A method for precise pressure regulation in injection molding holding process based on iterative learning control

Abstract

This invention relates to the field of industrial control technology, and more specifically, to a method for precise pressure adjustment in the injection molding holding process based on iterative learning control. The method includes: acquiring a reference holding pressure curve, an actual feedback pressure sequence, and injection molding holding control commands for the current holding stage; calculating thermal hysteresis characteristics characterizing changes in physical state; calculating a thermal drift evolution index characterizing the thermal drift rate; generating a nonlinear adaptive gain adjustment factor constrained within a safe range; and calculating and updating the injection molding holding control commands for the next holding stage. This invention more accurately reflects the comprehensive impact of changes in physical state, scientifically assesses the thermal drift rate and the difference between current and historical data, dynamically adapts to thermal drift to ensure control stability under different thermal conditions, effectively solves the problems of holding pressure overshoot, oscillation, and slow response, improves the dimensional consistency of injection molded products and the control accuracy of the holding process, and ensures production quality and efficiency.

Description

Technical Field

[0001] This invention relates to the field of industrial control technology. More specifically, this invention relates to a method for precise pressure regulation in the injection molding holding process based on iterative learning control. Background Technology

[0002] In the injection molding process of specialty engineering plastics and new materials, the holding pressure stage is a critical step that determines the dimensional stability, density, and internal stress distribution of the part. To achieve precise pressure adjustment during the holding pressure process, iterative learning control algorithms are often introduced in the industry. These algorithms utilize the highly repetitive cyclical nature of the injection molding process, recording the pressure tracking error of the holding pressure stage in the current injection cycle to correct the injection pressure control command for the holding pressure stage in the next injection cycle, thereby continuously approaching the ideal holding pressure curve.

[0003] Traditional iterative learning control algorithms typically use a fixed learning gain to update and iterate injection molding pressure holding control commands. In practical applications of injection molding pressure holding control, the control system uses a fixed learning gain to proportionally amplify and compensate for pressure errors.

[0004] However, in actual continuous batch production, traditional iterative learning control algorithms suffer from slow thermal drift due to the rising hydraulic oil temperature or servo system heat during continuous operation of the injection molding machine. This causes changes in the damping characteristics and bulk modulus of the hydraulic fluid as the system transitions from a cold to a hot state. Traditional iterative learning control algorithms use a fixed learning gain, which cannot adapt to the time-varying dynamic physical parameters of the controlled system caused by thermal drift. As a result, the fixed learning gain converges quickly and provides precise control in the initial cold state of production, but after the equipment heats up significantly, the fixed learning gain no longer matches the softened physical response characteristics of the current system. This leads to problems such as pressure overshoot, oscillation, or slow response, ultimately affecting the dimensional consistency of the injection molded products. Summary of the Invention

[0005] To address the problem that a fixed learning gain cannot adapt to the time-varying dynamic physical parameters of the controlled system caused by thermal drift, leading to overshoot and oscillation in the holding pressure and ultimately affecting the dimensional consistency of injection-molded products, this invention proposes a method for precise pressure adjustment in the injection molding holding process based on iterative learning control. This method includes the following steps:

[0006] The holding pressure stage within the current injection cycle of the injection molding machine is designated as the current holding pressure stage. The reference holding pressure curve, actual feedback pressure sequence, and injection holding pressure control command for the current holding pressure stage are obtained. Based on the cumulative pressure error between the reference holding pressure curve and the actual feedback pressure sequence, and the real-time rate of change of the actual pressure, the thermal hysteresis characteristic representing the change in physical state is calculated. Multiple historical thermal hysteresis characteristics within the historical cycle observation window are obtained, and based on the difference between the thermal hysteresis characteristic of the current holding pressure stage and the mean of multiple historical thermal hysteresis characteristics, the thermal drift evolution index representing the thermal drift rate is calculated. Based on the thermal hysteresis characteristic and thermal drift evolution index of the current holding pressure stage, an adaptive gain adjustment factor that is nonlinear and constrained within the safe range is generated by introducing an exponential decay operation with a safety lower limit. The adaptive gain adjustment factor is applied to the update law of the iterative learning control algorithm, and combined with the injection holding pressure control command of the current holding pressure stage, the injection holding pressure control command for the next holding pressure stage is calculated and updated.

[0007] This invention achieves the assessment of physical state changes by calculating thermal hysteresis characteristics, accurately reflecting the combined impact of accumulated pressure error and real-time pressure change rate, providing a reliable physical state characteristic basis for subsequent thermal drift evolution index calculation. By calculating the thermal drift evolution index, it achieves the analysis of thermal drift rate, accurately reflecting the difference between current thermal hysteresis characteristics and historical averages, and the generation of the gain adjustment factor provides a reliable basis for thermal drift assessment. By generating an adaptive gain adjustment factor, it improves upon the traditional iterative learning control algorithm, dynamically adjusting according to thermal hysteresis characteristics and the thermal drift evolution index, effectively adapting to the time-varying dynamic physical parameters caused by system thermal drift, and ensuring the stability of control performance under different thermal conditions. The control command update based on the adaptive gain adjustment factor effectively solves the problems of pressure overshoot, oscillation, or slow response during pressure holding, improving the dimensional consistency of injection molded products and the control accuracy of the pressure holding process.

[0008] Furthermore, the thermal hysteresis characteristics characterizing changes in physical state are calculated, including:

[0009] In the formula, For the first Thermal hysteresis characteristics during the pressure holding stage This is the total duration of the pressure holding phase. For the first During each pressure holding stage Reference holding pressure at all times For the first During each pressure holding stage Real-time feedback pressure This represents the real-time rate of change of actual pressure. This is a characteristic time constant used to offset dimensions and adjust sensitivity. To prevent extremely small constants with denominators of 0, It is the natural logarithm function. It is a function for maximizing the value.

[0010] This invention achieves a scientific assessment of thermal hysteresis characteristics by constructing a time integral model that includes a logarithmic term of the ratio of pressure error to pressure change rate. It more accurately reflects the combined influence of pressure tracking error and dynamic pressure change. The logarithmic function performs nonlinear normalization on the relative error, thereby effectively characterizing the degree of change in the physical state of the system during the pressure holding stage.

[0011] Furthermore, the thermal drift evolution index, which characterizes the thermal drift rate, is calculated, including:

[0012] In the formula, For the first The thermal drift evolution index during the pressure holding stage For the first Thermal hysteresis characteristics during the pressure holding stage The length of the preset historical period observation window, and The first The pressure holding stage and the first Thermal hysteresis characteristics during the pressure holding stage To prevent extremely small constants with denominators of 0, It is a natural exponential function. It is a function for maximizing the value.

[0013] This invention achieves a scientific assessment of the thermal drift evolution index by constructing an exponential function that includes the ratio of the difference between the current thermal hysteresis characteristics and the historical average. This more accurately reflects the degree of deviation between the current physical state and the historical average level. The normalization of the maximum value of the denominator ensures the numerical stability of the exponent, thereby effectively assessing the evolution rate of the system's thermal drift.

[0014] Furthermore, an adaptive gain adjustment factor that is nonlinear and constrained within a safe range is generated, including:

[0015] In the formula, For the first The adaptive gain adjustment factor for each voltage holding stage. This is the preset minimum safety gain coefficient. The preset attenuation sensitivity adjustment parameters are... For the first The thermal drift evolution index during the pressure holding stage For the first Thermal hysteresis characteristics during the pressure holding stage It is a natural exponential function. It is a function for maximizing the value.

[0016] This invention achieves a scientific evaluation of the adaptive gain adjustment factor by constructing an exponential decay model that includes the product of thermal drift evolution index and thermal hysteresis characteristic. This ensures that the gain adjustment factor varies between the minimum safe gain coefficient and 1. When the thermal drift evolution index and thermal hysteresis characteristic are large, the adjustment factor decreases accordingly, thereby effectively balancing the requirements of control accuracy and system stability.

[0017] Furthermore, the injection pressure holding control command for the next pressure holding stage is calculated and updated, including: using the adaptive gain adjustment factor as the weight of the learning gain in the iterative learning control algorithm to reconstruct the iterative learning control algorithm; and calculating and updating the injection pressure holding control command for the next pressure holding stage based on the reconstructed iterative learning control algorithm.

[0018] This invention achieves dynamic reconstruction of the iterative learning control algorithm by using an adaptive gain adjustment factor as the weight of the learning gain, ensuring that the learning gain is adaptively adjusted according to the system's thermal state. This effectively solves the matching problem of fixed gain under thermal drift conditions and improves the adaptability and control accuracy of iterative learning control in continuous batch production.

[0019] Furthermore, calculating and updating the injection pressure holding control command for the next pressure holding stage also includes a boundary condition processing step: adding the current time within the current pressure holding stage in the iterative learning control algorithm to the preset advance compensation time step to obtain the advance time corresponding to the current pressure holding stage for the next pressure holding stage; in response to the advance time being greater than the total duration of the pressure holding stage, setting the pressure error in the iterative learning control algorithm to the pressure error between the reference pressure holding pressure corresponding to the end time of the current pressure holding stage and the actual feedback pressure.

[0020] Furthermore, the reference holding pressure curve and the actual feedback pressure sequence are obtained as follows: the set reference holding pressure curve is obtained through the equipment controller; the actual feedback pressure sequence of the current holding pressure stage is collected in real time through the pressure sensor installed in the mold cavity or screw end.

[0021] Furthermore, the actual feedback pressure sequence is the actual feedback pressure sequence after low-pass filtering.

[0022] Furthermore, the real-time rate of change of actual pressure is obtained by dividing the difference in actual feedback pressure between adjacent sampling points during the current pressure holding phase by the sampling period and then performing differential filtering.

[0023] Furthermore, multiple historical thermal hysteresis features within the historical periodic observation window are obtained, including: retrieving historical thermal hysteresis features from multiple consecutive historical holding stages prior to the current holding stage from system memory, and using the historical thermal hysteresis features of multiple consecutive historical holding stages as the data basis within the historical periodic observation window.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] Firstly, addressing the issue that traditional iterative learning control with fixed learning gain cannot adapt to the thermal drift during continuous operation of injection molding machines, leading to time-varying physical parameters such as hydraulic damping and bulk modulus, and inaccurate control performance, this invention calculates thermal hysteresis characteristics by combining the accumulated pressure error during the holding pressure stage with the real-time change rate of the actual pressure. This more accurately characterizes the changes in the physical state of the current system caused by thermal drift. Then, it calculates the thermal drift evolution index by combining historical periodic observation window data to assess the dynamic rate of change of thermal drift. Finally, through exponential decay calculation with a safety lower limit, it generates an adaptive gain adjustment factor constrained within a safe range. This allows the iterative learning gain to no longer remain fixed but to be optimized in real-time according to the system's thermal dynamic characteristics. In cold operation, it ensures rapid convergence and precise control. When the equipment heats up and the system response characteristics soften, it automatically adjusts the gain to match the current physical response characteristics, effectively avoiding problems such as overshooting, oscillation, or slow response during holding pressure.

[0026] Secondly, this invention integrates the adaptive gain adjustment factor into the update law of iterative learning control, so that the iterative update of the holding pressure control command always fits the dynamic response characteristics of the current system. Whether in the cold state at the beginning of production, the thermal steady state during continuous operation, or the transition stage of thermal drift, it can maintain a stable iterative convergence speed and extremely high pressure tracking accuracy, so that the actual holding pressure curve accurately approaches the ideal reference curve. Precise holding pressure control can effectively ensure the dimensional stability and density of injection molded parts, optimize the internal stress distribution of parts, and meet the stringent requirements of special engineering plastics and new materials injection molding for the holding pressure process.

[0027] Thirdly, addressing the issues of batch-to-batch holding pressure fluctuations and poor product dimensional consistency caused by traditional fixed gain, this invention eliminates the impact of equipment thermal drift on holding pressure control through adaptive gain adjustment. This ensures that the holding pressure control accuracy remains stable in each injection molding cycle during continuous multi-batch production, effectively avoiding quality defects such as out-of-tolerance dimensions, shrinkage, warping, and excessive internal stress caused by uncontrolled holding pressure. This improves the quality consistency between injection molded product batches, reduces the defect rate, reduces the scrap and loss of high-value raw materials such as special engineering plastics, and improves production efficiency. Attached Figure Description

[0028] Figure 1This is a flowchart illustrating the steps of a method for precise pressure adjustment in the injection molding holding process based on iterative learning control, according to an embodiment of the present invention.

[0029] Figure 2 This is a comparative schematic diagram of the pressure curves during the injection molding holding stage when the injection molding holding pressure control command of this invention is not used.

[0030] Figure 3 This is a schematic diagram comparing the injection molding pressure holding control command curves of the present invention. Detailed Implementation

[0031] The technical solutions in the embodiments of the present invention will be clearly and completely described below. The described embodiments are only a part of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0033] Please see Figure 1 The diagram illustrates a flowchart of a method for precise pressure adjustment in an injection molding holding process based on iterative learning control, according to an embodiment of the present invention. The method includes the following steps:

[0034] S001: Record the holding pressure stage of the injection molding machine in the current injection cycle as the current holding pressure stage, and obtain the reference holding pressure curve, actual feedback pressure sequence and injection holding pressure control command of the current holding pressure stage.

[0035] Specifically, the set reference holding pressure curve is obtained through the equipment controller; at the same time, the actual feedback pressure sequence of the current holding pressure stage is collected in real time through the pressure sensor installed in the mold cavity or screw end. Considering the complex electromagnetic interference in the industrial site, the actual feedback pressure sequence is the data after low-pass filtering to eliminate high-frequency burr noise; in addition, the injection holding pressure control command issued by the controller in the current holding pressure stage is acquired and saved simultaneously.

[0036] S002: Based on the cumulative pressure error between the reference pressure curve and the actual feedback pressure sequence during the current pressure holding stage, and the real-time rate of change of the actual pressure, calculate the thermal hysteresis characteristics that characterize the change of physical state.

[0037] It should be noted that during the continuous operation of the injection molding machine, changes in the thermal state of the hydraulic system directly alter the dynamic characteristics of the pressure response during the holding pressure phase. These changes cannot be accurately and stably characterized using a single pressure error or pressure change rate indicator. Therefore, this step constructs an integral model that integrates pressure tracking error and pressure dynamic change rate to calculate the thermal hysteresis characteristics that accurately map changes in the system's physical state. This provides a core physical characteristic basis for subsequent assessment of thermal drift and adaptive gain adjustment.

[0038] Specifically, the calculation of thermal hysteresis characteristics characterizing changes in physical state includes:

[0039] ;

[0040] In the formula, For the first Thermal hysteresis characteristics during the pressure holding stage This is the total duration of the pressure holding phase. For the first During each pressure holding stage Reference holding pressure at all times For the first During each pressure holding stage Real-time feedback pressure The real-time rate of change of actual pressure is obtained by dividing the difference of actual feedback pressure at adjacent sampling points within the current pressure holding stage by the sampling period and then performing differential filtering. To compensate for the dimensionality and sensitivity adjustment, a characteristic time constant is obtained through a step response test during the factory commissioning of the injection molding machine or the introduction of a new process: the control proportional valve issues a pressure step command, and the time required for the actual pressure to rise from 10% to 90% is recorded. This time constant is used as the characteristic time constant. The baseline reference value is used, and fine-tuned based on the subsequent calculation results of the thermal hysteresis index. In this embodiment, The time constant for hydraulic pressure build-up is typically 0.5 to 1 times the hydraulic system pressure response rise time. For most small and medium-sized injection molding machines, the hydraulic pressure build-up time constant is usually between 0.5 and 2 seconds. The second coverage extends to the vast majority of operating conditions; To prevent extremely small constants with a denominator of 0, in this embodiment, The unit is the same as the real-time rate of change of the actual pressure; It is the natural logarithm function. It is a function for maximizing the value.

[0041] Specifically, when the system is heated, causing the hydraulic oil to thin and pressure transmission to slow down, the actual pressure dynamic response rate will decrease, and the deviation between the actual pressure and the reference holding pressure will increase. The increase in the independent variable deviation and the decrease in the rate of change together lead to a decrease in the dependent variable... The value increases significantly, thus accurately mapping the phenomenon of damping characteristic decay and physical softening caused by thermal drift of the system.

[0042] The above relationship is based on the dynamic response and error integral evaluation model in fluid mechanics and automatic control principles, aiming to evaluate the damping characteristic attenuation and physical state change of the controlled object under thermal changes. Macroscopic fluid mechanics and thermodynamic laws show that after continuous operation and heating of the injection molding machine, the hydraulic oil becomes thinner and the bulk modulus of elasticity decreases, leading to a decrease in system damping, slower pressure transmission, and an increase in steady-state deviation. Traditional error integrals only consider the accumulation of absolute deviation and cannot reflect the physical nature of this slowed dynamic response. Therefore, this invention follows the above natural laws and makes modifications: the pressure tracking error, which characterizes the increase in deviation, is used... As a molecule, it will characterize the real-time rate of change of actual pressure, which represents the slowing down of pressure transmission. The ratio term is constructed as the denominator; and a characteristic time constant is introduced. With the natural logarithm function After performing nonlinear normalization, the above construction logic results in a higher error rate as the system gets hotter, the slower the response, and the larger the error. The larger the value, the more accurately and objectively it characterizes the physical hindrance effect of the system in the hot state.

[0043] The dimension of pressure tracking error is The real-time rate of change of pressure has the following dimensions: Characteristic time constant The dimension of time is denominator After the dimensions are canceled out, it becomes Therefore, the ratio of the numerator to the denominator is dimensionless, the constant 1 is dimensionless, and the logarithm... The result is dimensionless, and the external integral is... The unit of measurement is time, divided by the total duration of time. The time dimensions cancel each other out, resulting in the final... It is a dimensionless, purely numerical characteristic quantity.

[0044] S003: Obtain multiple historical thermal hysteresis features within the historical periodic observation window, and calculate the thermal drift evolution index, which characterizes the thermal drift rate, based on the difference between the thermal hysteresis features of the current pressure holding stage and the mean of multiple historical thermal hysteresis features.

[0045] It should be noted that single-cycle thermal hysteresis characteristics only reflect the physical state of the system in the current cycle. They cannot distinguish whether characteristic fluctuations are caused by random disturbances in the production process or by the systematic thermal drift evolution of the equipment, nor can they assess the dynamic rate of change and development trend of the thermal state. Therefore, this step introduces a historical cycle observation window to compare the statistical differences between the current thermal hysteresis characteristics and historical characteristics, calculating a quantifiable thermal drift evolution rate thermal drift evolution index. This more accurately identifies the development stage of system thermal drift and provides crucial dynamic trend information for nonlinear adaptive gain adjustment.

[0046] Specifically, multiple historical thermal hysteresis features within the historical periodic observation window are obtained, including:

[0047] The system retrieves historical thermal hysteresis characteristics from multiple consecutive historical holding pressure stages preceding the current holding pressure stage from system memory. These historical thermal hysteresis characteristics from multiple consecutive historical holding pressure stages serve as the data basis within the historical cycle observation window. In this embodiment, the number of historical holding pressure stages is [number missing]. This range is necessary because sufficient historical data is needed to smooth out random fluctuations in a single cycle, such as slight differences in raw materials, sensor noise, and environmental disturbances. However, the window cannot be too long to avoid historical data, especially chiller data, dragging down the current judgment and causing the algorithm to lag in response to thermal changes.

[0048] Specifically, the thermal drift evolution index, which characterizes the thermal drift rate, is calculated, including:

[0049] ;

[0050] In the formula, For the first The thermal drift evolution index during the pressure holding stage For the first Thermal hysteresis characteristics during the pressure holding stage The length of the preset historical period observation window, and The first The pressure holding stage and the first Thermal hysteresis characteristics during the pressure holding stage To prevent extremely small constants with denominators of 0, It is a natural exponential function. It is a function for maximizing the value.

[0051] Specifically, the thermal drift rate of the system is assessed by comparing the current holding pressure stage characteristics with the historical window average. When the injection molding machine rapidly heats up from a cold state, the stagnation characteristics of the current holding pressure stage are much greater than the historical average, causing the strain to increase exponentially. When the machine reaches thermal equilibrium, the characteristic difference approaches zero, and the strain approaches zero, thus reflecting the entire physical process of thermal drift evolving from drastic change to gradual saturation.

[0052] The above relationship draws on the sliding window trend assessment and maximum value normalization concepts from time series analysis and statistics, aiming to scientifically evaluate the dynamic evolution rate of system thermal drift and the degree of deviation from historical average levels. The state evolution of thermodynamic systems exhibits strong inertia and nonlinear saturation characteristics; that is, the initial stage of operation is characterized by intense heat generation leading to abrupt parameter changes, while the later stage, with heat dissipation reaching equilibrium, causes the parameter drift rate to decay towards zero. Single-period characteristics are highly susceptible to random disturbances, making it impossible to accurately determine this systemic thermal evolution. Therefore, this invention extracts current characteristics. With historical window The mean within the range is used to evaluate the system evolution offset, and the historical maximum value is used for normalization to adapt to different injection molding processes. Then, the offset is used as the independent variable of the natural exponential function for calculation, so that the exponential term is rapidly amplified when the machine is heated from a cold state, and the exponential result is stable when the system tends to thermal equilibrium. This replicates and quantifies the physical evolution law of the system thermal drift from occurrence, development to saturation.

[0053] Both the mean difference and its historical terms are dimensionless. The numerator of the mean difference and the denominator of the maximum value are both dimensionless, and the result of their division is dimensionless. Subtracting the constant 1 also leaves it dimensionless. As the operational variable for the natural index, it conforms to mathematical rules, ultimately yielding... It is a dimensionless index.

[0054] S004: Based on the thermal hysteresis characteristics and thermal drift evolution index of the current pressure holding stage, an adaptive gain adjustment factor that is nonlinear and constrained within the safe range is generated by introducing an exponential decay calculation of the safety lower limit.

[0055] It should be noted that the learning gain of iterative learning control directly determines the correction magnitude of the control command and the convergence stability of the system. When the gain does not match the dynamic characteristics of the system, it will directly affect the control accuracy and operational stability of the holding pressure. Therefore, it is necessary to combine the current thermal state and thermal drift trend of the system to generate gain adjustment parameters that balance control accuracy and operational stability. Thus, this step, based on thermal hysteresis characteristics and thermal drift evolution exponent, generates an adaptive gain adjustment factor constrained within a safe range through exponential decay calculation with a safety lower limit, thereby achieving dynamic adaptation of the learning gain to the system's thermal state.

[0056] Specifically, generating a nonlinear adaptive gain adjustment factor constrained within a safe range includes:

[0057] ;

[0058] In the formula, For the first The adaptive gain adjustment factor for each voltage holding stage. The minimum safe gain coefficient is preset. After the injection molding machine reaches thermal equilibrium, a trial run is performed, and the lower limit of the gain is gradually reduced. It is observed whether the system exhibits an increase in steady-state error or oscillation of the pressure curve. The minimum value that ensures system stability while still possessing basic correction capability is selected. In this embodiment, This range ensures both basic correction capability under hot conditions and sufficient attenuation. The preset attenuation sensitivity adjustment parameters were observed during continuous production debugging. When rising The rate of decrease is important. If the pressure curve converges slowly and there is still overshoot in the hot state, it indicates that the gain decreases too slowly and should be reduced. If the adaptive gain adjustment factor reacts excessively to temperature fluctuations, causing the gain curve to jump drastically, it indicates that it is too sensitive and should be increased. In this embodiment, This range enables a smooth and timely response to changes in gain with thermal state, balancing dynamic tracking speed and noise immunity. For the first The thermal drift evolution index during the pressure holding stage For the first Thermal hysteresis characteristics during the pressure holding stage It is a natural exponential function. It is a function for maximizing the value.

[0059] By introducing a nonlinear function, it is ensured that the output of the adjustment factor remains smooth and strictly constrained. Within the safe range; when the thermal drift evolution index and thermal hysteresis characteristics increase synchronously, it indicates that the system has entered a severe softening thermal state, leading to strain. Nonlinear smoothing is reduced; thus, in a physical sense, a conservative adjustment strategy is implemented that weakens the intensity of subsequent control corrections as the system gets hotter and its volumetric elasticity decreases.

[0060] The above relationship is based on adaptive control theory and the exponential decay regulation model in nonlinear systems. It aims to generate dynamic gain weights that balance control accuracy and operational stability by combining the system's current thermal hysteresis state and drift trend. According to the principles of control engineering and physical system stability, when a hydraulic system experiences a decrease in volumetric elasticity due to heating, its ability to withstand strong control command corrections is significantly weakened. Maintaining a fixed gain would inevitably disrupt the energy balance and cause high-frequency oscillations. To follow this natural law, this invention abandons the traditional fixed learning gain mechanism and constructs a gain coefficient with the minimum safety gain. The exponential decay function is the lower bound, expressed as the drift rate. With the degree of obstruction Multiplication serves as the attenuation kernel, causing the exponential term to approach 0 as the system heats up more severely and deteriorates faster. This achieves, in a physical sense, a conservative adjustment strategy that adaptively and smoothly weakens the control correction strength as the system heats up and its volumetric elasticity decreases, ensuring that the gain is strictly constrained within a safe range.

[0061] Preset minimum security gain The coefficients are dimensionless, and the obtained values ​​are... and All are dimensionless, with preset attenuation sensitivity adjustment parameters. It is dimensionless; therefore, the exponential part Since it is dimensionless, the output of the natural exponent is dimensionless, and the final result is... It is also a dimensionless, purely numerical scaling factor that can be directly multiplied into the learning update law as a gain weight.

[0062] S005: Apply the adaptive gain adjustment factor to the update law of the iterative learning control algorithm, and calculate and update the injection pressure control command for the next pressure holding stage in combination with the injection pressure holding control command of the current pressure holding stage.

[0063] It should be noted that the core of the adaptive gain adjustment factor is to be implemented in the iterative learning control instruction iteration process to achieve precise optimization of the pressure holding control command. Simultaneously, it needs to consider the inherent delay characteristics of the hydraulic system and the boundary stability of the algorithm's engineering operation, avoiding practical application problems such as time index out-of-bounds errors and command overshoot. Therefore, this step integrates the adaptive gain adjustment factor into the update law of the iterative learning control algorithm, along with advance compensation and boundary condition handling, to complete the calculation and update of the injection pressure holding control command for the next pressure holding stage, ultimately achieving adaptive adjustment of the pressure during the pressure holding process.

[0064] Specifically, the adaptive gain adjustment factor is first used as the weight of the learning gain in the iterative learning control algorithm to reconstruct the iterative learning control algorithm.

[0065] During the calculation process, the current time in the current holding phase of the iterative learning control algorithm is added to the preset advance compensation time step to obtain the advance time of the next holding phase in the current holding phase.

[0066] In response to a situation where the lead time exceeds the total duration of the holding phase, to prevent system crashes due to time index out-of-bounds errors, the pressure error in the iterative learning control algorithm is set to the pressure error between the reference holding pressure corresponding to the end time of the current holding phase and the actual feedback pressure.

[0067] Through the above reconstruction and boundary processing, the injection pressure holding control command for the next pressure holding stage is finally calculated and updated, which can not only offset the inherent delay of the hydraulic system, but also avoid aggressive overshoot under hot softening conditions.

[0068] Specifically, a factor for dynamically mapping the thermal state of the device is injected into the learning law of the iterative learning control algorithm. When the equipment is in a cold state, Approaching 1, the system maintains its original strong learning gain to achieve rapid convergence; however, thermal drift occurs as production batches increase. The adaptive reduction automatically lowers the overall learning gain, preventing the controller from over-amplifying the pressure error of completed cycles when facing hydraulic systems with slowed physical responses due to high temperatures. This effectively eliminates command oversaturation and severe pressure oscillation caused by gain mismatch.

[0069] like Figure 2 As shown, the horizontal axis represents the time of the holding pressure stage in seconds; the vertical axis represents the relative holding pressure of the mold cavity in %, which is a normalized display based on the rated value of the reference holding pressure. In the first injection cycle during the initial production phase, i.e., the initial state of the cold machine, the actual pressure curve shows that the equipment has not experienced thermal drift, the system dynamic characteristics match the preset fixed gain, and the actual pressure can accurately track the reference curve. However, in the actual pressure curve after the 15th continuous production cycle, the hydraulic system has experienced significant thermal drift, and core physical parameters such as system damping and bulk modulus of elasticity have become time-varying. The fixed gain and system dynamic characteristics are severely mismatched, resulting in serious pressure tracking lag and steady-state deviation. The holding pressure deviates significantly from the reference curve, failing to meet the process requirements of special engineering plastic injection molding.

[0070] like Figure 3As shown in the figure, the horizontal axis represents the time of the pressure holding stage in seconds; the vertical axis represents the output value of the injection molding pressure holding control command. It can be seen from the figure that, as thermal drift occurs in the equipment, this invention can dynamically adjust the learning gain through real-time evaluation of thermal hysteresis characteristics and thermal drift evolution index, synchronously optimizing the output amplitude and dynamic characteristics of the control command to offset the changes in response characteristics caused by system thermal drift, thus providing a reliable command basis for accurate tracking of the pressure holding pressure throughout the entire cycle.

[0071] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for precise pressure adjustment in the injection molding holding process based on iterative learning control, characterized in that, include: The holding pressure stage of the injection molding machine in the current injection cycle is recorded as the current holding pressure stage. The reference holding pressure curve, actual feedback pressure sequence and injection holding pressure control command of the current holding pressure stage are obtained. Based on the cumulative pressure error between the reference pressure curve and the actual feedback pressure sequence during the current pressure holding stage, as well as the real-time rate of change of the actual pressure, the thermal hysteresis characteristics characterizing the change of physical state are calculated. Multiple historical thermal hysteresis features within the historical period observation window are obtained, and the thermal drift evolution index, which characterizes the thermal drift rate, is calculated based on the difference between the thermal hysteresis features of the current pressure holding stage and the mean of multiple historical thermal hysteresis features. Based on the thermal hysteresis characteristics and thermal drift evolution index of the current pressure holding stage, an adaptive gain adjustment factor that is nonlinear and constrained within the safe range is generated by introducing an exponential decay calculation of the safety lower limit. The adaptive gain adjustment factor is applied to the update law of the iterative learning control algorithm. Combined with the injection pressure holding control command of the current holding pressure stage, the injection pressure holding control command of the next holding pressure stage is calculated and updated.

2. The method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, Calculate the thermal hysteresis characteristics that characterize changes in physical state, including: ； In the formula, For the first Thermal hysteresis characteristics during the pressure holding stage This is the total duration of the pressure holding phase. For the first During each pressure holding stage Reference holding pressure at all times For the first During each pressure holding stage Real-time feedback pressure This represents the real-time rate of change of actual pressure. This is a characteristic time constant used to offset dimensions and adjust sensitivity. To prevent extremely small constants with denominators of 0, It is the natural logarithm function. It is a function for maximizing the value.

3. The method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, Calculating the thermal drift evolution index, which characterizes the thermal drift rate, includes: ； In the formula, For the first The thermal drift evolution index during the pressure holding stage For the first Thermal hysteresis characteristics during the pressure holding stage The length of the preset historical period observation window, and The first The pressure holding stage and the first Thermal hysteresis characteristics during the pressure holding stage To prevent extremely small constants with denominators of 0, It is a natural exponential function. It is a function for maximizing the value.

4. The method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, Generate a nonlinear adaptive gain adjustment factor constrained within a safe range, including: ； In the formula, For the first The adaptive gain adjustment factor for each voltage holding stage This is the preset minimum safety gain coefficient. The preset attenuation sensitivity adjustment parameters are... For the first The thermal drift evolution index during the pressure holding stage For the first Thermal hysteresis characteristics during the pressure holding stage It is a natural exponential function. It is a function for maximizing the value.

5. The method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, Calculate and update the injection pressure holding control instructions for the next pressure holding stage, including: The adaptive gain adjustment factor is used as the weight of the learning gain in the iterative learning control algorithm to reconstruct the iterative learning control algorithm. Based on the reconstructed iterative learning control algorithm, the injection pressure holding control command for the next pressure holding stage is calculated and updated.

6. The method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, The calculation and updating of the injection pressure holding control instructions for the next holding stage also includes boundary condition handling steps: Add the current time in the current holding phase of the iterative learning control algorithm to the preset advance compensation time step to obtain the advance time of the next holding phase in the current holding phase; When the lead time exceeds the total duration of the holding phase, the pressure error in the iterative learning control algorithm is set to the pressure error between the reference holding pressure and the actual feedback pressure at the end time of the current holding phase.

7. The method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, The reference holding pressure curve and the actual feedback pressure sequence are obtained as follows: The set reference holding pressure curve is obtained through the equipment controller; The actual feedback pressure sequence during the current pressure holding stage is collected in real time by a pressure sensor installed in the mold cavity or at the end of the screw.

8. A method for precise pressure adjustment in the injection molding holding process based on iterative learning control, as described in claim 1 or 7, characterized in that, The actual feedback pressure sequence is the actual feedback pressure sequence after low-pass filtering.

9. The method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, The real-time rate of change of actual pressure is obtained by dividing the difference in actual feedback pressure between adjacent sampling points during the current pressure holding phase by the sampling period and then processing it through differential filtering.

10. A method for precise pressure adjustment in the injection molding holding process based on iterative learning control according to claim 1, characterized in that, Obtain multiple historical thermal hysteresis features within the historical periodic observation window, including: Retrieve historical thermal hysteresis features from multiple consecutive historical holding pressure stages prior to the current holding pressure stage from system memory, and use these features as the data basis within the historical cycle observation window.

Images