjustment, viscosity data is collected for 20 seconds to verify the adjustment effect until the viscosity deviation falls back to the threshold range. After the adjustment is completed, the viscosity data of the entire homogenization final refining stage is continuously monitored until the end of this stage.

Images