Adaptive energy efficiency steady state control method and system for plastic extrusion molding process

Abstract

The application relates to the technical field of plastic extrusion molding control, and particularly discloses a self-adaptive energy efficiency steady-state control method and system for a plastic extrusion molding process, which comprises the following steps: collecting multidimensional data of the plastic extrusion molding process, wherein the multidimensional data comprises process data representing a current production state, aging characteristic data representing the health degree of equipment, and auxiliary data; pre-processing and feature fusing the multidimensional data, calculating an equipment aging comprehensive index, and constructing an aging trend prediction model based on a long short-term memory network to output an aging trend prediction value and the aging rate of a screw rod and a cylinder. Through real-time collection of multidimensional data and construction of the equipment aging comprehensive index and the trend prediction model, the application dynamically corrects a multivariable coupling coefficient and a multi-objective optimization function, and realizes self-adaptive energy efficiency steady-state control under the condition of equipment aging.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of plastic extrusion molding control, in particular to a self-adaptive energy efficiency steady state control method and system for plastic extrusion molding process. BACKGROUND

[0002] In the plastic extrusion molding process, in order to realize stable and efficient production, the control strategy based on fixed model or multivariable decoupling is usually adopted to adjust the extruder. However, with the continuous operation of the equipment, the key components such as screw and cylinder gradually wear out, and the response performance of the heating system decreases, which leads to the mismatch between the original control model and the actual process, that is, the coupling relationship between the key variables such as speed, pressure and temperature drifts, the control accuracy decreases, the energy consumption increases, and even the product quality fluctuates. The existing control method does not systematically consider the time-varying characteristics brought by equipment aging, still uses fixed coupling coefficients and optimization targets, and it is difficult to maintain the balance between energy efficiency and steady state in the whole life cycle of the equipment, forming the inherent contradiction that the aging of the equipment inevitably leads to the decline of the control performance.

[0003] In addition, the traditional control strategy often lacks prediction and self-adaptive compensation mechanism for aging trend, and cannot adjust the control parameters in advance in the accelerated aging stage. It can only be passively corrected after the fault or performance decreases obviously. This lag response not only affects the continuity and stability of production, but also leads to increased maintenance cost and low energy efficiency. How to realize long-term and self-adaptive energy efficiency and steady state control under the condition of continuous equipment aging has become a technical problem to be solved in this field. SUMMARY

[0004] The present application aims to at least solve one of the technical problems in the related art. To this end, the purpose of the present application is to propose a self-adaptive energy efficiency steady state control method and system for plastic extrusion molding process to improve the stability and energy efficiency of production.

[0005] To achieve the above-mentioned purpose, the first aspect of the present application proposes a self-adaptive energy efficiency steady state control method for plastic extrusion molding process, comprising the following steps:

[0006] Collecting multi-dimensional data of the plastic extrusion molding process, the multi-dimensional data including process data representing the current production state, aging characteristic data representing the health degree of the equipment, and auxiliary data;

[0007] Preprocessing and feature fusion are performed on the multi-dimensional data, the equipment aging comprehensive index is calculated, and the aging trend prediction model is constructed based on long short-term memory network, and the aging trend prediction value and the aging rate of the screw and the cylinder are outputted;

[0008] based on the equipment aging comprehensive index and the aging rate, a preset multivariate coupling coefficient matrix and a multi-objective optimization function are dynamically compensated and corrected to generate a corrected coupling coefficient matrix and a corrected multi-objective optimization function;

[0009] based on the corrected coupling coefficient matrix, a multivariate dynamic decoupling is performed on the extrusion molding process, and a prediction-feedback double closed loop correction mechanism is combined to perform hierarchical control according to a preset time scale, so as to realize energy efficiency steady state control of the plastic extrusion molding process.

[0010] To achieve the above purpose, the second aspect embodiment of the present application provides a kind of self-adapting energy efficiency steady state control system of plastic extrusion molding process, comprising:

[0011] Multidimensional data acquisition module is configured to synchronously acquire the process data of each sub-domain of extruder, the aging characteristic data of equipment component and production auxiliary data;

[0012] Data processing and analysis module is configured to preprocess the collected data, calculate equipment aging comprehensive index, and run long short-term memory network to output aging trend prediction value and aging rate;

[0013] Parameter dynamic compensation module is configured to correct multivariate coupling coefficient matrix and multi-objective optimization function in real time based on the equipment aging comprehensive index and the aging rate;

[0014] Hierarchical collaborative control module is configured to perform multivariate dynamic decoupling based on the corrected coupling coefficient matrix, and combine prediction-feedback double closed loop correction mechanism to send hierarchical control instruction to extruder actuator according to the time scale of different levels.

[0015] To achieve the above purpose, the third aspect embodiment of the present application provides an electronic device, which includes a memory, a processor and a computer program stored in the memory, and the computer program is executed by the processor to realize the self-adapting energy efficiency steady state control method of plastic extrusion molding process.

[0016] The self-adapting energy efficiency steady state control method and system of plastic extrusion molding process of the embodiment of the present application realize adaptive energy efficiency steady state control under the condition of equipment aging by real-time acquisition of multidimensional data and construction of equipment aging comprehensive index and trend prediction model, dynamic correction of multivariate coupling coefficient and multi-objective optimization function;The system can automatically adjust the control strategy according to the aging state, ensure the response speed and control accuracy, optimize energy consumption and delay equipment aging, support predictive maintenance, so as to improve the stability, energy efficiency and economy of overall production. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1is a flowchart of the adaptive energy efficiency steady state control method of the plastic extrusion molding process provided by the present application;

[0018] Figure 2 is a schematic diagram of the comparison curve between the predicted value and the actual value of the aging trend based on the LSTM network in the adaptive energy efficiency steady state control method of the plastic extrusion molding process provided by the present application;

[0019] Figure 3 is a comparison diagram of the control action intensity and system response of the Pareto frontier under different aging weight settings in the adaptive energy efficiency steady state control method of the plastic extrusion molding process provided by the present application;

[0020] Figure 4 is a time-domain waveform diagram of the pulse start-up torque instruction and the screw speed response in the transient overload penetration mode in the adaptive energy efficiency steady state control method of the plastic extrusion molding process provided by the present application;

[0021] Figure 5 is a comparison diagram of the actual feedback weight adjustment smoothing degree after fidelity discrimination in the high-entropy noise interference environment in the adaptive energy efficiency steady state control method of the plastic extrusion molding process provided by the present application;

[0022] Figure 6 is an implementation execution schematic diagram of the adaptive energy efficiency steady state control system of the plastic extrusion molding process provided by the present application;

[0023] Figure 7 is a structural schematic diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0024] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0025] The adaptive energy efficiency steady state control method, system and electronic device of the plastic extrusion molding process of the embodiments of the present application are described below with reference to the accompanying drawings.

[0026] Embodiment one:

[0027] The embodiment details a kind of adaptive energy efficiency steady state control method of plastic extrusion forming process based on data driving and physical model fusion.The method is configured for various single screw or twin screw extruder systems, especially for those long-running in complex working conditions, face the challenge of mechanical wear and thermal performance attenuation of precision extrusion equipment.The core logic of the embodiment is to solve the technical pain points that the traditional fixed parameter control model gradually fails in the whole life cycle of equipment by constructing and analyzing the aging characteristics of equipment, dynamically reconstructing the core parameters of control system.

[0028] As Figure 1 The adaptive energy efficiency steady state control method of plastic extrusion forming process described in the embodiment, the specific steps include:

[0029] Step S1: the control system first initializes a high-frequency, multi-channel data acquisition interface, which establishes real-time communication link with various sensors and actuators on the extruder production line through industrial field bus, such as EtherCAT or Profinet protocol.In order to build a digital twin model that can accurately map the real state of the physical world, this step performs multi-dimensional data acquisition operation of plastic extrusion forming process.The multi-dimensional data is strictly divided into three levels in logic, which are process data representing current production state, aging characteristic data representing equipment health degree and auxiliary data.

[0030] For example, the process data representing current production state refers to real-time variables directly reflecting the physical field state of extrusion rheological process.The system synchronously reads the real-time temperature values of multiple heating temperature zones distributed along the axial direction of extruder barrel, the instantaneous values of melt pressure in die flow channel, melt core temperature, actual rotation speed of screw, input current and terminal voltage of main drive motor, and feeding rate of feeding system with millisecond level sampling period.These data constitute the basic state space of control system, which reflects the thermodynamic and kinetic behavior of material in plasticizing, conveying and extruding process.

[0031] For example, the aging characteristic data representing equipment health degree refers to those characteristic physical quantities that can quantitatively reveal the wear state of mechanical parts, the attenuation degree of heater performance and the drift condition of sensor.In this embodiment, the system focuses on and extracts two kinds of key aging characteristic data, namely screw wear characteristic and heating response characteristic, because the two directly determine the material conveying efficiency and thermal control accuracy of extrusion process.

[0032] Specifically, the screw wear feature is characterized by the torque long-term drift. Physically, as the gap between the screw land and the barrel inner wall gradually increases due to long-term mechanical friction and corrosion, the melt leakage increases and the shear efficiency decreases. This means that the torque output required to maintain the same production or pressure under the same process setting conditions will irreversibly deviate. The torque long-term drift is defined as the percentage change of the current measured torque value relative to the initial baseline torque value of the equipment under the standard test condition. To quantify this feature, the system presets a standard test condition, which includes a specific rotational speed, temperature setting, and raw material formulation. The system periodically guides the equipment to run in this condition and calculates the torque long-term drift , whose calculation formula is:

[0033] ;

[0034] wherein, represents the average torque value measured by the high-precision torque sensor when the equipment is running under the standard test condition at the current time point; represents the initial baseline torque value of the equipment recorded and stored in the non-volatile memory under the same standard test condition at the initial stage of use or after completing the overhaul of the core components.

[0035] Specifically, the heating response feature is characterized by the heating warm-up time. As the heating elements, such as ceramic heating rings or cast aluminum heaters, age, their resistance values will drift due to oxidation, and the contact thermal resistance between the heating elements and the barrel wall will significantly increase due to the thickening of the oxidation layer, both of which will cause the electric-thermal conversion efficiency and heat conduction efficiency of the heating system to decrease. The heating warm-up time is defined as the time difference required for the heating component to rise from the preset starting temperature to the preset target temperature. The system automatically performs the warm-up test and records the time during each equipment startup preheating stage or special test period. The calculation formula of the heating warm-up time is:

[0036] ;

[0037] wherein, represents the time when the control system issues a full-power heating instruction and the current temperature zone temperature is equal to the preset starting temperature; represents the time when the temperature zone temperature first reaches the preset target temperature in the continuous rising process. This time difference directly reflects the dynamic response capability of the temperature control system.

[0038] For example, the auxiliary data includes, but is not limited to, the ambient temperature and relative humidity of the extrusion workshop, the inlet temperature and flow rate of the cooling water circulation system, and the voltage fluctuation rate of the industrial power grid. The purpose of collecting this data is to eliminate the influence of external environmental interference on equipment performance evaluation in subsequent algorithm processing, ensuring that the calculated aging index truly reflects the physical condition of the equipment itself, rather than environmental noise.

[0039] Step S2: Since the raw data collected from the industrial site inevitably contains electromagnetic noise, high-frequency spikes, and outliers caused by packet loss during signal transmission, preprocessing and feature fusion of the multi-dimensional data are necessary before performing core calculations. The system first applies a moving average filtering algorithm to smooth the high-frequency noise and uses the 3-Sigma criterion to identify and remove statistically significant outliers. Subsequently, the system performs data normalization, mapping physical quantities with different dimensions and orders of magnitude (such as Celsius, megapascals, and revolutions per minute) to a dimensionless range of zero to one, thus eliminating the impact of dimensional differences on subsequent weighted calculations.

[0040] Optionally, after data cleaning and standardization, the system executes a feature fusion algorithm to comprehensively quantify the health status of the equipment using a unified indicator. This step calculates the comprehensive equipment aging index, a highly integrated scalar that intuitively reflects the degree of deviation of the equipment from its design performance benchmark. Specifically, the step of calculating the comprehensive equipment aging index involves: using preset weighting coefficients to perform a weighted summation of the normalized long-term torque drift, heating time, pressure sensor zero-drift error, and temperature sensor response delay to obtain the comprehensive equipment aging index. Its mathematical expression is as follows:

[0041] ;

[0042] in, This is the comprehensive equipment aging index, with a value ranging from zero to one. The closer the value is to one, the more severe the equipment aging. This is the normalized long-term torque drift. This is the normalized heating time. The normalized zero-drift error of the pressure sensor is obtained by reading the pressure sensor when the screw has stopped rotating and the melt has been emptied. This value is used to characterize the static drift of the sensing system. The normalized temperature sensor response delay, obtained through a step response test, is used to characterize the dynamic response degradation of the thermocouple. , , , are a set of preset weighting coefficients and satisfy a normalization condition These weight coefficients are assigned according to the sensitivity of each component to the final product quality, for example, given that screw wear has a decisive influence on the extrusion yield and plasticizing quality, Usually set to a large value.

[0043] Illustratively, in order to realize the transition of the control strategy from post-failure maintenance to predictive maintenance, the embodiment not only assesses the current state, but also strives to foresee the future. The system builds an aging trend prediction model based on a long short-term memory network (LSTM). As a special recurrent neural network, the LSTM network can effectively overcome the gradient disappearance problem of traditional RNNs by virtue of its unique forget gate, input gate and output gate mechanism, and is extremely suitable for capturing and learning the long-period time sequence features of the slow degradation of device performance over time. The system inputs the historical device aging comprehensive index sequence calculated in the past several production cycles as an input vector into the LSTM network. After deep learning training, the LSTM model can output the aging trend prediction value in the future period of time and the aging rate of the key components at the current time.

[0044] Specifically, the aging trend prediction value is a time series vector that changes with the future time step, depicting the trajectory of the continued deterioration of device performance in the future; the aging rate of the screw and the barrel is defined as the first-order derivative of the device aging comprehensive index with respect to time, that is:

[0045] ;

[0046] This rate parameter reflects the acceleration of device performance deterioration, and has important indicative significance for judging whether the device is in a rapid deterioration avalanche period.

[0047] As Figure 2 intuitively shows the performance of the aging trend prediction model built based on the long short-term memory network in the whole life cycle of the device. The horizontal coordinate in the figure represents the time step as the production process continues to advance, and the vertical coordinate represents the normalized device aging comprehensive index.

[0048] Figure 2The middle gray solid line represents the actual aging index calculated in real time based on multi-dimensional sensors and weighted calculation. It can be seen that, affected by industrial site electromagnetic noise and process fluctuations, the curve presents obvious high-frequency oscillation characteristics. If this data is directly used as the input of the control system, it is easy to cause the misadjustment of the control parameters. In contrast, the blue solid line in the figure represents the learning and fitting output of the LSTM network on historical data. The curve closely follows the trend of device performance degradation while effectively filtering out nonlinear random noise, showing excellent smoothing characteristics. Taking the current time as the dividing line, Figure 2 The middle red dashed line shows the reasoning result of the model for the future production cycle.

[0049] It is worth noting that as time goes on to the later stage, the aging curve does not show linear growth, but shows an accelerating upward trend with an increasing slope, which accurately corresponds to the avalanche effect of mechanical wear and heat resistance increase in the physical world. For example, when the prediction curve reaches the dynamic maintenance threshold value of 0.8 shown in the figure at a certain time in the future, the system will identify that the device is about to enter the high-risk area of failure. This prediction ability based on trend extrapolation enables the control system to adjust the coupling coefficient matrix for deep compensation or trigger preventive maintenance warnings before the actual performance of the device collapses completely, thereby effectively avoiding unplanned downtime accidents.

[0050] Step S3: In traditional extrusion control, the controller usually assumes that the gain relationship (coupling coefficient) between the control variable (such as speed, temperature) and the controlled variable (such as pressure, melt temperature) is constant. However, in actual physical processes, as the device ages, these relationships will shift significantly. For example, screw wear can cause its pressure building capability to decline, i.e., the gain of speed to pressure becomes smaller. In order to offset this physical layer variation, the system must reconstruct the kernel of the control algorithm online based on the results of step S2. This step performs dynamic compensation correction on the pre-set multivariate coupling coefficient matrix and multi-objective optimization function based on the device aging comprehensive index and the aging rate, generating a corrected coupling coefficient matrix and a corrected multi-objective optimization function.

[0051] Optionally, the step of dynamically compensating and correcting the multivariate coupling coefficient matrix follows a refined logical judgment and calculation rule. The system first performs threshold judgment: the equipment aging comprehensive index is compared with a preset wear threshold. If the wear threshold is exceeded, it indicates that the mechanical fit clearance between the screw and the cylinder has reached a degree that affects fluid dynamics, and the system automatically corrects the speed-pressure coupling coefficient in the multivariate coupling coefficient matrix. At the same time, the system compares the aging rate with a preset decay rate threshold. If the decay rate threshold is exceeded, it indicates that the thermal inertia of the heating system is changing rapidly, and the system automatically corrects the temperature-pressure coupling coefficient in the multivariate coupling coefficient matrix.

[0052] Specifically, the correction method of the speed-pressure coupling coefficient and the temperature-pressure coupling coefficient follows the following rules:

[0053] The corrected speed-pressure coupling coefficient is equal to the original speed-pressure coupling coefficient multiplied by a first correction factor, and the calculation formula is:

[0054] ;

[0055] wherein, is the original speed-pressure coupling coefficient when the equipment is factory-finished or in a calibrated state; is the first correction factor, which is positively correlated with the equipment aging comprehensive index . It should be noted that here the positive correlation refers to the absolute value of the correction degree and the aging index, and in actual numerical operation the value may be negative to reflect the physical fact of gain reduction.

[0056] And the corrected temperature-pressure coupling coefficient is equal to the original temperature-pressure coupling coefficient multiplied by a second correction factor, and the calculation formula is:

[0057] ;

[0058] wherein, is the original temperature-pressure coupling coefficient; is the second correction factor, which is positively correlated with the aging rate . This correction logic ensures that when the equipment ages rapidly, the control model can timely capture the change in the sensitivity of temperature changes to pressure fluctuations.

[0059] In addition, the update period of the modified coupling coefficient matrix is not fixed, but is adaptively adjusted according to the risk level. The specific strategy is that the update period of the modified coupling coefficient matrix is dynamically set according to the risk level corresponding to the aging trend prediction value. When the aging trend prediction value is in a first risk level, it means that the device is close to the failure edge or the performance inflection point, and the system updates at a first period frequency (high frequency) to ensure that the control model closely follows the rapid change of the device state; when the aging trend prediction value is in a second risk level, it means that the device is in a stable aging period, and the system updates at a second period frequency (low frequency) to save computing resources. The first risk level corresponds to a higher aging degree than the second risk level, and the first period frequency is higher than the second period frequency.

[0060] Optionally, in addition to modifying the process model parameters, the system also needs to modify the target orientation of control. The modified multi-objective optimization function is used to find the optimal control solution in subsequent hierarchical control. The design of this function aims to balance the contradictions between response speed, control accuracy, energy consumption, and device life protection. The multi-objective optimization function contains four weighted components, which have the following mathematical form:

[0061] ;

[0062] Among them, the first component is the ratio of response speed deviation to reference response time, corresponding to the response weight , which is used to constrain the dynamic tracking performance of the system; the second component is the ratio of control accuracy deviation to reference accuracy, corresponding to the accuracy weight , which is used to constrain the steady-state error of the system and ensure the consistency of product size; the third component is the ratio of energy consumption per qualified product to reference energy consumption, corresponding to the energy consumption weight , which is used to drive the system to find the most energy-saving working point; the fourth component is the aging trend prediction value , corresponding to the aging weight . This is an innovative penalty term, which changes the nature of the optimization problem.

[0063] In particular, the modified multi-objective optimization function reduces the intensity of control action when the device is severely aged by increasing the aging weight. When indicates that the device is severely aged, the system automatically increases the value of the weight . In the minimization objective function In the optimization process, the algorithm will tend to suppress those control instructions that can bring fast response or high accuracy but will significantly increase mechanical stress or thermal shock, so as to realize flexible control and achieve the purpose of delaying the further aging of the equipment and avoiding sudden failure.

[0064] As Figure 3 The adaptive optimization trajectory of the multi-objective optimization function in solving the physical conflict between response speed and equipment protection is intuitively revealed in the form of a Pareto frontier curve. The horizontal coordinate in the figure represents the system response time deviation, and the smaller the value is, the faster the response is. The vertical coordinate represents the intensity of the control action, and the larger the value is, the greater the instantaneous stress or thermal shock caused to the mechanical parts is.

[0065] Figure 3 The black solid curve in the figure constitutes the Pareto optimal frontier of the system performance, which represents the performance limit boundary that the system can achieve under the current physical conditions. The blue working condition point A in the figure shows that in the early stage of the health of the equipment, since the aging trend prediction value is low, the aging weight set by the system is small, and the optimization algorithm tends to run on the upper part of the curve, that is, by applying relatively intense control action to exchange for a small response deviation, high dynamic performance is achieved. The orange working condition point B in the figure shows that when the equipment enters the aging stage, the system automatically increases the aging weight in the multi-objective optimization function according to the aging trend prediction value, driving the optimal control solution to slide to the right and down along the curve, and actively sacrificing part of the response speed to exchange for a substantial decrease in the intensity of the control action.

[0066] This smooth transition from working condition point A to working condition point B vividly illustrates the flexible control strategy described in the present application, that is, under the premise of ensuring that the production efficiency meets the standard, the accumulation of mechanical stress and thermal fatigue is maximally limited, thereby effectively delaying the physical degradation process of the equipment.

[0067] Step S4: After obtaining the corrected model and the corrected target function that can accurately describe the current equipment characteristics, the system enters the real-time control execution phase. In order to solve the strong coupling interference problem between variables such as rotation speed, temperature and pressure in the extrusion process, this step performs multi-variable dynamic decoupling of the extrusion molding process based on the corrected coupling coefficient matrix.

[0068] Specifically, the specific steps of the multivariable dynamic decoupling are: the system constructs a state space equation containing all control inputs (such as screw speed instructions, heating power of each zone, back pressure valve opening) and controlled outputs (such as melt pressure, melt temperature, output). The system calculates the inverse matrix of the modified coupling coefficient matrix, and inputs the multivariable control instruction into the inverse matrix for operation. Through this linear algebra transformation, the originally physically entangled and interdependent coupling control variables are mathematically decomposed into several logically independent single-variable control variables. This allows the controller to design high-performance regulators for each decoupled virtual single loop.

[0069] Optionally, considering that any mathematical model cannot completely eliminate the deviation from the real physical process, the embodiment further combines a prediction-feedback double closed loop correction mechanism on the basis of decoupling control. The prediction-feedback double closed loop correction mechanism includes a feedforward prediction path and a feedback correction path. In the feedforward path, the system outputs a preset proportion of the total control instruction based on the melt flow trend prediction model. The prediction model directly calculates the theoretically required control amount according to the current set value change and disturbance input using the modified coupling parameters, and takes it as the main input, for example, 80% of the total control amount, to ensure the fast response capability of the system.

[0070] In the feedback path, the system calculates the deviation value between the actual value and the set value based on the real-time collected process data. In order to dynamically balance the prediction confidence of the model and the feedback correction of the reality, the system introduces an adaptive adjustment mechanism: when the deviation value exceeds the preset prediction deviation threshold, it indicates that the current prediction model may have a large instantaneous error or be affected by unmodeled external disturbances, at this time the system automatically increases the proportion of feedback correction in the total control amount, otherwise the preset basic feedback proportion is maintained. This mechanism ensures that the system mainly relies on feedforward to obtain high-speed response when the model is accurate, and mainly relies on feedback to ensure the robustness and final control accuracy of the system when the model is inaccurate.

[0071] For example, in order to solve the physical conflict of different control objectives in the time response scale, for example, the establishment of pressure is on the order of milliseconds, and the balance of temperature is on the order of minutes. The system performs hierarchical control according to the preset time scale. The steps of performing hierarchical control according to the preset time scale specifically include the cooperative operation of three levels:

[0072] At the first control time scale layer, such as a fast response layer set to 10ms to 500ms, the system focuses on equipment safety and transient suppression. When the aging trend prediction value is detected to be at a first risk level, or a sharp fluctuation in pressure is monitored, this layer has the highest priority and can directly issue instructions to the actuator to adjust the back pressure valve opening and local heating power to quickly relieve pressure or compensate for thermal shock, preventing mechanical overload.

[0073] At the second control time scale layer, such as a process control layer set to 1s to 10s, the system aims to find the local optimal operating point. Through a heuristic global search algorithm (such as an improved particle swarm algorithm or genetic algorithm) combined with model predictive control (MPC), the screw speed setpoint and temperature setpoint of each temperature zone are adjusted under the premise of meeting safety constraints. The second control time scale is greater than the first control time scale, and this layer uses the decoupled model for fine adjustment.

[0074] At the third control time scale layer, such as an energy efficiency optimization layer set to minutes to hours, the system focuses on long-term comprehensive benefits. Based on the modified multi-objective optimization function, steady-state coordinated control is performed. The third control time scale is greater than the second control time scale. At this level, the system does not pursue immediate elimination of every small fluctuation, but adjusts the long-term set of the operating point to achieve the best balance between energy consumption and quality for the entire production process.

[0075] In addition, to adapt to fluctuations in raw material properties, the layered control adopts an adaptive weight allocation strategy that dynamically adjusts the weight parameters in the multi-objective optimization function according to the raw material melt index change rate and pressure standard deviation. For example, when a sharp fluctuation in the raw material melt index is detected, the system automatically increases the precision weight to prioritize product quality; when the raw material state is stable, the system automatically increases the energy consumption weight to prioritize reducing production costs.

[0076] Step S5: As the final closed loop of the adaptive control strategy, the embodiment also integrates a full life cycle maintenance management function. The method further includes a predictive maintenance warning step aimed at eliminating unplanned downtime and extending equipment service life.

[0077] Specifically, the system divides the equipment state into multiple aging grades, such as healthy period, stable wear period, accelerated wear period, and impending failure period, according to the aging trend prediction value. And set the corresponding dynamic maintenance threshold for each aging grade. This dynamic threshold design avoids false positives or false negatives of fixed thresholds at different aging stages. When the monitoring data triggers the dynamic maintenance threshold, the system immediately generates a maintenance warning signal and recommends the best maintenance time window through the human-machine interface or enterprise MES system. For example, when it is detected that screw wear will cause the energy efficiency ratio to be lower than the economic critical point, the system suggests replacing the screw in the next production gap.

[0078] Finally, after detecting the completion of the maintenance action, for example, the maintenance personnel confirm that the new screw is replaced and perform system reset operation, then the system executes self-learning reset logic, automatically resets the multivariate coupling coefficient matrix and the equipment aging comprehensive index to the initial state. This marks the beginning of a new adaptive control cycle, and the system will start accumulating data and iterating the model based on the physical properties of the new components.

[0079] Through the organic synergy of the above five steps, the method described in this embodiment successfully transforms the equipment aging, which is usually considered as a negative factor of interference, into an explicit variable in the control algorithm, and realizes the energy efficiency steady-state control of the plastic extrusion molding process in the whole life cycle through dynamic compensation and multi-objective optimization.

[0080] Embodiment Two:

[0081] This embodiment further deepens the technology for the most risky non-steady state working condition in the plastic extrusion molding process based on embodiment one. This embodiment focuses on describing a transient overload penetration control step for the cold start or high viscosity raw material switching stage. This control strategy aims to solve the technical problem that when the equipment is in an old state, the extruder cannot overcome the huge static friction or shear yield stress at the moment of starting due to the traditional control logic overemphasizing equipment protection and energy consumption limitation, resulting in a deadlock state. At the same time, this embodiment introduces a stress gradient monitoring mechanism based on high-order differentiation, effectively preventing catastrophic mechanical accidents caused by screw fatigue brittle fracture during forced start.

[0082] For example, the transient overload penetration control step described in this embodiment is configured to control the core safety and execution logic of the system, which has a higher execution priority than the hierarchical collaborative control logic described in embodiment one. This step is strictly set before the hierarchical control is executed to ensure that the system has safely completed the transition of the working condition before entering the steady-state regulation. The control system first starts the real-time monitoring program of the extruder running state, which focuses on identifying whether there is a significant lag or decoupling phenomenon between the control instruction and the mechanical response.

[0083] Optionally, the system monitors the actual rotational speed of the screw in real time with the set rotational speed to obtain a deviation value . The deviation value is defined as the absolute value of the set rotational speed minus the actual rotational speed. In a physical sense, if the screw fails to rotate as expected after the motor is energized, the deviation value will quickly increase. The system compares the obtained deviation value with a preset start-up dead zone threshold value in real time. The start-up dead zone threshold value is set to a value representing the motor's locked-rotor or low-rotational-speed critical state, for example, fifty percent of the set rotational speed. At the same time, the system uses the device aging comprehensive index calculated in Embodiment One as an auxiliary basis for judgment.

[0084] If the system detects that the following two conditions are met at the same time: one, the deviation value exceeds the preset start-up dead zone threshold value ; and two, the device aging comprehensive index indicates that the device is in a high-aging state, the system determines that the current working condition has fallen into a deadlock state that cannot be solved by the regular control. In this state, if the steady-state control strategy designed for energy saving and long service life in Embodiment One is continued to be used, the controller will limit the current output, causing the device to fail to start. Therefore, the system immediately performs a logical switch, suspends the use of the modified multi-objective optimization function, and activates the transient overload penetration mode.

[0085] By way of example, after the transient overload penetration mode is activated, the control strategy changes from flexible to rigid. The system generates a pulse start-up torque instruction that allows the maximum torque limit of steady-state operation to be exceeded. This instruction is not a constant signal that lasts for a long time, but a series of high-energy pulse sequences that are adjusted, and the amplitude is allowed to break through the safety boundary of the rated torque of the device for a short time, for example, to one hundred and twenty percent to one hundred and fifty percent of the rated torque. The physical basis of this design is that the carbonized layer or wear groove often exists on the inner wall of the aging extruder barrel, causing the static friction coefficient between the material and the barrel and screw to be much larger than the dynamic friction coefficient. Only by applying a transient shear force that exceeds the regular limit can the constraint of static friction be broken, and the material can be converted from a solid or high-viscoelastic state to a flowable viscous state.

[0086] As Figure 4 intuitively shows the dynamic response process of the system after the transient overload penetration mode is activated, which effectively proves the technical feasibility of the invention in solving the deadlock problem. The blue solid line on the left vertical axis in the figure represents the pulse start-up torque instruction issued by the controller, and the orange solid line on the right vertical axis in the figure represents the actual rotational speed response of the screw.

[0087] Figure 4 The middle gray dashed line indicates the rated torque safety limit when the device is in regular steady state operation, usually set to 100%. As can be clearly observed from the figure, at the time of 0.2 seconds after the system identifies the deadlock state, the control logic automatically generates a high-energy pulse with an amplitude of up to 140%, which significantly breaks through the regular safety limit line, and its physical meaning is to overcome the super-strong static friction generated by the high-viscosity material in the aging barrel wall by applying a transient and huge shear stress.

[0088] With the continuous action of the pulse torque, the screw speed curve starts to rise rapidly after experiencing a short physical inertia lag corresponding to the interval of about 0.2 seconds to 0.45 seconds in the figure, indicating that the static friction has been successfully broken, and the rotor starts to accelerate. When the speed curve reaches the preset target speed near 0.6 seconds, the system immediately returns the torque command to the steady-state load level of 80%, thereby avoiding the thermal damage to the motor coil caused by long-time overload while completing the start breakthrough task, realizing the perfect connection of rigid breakthrough and flexible protection.

[0089] Optionally, in order to solve the sharp contradiction between applying an overload torque and preventing the screw from breaking, the embodiment introduces a high-frequency stress dynamics monitoring mechanism. When the system outputs the pulse start torque command, a microsecond-level monitoring thread is started to calculate the stress accumulation gradient of the current monitoring torque value in real time . The calculation of the stress accumulation gradient not only focuses on the size of the torque, but also more deeply combines the first-order derivative of the torque with respect to time and the second-order derivative of the torque with respect to time. Its mathematical and physical model is defined as follows:

[0090] ;

[0091] Wherein, represents the real-time monitoring torque value after high-frequency sampling and anti-aliasing filtering processing; represents the time variable represents the absolute value of the first-order derivative of the torque with respect to time, which physically represents the rate of stress loading, i.e. the speed of torque rising; represents the absolute value of the second-order derivative of the torque with respect to time, which physically represents the acceleration of torque change, i.e. the degree of stress fluctuation; is the material brittleness coefficient.

[0092] For example, the material brittleness coefficient It is not a static constant, but a dynamic variable positively correlated with the wear characteristics of the screw. As the equipment's service life increases, the alloy steel material used in the screw undergoes fatigue evolution in its crystal structure under long-term exposure to high temperature and alternating stress, resulting in a decrease in material toughness and a significant increase in brittleness. When the comprehensive equipment aging index in Example 1... When the temperature rises, the system will automatically adjust the adjustment. The value. Increase it. The physical significance lies in the significant increase in the second derivative term throughout the stress accumulation gradient. The weighting in the equation. Because, according to the principles of fracture mechanics, before macroscopic fracture occurs in brittle materials, the propagation of internal microcracks causes high-frequency nonlinear oscillations of stress waves, which mathematically manifest as violent fluctuations in the second derivative. This is achieved through dynamic adjustment. The system is highly sensitive to subtle signs of aging screws breaking.

[0093] Optionally, considering the digital nature of industrial control systems, the derivative calculations in the above formulas are discretized in the processor using high-order difference equations. To ensure the capture of transient stress oscillations, the system adjusts the sampling period in transient overload penetration mode. The time frame is shortened to the range of 100 to 500 microseconds. For the first derivative of torque over time, the system uses the following first-order backward difference formula for approximate calculation:

[0094] ;

[0095] For the second derivative of torque as a function of time, the system uses the following second-order central difference formula for approximate calculation:

[0096] ;

[0097] in, The current sampling time The torque value; The previous sampling time The torque value; The previous sampling time The torque value.

[0098] After calculating the real-time stress accumulation gradient, the system performs a critical safety threshold determination. The system has a preset structural failure safety limit. This limit is pre-calibrated based on the yield strength limit, section modulus, and fatigue life curve of the screw material through finite element analysis. If the accumulated stress gradient is detected... Exceeding the preset structural failure safety limit This indicates that although the absolute value of the torque may not have reached the fracture limit, the stress accumulation mode has already exhibited highly unstable high-frequency oscillation characteristics, suggesting that irreversible plastic deformation or crack propagation is occurring inside the screw. At this critical moment, the system does not wait for the torque to reach its maximum value, but immediately forces a stop to the output torque command, cuts off the energy supply to the main drive motor, and triggers a first-level mechanical fault alarm to prevent shaft breakage due to blindly forcing the screw.

[0099] Optionally, if the startup process proceeds smoothly, the static friction of the material is successfully overcome, and the aforementioned safety blocking logic is not triggered during monitoring, the system will enter the exit determination process. The system continuously monitors the actual rotational speed of the screw. If the actual rotational speed is detected... Within the preset safety time limit The preset steady-state speed threshold was reached. This signifies the end of the transient process during startup, and the extrusion process has entered a relatively rheologically stable viscous flow transport stage. At this point, to avoid motor overheating due to prolonged operation in overload mode, the system immediately exits the transient overload penetration mode and resumes using the modified multi-objective optimization function. Control is then smoothly transferred to the hierarchical collaborative control module described in Example 1, and the system restarts refined closed-loop regulation based on optimal energy efficiency and equipment life extension objectives.

[0100] Through the transient overload penetration control steps described in this embodiment, the present invention effectively solves the dilemma of aging extrusion equipment being unable to operate under cold start and high viscosity material change conditions, and the risk of shaft breakage. Through in-depth analysis of physical stress state using mathematical models, adaptive safety control under extreme conditions is achieved.

[0101] Example 3:

[0102] This embodiment further refines the robustness design of the feedback control loop. Specifically addressing the complex electromagnetic interference and unstructured noise environments commonly found in industrial extrusion production sites, this embodiment elaborates on a signal fidelity discrimination and suppression mechanism based on information entropy theory.

[0103] This mechanism aims to address the technical challenge of traditional control systems easily misjudging high-amplitude random noise, leading to control divergence or equipment oscillation. By introducing the concept of entropy from thermodynamics and information theory, this embodiment endows the controller with the intelligent ability to distinguish between real physical deviations and spurious noise interference, ensuring that the system only responds positively to physically meaningful trend deviations.

[0104] By way of example, the signal fidelity discrimination mechanism described in the present embodiment is seamlessly embedded in the feedback regulation path of the hierarchical collaborative control module described in Embodiment 1. In actual operation logic, when the control system monitors that the real-time deviation value of a process parameter (e.g. the melt pressure at the die or the temperature of the barrel) exceeds the preset prediction deviation threshold, the system should originally increase the proportion of the feedback correction in the total control to quickly eliminate the deviation. However, before performing this action, the present embodiment forcibly adds a signal fidelity discrimination mechanism based on the deviation information entropy for constraint. The core of this constraint step is to quantify the degree of confusion of the deviation signal, which serves as the basis for adjusting the reliability of the feedback gain.

[0105] Optionally, in order to accurately capture the statistical characteristics of the deviation signal, the system first opens an independent buffer area in the memory to establish a time sliding window. The time sliding window is configured as a data queue with a fixed length and following the first-in-first-out (FIFO) principle. The capacity of the window is not randomly set, but is matched according to the dynamic response constant of the controlled object. For example, for a pressure control loop, considering the rapidity of pressure fluctuation, the window capacity may be set to contain data points of the past fifty to one hundred sampling periods. The system real-time collects the historical deviation data sequence within the window, which faithfully records the numerical set of the difference between the set value and the actual value in the recent period of time.

[0106] By way of example, after obtaining the original deviation sequence, the system does not perform simple averaging in the time domain, but instead performs statistical analysis in the probability domain. This is because simple mean filtering cannot distinguish between constant deviation and high-frequency oscillation with zero mean. The system performs a probability density estimation step: first, determine the dynamic value range of the deviation data in the window, i.e. the difference between the maximum deviation value and the minimum deviation value. Then, the system divides the dynamic value range into a number of preset discrete state intervals. The number of discrete state intervals determines the resolution of the entropy value calculation, and is usually between ten and twenty. Next, the system traverses the entire historical deviation data sequence, counts the frequency of data falling into each preset discrete state interval, and divides the frequency by the total number of data points in the sequence , thereby calculating the probability of the deviation data falling into the th interval.

[0107] Optionally, based on the probability distribution vector calculated above, the system invokes the Shannon information entropy formula to quantify the uncertainty of the current deviation signal. According to this, the deviation information entropy in the time sliding window is calculated. The mathematical definition formula is as follows

[0108] ;

[0109] wherein, denotes the probability that the deviation data falls into the th discrete state interval; denotes the natural logarithm operator; denotes the total number of discrete state intervals.

[0110] The formula reveals the ordered degree of the signal. If the current deviation is caused by real equipment failure, the deviation data will show obvious trend and concentrate in one or several adjacent intervals, resulting in extremely uneven probability distribution. At this time, the calculated deviation information entropy will present a smaller value, indicating that the signal contains a large amount of information and has high certainty. Conversely, if the current deviation is caused by random electromagnetic noise of the frequency converter switching frequency or poor contact of the sensor line, the deviation data will jump out of order in the entire value range, resulting in uniform distribution in each interval. At this time, the calculated deviation information entropy will tend to the theoretical maximum value, indicating that the signal is extremely chaotic and belongs to invalid noise.

[0111] After the deviation information entropy which can represent the purity of the signal is calculated, the system needs to convert it into a regulation parameter that has a substantial impact on the control strategy. To this end, the system calculates a feedback gain confidence factor according to the deviation information entropy . The feedback gain confidence factor is defined as a dimensionless weight coefficient between zero and one, which directly reflects the degree of trust of the controller in the authenticity of the current deviation signal. According to the design principle, the feedback gain confidence factor is negatively correlated with the deviation information entropy, that is, the more chaotic the signal (the higher the entropy), the less the system trusts the signal (the lower the confidence). The mapping relationship adopts the following nonlinear attenuation formula:

[0112] ;

[0113] wherein, is a preset noise sensitivity coefficient. The coefficient is a positive adjustable engineering parameter that is used to adjust the immunity of the algorithm to noise. In an old workshop with extremely poor electromagnetic environment, the value of can be appropriately increased, so that even if the entropy value only rises slightly, the confidence factor will also rapidly decrease, thereby enhancing the anti-interference ability of the system.

[0114] Optionally, after calculating the confidence factor, the system formally intervenes in the feedback adjustment execution phase. When performing the operation of increasing the proportion of the feedback correction in the total control quantity, the system no longer directly adopts the theoretical increment calculated by conventional PID algorithms or fuzzy logic, but instead utilizes the feedback gain confidence factor. The increment of the feedback correction is weighted and attenuated.

[0115] Specifically, assuming the control system originally planned to adjust the feedback weight from the current baseline value based on the current deviation magnitude. Increase to target value Then the theoretical weight increment is At this point, the actual feedback weight, after being corrected by the mechanism in this embodiment, The calculation formula is:

[0116] ;

[0117] A thorough analysis of this formula reveals its ingenious control logic: when the deviation information entropy At lower levels, Approaching one Approaching The system fully responds to deviations, significantly increasing feedback to quickly correct errors; and when the deviation information entropy At higher levels, Approaching zero This term then approaches zero, making Forcibly platinum Nearby. This means that, faced with high-entropy noise, the system actually rejects the request to increase the feedback weight, and the smaller the actual increase in feedback weight, the more it avoids drastic fluctuations in the actuator (such as screw speed or heating power) caused by misjudging noise.

[0118] For example, in addition to the flexible suppression mechanism for conventional noise described above, this embodiment also includes hard circuit breaker logic for extreme fault conditions. When the calculated deviation information entropy... Exceeding the preset entropy threshold At this point, the system will no longer perform flexible weighted calculations, but will directly determine the current deviation as random noise interference. This extremely high entropy state is usually no longer a normal process fluctuation, but corresponds to abnormal situations such as sensor probe disconnection, data acquisition card hardware failure, or extremely strong lightning interference.

[0119] Under this trigger condition, the system performs two key operations: one is to force the actual feedback weight to be equal to the basic feedback proportion, that is, to completely shield any attempt to adjust the control parameters according to the current error signal, to rely on the system's feedforward model to maintain the inertial operation of the device, and to prevent fault expansion; the second is to immediately start the filtering program to smooth the collected signal. The filtering program can select a first-order lag filter algorithm or a Kalman filter algorithm, which uses the statistical characteristics of historical data to reconstruct the current observation value, until the calculated entropy value falls within the normal confidence interval, and the system is unlocked to restore normal closed-loop control function.

[0120] As Figure 5 The figure intuitively shows the smoothing suppression effect of the signal fidelity discrimination mechanism of the present application on the actual feedback weight adjustment process when facing sudden high-entropy noise interference environment. The horizontal axis represents the continuous control period, and the vertical axis represents the actual feedback weight value finally adopted by the system.

[0121] Figure 5 The blue dashed line located at the value 0.3 of the vertical axis represents the basic feedback proportion preset by the system, which is the safe reference weight that the system should maintain under stable state operation or when the signal is not reliable, symbolizing the stable point of the system operation. The gray background area in the figure marks the high-entropy noise interference period from the 30th to the 70th period. During this period, due to strong electromagnetic interference on the sensor, the theoretically calculated deviation value presents a dramatic random jump. The gray dashed line in the figure represents the weight response under the traditional control strategy, which can be seen that it completely loses the constraint of the blue reference line, deviates greatly from the safe reference value and follows the noise to oscillate at a high frequency. This oscillation, if directly applied to the actuator, will lead to unstable operation of the device.

[0122] In sharp contrast to this, Figure 5 The orange solid line in the figure represents the adjustment curve after adopting the information entropy discrimination strategy based on the present application. Due to the significant increase in the real-time calculation of the deviation information entropy in this interval, the feedback gain confidence factor decays rapidly, thereby driving the control algorithm to reject the false deviation increment and intelligently clamp the actual feedback weight near the basic feedback proportion shown by the blue dashed line. This characteristic of forced regression to the safe reference line in extreme noise environment avoids control divergence and effectively proves the significant advantage of the mechanism in improving the robustness of the system.

[0123] Through the deviation information entropy-based signal fidelity discrimination mechanism described in Example Three, the present method enables the system not only to perceive the size of the deviation, but also to deeply understand the nature of the deviation. This mechanism not only ensures a sharp response to real process drift, but also achieves almost perfect immunity to false environmental noise, greatly improving the steady-state accuracy and long-term reliability of the plastic extrusion molding process control system.

[0124] Embodiment Four

[0125] Based on the method flow elaborated in the foregoing Embodiment One to Embodiment Three, this embodiment further discloses a self-adaptive energy efficiency steady-state control system for a plastic extrusion molding process from the perspective of hardware architecture and functional modularization. The system is designed as a software and hardware deeply integrated industrial control solution, aiming to solve the technical pain points of control model mismatch, low energy efficiency and production stability decline caused by equipment component aging, wear and tear and thermal response attenuation as pointed out in the background technology. The system of this embodiment can perform the method described in Embodiments One to Three, and realizes the optimal control of the extruder throughout its life cycle through the organic combination of multi-dimensional perception, deep learning prediction and adaptive compensation.

[0126] As shown in Figure 6 , the self-adaptive energy efficiency steady-state control system for a plastic extrusion molding process described in this embodiment mainly includes four core functional modules in its logic architecture: a multi-dimensional data acquisition module, a data processing and analysis module, a parameter dynamic compensation module and a hierarchical collaborative control module. The modules interact in real time through a high-speed data bus, and collaboratively complete the perception, decision-making and execution of the extrusion process.

[0127] By way of example, the multi-dimensional data acquisition module, as the system's sense organ for the physical world, is configured to synchronously acquire process data of each sub-domain of the extruder, aging characteristic data of the equipment components and production auxiliary data. The module integrates a multi-channel high-precision analog input interface and a field bus communication card, such as an EtherCAT or Profinet interface, on the hardware to ensure the time synchronization and high-frequency characteristics of data acquisition. For process data representing the current production state, the module reads key variables including barrel zone temperature, melt pressure, screw speed and main machine current in real time to build a basic data set reflecting the extrusion rheological state. For aging characteristic data representing the health of the equipment, the module does not stop at simply receiving passively, but has built-in specific test sequence trigger logic that can capture and calculate the long-term drift of the screw torque and the heating duration of the heating components under standard test conditions. In addition, the module also synchronously acquires production auxiliary data including environmental temperature and humidity and power grid voltage fluctuations to provide data support for subsequent removal of environmental noise interference.

[0128] The data processing and analysis module is configured as a computing core of the system, and is configured to preprocess the collected data, calculate a device aging comprehensive index, and run a long short-term memory network to output an aging trend prediction value and an aging rate. In the data preprocessing stage, the module performs a moving average filtering and an outlier rejection algorithm, and maps heterogeneous data to a normalized space of zero to one. Subsequently, the module calls a preset feature fusion algorithm, uses a weighting coefficient The normalized torque long-term drift, heating open temperature duration, and sensor drift error are weighted and summed to obtain a device aging comprehensive index that quantifies the device health status . More importantly, the module integrates a pre-trained long short-term memory (LSTM) inference engine. The engine takes a historical aging index sequence as input, and uses a gating mechanism of a forgetting gate and an output gate to predict a device performance degradation trajectory in a future time window, output an aging trend prediction value , and an aging rate representing a current aging acceleration . This functional module enables the system to understand the device aging trend from the data, and realizes a technical leap from passive response to active prediction.

[0129] The parameter dynamic compensation module is configured as an adaptive adjustment engine of the system, and is configured to correct a multivariate coupling coefficient matrix and a multi-objective optimization function in real time based on the device aging comprehensive index and the aging rate. The module aims to eliminate the deviation between physical device aging and fixed control models. Specifically, when it is monitored that an abrasion threshold is exceeded, or a decay rate threshold is exceeded, the module uses a first correction factor and a second correction factor to dynamically adjust a rotational speed-pressure coupling coefficient and a temperature-pressure coupling coefficient , thereby reconstructing a transfer function matrix of the system. At the same time, the module dynamically redistributes each weight component in the multi-objective optimization function , and in particular, increases the aging weight , so that the corrected multi-objective optimization function can actively avoid aggressive control strategies that may exacerbate device mechanical fatigue or thermal shock in the optimization process, thereby maintaining production efficiency while maximizing the delay of device life decay,

[0130] For example, the hierarchical cooperative control module is configured as the final execution unit of the system, and is configured to perform multivariable dynamic decoupling based on the corrected coupling coefficient matrix, and to send hierarchical control instructions to the extruder actuator according to different time scales of different levels in combination with a prediction-feedback double closed-loop correction mechanism. The module internally runs complex decoupling operation logic, and calculates the inverse matrix of the corrected coupling coefficient matrix to decompose the strong coupling relationship among the multivariables such as rotation speed, temperature, and pressure into independent single-loop control channels. On this basis, the module integrates the signal fidelity discrimination mechanism described in Embodiment Three, and uses the deviation information entropy to calculate the feedback gain confidence factor , intelligently filters high-entropy random noise in the field, and ensures the accuracy of feedback regulation.

[0131] Optionally, the hierarchical cooperative control module also embeds the transient overload penetration control logic described in Embodiment Two. In extreme working conditions such as device cold start or high-viscosity raw material switching, if it is monitored that the system is in a deadlock state, the module can temporarily take control, activate the transient overload penetration mode, output high-energy pulse torque instructions, and monitor the stress accumulation gradient in real time to prevent screw breakage and ensure safe start of the device. In the steady-state running phase, the module strictly performs safety interlocking, process control, and energy efficiency optimization tasks according to three time dimensions of millisecond, second, and minute, respectively, to achieve full-time domain accurate coverage of the extrusion process.

[0132] In summary, the system provided in the embodiment solves the technical problem that the traditional control system cannot adapt to the time-varying characteristics of the device through close cooperation of the modules, and significantly improves the stability and energy efficiency of plastic extrusion production by fusing the aging prediction and dynamic compensation mechanisms, thereby providing strong hardware and algorithm support for predictive maintenance and intelligent management of factory equipment.

[0133] Embodiment Five

[0134] Corresponding to the above-mentioned embodiments, the application further provides an electronic device.

[0135] As Figure 7 shown is a structural schematic diagram of an electronic device in the application, the electronic device 100 comprises a processor 101 and a memory 103. The processor 101 and the memory 103 are connected, for example, connected through a bus 102. Optionally, the electronic device 100 can further comprise a transceiver 104. It should be noted that in actual application, the transceiver 104 is not limited to one, and the structure of the electronic device 100 does not constitute a limitation on the embodiments of the application.

[0136] The processor 101 can be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in conjunction with the present disclosure. The processor 101 can also be a combination that implements a computing function, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0137] The bus 102 can include a path that transmits information between the above-mentioned components. The bus 102 can be a PCI bus or an EISA bus, etc. The bus 102 can be divided into an address bus, a data bus, a control bus, etc. For convenience of representation, Figure 7 In the figure, only one thick line is used, but it does not mean that there is only one bus or only one type of bus.

[0138] The memory 103 is used to store a computer program corresponding to the adaptive energy efficiency steady state control method of a plastic extrusion molding process of the above-mentioned embodiments of the present application, which is controlled and executed by the processor 101. The processor 101 is used to execute the computer program stored in the memory 103 to realize the content shown in the above-mentioned method embodiments.

[0139] The electronic device 100 includes, but is not limited to, a mobile terminal such as a notebook computer, a PAD (tablet computer), and the like, and a fixed terminal such as a desktop computer, and the like. Figure 7 The electronic device 100 shown is only an example, and should not bring any limitation to the function and use range of the embodiments of the present application.

[0140] Although the embodiments of the present application have been shown and described above, it should be understood that the above-mentioned embodiments are exemplary and should not be understood as limiting the present application, and those skilled in the art can make changes, modifications, replacements, and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. An adaptive energy efficiency steady-state control method for a plastic extrusion molding process, characterized in that, Includes the following steps: Collect multi-dimensional data of the plastic extrusion molding process, including process data characterizing the current production status, aging characteristic data characterizing the health of the equipment, and auxiliary data; The multi-dimensional data is preprocessed and feature fusion is performed to calculate the comprehensive aging index of the equipment. An aging trend prediction model is constructed based on the long short-term memory network to output the aging trend prediction value and the aging rate of the screw and barrel. Based on the comprehensive aging index of the equipment and the aging rate, the preset multivariate coupling coefficient matrix and multi-objective optimization function are dynamically compensated and corrected to generate the corrected coupling coefficient matrix and the corrected multi-objective optimization function. Based on the modified coupling coefficient matrix, the extrusion molding process is dynamically decoupled in multiple variables, and combined with the prediction-feedback dual closed-loop correction mechanism, hierarchical control is executed according to the preset time scale to achieve energy efficiency steady-state control of the plastic extrusion molding process. The multivariate dynamic decoupling includes: The multivariable control command is input into the inverse matrix of the modified coupling coefficient matrix for calculation, thereby decomposing the originally coupled control variables into independent single-variable control quantities; The prediction-feedback dual closed-loop correction mechanism includes: outputting a pre-control command based on a melt flow trend prediction model, which accounts for a preset proportion of the total control quantity, and calculating the deviation value in combination with the real-time acquired process data; When the deviation value exceeds the preset prediction deviation threshold, a signal fidelity identification mechanism based on deviation information entropy is used to constrain and adjust the proportion of the feedback correction amount in the total control amount; otherwise, the preset basic feedback proportion is maintained. The signal fidelity discrimination mechanism specifically includes: A time-sliding window is established, and historical deviation data sequences within the window are collected. The probability that the historical deviation data falls into a preset discrete state interval is calculated, and the deviation information entropy within the time-sliding window is calculated accordingly. A feedback gain confidence factor is calculated based on the deviation information entropy, and the feedback gain confidence factor is negatively correlated with the deviation information entropy. The increment of the feedback correction is weighted and attenuated using the feedback gain confidence factor, so that when the deviation information entropy is higher, the actual increase in feedback weight is smaller.

2. The method according to claim 1, characterized in that, The aging characteristic data includes screw wear characteristics and heating response characteristics; The screw wear characteristics are characterized by long-term torque drift, which is defined as the percentage change in the current measured torque value relative to the initial reference torque value of the equipment under standard test conditions. The heating response characteristics are characterized by the heating time, which is defined as the time difference required for the heating element to rise from a preset initial temperature to a preset target temperature. The specific steps for calculating the comprehensive equipment aging index are as follows: using preset weighting coefficients, the normalized long-term torque drift, the heating time, the zero drift error of the pressure sensor, and the response delay of the temperature sensor are weighted and summed to obtain the comprehensive equipment aging index.

3. The method according to claim 1, characterized in that, The step of dynamically compensating and correcting the multivariable coupling coefficient matrix specifically includes: The comprehensive aging index of the equipment is compared with a preset wear threshold. If the wear threshold is exceeded, the speed-pressure coupling coefficient in the multivariate coupling coefficient matrix is ​​corrected. The aging rate is compared with a preset decay rate threshold. If the decay rate threshold is exceeded, the temperature-pressure coupling coefficient in the multivariable coupling coefficient matrix is ​​corrected. The update period of the modified coupling coefficient matrix is ​​dynamically set according to the risk level corresponding to the aging trend prediction value: When the predicted aging trend value is at the first risk level, it is updated at the first cycle frequency; when the predicted aging trend value is at the second risk level, it is updated at the second cycle frequency. The first risk level corresponds to a higher degree of aging than the second risk level, and the first cycle frequency is higher than the second cycle frequency.

4. The method according to claim 3, characterized in that, The correction methods for the speed-pressure coupling coefficient and the temperature-pressure coupling coefficient follow the following rules: The corrected speed-pressure coupling coefficient is equal to the original speed-pressure coupling coefficient multiplied by the first correction factor, which is positively correlated with the comprehensive aging index of the equipment. The corrected temperature-pressure coupling coefficient is equal to the original temperature-pressure coupling coefficient multiplied by a second correction factor, which is positively correlated with the aging rate.

5. The method according to claim 1, characterized in that, The modified multi-objective optimization function is used to find the optimal control solution in the hierarchical control, and the multi-objective optimization function contains four weighted components: The first component is the ratio of the response speed deviation to the reference response time, corresponding to the response weight; The second component is the ratio of control accuracy deviation to reference accuracy, corresponding to the accuracy weight; The third component is the ratio of energy consumption per unit of qualified product to benchmark energy consumption, corresponding to the energy consumption weight. The fourth component is the predicted aging trend value, corresponding to the aging weight; The modified multi-objective optimization function reduces the intensity of control actions when the equipment is severely aged by increasing the aging weight.

6. The method according to claim 1, characterized in that, The step of performing hierarchical control according to a preset time scale specifically includes: At the first control time scale layer, when the aging trend prediction value is detected to be at the first risk level, instructions to adjust the back pressure valve opening and local heating power are directly issued to the actuator. In the second control time scale layer, the screw speed setpoint and the temperature setpoint of each temperature zone are adjusted by combining a heuristic global search algorithm with model predictive control, wherein the second control time scale is greater than the first control time scale; At the third control time scale layer, steady-state cooperative control is performed based on the modified multi-objective optimization function, wherein the third control time scale is greater than the second control time scale. The hierarchical control adopts an adaptive weight allocation strategy, which dynamically adjusts the weight parameters in the multi-objective optimization function according to the rate of change of the raw material melt index and the pressure standard deviation.

7. The method according to claim 5, characterized in that, The method further includes a transient overload penetration control step for cold start or high-viscosity raw material switching phases, which makes a determination before performing the stratification control: The deviation between the actual speed of the screw and the set speed is monitored in real time. When the deviation exceeds the preset start-up dead zone threshold and the comprehensive aging index of the equipment indicates that the equipment is in a high aging state, the modified multi-objective optimization function is suspended and the transient overload penetration mode is activated. In the transient overload penetration mode: a pulse start torque command is generated that allows the torque to exceed the maximum torque limit of steady-state operation, and the stress accumulation gradient of the current monitored torque value is calculated in real time; The calculation of the stress accumulation gradient combines the first derivative of torque over time and the second derivative of torque over time. If the stress accumulation gradient is detected to exceed the preset structural failure safety limit, the output torque command will be forcibly stopped immediately. If the actual rotational speed is detected to have reached the preset steady-state rotational speed threshold within a preset safety time limit, the transient overload penetration mode is exited, and the modified multi-objective optimization function is resumed.

8. The method according to claim 1, characterized in that, The signal fidelity identification mechanism also includes: When the calculated deviation information entropy exceeds the preset entropy threshold, the current deviation is determined to be random noise interference. The actual feedback weight is kept equal to the basic feedback ratio, and a filtering program is started to smooth the acquired signal.

9. The method according to claim 1, characterized in that, The method also includes a predictive maintenance early warning step: The equipment status is divided into multiple aging levels based on the aging trend prediction value, and a corresponding dynamic maintenance threshold is set for each aging level. When the monitoring data triggers the dynamic maintenance threshold, a maintenance warning signal is generated and a maintenance time window is recommended; Upon detection that a maintenance action has been completed, the multivariate coupling coefficient matrix and the equipment aging comprehensive index are automatically reset to their initial states.

10. An adaptive energy-efficient steady-state control system for a plastic extrusion molding process, characterized in that, The system for performing the method according to any one of claims 1 to 9 comprises: The multi-dimensional data acquisition module is configured to simultaneously acquire process data from each subdomain of the extruder, aging characteristic data of equipment components, and production auxiliary data. The data processing and analysis module is configured to preprocess the collected data, calculate the comprehensive aging index of the equipment, and run the long short-term memory network to output the aging trend prediction value and aging rate. The parameter dynamic compensation module is configured to correct the multivariate coupling coefficient matrix and the multi-objective optimization function in real time based on the comprehensive aging index of the equipment and the aging rate. The hierarchical collaborative control module is configured to perform multivariable dynamic decoupling based on the modified coupling coefficient matrix, and combined with the prediction-feedback dual closed-loop correction mechanism, it sends hierarchical control commands to the extruder actuator according to different time scales.

Images