An adaptive control method for an insulation material extrusion molding process

Abstract

The present application belongs to the technical field of automatic control of thermal insulation material extrusion processing, and relates to a self-adaptive control method for the extrusion molding process of thermal insulation material, comprising: collecting screw drive frequency instructions and finished product thickness deviation data to generate discrete sequences; pressing them into corresponding drive frequency history queue and thickness deviation history queue; scanning the drive frequency history queue to lock the inflection point step of the variable rate; defining a search window in the thickness deviation history queue and extracting the maximum variable rate feature item to establish a dynamic time delay parameter; updating the feedforward gain to adjust the screw drive frequency and output the target control instruction; the present application converts the transmission delay into the relative displacement of the storage space, establishes the time domain alignment capability of the excitation and response, avoids the feedback regulation divergence oscillation caused by the timing misalignment, and improves the cooperative self-healing efficiency of the control loop.

Description

Technical Field

[0001] This invention belongs to the field of automatic control technology for thermal insulation material extrusion processing, and relates to an adaptive control method for the thermal insulation material extrusion molding process. Background Technology

[0002] Currently, adaptive control of the plastic extrusion molding process using high-frequency multidimensional data sensing and intelligent optimization algorithms has become a core approach in the intelligent plastics processing industry. However, existing technologies still have limitations in handling the synergistic constraints of multidimensional rheological transient fluctuations in materials, long-cycle mechanical wear and aging of extrusion equipment, and high-frequency, refined anti-disturbance adjustment.

[0003] First, there is the end-point local control technology based on real-time acquisition of multi-dimensional parameters and reconstruction of temperature field. For example, Chinese invention patent CN120552336B discloses a control method and system for extrusion molding of insulation board. It uses a detection section set at the exit section of the extruder die head to simultaneously acquire melt temperature distribution, apparent viscosity based on differential pressure principle, and foaming agent concentration data based on infrared spectroscopy, and then performs evaluation and distributed dynamic reconstruction of temperature field. Its drawback is that the scheme is essentially a reactive instantaneous rheological state feedback at the end of the die head. Its control and decision-making logic lacks rigid characterization of the long-term mechanical degradation and cumulative wear of the hardware throughout the extrusion equipment process. It fails to map the nonlinear time-varying parameter drift caused by the long-term aging of the machine itself into the boundary constraints of the subsequent control space, which may lead to the control system outputting aggressive control commands that violate the actual physical load-bearing limit of the hardware in the middle and later stages of operation.

[0004] Second, there is the steady-state optimization technology based on Long Short-Term Memory (LSTM) network for predicting equipment aging trends. For example, Chinese invention patent CN121541496B discloses an adaptive energy efficiency steady-state control method and system for plastic extrusion molding processes. It calculates the comprehensive aging index of the equipment by integrating on-site process data and long-term aging characteristic data of the equipment, and constructs an aging trend prediction model based on LSTM network to dynamically correct the multivariate coupling coefficient matrix online. Its shortcomings are: the technology focuses on learning the degradation trajectory of equipment performance from a macroscopic trend and hierarchical steady-state optimization under long-scale time resolution. Although it effectively improves the system energy efficiency throughout the entire life cycle, it fails to closely link the identified macroscopic aging state with the transient high-frequency rheological characteristics of the micro-melt in the die outlet section, such as the sudden change in foaming agent concentration and local instantaneous viscosity fluctuation caused by abnormal batches of raw materials. It cannot map in real time the physical compression of the current thermodynamic control path by the high-frequency variation of complex materials under high flow rate, and it is difficult to cope with the quality defects caused by micro-rheological instability under non-steady-state transient conditions.

[0005] Third, there is the anti-disturbance adaptive adjustment technology based on recursive parameter identification and nonlinear sliding mode. For example, Chinese invention patent application CN121454947A discloses a cable extrusion control method based on adaptive control. It identifies parameters in real time by introducing a least squares model with cointegrated residual compensation terms and input regularization constraints, and determines the total adjustment signal by combining a trend-aware sliding mode controller with a composite control structure. Its drawback is that existing anti-disturbance adaptive control models usually regard the extrusion molding process as a continuously differentiable ideal state that satisfies high-frequency and high-speed correction, ignoring the intense stress impact on the physical structure caused by the strong correction of sliding mode control actions over a long time. When optimizing for long-cycle operation or facing highly sensitive polymer melts, if there is a lack of forced pruning and flexible amplitude limiting protection mechanisms for the aging and decay state of long-cycle equipment, the intense feedback control trajectory is prone to high-frequency mechanical chattering and local stress accumulation deviation, making the final output control scheme lose its practical value because it is easy to exceed the safety threshold of the physical structure failure of the screw or barrel.

[0006] Therefore, the technical problem to be solved by this invention is how to construct an adaptive control architecture with the ability to perceive the aging state throughout the entire life cycle, optimize the macroscopic steady state, and coordinate the microscopic high-frequency rheological transient disturbance resistance, so that the regulation logic output by the controller always converges to the real physical and process safety boundary throughout the entire life cycle. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art, solve the technical problem of the oscillation of regulation gain and residual error in the same direction caused by timing misalignment in traditional feedback regulation, and provide an adaptive control method for the extrusion molding process of thermal insulation materials.

[0008] To achieve the above-mentioned objectives, the present invention provides an adaptive control method for the extrusion molding process of thermal insulation materials, comprising the following steps: Step S1: Synchronously collect drive frequency and thickness deviation to generate dual sequences: Use the data bus to collect screw drive frequency commands at 100ms intervals to generate the first sequence, and simultaneously use the digital-to-analog conversion channel to collect finished product thickness deviation data at 100ms intervals to generate the second sequence. Step S2, push sequence components into the history queue and maintain a depth of 500 levels: push the current components of the first sequence and the current components of the second sequence into the tail of the corresponding driving frequency history queue and thickness deviation history queue in the internal storage area, respectively, while discarding the historical components at the head of the two history queues, and keeping the data storage depth of each history queue constant at 500 levels. Step S3, Scan the historical queue to extract slope mutation terms and lock the inflection point time step: Scan the historical queue of driving frequency, extract mutation terms whose absolute value of slope change of driving frequency command exceeds 15%, and lock their time domain position as the inflection point time step of the rate of change. Step S4, Delineate the lag window to detect the largest feature item and establish the time delay parameter: Delineate the lag retrieval window in the thickness deviation history queue based on the inflection point time step of the rate of change, retrieve the feature item with the largest absolute value of the deviation rate of change in the lag retrieval window, and calculate the time step difference between the feature item and the inflection point time step of the rate of change to establish the dynamic equivalent time delay parameter. Step S5: Update the gain frequency adjustment and output control command according to the time delay parameter: Update the feedforward compensation control gain according to the dynamic equivalent time delay parameter, adjust the screw drive frequency and output the target control command.

[0009] Step S5 of the present invention includes the following sub-steps: Step S51, collecting the driving frequency command variation rate in the driving frequency history queue through the adaptive confidence gating unit; Step S52, when the driving frequency command variation rate is within 15%, maintaining the dynamic equivalent delay parameter as the final hedging step size of the current control cycle; Step S53, when the driving frequency command variation rate exceeds 15%, truncating the dynamic equivalent delay parameter and limiting it within the preset stepped safety boundary range, while reducing the gain weight of the feedforward hedging term by 50%.

[0010] Step S5 of the present invention further includes the following sub-steps: Step S54, calculating the current control confidence coefficient using the adaptive confidence gating unit; Step S55, multiplying the control confidence coefficient, the preset feedforward matrix gain weight coefficient, the final offset step size, and the latest component in the current thickness deviation history queue to generate the feedforward dynamic correction amount; Step S56, subtracting the feedforward dynamic correction amount from the basic feedback control amount output by the original regulator to generate the target control command.

[0011] The present invention also includes the following steps: Step S6, the generated target control command is converted into the corresponding speed adjustment voltage signal and the corresponding heating power current signal through the digital-to-analog conversion channel, and then output to the screw drive frequency converter control terminal and the die head thyristor power regulator input terminal of the thermal insulation material extruder.

[0012] Step S6 of the present invention includes the following sub-steps: Step S61, calculate the deviation gain sensitivity vector of the target control command within three consecutive sampling periods, and decompose the deviation gain sensitivity vector into mutually orthogonal temperature offset operator and speed correction operator in the transformation space using an orthogonal transformation matrix; adjust the output frequency of the screw drive inverter in steps from 1Hz to 5Hz through the control terminal of the screw drive inverter to respond to the speed correction operator, and at the same time adjust the opening of the thyristor regulator of the mold head through the input terminal of the thyristor regulator of the mold head in increments from 1% to 5% to respond to the temperature offset operator.

[0013] The present invention also includes the following steps: Step S7, real-time monitoring of the sign state of the second sequence; when the components of the second sequence maintain consistent positive and negative signs within 5 consecutive sampling periods, and the target control command reaches the maximum amplitude limit boundary of the control loop output, the cumulative weight of the integral term of the internal regulator is reduced to 0, and the anti-integral saturation constraint control flow is started.

[0014] The present invention also includes the following steps: Step S8, when it is detected that the components of the second sequence are all zero noise or in a signal blocking state for 50 consecutive sampling times, it is determined that the downstream thickness gauge has a physical material breakage fault, the current feedforward compensation control gain is frozen, and the output frequency of the screw drive frequency converter is maintained at the sliding average value of the 5 control cycles before the physical material breakage fault occurred.

[0015] In step S2 of the present invention, the data storage depth of the historical drive frequency queue and the historical thickness deviation queue is adjusted through the following sub-steps: Step S21, collect the real-time rotational speed value of the screw of the thermal insulation material extruder; Step S22, when the real-time rotational speed value drops to less than 50% of the normal rotational speed value, adjust the data storage depth from 500 levels to 1000 levels to expand the retrieval range of the historical time domain.

[0016] The present invention further includes the following steps before step S1: Step S0, during the system startup phase, reads the prior distribution matrix of heat conduction delay corresponding to the distribution ratio of each material group through an external interface, and injects the prior distribution matrix of heat conduction delay into the addressing operator as the initial time domain boundary constraint for defining the lag retrieval window in step S4.

[0017] Compared with the prior art, the present invention has at least the following beneficial effects: 1. In the adaptive control of the thermal insulation material extrusion molding process, the control loop uses the control command shift register array and the deviation response shift register array to maintain the data storage depth. The addressing operator locks the excitation inflection point pointer and the response inflection point pointer within the two register arrays. The logic shift addressing operator calculates the relative address unit difference based on the above pointers and determines the current one-dimensional position index offset to generate the equivalent pure time delay parameter. This method of converting the spatiotemporal transmission delay into the relative displacement of the memory address space establishes the time domain alignment capability of the control excitation and the process response in the logical storage space, avoiding the superposition fluctuation of the feedforward command and the positive feedback of the inertial link induced by the misalignment of the control timing in conventional closed-loop feedback.

[0018] 2. The adaptive confidence gating unit monitors the rate of change of the first one-dimensional time discrete sequence. When the rate of change exceeds the preset mutation threshold, the limiter truncates and locks the revision boundary of the equivalent pure time delay parameter to keep it within the step safety range. At the same time, it controls the gain weight of the confidence coefficient decay feedforward hedging term. The state recognition and behavior adjustment mechanism triggered by data mutation enables the control loop to actively intervene under sudden changes in material flow rate. It uses the existing processor in the system to decouple time-domain misalignment conflicts and avoids the periodic oscillation instability caused by phase misalignment in the adjustment loop. It provides stable convergence of the control system transient process without changing the molding mechanical structure.

[0019] 3. The control loop processes the second one-dimensional time discrete sequence through a differential filter to extract the slope of the error change at the current moment. Combined with the first-order linear calculation of the discrete state observer, the extracted trend and the control command flow are deeply integrated in the information space. The adaptive iteration of the dynamic equivalent time delay factor is jointly driven by the historical time delay forgetting factor and the time delay evolution sensitivity coefficient. This cross-signal dynamic linkage and multi-mechanism collaborative architecture improves the system's ability to track the state of nonlinear disturbances, thereby shortening the convergence time of the control loop under time-varying thermal inertial disturbances, reducing the maximum value of the steady-state residual in the stable operating cycle, and reducing the frequency of deep integral saturation of the actuator. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the discrete processing flow of the extrusion state parameters of the thermal insulation material of the present invention; Figure 2 This is a schematic diagram of the multi-dimensional verification of extrusion dynamic delay and deviation state of the present invention. Detailed Implementation

[0021] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings.

[0022] Example 1: This embodiment discloses an adaptive control method for the extrusion molding process of thermal insulation materials, including the following steps: Step S1: Synchronously collect drive frequency and thickness deviation to generate dual sequences: Use the data bus to collect screw drive frequency commands at 100ms intervals to generate the first sequence, and simultaneously use the digital-to-analog conversion channel to collect finished product thickness deviation data at 100ms intervals to generate the second sequence. Step S2, push sequence components into the history queue and maintain a depth of 500 levels: push the current components of the first sequence and the current components of the second sequence into the tail of the corresponding driving frequency history queue and thickness deviation history queue in the internal storage area, respectively, while discarding the historical components at the head of the two history queues, and keeping the data storage depth of each history queue constant at 500 levels. Step S3, Scan the historical queue to extract slope mutation terms and lock the inflection point time step: Scan the historical queue of driving frequency, extract mutation terms whose absolute value of slope change of driving frequency command exceeds 15%, and lock their time domain position as the inflection point time step of the rate of change. Step S4, Delineate the lag window to detect the largest feature item and establish the time delay parameter: Delineate the lag retrieval window in the thickness deviation history queue based on the inflection point time step of the rate of change, retrieve the feature item with the largest absolute value of the deviation rate of change in the lag retrieval window, and calculate the time step difference between the feature item and the inflection point time step of the rate of change to establish the dynamic equivalent time delay parameter. Step S5: Update the gain frequency adjustment and output control command according to the time delay parameter: Update the feedforward compensation control gain according to the dynamic equivalent time delay parameter, adjust the screw drive frequency and output the target control command.

[0023] Step S5 in this embodiment includes the following sub-steps: Step S51, collecting the driving frequency command variation rate in the driving frequency history queue through the adaptive confidence gating unit; Step S52, when the driving frequency command variation rate is within 15%, maintaining the dynamic equivalent delay parameter as the final hedging step size of the current control cycle; Step S53, when the driving frequency command variation rate exceeds 15%, truncating and limiting the dynamic equivalent delay parameter within the preset stepped safety boundary range, and simultaneously reducing the gain weight of the feedforward hedging term by 50%.

[0024] Step S5 in this embodiment further includes the following sub-steps: Step S54, calculate the current control confidence coefficient using the adaptive confidence gating unit; Step S55, multiply the control confidence coefficient, the preset feedforward matrix gain weight coefficient, the final hedging step size, and the latest component in the current thickness deviation history queue to generate the feedforward dynamic correction amount; Step S56, subtract the feedforward dynamic correction amount from the basic feedback control amount output by the original regulator to generate the target control command.

[0025] The embodiment also includes the following steps: Step S6, the generated target control command is converted into the corresponding speed adjustment voltage signal and the corresponding heating power current signal through the digital-to-analog conversion channel, and then output to the screw drive frequency converter control terminal and the die head thyristor power regulator input terminal of the thermal insulation material extruder.

[0026] Step S6 in this embodiment includes the following sub-steps: Step S61, calculate the deviation gain sensitivity vector of the target control command within three consecutive sampling periods, and decompose the deviation gain sensitivity vector into mutually orthogonal temperature offset operator and speed correction operator in the transformation space using an orthogonal transformation matrix; adjust the output frequency of the screw drive inverter in steps from 1Hz to 5Hz through the control terminal of the screw drive inverter to respond to the speed correction operator, and at the same time adjust the opening of the thyristor regulator of the mold head through the input terminal of the mold head thyristor regulator in increments from 1% to 5% to respond to the temperature offset operator.

[0027] The embodiment also includes the following steps: Step S7, real-time monitoring of the sign state of the second sequence; when the components of the second sequence maintain consistent positive and negative signs within 5 consecutive sampling periods, and the target control command reaches the maximum amplitude limit of the control loop output, the cumulative weight of the integral term of the internal regulator is reduced to 0, and the anti-integral saturation constraint control flow is started.

[0028] The embodiment also includes the following steps: Step S8, when it is detected that the components of the second sequence are all zero noise or in a signal blocking state for 50 consecutive sampling times, it is determined that the downstream thickness gauge has a physical material breakage fault, the current feedforward compensation control gain is frozen, and the output frequency of the screw drive frequency converter is maintained at the sliding average value of the 5 control cycles before the physical material breakage fault occurred.

[0029] In step S2 of this embodiment, the data storage depth of the historical drive frequency queue and the historical thickness deviation queue is adjusted through the following sub-steps: Step S21, collect the real-time rotational speed value of the screw of the thermal insulation material extruder; Step S22, when the real-time rotational speed value drops to less than 50% of the normal rotational speed value, adjust the data storage depth from 500 levels to 1000 levels to expand the retrieval range of the historical time domain.

[0030] Before step S1 in this embodiment, the following steps are also included: Step S0, during the system startup phase, the heat conduction delay prior distribution matrix corresponding to the allocation ratio of each material group is read through the external interface, and the heat conduction delay prior distribution matrix is ​​injected into the addressing operator as the initial time domain boundary constraint for defining the lag retrieval window in step S4.

[0031] Example 2: In this embodiment, during the operation of the control loop for thermal insulation material extrusion molding, the multiphase mixed material generates variations in flow resistance within the extrusion channel. When heat accumulates inside the barrel due to screw shearing or when the recycled material ratio undergoes time-varying drift, the melt viscosity fluctuates nonlinearly. This causes the material transport delay and heat conduction inertia between the extruder die outlet and the downstream thickness measuring and transmitting unit to exhibit time-varying delay characteristics. At this time, the fixed parameter control, due to time-domain misalignment in the adjustment action, results in control gain saturation or feedforward compensation phase inversion, causing periodic divergent oscillations in the control loop and leading to thickness deviations in the thermal insulation material product. Therefore, the control and adjustment system is connected to the thermal insulation material extrusion molding production line and a target process thickness reference vector is set. The industrial control computer uses a fixed sampling period of 100ms to collect the actual thickness output sequence from the thickness gauge transmitter at the process terminal and uses an internal differential subtractor to calculate the current sampling time. Thickness deviation time series The current spindle speed feedback characteristics are read synchronously through the inverter bus communication interface. The differential filter is invoked to extract the slope of the error change at the current moment using differential subtraction. Error variation slope The relation is satisfied as follows ,in, The slope of the error variation. The current sampling time Thickness deviation, This represents the thickness deviation at the previous sampling time. The sampling period is 0.1s, and the system will calculate the slope of the error variation. Absolute value conversion, along with spindle speed feedback characteristics The common inputs are fed into a preset discrete state observer, and the current sampling time is calculated through first-order linear evolution. Dynamic equivalent time delay factor Dynamic equivalent time delay factor in discrete state observer The update logic satisfies the following relation: ,in, The current sampling time The dynamic equivalent time delay factor, This represents the dynamic equivalent time delay factor at the previous sampling time. The dimensionless historical delay forgetting factor has a value of 0.85. The time delay evolution sensitivity coefficient has dimensionally normalized properties, and its value is... Its physical dimensions are This algorithm linearly maps the geometric-dynamic ratio of the thickness deviation variation rate to the spindle rotation angular velocity to an equivalent time dimension, ensuring complete closure of all terms in the formula at the physical dimension level. To guarantee the numerical stability of the microprocessor in the full lifecycle control calculations and eliminate division-by-zero anomalies, the algorithm incorporates a dynamic lower limit clamping mechanism for the spindle speed and a working condition diversion mechanism at the discrete state observer input: during normal extrusion control of the system, the spindle speed feedback characteristics... Constrained within the rated physical operating range of the extruder and satisfying ,in The minimum critical operating reference speed for maintaining melt shearing and plasticizing by the screw is set to . When the equipment is in a stopped / reduced speed or initially not started speed state, i.e., when detected... At that time, the control system automatically locks the division update channel of the discrete state observer and maintains the time delay value of the previous cycle of the output. Alternatively, the initial pure time delay parameter preset at startup can be directly called. This avoids the risk of numerical overflow due to division by zero in the denominator at the control hardware execution and underlying algorithm levels. In actual operation, the quantized values ​​of the above discrete control parameters and safety boundaries are calibrated through discrete homogeneous Lyapunov stability functional analysis of the closed-loop control loop and actual machine convergence stress testing. Specifically, a historical delay forgetting factor of 0.85 is selected to ensure that the system eliminates sudden random noise interference while maintaining a fast tracking response to time-varying pure delay parameters. If it is lower than 0.75, the observer will become sensitive to high-frequency parasitic glitches, causing parameter drift. If it is higher than 0.95, it will cause filter phase shift. The delay evolution is sensitive. The degree coefficient is calibrated to 1.25, which is used to match the unit conversion factor of the thickness deviation and speed ratio to ensure that when the dynamic equivalent time delay factor evolves in a first-order linear manner, its step size does not exceed the stable circle boundary of the closed-loop characteristic root of the system. The safety threshold of the spindle speed variation rate is set at 15% because actual measurements show that when the speed variation exceeds 15%, the shear thermal distortion of the extrusion channel will trigger nonlinear viscosity collapse. At this time, the time delay change rate shows an unbounded divergent trend. Therefore, the safety boundary is used to cut off the dynamic equivalent time delay factor and rigidly lock it at the physical extreme point of 1.40s, thereby forcibly maintaining the polarity self-consistency of the feedforward phase inside the transition period control loop.

[0032] The spindle speed feedback characteristics are calculated in real time using an adaptive confidence gating unit. The dynamic rate of change, when the spindle speed feedback characteristic When the dynamic change rate is within 15% of the preset safety threshold, the current dynamic equivalent delay factor is maintained. Output, when the spindle speed feedback characteristic When the dynamic change rate exceeds the preset safety threshold range of 15%, the system determines that it has entered a transitional operating condition and uses a limiter to adjust the dynamic equivalent delay factor. The cutoff is limited to a stable boundary range of a fixed value of 1.40s. Simultaneously, the gain weight of the feedforward offset term is reduced by 50% through a confidence-gated unit to mitigate the phase misalignment caused by large hysteresis feedback. Correspondingly, the feedforward compensation matrix is ​​based on the dynamic equivalent time delay factor. and thickness deviation time series The basic feedback control quantity output by the original controller is corrected, and the target control command is calculated. Target control commands The operational logic satisfies the following relation: ,in, For target control commands, Based on the feedback control variable To control the confidence coefficient, when the spindle speed changes abruptly by more than 15%, the coefficient is changed from the normal value of 0.88 to 0.45. The gain weighting coefficient of the feedforward matrix has a value of 2.5, and its physical dimensions are defined as follows: To enable feedforward dynamic correction Normalized to the basic feedback control quantity The dimensionless control increment with consistent dimensions is further calculated by the system for the target control command over three consecutive sampling periods. The deviation gain sensitivity vector is decoupled and decomposed into mutually orthogonal temperature offset operators and velocity correction operators in the transformation space using an orthogonal transformation matrix. Specifically, the orthogonal transformation matrix adopts a 2x2 normalized orthogonal matrix structure, with its basis vectors mapped to the velocity adjustment axis and temperature adjustment axis, respectively. In the specific floating-point operation, the microprocessor takes a 2-dimensional row vector consisting of the target control command change and thickness deviation change over three consecutive sampling periods as input and performs matrix multiplication with the 2x2 orthogonal transformation matrix. The main diagonal elements of the orthogonal transformation matrix represent the same-direction coupling. The combination coefficient, with anti-diagonal elements representing orthogonal decoupling coefficients, is calculated using a preset fixed-point sine and cosine mapping function. This causes the deviation gain sensitivity vector to undergo orthogonal rotation within the geometric transformation space. The rotated horizontal axis output is directly defined as the speed correction operator, used to isolate speed delay disturbances caused by temperature. The rotated vertical axis output is directly defined as the temperature offset operator, used to counteract surface shear heat accumulation caused by speed fluctuations. This transforms the multivariable strong coupling relationship into two independent single-input single-output channels. The control unit drives the inverter control terminal corresponding to the screw drive unit according to the speed correction operator. The speed is adjusted in 1Hz to 5Hz increments, while the heating power opening is adjusted in 1% to 5% increments based on the thyristor power regulator input of the temperature offset operator. To overcome the asymmetric physical characteristics of the slow-changing thermodynamic system and the fast-changing mechanical dynamic system, which differ by several orders of magnitude in time scale, this invention introduces a time-scale smoothing bridging mechanism in the specific control timing. The speed correction operator directly acts on the high-frequency refreshed inverter control cycle, and its control period is maintained at 100ms to provide transient dynamic tracking efficiency; while the temperature offset... Before the operator is output to the thyristor power regulator, it undergoes time-scale stretching through a first-order inertial low-pass digital filter inside the microprocessor. Specifically, the time constant of the filter is set to 45s to match the macroscopic thermal conduction physical hysteresis of the mold head metal body. This allows the transient step energy of the temperature counteracting operator to be converted into a slowly released continuous heat flow increment. Under this filtering mechanism, the thyristor power regulator receives only a small, gradual adjustment amount within a 100ms sampling period, avoiding overshoot of the mold head thermal conduction caused by high-frequency power regulation. Thus, physical self-consistency and stable convergence of cross-scale time characteristics are achieved at the actuator execution level.

[0033] During the continuous operation of the control loop, the monitoring unit monitors the thickness deviation time series in real time. The positive or negative sign state of the thickness deviation time series within 5 consecutive sampling periods. Maintain consistency between positive and negative signs and target control commands When the maximum amplitude limit of the control loop output is reached (100%), the system automatically reduces the cumulative weight of the integral term of the internal regulator to 0 to operate an anti-integral saturation constraint control flow. If the monitoring unit detects a thickness deviation time series... If the data from 50 consecutive sampling times are all zero noise or in a signal blocking state, the current feedforward compensation control gain is automatically frozen and the output frequency of the screw drive inverter is maintained at the sliding average of the 5 control cycles before the abnormal state occurs. Under this operating mechanism, the control loop converts the spatial propagation delay into discrete-time term iterations in the logic control space, aligning the control excitation and feedback response in the storage space. At this time, the closed-loop control loop maintains time-domain synchronization of the control phase and inherent hysteresis under the interference of multiphase material rheological fluctuations. The parameter convergence time of the control loop is within 35s, the maximum dynamic overshoot is controlled below 4.2%, and the maximum steady-state residual of the overall operating cycle is not greater than 0.01mm. This eliminates the oscillation condition of the regulation gain and residual error superimposed in the same direction caused by the timing misalignment of traditional feedback regulation. The control system enters a stable convergence and steady-state maintenance state under the interference of temperature drift and viscosity distortion.

[0034] Example 3: In this embodiment, when the system faces the adaptive control closed-loop verification test condition of the thermal insulation material, the verification test platform is built with a physical hardware closed-loop circuit. It includes a microprocessor with floating-point arithmetic capabilities and a clock frequency of no less than 120MHz, and a laser thickness gauge with a measurement resolution of 0.001mm, a sampling frequency of 50Hz, and the ability to output quantified actual thickness values. To simulate the measurement system state under an imperfect electromagnetic environment on the production line, Gaussian white noise with a signal-to-noise ratio of 20dB and power frequency interference harmonics at a frequency of 50Hz are superimposed on the acquired actual thickness values. A sampling period is set. The 100ms process satisfies the following quantization control mechanism. The main factors affecting the discrete sampling step size include the upper limit of the cutoff frequency of the thickness measurement transmitter signal and the overhead of the microprocessor's single instruction cycle. The system needs to maintain a balance between the real-time performance of the control demodulation and reducing the data throughput dissipation of the controller. Specifically, since the thickness measurement transmitter output waveform is accompanied by high-frequency noise components, to prevent control polarity distortion caused by Nyquist sampling aliasing during discrete differential truncation, the sampling period... The sampling period tends towards the lower limit of the preset control frequency band period. Under this constraint, the microprocessor calculates the cutoff frequency of the input signal in real time to fix the sampling period. 100ms is used as the discrete control reference value to complete the physical anchoring of the test input.

[0035] The experimental group includes four control groups, configured as follows: an experimental group using complete control logic, a control group using a fixed-parameter proportional-integral-derivative (PI-DE) control method, and a control group where the spindle speed feedback feature is removed from the discrete-state observer evolution operator. To examine the dynamic evolution trajectory of the system under different rheological time delay disturbances, the variable frequency of the main shaft drive unit was adjusted to trigger three step change amplitude gradients of the main shaft speed at 5%, 12%, and 25% in sequence, in order to discretize and simulate the intensity of sudden changes in material flow rate caused by fluctuations in the proportion of recycled insulation material. The control group with missing components of the dynamic divisor coupling term and the out-of-range control group with the safety threshold for state switching increased from 15% to 40% were used.

[0036] When the system is subjected to a low-intensity spindle speed variation disturbance of 5%, the thickness gauge collects and outputs an initial signal containing the aforementioned random noise fluctuations, which is represented by the current sampling time. Thickness deviation time series The slope of error variation extracted after processing by a differential filter. The dynamic equivalent time delay factor generated by the first-order linear iteration of the discrete state observer is 0.03 mm / s. The value is 1.12s. Since the spindle speed variation rate does not exceed the preset safety threshold of 15%, the control confidence coefficient calculated by the adaptive confidence gating unit is... Maintaining a steady-state median of 0.88, the feedforward compensation matrix is ​​based on the formula... Real-time correction of driving parameters, including the gain weight coefficients of the feedforward matrix. With a fixed setting of 2.5, due to time delay catching-up phase alignment, the convergence time of the experimental group was limited to 33.5s, and the maximum dynamic overshoot was less than 4.1%. When the amplitude of the spindle speed step change increased to a high-intensity disturbance condition of 25%, the convergence time of the control group deteriorated to 182.4s due to the phase misalignment of the control command, resulting in the superposition of positive feedback between the action gain and residual error. Furthermore, the maximum dynamic overshoot increased sharply to 24.6%. The component-deficient control group, because it lacked the real-time normalization constraint of spindle speed on the time delay evolution sensitivity, exhibited a lower dynamic equivalent time delay factor during the transition process. Unbounded divergence occurred, extending the convergence time to 126.8 s. Meanwhile, the out-of-range control group experienced safety gating protection failure due to an excessively wide safety threshold, resulting in a dynamic equivalent time delay factor. As the error accumulates and monotonically surges to the saturation inflection point of 2.15s in the overload region, it deviates from the actual time domain window, resulting in deep saturation of the feedback gain and deep phase deflection. The convergence time of the control loop reaches 165.2s and the steady-state residual reaches 0.076mm. Correspondingly, under the high-intensity spindle speed variation condition of 25% in the test group, the adaptive confidence gating unit is triggered by the sudden speed change condition to activate the boundary limit, forcibly increasing the dynamic equivalent time delay factor. Limiting the time to the 1.40s safety step boundary and simultaneously adjusting the control confidence coefficient. The value was automatically changed to 0.45 to reduce the weight of the feedforward offset term and smoothly fade out the compensation gain. The final measured convergence time of the experimental control loop remained at 34.6s, the maximum dynamic overshoot was limited to 4.1%, and the steady-state residual of the operating cycle was no more than 0.01mm. Thus, through causal verification of non-ideal physical parameters, it was confirmed that the 15% safety threshold and the 1.40s step boundary constitute the optimal working window of the adaptive control loop under strong disturbance and large time delay conditions. This achieved the goal of using discrete control methods to suppress pure time delay divergent oscillations and enable the closed-loop system to converge quickly.

[0037] Example 4: This embodiment combines Figures 1 to 2 An adaptive control method for the extrusion molding process of a thermal insulation material is described, such as... Figure 1 As shown, step S1 synchronously collects the driving frequency and thickness deviation to generate a dual sequence. Then, step S2 pushes the sequence components into the historical queue and maintains a depth of 500 levels. Step S3 scans the historical queue to extract the slope mutation term and lock the inflection point time step. Then, step S4 delineates the lag window to detect the maximum feature term and establishes the time delay parameter. Finally, step S5 updates the gain frequency according to the time delay parameter and outputs the control command.

[0038] like Figure 2 As shown, the adaptive state verification of extrusion molding includes verification in two dimensions: the spindle speed feedback characteristic variation rate and the thickness deviation time series state. The spindle speed feedback characteristic variation rate is divided into two cases: the variation rate is within 15% and the variation rate is greater than 15%. When the variation rate is within 15%, the dynamic equivalent time delay factor output is maintained. When the variation rate is greater than 15%, the time delay factor is truncated and limited to the stable boundary. The thickness deviation time series state is divided into two cases: five consecutive cycles with the same sign and reaching the boundary and five consecutive times with zero value or blockage. When five consecutive cycles with the same sign and reaching the boundary, the integral term weight is reduced to 0. When five consecutive times with zero value or blockage, the downstream thickness gauge is judged to have a physical material breakage fault.

[0039] Example 5: In this embodiment, during material batch switching adjustments, the rheological resistance and flow rate of the extrusion channel undergo a sudden change, causing the heating transfer delay to overlap with the material conveying delay, thus increasing the dynamic equivalent time delay factor. The tracking trajectory exhibits divergent fluctuations during the state transition period because the flow characteristics of the multiphase compound fluctuate drastically, and the slope of the historical tracking deviation change in the previous operating cycle undergoes a non-monotonic jump. This is due to the historical delay forgetting factor in the discrete state observer. and time delay evolution sensitivity coefficient Maintaining a fixed value can easily lead to gain saturation in the observation loop or phase discontinuity in the feedforward compensation during the time delay and timing adjustment phase, resulting in changes to the target control command. The parameter adjustments deviated from the design boundary, resulting in an out-of-tolerance state of the thickness dimension of the insulation material products at the end of the process.

[0040] For the aforementioned transitional operating conditions, the control and regulation system utilizes a microprocessor to calculate the dynamic equivalent time delay factor. During the process, a time delay transition damping ratio parameter is introduced. A constraint relationship is established by multiplying the material conveying status feedback and the time-series buffer deviation data, which is used to adjust the historical time delay forgetting factor in real time. and time delay evolution sensitivity coefficient The stepping trajectory, in specific operation, involves the microprocessor reading the thickness deviation time series within five consecutive sampling periods. The thickness deviation time series was calculated. The variance of the discrete tracking residual is then compared with the current moment's variable frequency spindle speed feedback characteristics. The microprocessor performs a product operation to generate a dimensionless time-delay transition damping ratio parameter. Specifically, to eliminate the dimensional mismatch caused by multiplying square millimeters by revolutions per minute, the microprocessor introduces a dimension normalization conversion interface during the product operation. Before performing the product operation, the microprocessor retrieves a reference physical dimension factor preset in the read-only memory. The value of the reference physical dimension factor is fixed as the reciprocal of the product of the square of the maximum allowable thickness deviation specified by the process and the rated maximum spindle speed. Its corresponding physical dimension is defined as revolutions per minute per square millimeter. The microprocessor uses an internal digital multiplier to calculate the discrete tracking residual... The variance, the frequency conversion spindle speed feedback characteristics, and the reference physical dimension factor are multiplied together. Through the reverse elimination effect of the reference physical dimension factor, the characteristic product of macroscopic mechanical quantities and microscopic geometric quantities is normalized to the same digital space level, ultimately generating a dimensionless constant with a purely digital dimension. This establishes a completely self-consistent causal logic mapping path in discrete control logic. In an example of discontinuous material collapse and flow velocity variation, the microprocessor collects thickness deviation values ​​for five consecutive cycles. After discrete calculation, the residual variance value is determined to be 0.025 square millimeters. Simultaneously, the spindle speed feedback characteristics are read from the bus interface. The speed is 1450 r / min. After processing by the internal floating-point unit, the output time delay transition damping ratio parameter at the current sampling time is 0.36. Under this parameter constraint, the microprocessor automatically adjusts the weights of the time-domain filter and incorporates the historical time delay forgetting factor. The initial normal value of 0.85 was adjusted in steps to 0.62, and the time delay evolution sensitivity coefficient was adjusted simultaneously. The initial normal value of 1.25 is gradually adjusted to 1.45, thereby increasing the weight of the current slope in the time delay adaptive update operator and narrowing the error phase blind zone of time delay state tracking. During the actual automatic weight adjustment process, the microprocessor performs piecewise logic mapping based on the calculated range of the time delay transition damping ratio parameter. When the time delay transition damping ratio parameter is in the normal range of 0 to 0.25, the microprocessor maintains the default time-domain filter parameters and does not intervene in the weighting. When the time delay transition damping ratio parameter monotonically increases and crosses the first trigger threshold of 0.25, it indicates that the material flow rate anomaly is beginning to intensify. The processor initiates a linear interpolation decay program, decreasing the historical delay forgetting factor from its initial normal value of 0.85 in steps of 0.05 per second, while simultaneously increasing the delay evolution sensitivity coefficient from its initial normal value of 1.25 in steps of 0.04 per second. If the damping ratio parameter continues to rise to exceed the second safety threshold of 0.5, hardware-level protection is triggered. The microprocessor forcibly locks the historical delay forgetting factor at the limit of 0.5 and keeps the delay evolution sensitivity coefficient constant at 1.5, thereby breaking the positive feedback superposition chain of historical outdated delay data and achieving rigid constraint on the transition period state vector trajectory.

[0041] As the time delay transition damping ratio parameter calibration program runs, the state vector inside the discrete state observer undergoes a smooth transition, resulting in a dynamic equivalent time delay factor. The feedforward compensation control gain output converges to within 1.38s within 12.5s after the transient condition occurs, ensuring synchronization between the feedforward compensation control gain output and the macroscopic transport inertia of the physical fluid in terms of storage address timing. The feedforward compensation matrix is ​​based on the corrected dynamic equivalent delay factor. and the thickness deviation time series at the current sampling time Adjusting the gain, the final output target control command It effectively offsets the heterogeneous disturbances caused by fluctuations in material rheological properties, avoids divergent oscillation conditions where adjustment gain and residual error are superimposed in the same direction. Under the above-mentioned parameter linkage adjustment mode, the system maintains parameter steady state under the combined disturbances of shear heat accumulation and recycled material fluctuations. The convergence time of the control closed loop is within 35s, the maximum dynamic overshoot is controlled below 4.2%, and the maximum steady-state residual value of the overall stable operation cycle is not greater than 0.01mm. Finally, the extrusion molding process is restored to a balanced operating state within the process specification range.

[0042] Example 6: In this embodiment, when the system faces the conditions of no-load start-up and initial state establishment, there is a spatial physical span between the die head outlet and the detection end, resulting in an initial state measurement blind zone. This causes the adaptive control loop to oscillate in control gain during the start-up phase due to initial value mismatch. To address this, the system retrieves the physical data of the flow channel before the screw starts up. Using the ranging component, the span distance between the die head outlet and the laser thickness gauge is measured to be 3.5m, and the cross-sectional area of ​​the flow channel is obtained as 0.008 square meters. This is combined with the feedback characteristics of the drive motor at the spindle speed. The volumetric flow rate corresponding to 1450 r / min is 0.0022 cubic meters per second. This is used to calculate the material flow delay baseline value and serve as the initial state parameter for the discrete state observer. The initial pure time delay satisfies the following formula: ,in, This represents the initial pure time delay. For span distance, The cross-sectional area of ​​the flow channel. This refers to volumetric flow rate.

[0043] After the microprocessor writes the initial pure time delay values ​​online into the initial state register, the discrete state observer monotonically reduces the parameter identification divergence rate to less than 0.5% during the initial feeding transition period, and the target control instructions... The feedback gain remains within 5% of the rated control threshold during the period when the material in the shaping section is not yet fully filled at the thickness measurement point. This eliminates the start-up divergence condition caused by unknown initial values, allowing the adjustment loop to directly enter a stable tracking and parameter convergence state when the material reaches the detection terminal. In actual operation, to further ensure the time-domain boundary convergence of the system during the startup phase, the addressing operator inside the microprocessor acts as a memory address mapping pointer controller. After reading the injected heat conduction delay prior distribution matrix, the operator extracts the set of time delay discrete indicators corresponding to the material grouping ratio code set at the current startup. The addressing operator divides the discrete scalar value of the time delay by a fixed sampling period of 100ms to obtain the initial memory address offset step. The addressing operator takes the reference memory address corresponding to the current inflection point of the rate of change in the internal memory as the origin, subtracts 15 steps from the initial memory address offset step to obtain the left boundary pointer address of the search window, and adds 15 steps to the initial memory address offset step to obtain the right boundary pointer address of the search window. In this way, the abstract matrix boundary constraint is transformed into a specific physical addressing step range on the hardware memory bus, realizing the deterministic logical division of the initial time domain boundary of the lag search window.

Claims

1. An adaptive control method for the extrusion molding process of thermal insulation materials, characterized in that, Includes the following steps: Step S1: Use the data bus to collect screw drive frequency commands at 100ms intervals to generate the first sequence, and simultaneously use the digital-to-analog conversion channel to collect finished product thickness deviation data at 100ms intervals to generate the second sequence. Step S2: Push the current component of the first sequence and the current component of the second sequence into the tail of the corresponding driving frequency history queue and thickness deviation history queue in the internal storage area, respectively, and discard the historical components at the head of the two history queues, keeping the data storage depth of each history queue constant at 500 levels. Step S3: Scan the historical queue of driving frequency, extract the mutation terms whose absolute value of the slope change of driving frequency command exceeds 15%, and lock their time domain position as the inflection point of the rate of change. Step S4: Based on the inflection point time step of the rate of change, define a lag retrieval window in the thickness deviation history queue, retrieve the feature item with the largest absolute value of the deviation rate of change within the lag retrieval window, and calculate the time step difference between the feature item and the inflection point time step of the rate of change to establish the dynamic equivalent time delay parameter. Step S5: Update the feedforward compensation control gain based on the dynamic equivalent time delay parameter, adjust the screw drive frequency, and output the target control command.

2. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 1, characterized in that, Step S5 includes the following sub-steps: Step S51, the driving frequency command variation rate in the driving frequency history queue is collected through the adaptive confidence gating unit; Step S52, when the driving frequency command variation rate is within 15%, the dynamic equivalent delay parameter is maintained as the final hedging step size of the current control cycle. Step S53: When the driving frequency command variation rate exceeds 15%, the dynamic equivalent delay parameter is truncated and limited to the preset stepped safety boundary range, and the gain weight of the feedforward offset term is reduced by 50%.

3. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 2, characterized in that, Step S5 further includes the following sub-steps: Step S54, calculate the current control confidence coefficient using the adaptive confidence gating unit; Step S55, multiply the control confidence coefficient, the preset feedforward matrix gain weight coefficient, the final hedging step size, and the latest component in the current thickness deviation history queue to generate the feedforward dynamic correction amount; Step S56, subtract the feedforward dynamic correction amount from the basic feedback control amount output by the original regulator to generate the target control command.

4. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 3, characterized in that, It also includes the following steps: Step S6: The generated target control command is converted into the corresponding speed adjustment voltage signal and the corresponding heating power current signal through the digital-to-analog conversion channel, and then output to the screw drive frequency converter control terminal and the die head thyristor power regulator input terminal of the thermal insulation material extruder, respectively.

5. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 4, characterized in that, Step S6 includes the following sub-steps: Step S61, calculate the deviation gain sensitivity vector of the target control command within three consecutive sampling periods, and decompose the deviation gain sensitivity vector into mutually orthogonal temperature offset operator and speed correction operator in the transformation space using an orthogonal transformation matrix; adjust the output frequency of the screw drive inverter in steps from 1Hz to 5Hz through the control terminal of the screw drive inverter to respond to the speed correction operator, and at the same time adjust the opening of the thyristor regulator of the mold head through the input terminal of the thyristor regulator of the mold head in increments from 1% to 5% to respond to the temperature offset operator.

6. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 3, characterized in that, It also includes the following steps: Step S7: Monitor the sign state of the second sequence in real time; when the components of the second sequence maintain the same positive and negative signs for 5 consecutive sampling periods, and the target control command reaches the maximum amplitude limit of the control loop output, reduce the cumulative weight of the integral term of the internal regulator to 0 and start the anti-integral saturation constraint control flow.

7. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 4, characterized in that, It also includes the following steps: Step S8: When it is detected that the components of the second sequence are all zero noise or in a signal blocking state for 50 consecutive sampling times, it is determined that the downstream thickness gauge has a physical material breakage fault, the current feedforward compensation control gain is frozen, and the output frequency of the screw drive frequency converter is maintained at the sliding average value of the 5 control cycles before the physical material breakage fault occurred.

8. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 1, characterized in that, In step S2, the data storage depth of the drive frequency history queue and the thickness deviation history queue is adjusted through the following sub-steps: Step S21, collect the real-time rotational speed value of the screw of the thermal insulation material extruder; Step S22, when the real-time rotational speed value drops to less than 50% of the normal rotational speed value, adjust the data storage depth from 500 levels to 1000 levels to expand the retrieval range of the historical time domain.

9. The adaptive control method for the extrusion molding process of thermal insulation material according to claim 1, characterized in that, Before step S1, the system includes the following steps: In step S0, during the system startup phase, the heat conduction delay prior distribution matrix corresponding to the allocation ratio of each material group is read through the external interface, and the heat conduction delay prior distribution matrix is ​​injected into the addressing operator as the initial time domain boundary constraint for defining the lag retrieval window in step S4.

Images