A multi-parameter collaborative control system of an extruder production line

Abstract

The application relates to the technical field of industrial control and discloses a multi-parameter cooperative control system of an extruder production line, which comprises the following steps: synchronously collecting operation parameters to peel off transient viscous dissipation power, combining with the internal effective volume of equipment and the transient flow tracking of fluid to track the space-time delay time of convection mass transfer; using the delay time and the dissipation power to deduce predicted temperature and pressure, extracting a macro energy flow ratio to construct a preset target functional; then, under the driving of the functional, combining with the damping inertia weight, spontaneous iteration optimization is carried out to obtain optimal speed and power double-instruction increments; finally, based on the mechanical-electrical-thermal boundary energy conservation, overload prevention is carried out, and the time difference value of the transient thermodynamic entropy production rate of an isolated system is calculated to close-loop trigger rectification.

Description

Technical Field

[0001] This invention relates to the field of industrial control technology, and more specifically, to a multi-parameter collaborative control system for an extruder production line. Background Technology

[0002] Extruders are widely used in the processing of polymer materials. Their production lines typically consist of a main motor, heaters, a barrel, screw, metering section, and forming die, connected in series. In actual processing, the polymer melt follows the fundamental laws of non-Newtonian pseudoplastic fluids and thermodynamics, exhibiting extremely complex spatiotemporal delays and multivariable interference phenomena in the high-shear confined region of the barrel and screw and the high-damping region of the die. Traditional control systems often employ a single-variable independent control strategy, assuming that screw speed is only positively correlated with extrusion pressure and heater power is only positively correlated with temperature. However, while increasing screw speed increases mechanical thrust, it also triggers intense friction between macromolecular chains, generating significant viscous heat dissipation, leading to a localized temperature spike and a rapid decrease in the apparent viscosity of the melt. This thinned, high-temperature, low-viscosity melt reaches the end channel after a long convective delay, causing a secondary, delayed pressure drop. Faced with this spatiotemporal asymmetry in interference and pressure reduction, conventional experience-based controllers are prone to making the wrong judgment that the pressure is insufficient, and then blindly increasing the speed, inducing more severe viscosity reduction and rheological oscillation, causing the machine to fall into a vicious cycle of increasing speed and decreasing pressure. Summary of the Invention

[0003] This invention provides a multi-parameter collaborative control system for an extruder production line, which solves the technical problems mentioned in the background art.

[0004] This invention provides a multi-parameter collaborative control system for an extruder production line, applicable to extrusion equipment including a main motor, heater, barrel, screw, metering section, and forming die, comprising: Collect system operating parameters, which include at least the actual heating power; The transient volumetric flow rate and transient viscous dissipation power are calculated based on the system operating parameters. The time-space delay is calculated by combining the internal effective volume of the extrusion equipment with the transient volumetric flow rate. By using the aforementioned spatiotemporal delay time and the aforementioned transient viscous dissipation power, forward extrapolation is performed to obtain the predicted temperature and predicted pressure; Based on the transient viscous dissipation power and the actual heating power, a target functional is constructed by combining the predicted temperature and the predicted pressure. The damping inertia weight is generated using the spatiotemporal delay time, and the optimal control increment is obtained through iterative optimization under the drive of the target functional. Based on the physical boundaries of the equipment, the optimal control increment is truncated to prevent overload, and actual power command and actual speed command are generated and controlled to be executed. The transient thermodynamic entropy yield of the system and its time difference value are calculated, and the operating status is evaluated based on the time difference value to trigger closed-loop correction.

[0005] The beneficial effects of this invention include: by stripping transient viscous dissipation power in real time and tracking the spatiotemporal delay of convective mass transfer, a forward evolution model spanning time and space is constructed. Then, by combining the macroscopic energy flow ratio of endogenous heat and exogenous heat, a collaborative optimization of functionals without a preset target is performed, effectively overcoming the hysteresis and asymmetric interference problems caused by the alternation of high shear confined space and high damping throttling space. At the same time, overload prevention is achieved by relying on the objective limit of boundary energy conservation, and the time difference value of the transient thermodynamic entropy yield of the system is introduced as an evolution trend evaluation index for closed-loop correction, thereby avoiding system oscillation and runaway caused by traditional empirical parameter tuning, and significantly improving the stability and multivariate control accuracy of the extrusion process. Attached Figure Description

[0006] Figure 1 This is a flowchart of a multi-parameter collaborative control system for an extruder production line according to the present invention. Detailed Implementation

[0007] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0008] like Figure 1 As shown, a multi-parameter collaborative control system for an extruder production line is applied to an extrusion equipment including a main motor, heater, barrel, screw, metering section, and forming die, comprising: Collect system operating parameters, which include at least the actual heating power; The transient volumetric flow rate and transient viscous dissipation power are calculated based on the system operating parameters. The time-space delay is calculated by combining the internal effective volume of the extrusion equipment with the transient volumetric flow rate. By using the aforementioned spatiotemporal delay time and the aforementioned transient viscous dissipation power, forward extrapolation is performed to obtain the predicted temperature and predicted pressure; Based on the transient viscous dissipation power and the actual heating power, a target functional is constructed by combining the predicted temperature and the predicted pressure. The damping inertia weight is generated using the spatiotemporal delay time, and the optimal control increment is obtained through iterative optimization under the drive of the target functional. Based on the physical boundaries of the equipment, the optimal control increment is truncated to prevent overload, and actual power command and actual speed command are generated and controlled to be executed. The transient thermodynamic entropy yield of the system and its time difference value are calculated, and the operating status is evaluated based on the time difference value to trigger closed-loop correction.

[0009] This embodiment is applied to an extrusion equipment including a main motor, heater, barrel screw metering section, and forming die. The system adopts a fixed-cycle real-time task scheduling architecture, with the scheduling cycle consistent with the system sampling cycle. The sampling cycle can be set from 1 millisecond to 1 second to adapt to the dynamic response requirements of different extrusion equipment. The system's real-time tasks are prioritized from high to low as follows: sensor data acquisition and preprocessing, core parameter calculation, iterative optimization, control command generation and output, operating condition assessment and closed-loop correction, and fault degradation processing. All tasks are synchronized within a single scheduling cycle, and a double-buffering mechanism is used for data interaction between tasks to avoid data contention and timing errors. All physical quantity calculations in the system are enforced to use the International System of Units (SI), and all input and output parameters of formulas follow the SI dimension matching rules. Data acquired in non-SI formats must be converted to units before being substituted into the formulas for calculation.

[0010] The system operating parameters are collected, including main motor torque, measured screw angular velocity, head pressure differential, mass flow rate, melt density, and actual heating power. The acquisition and preprocessing procedures for each parameter are as follows: The main motor torque is collected; the main motor torque is the real-time output torque of the extruder's main drive motor output shaft, denoted by [symbol missing]. The International System of Units (SI) unit is the newton-meter (N / m), with the symbol 1. This parameter is synchronously acquired via the motor's built-in torque sensor or an external high-precision torque meter. The acquisition period is completely consistent with the system scheduling period. The acquired data is synchronized via the NTP network time protocol, with a synchronization error not exceeding 10% of the sampling period. The acquired raw torque signal must first undergo first-order low-pass filtering. The filtering formula is: ; in, This represents the original torque value collected at time t. This is the filtered torque value from the previous sampling time. The system sampling period is The filter coefficient has a value range of [0.1, 1] and can be adjusted according to the on-site noise level. The filtered torque value needs to be checked for boundary validity. When the collected value is less than 0 or greater than the rated maximum torque of the motor, it is determined to be invalid data, and the valid filtered value from the previous moment is used to replace it, so as to avoid invalid data causing errors in subsequent calculations.

[0011] The measured screw angular velocity is the real-time rotational angular velocity of the extruder screw, denoted by . The International System of Units (SI) unit is the radian per second, with the symbol . This parameter is synchronously acquired via the incremental encoder built into the main motor, with the acquisition period matching the system sampling period and strictly aligned with the torque acquisition clock. The acquired raw angular velocity signal undergoes the same first-order low-pass filtering as the torque signal, with consistent filter coefficients to ensure matching dynamic response characteristics between the two parameters. The filtered angular velocity value undergoes boundary validity verification; if the acquired value is less than 0 or greater than the motor's rated maximum angular velocity, it is considered invalid data and replaced with the valid filtered value from the previous moment.

[0012] The die head pressure differential is the difference between the melt pressure at the extruder die inlet and the ambient pressure at the outlet, denoted by [symbol missing]. The International System of Units (SI) unit is the Pascal, with the symbol pascal. The pressure differential is measured using a gauge pressure method to ensure that the physical meaning of the pressure difference is fully matched with the calculation logic of the effective pumping power. This parameter is synchronously acquired through a high-temperature melt pressure sensor installed at the inlet of the pump head, with the acquisition period consistent with the system sampling period and strictly aligned with the torque and angular velocity acquisition clocks. The acquired raw pressure differential signal undergoes the same first-order low-pass filtering process as the torque signal, with consistent filter coefficients. The filtered pressure differential value undergoes boundary validity verification. When the acquired value is less than 0 or greater than the equipment's rated maximum melt pressure, it is determined to be invalid data, and the valid filtered value from the previous moment is used instead.

[0013] Mass flow rate is the mass of polymer melt output from the extruder through the die per unit time, denoted by the symbol . The International System of Units (SI) unit is the kilogram per second, with the symbol . This parameter is synchronously acquired via a Coriolis mass flow meter installed in the melt pipeline, or calculated using the unit-time weight change from a weighing metering device. The acquisition period is consistent with the system sampling period and strictly aligned with the clocks of other acquired parameters. The acquired raw mass flow signal undergoes the same first-order low-pass filtering as the torque signal, with consistent filter coefficients. The filtered mass flow value undergoes boundary validity verification. When the acquired value is less than 0, it is considered invalid data and replaced with 0; when the acquired value is greater than the equipment's rated maximum extrusion flow rate, it is considered invalid data and replaced with the valid filtered value from the previous moment.

[0014] The melt density is collected; the melt density is the density of the polymer melt under processing conditions, denoted by the symbol . The SI unit is kilogram per cubic meter, with the symbol . This parameter can be obtained in two ways. The first way is to look up the melt density value within the processing temperature range in the corresponding polymer material's property handbook, and then use the accompanying temperature correction formula to adapt to the working conditions. The temperature correction formula is as follows: ; in, Reference temperature The melt density is constant at the following values. The standard processing reference temperature for polymer materials, in degrees Celsius. The coefficient of volumetric expansion of the melt, expressed in degrees Celsius, is represented by the symbol . The first method is to obtain the density from a material property handbook. The second method is to collect the melt density in real time using an online density measurement device. The acquisition period is consistent with the system sampling period and strictly aligned with the clock of other acquisition parameters. The acquired real-time density signal needs to undergo the same first-order low-pass filtering process as the torque signal, with the filtering coefficients kept consistent. The filtered value needs to be checked for boundary validity. If it exceeds the density range given in the material property handbook, the valid value at the previous moment is used instead.

[0015] The actual heating power is collected, which is the sum of the real-time output power of the heaters in each section of the extruder barrel, denoted by . The International System of Units (SI) unit is the watt, with the symbol . This parameter is synchronously acquired by the power acquisition modules of each heater drive circuit, with the acquisition period consistent with the system sampling period and strictly aligned with the clocks of other acquired parameters. The acquired raw power signal needs to undergo the same first-order low-pass filtering process as the torque signal, with the filtering coefficients remaining consistent. The filtered power value needs to be checked for boundary validity. When the acquired value is less than 0, it is determined to be invalid data and replaced with 0; when the acquired value is greater than the rated maximum total power of the heater, it is determined to be invalid data and replaced with the valid filtered value from the previous moment.

[0016] Dividing the mass flow rate by the melt density yields the transient volumetric flow rate, calculated using the following formula: ; in, The transient volumetric flow rate at time t, expressed in cubic meters per second (SI), with the symbol . This characterizes the real-time volumetric flow rate of the polymer melt within the barrel flow channel. When it is 0, Take the value 0 directly to avoid division by zero errors.

[0017] Multiply the main motor torque by the measured screw angular velocity to obtain the total mechanical power. The calculation formula is as follows: ; in, The total mechanical power at time t is expressed in watts (SI), with the symbol t. This represents the total mechanical power input from the main motor to the screw system. When or If the value is invalid, the effective total mechanical power value from the previous moment shall be used instead.

[0018] Multiplying the head pressure differential by the transient volumetric flow rate, the effective pumping power is obtained using the following formula: ; in, The effective pumping power at time t is expressed in watts (SI), with the symbol t. This characterizes the effective mechanical power used to propel the melt forward. When or When it is 0, Take the value 0 directly.

[0019] Subtracting the effective pumping power from the total mechanical power yields the transient viscous dissipation power, calculated using the following formula: ; in, The transient viscous dissipation power at time t, with the SI unit being the watt and the symbol . This represents the real-time power converted into heat energy due to friction and viscous dissipation within polymer chains during the screw shearing of the melt, and is the core source of endogenous heat in the extrusion process. When the calculated result is less than 0, We directly set the value to 0 to avoid negative power dissipation results that have no physical meaning.

[0020] The effective internal volume formed by the extruder barrel, screw metering section, and forming die is obtained. This effective internal volume is the sum of the cavity volume of the extruder barrel, screw metering section, and the flow channel volume of the forming die, denoted by [symbol missing]. The International System of Units (SI) unit is the cubic meter, with the symbol . The cavity volume of the metering section of the barrel screw is the total volume of the screw groove in the metering section. The starting and ending points of the metering section are from the end of the melting section of the screw to the front end of the screw head. The volume of the forming die flow channel is the total flow channel volume from the die inlet flange to the die outlet. This parameter can be calculated from the geometric dimensions in the extrusion equipment design drawings or obtained directly from the equipment's factory parameters. It is a fixed structural parameter of the equipment, and this parameter value needs to be updated synchronously when the die is replaced.

[0021] Dividing the effective internal volume by the transient volumetric flow rate yields the spatiotemporal delay time, calculated using the following formula: ; in, Let be the spacetime delay at time t, with the SI unit being the second and the symbol . This characterizes the theoretical time required for the polymer melt to flow from the metering section inlet to the die outlet, reflecting the inherent time lag between shear heat generation and pressure / temperature response during extrusion. When it is 0, The preset maximum delay time is taken, which is the theoretical delay time under the minimum extrusion flow rate of the equipment, to avoid the calculation error of dividing by zero. The calculated... Boundary truncation is required. When the calculation result is greater than the preset maximum delay time, the preset maximum delay time is used; when the calculation result is less than the sampling period, the sampling period is used to ensure the effectiveness of the subsequent historical data window delineation.

[0022] A historical data cache sequence is constructed. The system pre-establishes a ring-shaped historical data cache sequence. The storage depth of the cache sequence is not less than the number of sampling points corresponding to the preset maximum delay time. Each storage unit in the cache sequence corresponds to a timestamp of a sampling moment and the transient viscous dissipation power, main motor torque, measured screw angular velocity, head pressure differential, mass flow rate, melt density, actual heating power, current measured temperature, and current measured head pressure parameters at that moment. The cache sequence adopts an overwrite update rule. In each sampling period, the latest valid parameters are written to the latest storage unit of the cache sequence. When the cache sequence is full, the oldest storage unit is automatically overwritten to ensure that the cache sequence always retains the latest historical data that meets the maximum delay time requirement.

[0023] Based on the spatiotemporal delay, a historical data extraction window is defined within the historical data cache sequence. The time range of the historical data extraction window is [missing information]. Where t is the current time. During window delineation, if the window start timestamp is not perfectly aligned with the sampling timestamps stored in the cache sequence, linear interpolation is used to obtain the parameter values ​​at the corresponding time to ensure the temporal continuity of the data within the window. When the system starts for the first time and the length of the data in the cache sequence is less than the spatiotemporal delay time, the missing data in the window is filled with the valid parameters at the current time to ensure that the window delineation logic can be executed normally during the first startup.

[0024] Obtain the basic parameters for forward inference. These parameters include candidate heating power, candidate screw speed, prediction step size, current measured temperature, current measured head pressure, specific heat capacity, rheological activation energy, and ideal gas constant. The candidate heating power is the measured value of the heater output power to be verified during the iterative optimization process, denoted by [symbol missing]. The International System of Units (SI) unit is the watt, with the symbol . , is one of the optimization variables in the iterative optimization process. The candidate screw speed is the test value of the screw rotational angular velocity to be verified during the iterative optimization process, denoted by . The International System of Units (SI) unit is the radian per second, with the symbol . , is one of the optimization variables in iterative optimization. The prediction step size is the time span of a single forward inference, denoted by . The International System of Units (SI) unit is the second, with the symbol . The value is taken exactly in line with the system sampling period to ensure that the deduction process is aligned with the data acquisition sequence and to avoid time lag deviation. The current measured temperature is the real-time melt temperature at the end of the metering section of the extruder barrel, denoted by . The International System of Units (SI) unit is Celsius, with the symbol . The temperature signal is synchronously acquired via armored thermocouples installed on the inner wall of the barrel, with the acquisition period consistent with the system sampling period and strictly aligned with the clocks of other acquired parameters. The acquired raw temperature signal undergoes the same first-order low-pass filtering as the torque signal, with consistent filtering coefficients. The filtered value undergoes boundary validity verification; if it exceeds the processing temperature range of the polymer material, the valid value from the previous moment is used instead. The currently measured die head pressure is the gauge pressure of the melt at the die inlet corresponding to the die head pressure differential at time t, denoted by [symbol missing]. The International System of Units (SI) unit is the Pascal, with the symbol pascal. The same data acquisition and preprocessing methods are used as those for differential pressure establishment at the die head. Specific heat capacity is the isobaric specific heat capacity of the polymer melt under processing conditions, denoted by [symbol missing]. The SI unit is joule per kilogram (Kelvin), with the symbol . The specific heat capacity can be obtained as a constant value within the processing temperature range from polymer material property handbooks, or it can be acquired in real time using an online specific heat capacity measuring device. When using real-time acquisition, the acquisition cycle should be consistent with the system sampling cycle, and the data should undergo the same filtering and validity verification processing. The rheological activation energy is the viscous flow activation energy of the polymer melt, denoted by . The SI unit is joule per mole, with the symbol joule. These are inherent physical properties of the corresponding polymer materials, which can be obtained from the material rheological performance test report. These parameters need to be updated when changing the processed material. The ideal gas constant is a universal thermodynamic constant, denoted by . The value is 8.314 joules per mole Kelvin, with the symbol . , is a fixed constant.

[0025] The heat of conduction is obtained by multiplying the candidate heating power by the prediction step size, and the calculation formula is as follows: ; in, To predict the total amount of conductive heat input from the heater within a step length, the SI unit is the joule, with the symbol joule. It represents the total energy input from external heat sources.

[0026] The cumulative shear heat is obtained by integrating the transient viscous dissipation power within the historical data extraction window over time. For discretely sampled time-series data, the trapezoidal integral method is used for numerical integration calculation, and the calculation formula is as follows: ; in, This refers to the total cumulative shear heat generated by viscous dissipation within the historical data extraction window, measured in joules (J) using the International System of Units (SI). This represents the total amount of endogenous heat absorbed by the melt that arrives at the die at the current moment during its flow. This is the integration time variable, and its value range is the time interval of the historical data extraction window. This represents the total number of sampling points within the historical data extraction window. This is the timestamp corresponding to the kth sampling point within the window; This represents the transient viscous dissipation power value corresponding to the k-th sampling point within the window.

[0027] Add the conductive heat and the accumulated shear heat, divide by the product of the effective internal volume, melt density, and specific heat capacity, and then add the current measured temperature to obtain the predicted temperature. The calculation formula is as follows: ; in, The predicted melt temperature at the end of the predicted step is expressed in degrees Celsius, with the symbol . This represents the melt temperature evolution after a predicted step size, under candidate heating power and candidate screw speed. The unit of total heat in the numerator is joules, the unit of total heat capacity in the denominator is joules per Kelvin, and the dimension of temperature change is Kelvin. The temperature change value under the Kelvin scale is completely consistent with the temperature change value under the Celsius scale, therefore it can be directly superimposed to the current measured temperature under the Celsius scale. The calculated... Physical boundary truncation is required. The upper and lower limits of the truncation are the decomposition temperature and viscous flow temperature of the polymer material, respectively. When the temperature exceeds the boundary, the corresponding boundary value is taken to avoid temperature results that are not physically meaningful.

[0028] Divide the rheological activation energy by the ideal gas constant, then multiply by the difference between the reciprocal of the predicted temperature and the reciprocal of the current measured temperature. Calculate the natural exponent of this product to obtain the viscosity decay factor. The formula is as follows: ; in, is the viscosity decay factor, a dimensionless parameter that characterizes the relative change in apparent viscosity of the polymer melt caused by temperature changes. An addition of 273.15 to the temperature value in the formula converts the Celsius temperature to a thermodynamic Kelvin scale, satisfying the dimensionality and calculation specifications of the Arrhenius viscosity equation. The independent variable of the exponential function needs to be truncated within the range [-20, 20]. Values ​​exceeding the boundary are taken to avoid invalid calculation results due to exponential function overflow or underflow.

[0029] The momentum response factor is obtained by dividing the candidate screw rotation speed by the measured screw angular velocity. The calculation formula is as follows: ; in, is the momentum response factor, a dimensionless parameter that characterizes the relative change in melt shear rate and flow momentum caused by changes in screw speed. When When it is 0, Use the value 1 to avoid division by 0 errors.

[0030] The predicted pressure is obtained by multiplying the current measured head pressure by the momentum response factor and the viscosity decay factor in sequence. The calculation formula is as follows: ; in, The predicted nose pressure at the end of the step is given by the SI unit Pascal, with the symbol pascal. This characterizes the evolution of the die head pressure after a predicted step size, under candidate heating power and candidate screw speed. The calculated... Physical boundary truncation is required, with a lower limit of 0 and an upper limit of the equipment's rated maximum melt pressure. When the pressure exceeds the boundary, the corresponding boundary value is taken to avoid pressure results that are not physically meaningful.

[0031] Obtain the target process parameters, including the target pressure and target temperature. The target pressure is the die melt pressure setpoint required by the extrusion process, denoted by [symbol missing]. The International System of Units (SI) unit is the Pascal, with the symbol pascal. The target temperature is the melt temperature setpoint required by the extrusion process, denoted by the symbol . The International System of Units (SI) unit is Celsius, with the symbol . Both parameters are preset based on the processing requirements of the material and the molding requirements of the product. They are the target values ​​for system control and need to be updated synchronously when the processing material or product is changed.

[0032] Dividing the transient viscous dissipation power by the sum of its value and the actual heating power yields the endogenous heat ratio, calculated using the following formula: ; in, The endogenous heat ratio is a dimensionless parameter with a value range of [0,1]. It characterizes the proportion of endogenous heat generated by viscous dissipation in the total heat input of the system, reflecting the weight of the screw shearing action on the rheological properties of the melt. and When both are 0 Use a value of 0.5 to avoid errors caused by dividing by zero.

[0033] Divide the actual heating power by the sum of its actual and transient viscous dissipation power to obtain the external heat source ratio. The calculation formula is as follows: ; in, The external heat ratio is a dimensionless parameter, ranging from [0,1], representing the proportion of external heat input from the heater in the total system heat input, reflecting the weight of external heating on the melt rheological properties. The sum of the endogenous heat ratio and the external heat ratio is 1, achieving weight normalization of the total heat input. When... and When both are 0 Use a value of 0.5 to avoid errors caused by dividing by zero.

[0034] The pressure residual is obtained by dividing the difference between the predicted pressure and the target pressure by the target pressure and then squaring the result. The calculation formula is as follows: ; in, The pressure residual is a dimensionless parameter that characterizes the relative deviation between the predicted and target pressures. A squaring operation is used to eliminate the influence of the sign of the deviation and amplify the weight of larger deviations. During the calculation, the predicted and target pressures must use the same units and measurement methods to ensure the accuracy of the relative deviation calculation.

[0035] The temperature residual is calculated by dividing the difference between the predicted temperature and the target temperature by the target temperature and then squaring the result. The formula is as follows: ; in, The temperature residual is a dimensionless parameter that characterizes the relative deviation between the predicted and target temperatures. A squaring operation is used to eliminate the influence of the sign of the deviation and amplify the weight of larger deviations. During the calculation, the predicted and target temperatures must use the same temperature scale and units to ensure the accuracy of the relative deviation calculation.

[0036] Multiplying the endogenous heat ratio by the pressure residual and adding the exogenous heat ratio by the temperature residual, we obtain the objective functional, calculated as follows: ; in, Let be the objective functional and be a dimensionless parameter. This objective functional represents the optimization target in the iterative optimization process; a smaller value indicates a smaller deviation between the predicted operating conditions and the target process requirements. The objective functional is updated synchronously after each particle iteration. During the update process, the endogenous heat ratio, exogenous heat ratio, predicted temperature, and predicted pressure values ​​are simultaneously refreshed to ensure a perfect match between the optimization process and real-time operating conditions. Through the weighted allocation of the endogenous heat ratio and the exogenous heat ratio, the collaborative weights of pressure control and temperature control are adaptively adjusted. When the proportion of endogenous heat is high, the system automatically increases the weight of pressure control; when the proportion of exogenous heat is high, the system automatically increases the weight of temperature control, thus resolving the collaborative control problem of multi-variable coupling interference.

[0037] Divide the negative spatiotemporal delay by the prediction step size and then perform a natural exponential calculation to obtain the damped inertia weight. The calculation formula is as follows: ; in, The damped inertia weight at time t is a dimensionless parameter with a value range of (0,1], used to adjust the degree of inheritance of historical iteration step sizes during iterative optimization. The independent variable of the exponential function needs to be truncated, with a truncation range of [-10,0]. When it exceeds the boundary, the corresponding boundary value is taken to avoid invalid calculation results due to exponential function underflow. When the time-space delay time increases, the damped inertia weight decreases accordingly to reduce the inertia of the iteration process, avoid iterative oscillations caused by the large lag characteristics of the system, and improve the stability of the optimization process.

[0038] The cognitive learning factor is calculated by dividing the absolute value of the difference between the current measured head pressure and the target pressure by the target pressure, taking a negative sign, performing natural exponentiation, and then subtracting it from the result. The formula is as follows: ; in, Let be the cognitive learning factor at time t, a dimensionless parameter with a value range of [0,1]. It is used to adjust the step size weight when learning from the individual's historical best solution during iterative optimization. As the pressure bias increases, the cognitive learning factor increases accordingly, increasing the weight of the individual's learning and accelerating the convergence speed of the pressure bias.

[0039] The social learning factor is calculated by subtracting the target temperature from the current measured temperature, dividing the absolute value by the target temperature, taking a negative sign, performing natural exponentiation, and then subtracting it from the result. The formula is as follows: ; in, Let be the social learning factor at time t, a dimensionless parameter with a value range of [0,1]. It is used to adjust the step size weight for learning from the global optimal solution of the population during the iterative optimization process. As the temperature deviation increases, the social learning factor increases accordingly, enhancing the weight of the population learning and accelerating the convergence speed of the temperature deviation.

[0040] After completing the initial configuration of the particle swarm optimization (PSO) algorithm, the optimization variables are two-dimensional command vectors. The two dimensions correspond to screw speed and heating power, respectively. The two dimensions employ completely independent iterative operation rules, with independent parameter constraints and step size clamps set to avoid interference from parameters of different dimensions. The number of particles in the PSO algorithm is set from 20 to 100, and can be adjusted according to the equipment's computing power and real-time control requirements. The initial command vector values ​​are as follows: the initial value for the screw speed dimension is uniformly and randomly generated between 0 and the motor's rated maximum angular velocity; the initial value for the heating power dimension is uniformly and randomly generated between 0 and the heater's rated maximum total power. The initial iteration step size vector values ​​are as follows: the initial step size for the screw speed dimension is uniformly and randomly generated between -0.1 times the rated maximum angular velocity and 0.1 times the rated maximum angular velocity; the initial step size for the heating power dimension is uniformly and randomly generated between -0.1 times the rated maximum total power and 0.1 times the rated maximum total power. The velocity clamping boundaries for the iteration step size are as follows: the maximum allowable step size for a single step in the screw speed dimension is 0.2 times the rated maximum angular velocity, and the maximum allowable step size for a single step in the heating power dimension is 0.2 times the rated maximum total power. During iteration, step sizes exceeding the clamping boundaries are directly taken from the corresponding boundary values ​​to avoid excessively large iteration steps that exceed the optimal solution or excessively small steps that fall into local optima. Iteration convergence conditions include: the number of iterations reaches a preset maximum number of iterations (range 50 to 200); the globally optimal objective functional value does not decrease for 10 consecutive iterations; and the globally optimal objective functional value is less than a preset convergence threshold (range 1e-6 to 1e-4). Iteration terminates when any of these conditions are met. The update rules for the individual optimal position vector and the globally optimal position vector are as follows: the corresponding optimal position vector is updated only when the objective functional value obtained in the new iteration is strictly less than the historical optimal value; if it is equal to the historical optimal value, the original optimal position vector remains unchanged to avoid invalid updates. During the iteration process, the candidate screw speed and candidate heating power need to be subject to physical boundary constraints. The constraint range of screw speed is 0 to the rated maximum angular velocity of the motor, and the constraint range of heating power is 0 to the rated maximum total power of the heater. Candidate values ​​that exceed the boundaries are directly taken from the corresponding boundary values ​​to ensure that the candidate instructions generated during the iteration process are physically executable.

[0041] Obtain the basic vectors and parameters for iterative optimization. These include the current iteration step size vector, the current command vector, the local extremum position vector, the global extremum position vector, the first random number, and the second random number. The current iteration step size vector is the control command iteration step size corresponding to the i-th particle in the k-th iteration, denoted by [symbol missing]. Let be a two-dimensional column vector, with the two dimensions corresponding to the iteration step size of the screw speed and heating power, respectively. The SI units are radians per second per iteration and watts per iteration, where iteration is a dimensionless counting unit to ensure that the dimensions of the step size vector and the command vector can be matched for computation. The current command vector is the test value of the control command corresponding to the i-th particle in the k-th iteration, denoted by . Let be a two-dimensional column vector, with the two dimensions corresponding to the candidate screw speed and the candidate heating power, respectively, in radians per second and watts (SI units). The local extremum position vector is the command vector corresponding to the i-th particle obtaining the minimum objective functional value in the iteration history, denoted by . , which is a two-dimensional column vector, where each particle independently maintains its own local extremum position vector, which is updated in real time during the iteration process. The global extremum position vector is the instruction vector corresponding to the minimum objective functional value obtained by the entire particle swarm in the iteration history, denoted by . is a two-dimensional column vector. The entire particle swarm maintains a unique global extremum position vector, which is updated in real time during the iteration process. The first random number and the second random number are uniformly distributed random numbers with values ​​in the range [0,1], denoted as and respectively. and Each particle is independently regenerated during each iteration to increase the randomness of the iteration process and prevent the optimization process from getting stuck in a local optimum.

[0042] The damped inertia weight is multiplied by the current iteration step size vector to obtain the inertia vector components. The calculation formula is as follows: ; in, In the (k+1)th iteration, the inertial vector component of the i-th particle is represented by a two-dimensional column vector. This vector inherits the iteration step size from the previous iteration, maintaining the search trend during the optimization process. The multiplication operation is a scalar multiplication of a scalar and a vector, with each dimension calculated independently to ensure the operational independence of parameters with different dimensions.

[0043] The cognitive vector component is obtained by multiplying the cognitive learning factor, the first random number, and the difference between the local extremum position vector and the current instruction vector. The calculation formula is as follows: ; in, In the (k+1)th iteration, the cognitive vector component of the i-th particle is a two-dimensional column vector used to drive the particle to search towards its own historical optimal solution, thereby achieving learning and optimization based on individual experience. The multiplication operation is a scalar multiplication of a scalar and a vector, with the two dimensions calculated independently to ensure the operational independence of parameters with different dimensions.

[0044] The social vector component is obtained by multiplying the social learning factor, the second random number, and the difference between the global extreme value position vector and the current instruction vector. The calculation formula is as follows: ; in, In the (k+1)th iteration, the social vector component of the i-th particle is a two-dimensional column vector used to drive the particle to search towards the global optimal solution of the swarm, achieving collaborative optimization of swarm experience. The multiplication operation is a scalar multiplication of a scalar and a vector, with the two dimensions calculated independently to ensure the independence of operations on parameters of different dimensions.

[0045] The update iteration step size vector is obtained by adding the inertial vector component, the cognitive vector component, and the social vector component. The calculation formula is as follows: ; in, Let be the update iteration step size vector for the i-th particle in the (k+1)-th iteration, which is a two-dimensional column vector. The addition operation is an element-wise addition of the vectors, with the two dimensions calculated independently. After the calculation is completed, the step size value of each dimension needs to be velocity clamped, and the step size value that exceeds the clamping boundary is taken as the corresponding boundary value.

[0046] The updated instruction vector is obtained by adding the update iteration step size vector to the current instruction vector. The calculation formula is as follows: ; in, Let be the update instruction vector for the i-th particle in the (k+1)-th iteration, which is a two-dimensional column vector. The addition operation is an element-wise addition of the vectors, with the two dimensions calculated independently. After the calculation is completed, the instruction values ​​of each dimension need to be subject to physical boundary constraints. Instruction values ​​that exceed the constraint range are taken from the corresponding boundary values.

[0047] The process involves iterative convergence judgment and optimal control increment extraction. During iteration, after each iteration update of the entire particle swarm, the objective functional value corresponding to the update command vector of each particle is calculated synchronously, and the individual optimal position vector of each particle and the global optimal position vector of the entire particle swarm are updated. Then, it is determined whether the iterative convergence condition is met. If not, the iteration count is incremented by 1, and the iteration step size and command vector update steps are repeated. If the condition is met, the iteration terminates, and the global optimal position vector is extracted as the optimal control command. The optimal control increment is the difference between the optimal control command and the current actual operating command. Specifically, the optimal rotational speed increment is the difference between the screw rotational speed dimension value in the optimal control command and the current measured screw angular velocity, and the optimal heating power increment is the difference between the heating power dimension value in the optimal control command and the current actual heating power. If the iteration reaches the preset maximum number of iterations without convergence, the global optimal position vector obtained during the iteration process is used as the optimal control command to ensure that the system always has executable control output and avoids control interruptions caused by optimization anomalies.

[0048] Obtain the basic parameters of the overload protection boundary, including ambient temperature, convective heat dissipation rate, equipment exposed area, rated maximum thermal power, and rated maximum mechanical power. Ambient temperature refers to the ambient temperature at the location of the extrusion equipment, denoted by [symbol missing]. The International System of Units (SI) unit is Celsius, with the symbol . The ambient temperature is collected by an on-site ambient temperature sensor, with the collection period matching the system sampling period. The data undergoes first-order low-pass filtering and validity verification. The convective heat transfer rate is the convective heat transfer coefficient between the outer surface of the extruder barrel and the ambient air, denoted by . The SI unit is the watt per square kelvin, with the symbol . This parameter was obtained through a standard flat plate steady-state heat transfer experiment. The test environment was a windless indoor environment, and the test temperature range covered the normal processing temperature range of the equipment, with a typical value range of 5 to 25 degrees Celsius. It can be adjusted according to the on-site installation environment. The exposed area of ​​the equipment is the total area of ​​the outer surface of the extruder barrel in contact with the external environment, denoted by . The International System of Units (SI) unit is the square meter, with the symbol . The rated maximum thermal power is calculated through the equipment's geometric dimensions and serves as a fixed structural parameter for the equipment. The rated maximum thermal power is the total rated maximum output power of the extruder heaters, denoted by [symbol missing]. The International System of Units (SI) unit is the watt, with the symbol . These are the rated parameters specified by the manufacturer for the heater. The rated maximum mechanical power is the rated maximum shaft output power of the extruder's main motor, denoted by [symbol missing]. The International System of Units (SI) unit is the watt, with the symbol . These are the rated parameters specified by the manufacturer for the main motor.

[0049] Subtracting the ambient temperature from the current measured temperature, multiplying by the convective heat dissipation rate and the exposed area of ​​the equipment, yields the natural heat dissipation. Subtracting the natural heat dissipation from the rated maximum thermal power yields the upper limit of the dynamic thermal power. The calculation formula is as follows: ; in, The upper limit of dynamic thermal power at time t, the SI unit is watt, symbol . This characterizes the maximum effective output power that the heater can use to increase the melt temperature, taking into account natural heat dissipation losses. When Less than At that time, the natural heat dissipation is set to 0 to avoid invalid results where the dynamic heat power limit exceeds the rated maximum heat power. The calculated... Boundary truncation is required, with a lower limit of 0 and an upper limit of the rated maximum thermal power. When the limit is exceeded, the corresponding boundary value is taken.

[0050] The absolute heating power command mapped from the optimal heating power increment is compared with the dynamic thermal power upper limit, and the smaller value is taken. Then, this smaller value is compared with zero, and the larger value is taken to generate the actual power command. The calculation formula is as follows: ; in, The power command to be output at the next moment, the SI unit is watt, and the symbol is . ; This is the absolute heating power command value mapped to the optimal heating power increment. By comparing two extreme values, the upper and lower limits of the heating power command are truncated, ensuring that the output command does not exceed the upper limit of the equipment's dynamic thermal power, and is not lower than zero. This avoids invalid commands output by the heater in the opposite direction, ensuring the safe operation of the equipment.

[0051] The dynamic speed limit is obtained by dividing the rated maximum mechanical power by the main motor torque. The calculation formula is as follows: ; in, This represents the upper limit of the dynamic rotational speed at time t, with the SI unit being radians per second, and the symbol being . This represents the maximum permissible speed of the main motor under the current motor torque, where the rated shaft power is not exceeded. When the torque is less than the motor's rated minimum torque, Use the motor's rated maximum angular velocity to avoid calculation errors such as dividing by zero or setting the upper speed limit to infinity. The calculated value is... A boundary cutoff is required, with the upper limit being the motor's rated maximum angular velocity. If the angular velocity exceeds the boundary, the rated maximum angular velocity will be used.

[0052] The actual speed command is generated by comparing the absolute speed command mapped by the optimal speed increment with the dynamic speed upper limit and taking the smaller value. The calculation formula is as follows: ; in, This is the actual rotational speed command to be output at the next moment, with the SI unit being radians per second, and the symbol being [symbol missing]. ; This is the absolute speed command value mapped to the optimal speed increment. The calculated actual speed command needs to be truncated to a lower limit of 0 to avoid invalid commands with negative speeds.

[0053] The generated actual power command and actual speed command are subjected to ramp smoothing to avoid drastic fluctuations in operating conditions caused by step commands. The smoothing formula is as follows: ; in, This is the smoothed actual power command. This is the smoothed actual speed command; The ramp smoothing coefficient for the power command has a value range of [0.05, 1]. The ramp smoothing coefficient for the speed command ranges from [0.05, 1]. The smoothing coefficient can be adjusted according to the dynamic response characteristics of the equipment; the smaller the coefficient, the smoother the command change. The smoothed command needs to be truncated again to ensure that it does not exceed the previously set upper and lower limit constraints.

[0054] The smoothed actual power command and actual speed command are sent to the heater's drive controller and the main motor's frequency converter via the industrial fieldbus, respectively, driving the corresponding actuators to complete the command actions. The command output cycle is completely consistent with the system sampling cycle, and the power command and speed command are output synchronously to ensure that the action sequence of the two actuators is completely aligned, achieving multi-parameter coordinated control. After the command is output, the valid command value of this output is stored in the historical data cache sequence for subsequent operating condition evaluation and closed-loop correction. When the command is in a truncated state, the system latches the truncated state for 3 sampling cycles. If the truncated condition persists during the latch period, the current truncated command output is maintained to avoid command oscillation caused by frequent truncation. After the latch period ends, the truncated condition is re-evaluated, and the command output is updated.

[0055] Add the transient viscous dissipation power to the actual heating power, divide by the current measured temperature converted to the Kelvin scale, and obtain the transient thermodynamic entropy production rate. The calculation formula is as follows: ; in, The transient thermodynamic entropy yield at time t, with the SI unit of watt per Kelvin, symbol . The entropy production rate, characterizing the irreversible process of an extrusion system as a thermodynamically isolated system, is a core thermodynamic indicator for evaluating the system's energy utilization efficiency and operational stability. The formula adds 273.15 to the temperature value to convert Celsius to the thermodynamic Kelvin scale, satisfying the dimensional specifications and physical definitions required for calculating the thermodynamic entropy yield.

[0056] Obtain the historical thermodynamic entropy yield at the previous sampling time of the system. Subtract the historical thermodynamic entropy yield from the current transient thermodynamic entropy yield, and then divide by the system sampling period to obtain the entropy yield time difference value. The calculation formula is as follows: ; in, The time difference fraction of entropy production rate, with the SI unit of watts per square Kelvin second, symbol . It characterizes the rate of change of the system's entropy productivity and reflects the direction of evolution of the system's work state; The historical thermodynamic entropy yield at the previous sampling time is stored in the system's historical data cache sequence; This is the system sampling period. This definition perfectly aligns with the physical meaning of time difference; changes in the sampling period will not cause orders of magnitude deviations in the difference values.

[0057] After completing the operational status assessment and closed-loop correction triggering, the system sets the dead zone threshold for the entropy productivity time difference value, denoted as . The value ranges from 1e-6 to 1e-3. The time difference can be adjusted according to the stability requirements of equipment operation. When the absolute value of the time difference is less than the dead zone threshold, it is determined to be an invalid fluctuation caused by sensor noise, and no correction action is triggered, maintaining the current command output. When the time difference is greater than the dead zone threshold, it is determined that the system's working state has deteriorated, the irreversible entropy production rate continues to accelerate, energy utilization efficiency decreases, and the melt rheological state tends to be unstable. The system immediately triggers a new round of iterative optimization process, updates control commands, suppresses the increase in entropy production rate, and restores the system to a stable operating state. When the time difference is less than or equal to a negative dead zone threshold, it is determined that the system's working state has optimized, the irreversible entropy production rate continues to decrease, energy utilization efficiency improves, and the melt rheological state tends to be in a more stable range. The system maintains the current control command output and does not trigger additional optimization actions. When the time difference is less than or equal to the dead zone threshold and greater than or equal to a negative dead zone threshold, it is determined that the system is in a stable operating region, the entropy production rate remains stable, and the melt rheological state is in the stable range required by the process. The system maintains the current control command output to avoid unnecessary command adjustments that could cause system oscillations.

[0058] After completing the cold start adaptation, during the cold start phase, when the barrel temperature has not reached the viscous flow temperature of the polymer material, the melt has not filled the flow channel, and the mass flow rate is 0, the system executes the cold start control logic. During the cold start phase, the system first controls the heater output according to the preset temperature rise curve to heat each section of the barrel to the preset preheating temperature, which is not lower than the viscous flow temperature of the material. After preheating, the system controls the screw rotation according to the preset low-speed start curve, gradually increasing the screw speed and extrusion flow rate. Once the melt fills the flow channel and the mass flow rate stabilizes, it automatically switches to the normal operation collaborative control logic. During the cold start phase, the time-space delay time is the preset maximum delay time, and the historical data extraction window is filled with the preheating parameters at the current moment to ensure that the system logic can execute normally.

[0059] After completing the material and die change adaptation, when changing the processing material or forming die, the system first executes a shutdown and material clearing process to empty the original melt in the barrel and die. After the change is completed, the system automatically updates the physical properties of the corresponding material, including melt density, specific heat capacity, rheological activation energy, viscous flow temperature, and decomposition temperature, while also updating the internal effective volume parameters of the die. After the parameters are updated, the system executes the startup process according to the new process target parameters. Once the operating conditions stabilize, it automatically switches to the normal operation collaborative control logic.

[0060] The system implements fault degradation handling for sensor anomalies. When a core data acquisition sensor experiences continuous invalid data, exceeds its measurement range, or experiences a disconnection, the system automatically triggers fault degradation logic. If a temperature sensor fails, the system switches to a soft-sensor model based on viscous dissipation power and heating power to replace the measured temperature value in the control logic, while simultaneously triggering a fault alarm. If a pressure sensor fails, the system switches to open-loop process control mode, operating according to preset speed and power process parameters, while also triggering a fault alarm. If torque, speed, or flow sensors fail, the system executes an orderly shutdown procedure, gradually reducing the screw speed and heating power until shutdown, while simultaneously triggering a fault alarm to prevent equipment damage.

[0061] The quantitative verification indicators for system control performance include temperature control steady-state error, pressure control steady-state error, convergence time of operating condition fluctuations, and system anti-interference capability. Temperature control steady-state error is the average absolute value of the difference between the measured temperature and the target temperature during stable operation, and it is required to be no greater than [value missing]. The steady-state error of pressure control, during the stable operation phase, is the average of the absolute values ​​of the differences between the measured pressure and the target pressure, which must not exceed [a certain value]. Target pressure. The convergence time for operating condition fluctuations is the time required for the system's measured values ​​to enter the allowable steady-state error range after the target process parameters are adjusted, and it must not exceed 3 times the space-time delay time. The system's anti-interference capability is its ability to maintain stable process parameters when the characteristics of incoming materials and ambient temperature fluctuate, requiring the fluctuation amplitude to not exceed 2 times the allowable steady-state error range, and the convergence time to not exceed 5 times the space-time delay time.

[0062] The system is designed for the multi-section barrel structure of extruders and incorporates a power coordination and distribution control logic for multi-section heaters. The extruder barrel is divided axially into three independent heating zones: the feeding zone, the melting zone, and the metering zone. Each zone is equipped with an independent heater, temperature sensor, and power acquisition module. The coordinated distribution of heating power across each zone is based on the temperature rise requirements of the melt flow and the viscous heat dissipation distribution. The sum of the actual heating power of each heater zone represents the total actual heating power of the system, calculated using the following formula: ; in, The actual output power of the nth segment heater at time t, expressed in watts (SI). ; n represents the barrel heating section number, where 1 corresponds to the feeding section, 2 to the melting section, and 3 to the metering section. The weighting of the heating power for each section is determined based on the theoretical melt temperature rise requirement and the proportion of viscous heat dissipation for that section. The weighting calculation formula is as follows: ; in, The power allocation weight for the nth segment heater is a dimensionless parameter, and the sum of the weights of all segments is 1. The target temperature rise setting for the nth segment, in degrees Celsius; This represents the viscous dissipation power of the melt corresponding to the nth screw segment, in watts. and These are the external heat source weighting coefficient and the internal heat source weighting coefficient, respectively, both dimensionless parameters, and their sum is 1. They can be adjusted according to the rheological properties of the processed material. The overall optimal heating power command obtained through iterative optimization is decomposed into independent heating power commands for each segment according to the power allocation weight of each segment. The decomposition formula is as follows: ; in, This represents the optimal heating power command for the nth stage heater, expressed in watts. Each stage heating power command undergoes independent overload protection and ramp smoothing to ensure coordinated and stable temperature control across stages, preventing rheological instability caused by uneven melt temperature distribution along the process.

[0063] The system is equipped with a soft-sensing model for melt temperature, used as a substitute measurement in case of temperature sensor failure, and also for redundancy verification of measured temperatures. The soft-sensing model is constructed based on the energy conservation equation for melt flow along the flow path. The model input parameters include the actual output power of each heater section, screw speed, main motor torque, headstock pressure differential, mass flow rate, and material properties. The model output is the predicted melt temperature at the end of the metering section. The soft-sensing calculation formula is as follows: ; in, The value of the soft-measured melt at time t is expressed in degrees Celsius. The feed temperature of the polymer raw material is measured in degrees Celsius and is collected by a temperature sensor installed at the feed inlet. This is the theoretical time for the melt to flow from the inlet to the end of the metering section, in seconds; The total power loss due to natural heat dissipation of the barrel is expressed in watts, and the calculation formula is as follows: ; in, The measured temperature value of the nth barrel segment is given in degrees Celsius. The output value of the soft-sensor model needs to undergo the same first-order low-pass filtering as the measured temperature, with consistent filtering coefficients. Boundary validity checks are also performed; if the temperature exceeds the material processing temperature range, the corresponding boundary value is used. During normal operation, the system compares the soft-sensor temperature value with the measured temperature value. When the difference exceeds a preset deviation threshold, a temperature sensor malfunction warning is triggered. If the temperature sensor experiences a persistent fault, the system automatically switches to the soft-sensor temperature value to replace the measured temperature value, maintaining the continuous and stable operation of the control logic.

[0064] The system improves the fallback rules for handling iteration anomalies in the particle swarm optimization algorithm. It sets up a multi-level fallback mechanism to address potential anomalies such as non-convergence, oscillation, and local optimum stagnation during iteration. The first fallback is for iteration stagnation handling: when the globally optimal objective functional value does not decrease after a preset number of consecutive stagnation iterations, it is determined to be iteration stagnation. The system automatically performs partial reinitialization of the particle swarm, with 30% to 50% of the particles being reinitialized. These reinitialized particles randomly generate initial instruction vectors and step size vectors within the feasible region, while retaining the historical globally optimal position vectors, thus breaking the local optimum stagnation state. The second fallback is for iteration divergence handling: when the globally optimal objective functional value continues to rise after 5 consecutive iterations, it is determined to be iteration divergence. The system immediately terminates the current iteration, resets the particle swarm to its initial state, reduces the clamping boundary of the iteration step size to 50% of the original boundary, and restarts the iteration optimization process. The third fallback layer is iteration timeout handling. When the convergence condition is not met even after reaching the preset maximum number of iterations, the system extracts the globally optimal position vector obtained during the iteration process as the optimal control command. Simultaneously, it records any abnormal iteration states and automatically adjusts parameters such as the number of particles, maximum number of iterations, step size, and clamping boundaries in the next iteration to improve convergence performance. The fourth fallback layer is algorithm failure handling. If the optimal control command fails to converge effectively in three consecutive iterations, the algorithm is deemed to have failed. The system automatically switches to the preset open-loop process control mode, outputting control commands according to the standard process parameters corresponding to the current processing material, and triggering a fault alarm to ensure safe equipment operation.

[0065] The system details the implementation of a dual-buffering mechanism for historical data caching. This mechanism employs a dual-buffered circular queue structure, comprising a working cache queue and a backup cache queue. The two queues have identical storage depth and data structure. The working cache queue is used to write valid parameter data for the current sampling period in real time, while also providing data reading support for parameter calculation, forward extrapolation, and iterative optimization for the current control period. The backup cache queue stores all cached data from the previous complete control period. When a data write anomaly or read conflict occurs in the working cache queue, the system can immediately switch to the backup cache queue to read historical data, ensuring uninterrupted control logic. Each cache queue's storage unit is structured, containing 11 data fields: sampling timestamp, transient viscous dissipation power, main motor torque, measured screw angular velocity, head pressure differential, mass flow rate, melt density, actual heating power, measured temperature of each barrel section, current measured head pressure, and output control commands. All fields are stored in International System of Units (SI), ensuring no additional unit conversion is required during data reading. The write operation of the cache queue adopts atomic operation, and a complete write is completed in each sampling period. After the write is completed, the read and write pointers of the queue are updated to avoid data races and read and write conflicts. The read operation of the cache queue supports random address access and continuous range access. It can quickly locate the corresponding storage unit according to the time range of the historical data extraction window, and also supports linear interpolation calculation to achieve accurate extraction of data without sampling timestamps.

[0066] The system refines the timing synchronization and scheduling of real-time control tasks. It employs a priority-based preemptive scheduling architecture based on a real-time operating system, assigning fixed priorities and execution time slices to all tasks to ensure the determinism and real-time performance of task execution. The highest priority task is sensor data acquisition and preprocessing. Its execution cycle is identical to the system sampling cycle, with an execution time slice not exceeding 20% ​​of the sampling cycle. Within this task, all sensor signals are synchronously acquired, clocked, filtered, and validated, ensuring that the time synchronization error of all acquired data does not exceed 10% of the sampling cycle. The next highest priority task is fault diagnosis and degradation processing. Its execution cycle is identical to the sampling cycle, with an execution time slice not exceeding 10% of the sampling cycle. Within this task, all sensors and actuators are fault-diagnosed, and abnormal conditions are identified and degradation processing logic is triggered. The third highest priority task is core parameter calculation. Its execution cycle is identical to the sampling cycle, with an execution time slice not exceeding 20% ​​of the sampling cycle. Within this task, all core parameter calculations are performed, including transient volumetric flow rate, viscous dissipation power, spatiotemporal delay, historical data window definition, forward extrapolation, and objective functional calculation. The fourth priority task is the iterative optimization task, with an execution cycle that is an integer multiple of the sampling cycle. This can be adjusted to 1 to 5 times the sampling cycle depending on the device's computing power, and the execution time slice cannot exceed 30% of the corresponding cycle. Within this task, the entire process of particle swarm optimization is iteratively optimized, outputting the optimal control increment. The fifth priority task is the control command generation and output task, with an execution cycle consistent with the sampling cycle. The execution time slice cannot exceed 15% of the sampling cycle, and the task performs overload truncation of the optimal control command, ramp smoothing, multi-segment power allocation, and synchronous command output. The lowest priority task is the operating condition assessment, data storage, and communication task, with an execution cycle of 5 to 10 times the sampling cycle. The execution time slice cannot exceed 20% of the corresponding cycle, and the task performs non-real-time operations such as entropy production rate calculation, operating condition determination, closed-loop correction triggering, historical data storage, and communication with the host computer. All tasks are executed using a time-triggered mechanism, with a fixed start time for each task to avoid execution conflicts and ensure the timing synchronization and determinism of the entire control process.

[0067] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A multi-parameter collaborative control system for an extruder production line, applied to extrusion equipment including a main motor, heater, barrel, screw, metering section, and forming die, characterized in that, include: Collect system operating parameters, which include at least the actual heating power; The transient volumetric flow rate and transient viscous dissipation power are calculated based on the system operating parameters. The time-space delay is calculated by combining the internal effective volume of the extrusion equipment with the transient volumetric flow rate. By using the aforementioned spatiotemporal delay time and the aforementioned transient viscous dissipation power, forward extrapolation is performed to obtain the predicted temperature and predicted pressure; Based on the transient viscous dissipation power and the actual heating power, a target functional is constructed by combining the predicted temperature and the predicted pressure. The damping inertia weight is generated using the spatiotemporal delay time, and the optimal control increment is obtained through iterative optimization under the drive of the target functional. Based on the physical boundaries of the equipment, the optimal control increment is truncated to prevent overload, and actual power command and actual speed command are generated and controlled to be executed. The transient thermodynamic entropy yield of the system and its time difference value are calculated, and the operating status is evaluated based on the time difference value to trigger closed-loop correction.

2. The multi-parameter collaborative control system for an extruder production line according to claim 1, characterized in that, The operating parameters of the acquisition system include: main motor torque, measured screw angular velocity, head pressure difference, mass flow rate, melt density, and actual heating power; The calculation of transient volumetric flow rate and transient viscous dissipation power based on the system operating parameters includes: Divide the mass flow rate by the melt density to obtain the transient volumetric flow rate; The total mechanical power is obtained by multiplying the main motor torque by the measured screw angular velocity. The effective pumping power is obtained by multiplying the pressure difference at the pump head by the transient volumetric flow rate. The transient viscous dissipation power is obtained by subtracting the effective pumping power from the total mechanical power.

3. The multi-parameter collaborative control system for an extruder production line according to claim 2, characterized in that, The calculation of the spatiotemporal delay time by combining the internal effective volume of the extrusion equipment with the transient volumetric flow rate includes: Obtain the effective internal volume formed by the metering section of the barrel screw and the forming die; Divide the internal effective volume by the transient volumetric flow rate to obtain the spatiotemporal delay time; Based on the aforementioned spatiotemporal delay time, a historical data extraction window is defined in the historical data cache sequence.

4. The multi-parameter collaborative control system for an extruder production line according to claim 3, characterized in that, The step of using the spatiotemporal delay time and the transient viscous dissipation power to perform forward extrapolation to obtain the predicted temperature and predicted pressure includes: Obtain candidate heating power, candidate screw speed, prediction step size, current measured temperature, current measured head pressure, specific heat capacity, rheological activation energy, and ideal gas constant; The candidate heating power is multiplied by the prediction step size to obtain the conduction heat; the transient viscous dissipation power within the historical data extraction window is integrated over time to obtain the cumulative shear heat; The predicted temperature is obtained by adding the conductive heat and the accumulated shear heat, dividing by the product of the internal effective volume, the melt density and the specific heat capacity, and then adding the current measured temperature. Divide the rheological activation energy by the ideal gas constant, then multiply by the difference between the reciprocal of the predicted temperature and the reciprocal of the current measured temperature, and calculate the natural exponent of the product to obtain the viscosity decay factor. The momentum response factor is obtained by dividing the candidate screw rotation speed by the measured screw angular velocity. The predicted pressure is obtained by multiplying the current measured head pressure by the momentum response factor and the viscosity decay factor in sequence.

5. The multi-parameter collaborative control system for an extruder production line according to claim 4, characterized in that, The construction of the objective functional based on the transient viscous dissipation power and the actual heating power, combined with the predicted temperature and the predicted pressure, includes: Obtain the target pressure and target temperature; Divide the transient viscous dissipation power by the sum of its power and the actual heating power to obtain the endogenous heat ratio; Divide the actual heating power by the sum of its power and the transient viscous dissipation power to obtain the external heat ratio; The pressure residual is obtained by dividing the difference between the predicted pressure and the target pressure by the target pressure and then squaring the result. The temperature residual is obtained by dividing the difference between the predicted temperature and the target temperature by the target temperature and then squaring the result. The objective functional is obtained by multiplying the endogenous heat ratio by the pressure residual and adding the exogenous heat ratio by the temperature residual.

6. The multi-parameter collaborative control system for an extruder production line according to claim 5, characterized in that, The process of generating damped inertial weights using the spatiotemporal delay time and iteratively optimizing the optimal control increment under the drive of the objective functional includes: The damping inertia weight is obtained by dividing the negative spatiotemporal delay time by the prediction step size and then performing a natural exponential operation. The cognitive learning factor is obtained by dividing the absolute value of the difference between the current measured head pressure and the target pressure by the target pressure, taking a negative sign, performing natural exponentiation, and then subtracting one. The social learning factor is obtained by dividing the absolute value of the difference between the current measured temperature and the target temperature by the target temperature, taking a negative sign, performing natural exponentiation, and then subtracting one. Obtain the current iteration step size vector, the current instruction vector, the local extremum position vector, the global extremum position vector, the first random number, and the second random number; Multiply the damping inertia weight by the current iteration step size vector to obtain the inertia vector component; multiply the cognitive learning factor, the first random number, and the difference between the local extreme value position vector and the current instruction vector to obtain the cognitive vector component; multiply the social learning factor, the second random number, and the difference between the global extreme value position vector and the current instruction vector to obtain the social vector component; The inertial vector component, the cognitive vector component, and the social vector component are added together to obtain the update iteration step size vector, and then added to the current instruction vector to obtain the update instruction vector; Based on the principle of minimizing the objective functional, the optimal control increment is extracted from the update command vector after iterative convergence. The optimal control increment includes the optimal speed increment and the optimal heating power increment.

7. The multi-parameter collaborative control system for an extruder production line according to claim 6, characterized in that, The process of truncating the optimal control increment based on the physical boundaries of the equipment to prevent overload, generating actual power commands and actual speed commands, and controlling their execution includes: Obtain ambient temperature, convective heat dissipation rate, equipment exposed area, rated maximum thermal power, and rated maximum mechanical power; Subtract the ambient temperature from the current measured temperature, multiply by the convective heat dissipation rate and the exposed area of ​​the equipment to obtain the natural heat dissipation, and subtract the natural heat dissipation from the rated maximum thermal power to obtain the dynamic thermal power upper limit. The optimal heating power increment is compared with the dynamic thermal power upper limit and the smaller value is taken. Then, the smaller value is compared with zero and the larger value is taken to generate the actual power command. The rated maximum mechanical power is divided by the main motor torque to obtain the dynamic speed limit. The optimal speed increment is compared with the dynamic speed limit and the smaller value is taken to generate the actual speed command. The actual power command and the actual speed command are sent to the heater and the main motor respectively for execution.

8. The multi-parameter collaborative control system for an extruder production line according to claim 7, characterized in that, The calculation system calculates the transient thermodynamic entropy yield and its time difference value, and evaluates the operating status based on the time difference value to trigger closed-loop correction, including: The transient viscous dissipation power is added to the actual heating power and divided by the current measured temperature to obtain the transient thermodynamic entropy production rate. Obtain the historical thermodynamic entropy production rate of the system at the previous moment, and subtract the historical thermodynamic entropy production rate from the transient thermodynamic entropy production rate to obtain the time difference value; When the time difference score is greater than zero, it is determined that the system's work state has deteriorated, and the next round of optimization evolution is carried out to update the instructions; When the time difference value is less than or equal to zero, the system is determined to be in a stable operating region, and continues to send the current actual power command and the actual speed command.

Images