Intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting
By using an intelligent control system based on variable mold temperature and coordinated venting, the problems of increased flow resistance and poor venting caused by rapid melt cooling in ultra-thin injection molding are solved. This achieves stability and precision control of the injection molding process, avoids defects such as flash, overflow and scorching, and ensures high-quality molding of injection molded parts.
Patent Information
- Application Number
- CN202611266043.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-20
- Publication Date
- 2026-11-10
AI Technical Summary
Existing ultra-thin injection molding control systems lack feedback on the actual evolution of the physical field, resulting in extremely rapid melt cooling and a sharp increase in flow resistance. Poor venting can easily lead to air blockage, entrapment, and scorching defects. The timing of the venting valve's action is difficult to control precisely; closing it too early can cause overflow, while closing it too late can cause blockage. Furthermore, the control logic is difficult to match the actual energy decay process of the system.
An intelligent control system based on variable mold temperature and coordinated venting is adopted. The transient temperature distribution matrix is constructed by the thermal field sensing module, the dynamic impedance sealing module generates an adaptive impedance change judgment threshold, the proxy prediction and interception module predicts the arrival time of the melt wavefront, the gas resistance feedback compensation module adjusts the injection speed, and the dual-mode relaxation module controls the cooling water flow, so as to achieve dynamic coordination of mold temperature, injection speed and venting process.
Ensure that the high-speed filling action achieves physical sealing in the microscopic gaps of the mold parting surface, avoids flash and overflow, prevents melt overflow and blockage of the venting structure, controls the solidification process of the injection molded part, and ensures the dimensional accuracy and structural morphology of the injection molded part.
Smart Images

Figure CN123060883A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of injection molding control technology, specifically to an intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting. Background Technology
[0002] Ultra-thin part injection molding technology is widely used in precision manufacturing. These products have extremely small cavity spaces and a very high surface area to volume ratio. When high-temperature polymer melt fills the extremely narrow cavity, the rate of heat transfer to and dissipation from the mold is extremely rapid. The entire molding cycle is short, and the melt flow resistance changes drastically. The molding process places extremely high physical requirements on the distribution of the mold temperature field, the conversion efficiency of injection kinetic energy, and the rate of gas discharge from the cavity.
[0003] Existing injection molding control processes for ultra-thin parts largely rely on fixed time or position nodes for stage switching. Equipment typically employs constant mold temperature or simple variable mold temperature logic based on time presets. Gas venting from the mold cavity mainly relies on passive venting channels or basic mechanical venting valves. The control logic generally involves the system issuing a valve-closing command to the actuator after the front-end sensor detects that the melt has reached a predetermined position. At the end of the injection molding process, after filling, the system directly switches to multi-stage stepped pressure holding according to a fixed program, and cuts off the mold heating medium based on time parameters set by manual experience. Subsequently, cooling water is introduced to complete the solidification and shaping of the product.
[0004] However, existing control systems lack feedback on the actual evolution of the physical field. High-speed filling is initiated before the local temperature reaches the threshold, and the rapid thickening of the condensation layer can easily cause cold material blockage. Forcibly increasing the injection pressure can easily cause the melt to break through the physical boundary of the mold and form flash defects. The mechanical vent valve has an inherent delay in action, and closing the valve when the material is seen often causes the melt to rush into the venting channel and cause structural blockage. If the valve is closed in advance at a fixed time, the gas trapped in the cavity is violently compressed, causing the back pressure to rise sharply and the melt flow rate to drop significantly, which can easily lead to material shortage or adiabatic compression scorching of the product. Moreover, simply relying on the time dimension to switch between holding pressure and cooling is difficult to match the actual energy decay process of the system. The continuous holding pressure shearing action in the runner causes the residual stress to be directly frozen inside the plastic part. Therefore, this invention provides an intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting to solve the shortcomings of the existing technology. Summary of the Invention
[0005] To address the shortcomings of existing technologies, the purpose of this application is to provide an intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting. This system solves the problems of extremely small cavity space, rapid melt cooling leading to a sharp increase in flow resistance, and poor venting easily causing air blockage, entrapment, and scorching defects during ultra-thin part injection molding; the difficulty in precisely controlling the timing of the venting valve's action, with premature closure resulting in incomplete venting and delayed closure causing melt overflow and blockage of the venting structure; and the difficulty of achieving dynamic coordination of the physical quantities of mold temperature, transient injection speed, and active venting process in existing control logic.
[0006] To achieve the above objectives, the present invention provides the following technical solution: an intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting, comprising:
[0007] The thermal field sensing module is used to construct a transient temperature distribution matrix. When the global minimum value of the transient temperature distribution matrix exceeds the glass transition temperature threshold, a negative pressure environment is established and initial feedforward parameters are output.
[0008] The dynamic impedance sealing module is used to generate an adaptive impedance change judgment threshold based on the initial feedforward parameters, and outputs a high-speed filling stage trigger command when the transient rheological impedance is greater than the adaptive impedance change judgment threshold.
[0009] The proxy prediction and interception module is used to respond to the high-speed filling stage trigger command to drive the preset neural network model to output the prediction of the remaining time and issue a servo exhaust valve closing command.
[0010] The air resistance feedback compensation module is used to synchronously monitor the back pressure of the gas at the front end of the servo exhaust valve, and to adjust the transient injection speed based on the partial derivative of the pressure difference between the main pressure of the mold cavity and the back pressure of the gas at the front end of the servo exhaust valve by generating a speed compensation signal.
[0011] The dual-mode relaxation module is used to introduce cooling water when the integral of the total enthalpy of the mold cavity reaches the critical enthalpy threshold, to perform negative pressure extraction when the derivative of pressure and temperature drop meets the preset threshold conditions, and to control positive pressure gas according to the dynamic gas pressure function.
[0012] Preferably, the thermal field sensing module includes:
[0013] The data acquisition unit is used to control the flow of the variable mold temperature medium into the mold conformal water channel, and to collect discrete temperature data through an array of temperature sensors distributed on the surface of the mold cavity.
[0014] The matrix construction unit is used to drive the spatial interpolation algorithm to perform continuous processing based on the discrete temperature data, and reconstruct and generate a transient temperature distribution matrix representing the global temperature gradient information of the mold cavity surface.
[0015] The readiness determination unit and the collaborative triggering unit are used to extract the effective high-temperature region area ratio based on the transient temperature distribution matrix. When it is determined that the global minimum value of the transient temperature distribution matrix exceeds the glass transition temperature threshold, the collaborative triggering unit extracts the current thermal field data to construct the initial feedforward parameters and outputs a vacuum suction command to the microporous breathable steel array to drive the establishment of a negative pressure environment.
[0016] Preferably, the dynamic impedance sealing module includes:
[0017] The impedance calculation unit is used to collect the transient injection pressure and transient injection speed of the injection mechanism. By calculating the ratio of the derivative of the transient injection pressure with respect to time to the transient injection speed, the transient rheological impedance in the generated cavity is constructed.
[0018] The threshold adaptive unit is used to receive initial feedforward parameters to extract representative interface contact temperatures, and combine the viscosity-temperature sensitivity characteristics of polymer materials with the reference impedance constant to generate an adaptive impedance change judgment threshold representing the dynamic leak-proof boundary.
[0019] The stage switching determination unit is used to perform high-frequency cross-comparison between the transient rheological impedance and the adaptive impedance change determination threshold. When it is determined that the transient rheological impedance is greater than the adaptive impedance change determination threshold and the set observation time window is maintained, it confirms that the micro gap has achieved physical sealing and outputs a high-speed filling stage trigger command.
[0020] Preferably, the proxy prediction and interception module includes:
[0021] The feature construction unit is used to respond to the high-speed filling stage trigger command, extract historical cycle data of the main pressure of the mold cavity and the injection speed, perform time axis alignment preprocessing, and construct a two-dimensional feature vector sequence matrix with time sliding window characteristics;
[0022] The timing prediction unit is used to drive a pre-built neural network model deployed on the edge computing node to perform forward nonlinear inference using the two-dimensional feature vector sequence matrix as the underlying feature input, and output the predicted remaining time for the melt wavefront to reach the servo exhaust valve.
[0023] The interception execution unit is used to construct a feedforward trigger interception boundary by combining the predicted remaining time, the mechanical action delay constant of the servo exhaust valve, and the control safety redundancy time. At the feedforward time node that meets the interception boundary, the servo exhaust valve closing command is issued.
[0024] Preferably, the air resistance feedback compensation module includes:
[0025] The data acquisition and filtering unit is used to obtain the back pressure of the gas at the front end of the servo exhaust valve and the main pressure of the mold cavity, and to clean up high-frequency aerodynamic noise through the Kalman filter algorithm to generate a smooth pressure time series with strict time alignment.
[0026] The differential pressure trend analysis unit is used to calculate the net pressure difference between the main pressure of the mold cavity and the back pressure of the gas at the front end of the servo exhaust valve based on the smoothed pressure time series, and to use the finite difference algorithm to approximately derive the partial derivative of the differential pressure as the net pressure difference decays over time.
[0027] The compensation signal generation unit is used to generate a speed compensation signal that is limited by the maximum safe physical injection speed when the differential pressure partial derivative exceeds the dead zone threshold and decays negatively, and to superimpose the speed compensation signal into the instruction to dynamically adjust the transient injection speed.
[0028] Preferably, the dual-modal relaxation module includes an enthalpy calculation unit;
[0029] The enthalpy calculation unit is used to receive the physical blockage state where the exhaust mass flow rate is zero after the servo exhaust valve is closed. Based on the continuous accumulation of the differential substitution between the transient fluid convection heat transfer rate and the work done by the microscopic compression of the gas trapped by the exhaust valve, it generates the integral of the total enthalpy of the mold cavity, which represents the energy decay state of the control system. When the integral of the total enthalpy reaches the critical enthalpy threshold, it issues a command to cut off the variable mold temperature medium and introduce cooling water.
[0030] Preferably, the dual-modal relaxation module further includes:
[0031] The dual-modal capture unit is used to synchronously execute discrete differential operations in response to the cooling water supply command. When the attenuation derivative of the cavity holding pressure is less than the preset pressure attenuation threshold and the decrease derivative of the cavity surface temperature is less than the preset temperature decrease threshold, a dual-modal trigger signal is generated to send a command to the control system to stop the positive pressure holding and execute the negative pressure withdrawal action to cut off the flow channel shear stress.
[0032] The pneumatic pressing actuator is used to respond to the dual-mode trigger signal to simultaneously cut off the vacuum suction air path of the microporous breathable steel array and introduce positive pressure gas. It extracts the ratio of the current total enthalpy to the enthalpy reference of the completely solid state to construct the solidification layer evolution index, and drives the dynamic gas pressure function to control the positive pressure gas pressure to be applied exponentially smoothly until the injection molded part is solidified and shaped.
[0033] Preferably, before driving the spatial interpolation algorithm, the matrix construction unit performs overlapping feature detection on the discrete coordinate point set to actively remove redundant physical overlapping points, and introduces a preset regularization perturbation term on the diagonal of the spatial coordinate matrix.
[0034] The readiness determination unit simultaneously compares the effective high-temperature region area ratio and the local minimum temperature after filtering and truncating abnormally low-value data with the glass transition temperature threshold and the system tolerance temperature constant through an intersection constraint comparison.
[0035] Preferably, the feature construction unit uses a normalization preprocessing algorithm that introduces a steady-state prevention and elimination of zero-redundancy constant standard deviation to standardize the data in real time, eliminate the offset of physical dimension data from heterogeneous sensors, and generate a standard tensor input to the preset neural network model.
[0036] The interception execution unit has built-in feedforward verification logic, which synchronously calculates the first-order time derivative of the current main pressure of the mold cavity. Only when the first-order time derivative is greater than zero and the system is in a continuous filling state, and the predicted remaining time meets the interception boundary, is it confirmed as a valid trigger and a servo exhaust valve closing command is issued.
[0037] Preferably, when calculating the differential pressure partial derivative, the differential pressure trend analysis unit introduces a time constraint constant into the time denominator of the differential algorithm to prevent division by zero overflow caused by the jitter of the underlying communication network clock.
[0038] The compensation signal generation unit has built-in multi-dimensional cross-validation logic. Before outputting the speed compensation signal, it verifies that the absolute net pressure difference between the main pressure of the mold cavity and the back pressure of the gas at the front end of the servo exhaust valve at the current moment is greater than the set minimum maintenance work threshold, so as to ensure that the intervention action is applied to the microscopic gas resistance stage.
[0039] In summary, this application includes at least one of the following beneficial technical effects:
[0040] 1. This invention constructs a transient temperature distribution matrix and calculates transient rheological impedance. The system only outputs a high-speed filling command when it determines that the lowest local temperature in the mold cavity exceeds the glass transition temperature and the rheological impedance is greater than the adaptive judgment threshold. This ensures that the high-speed filling action is only started after the cold material blocking boundary is eliminated and the micro gaps of the mold parting surface are confirmed to be physically sealed. This prevents flash and overflow defects in ultra-thin parts during the early stage of high-speed and high-pressure injection molding and improves the system stability in the early stage of filling.
[0041] 2. This invention introduces a servo exhaust valve control and dynamic air resistance feedback compensation mechanism based on time-series feature prediction. It uses a pre-set neural network model to predict the arrival time of the melt wavefront and combines the mechanical action delay constant to issue the closing command in advance, avoiding melt overflow and blockage of the exhaust structure caused by the exhaust valve closing too late. The system extracts the partial derivative of the pressure difference between the main pressure of the mold cavity and the back pressure at the front end of the exhaust valve to generate a speed compensation signal, which dynamically increases the transient injection speed during the air resistance rise stage, overcomes the pneumatic resistance in the extremely narrow cavity, and reduces the risk of material shortage at the end of the injection molded part and local burning.
[0042] 3. This invention generates an integral of enthalpy by continuously accumulating the heat transfer rate of fluid convection and the work done during compression. This integral serves as the physical basis for determining when to cut off the heating medium and introduce cooling water. At the end of the molding process, the positive pressure holding is stopped by combining the attenuation derivative of the holding pressure and the surface temperature. Simultaneously, negative pressure is withdrawn to cut off the shear stress in the flow channel. Furthermore, a dynamic gas pressure function is used to smoothly apply positive pressure gas, thereby controlling the solidification evolution boundary of the product and ensuring the dimensional accuracy and structural morphology of the final injection molded part. Attached Figure Description
[0043] Figure 1 This is the system architecture diagram of this application;
[0044] Figure 2 This is a flowchart of the method in this application;
[0045] Figure 3 This is a schematic diagram of the air resistance feedback compensation module of this application;
[0046] Figure 4 These are experimental verification comparison diagrams of the multiphysics coupling control method of this application. Among them, (a) is a comparison diagram of the microscopic air resistance shock wave elimination effect in the high-speed filling stage, and (b) is a dynamic air pressure adaptive loading curve in the pressure holding and cooling stage.
[0047] Among them, 10 is the thermal field sensing module; 20 is the dynamic impedance sealing module; 30 is the proxy prediction and interception module; 40 is the air resistance feedback compensation module; and 50 is the dual-mode relaxation module. Detailed Implementation
[0048] The following is in conjunction with the appendix Figure 1 - Appendix Figure 4 This application will be described in further detail below.
[0049] See attached document Figure 1 , Figure 1 This is a system architecture diagram according to an embodiment of the present invention. The present invention provides an intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting, including a thermal field sensing module 10, a dynamic impedance sealing module 20, a proxy prediction and interception module 30, an air resistance feedback compensation module 40, and a dual-modal relaxation module 50. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting is built on industrial real-time Ethernet and edge computing nodes. The thermal field sensing module 10, dynamic impedance sealing module 20, proxy prediction and interception module 30, air resistance feedback compensation module 40, and dual-modal relaxation module 50 interact with the main controller of the injection molding machine via a real-time bus to execute various control logics of the ultra-thin part injection molding cycle.
[0050] During the injection molding preheating stage, the thermal field sensing module 10 controls the variable mold temperature medium to be introduced into the mold conformal water channel. Temperature data is collected by the temperature sensor array distributed on the surface of the mold cavity and a transient temperature distribution matrix of the mold cavity surface is constructed. When the global minimum value in the transient temperature distribution matrix crosses the preset glass transition temperature threshold of the polymer material, the thermal field sensing module 10 extracts the current thermal field data as the initial feedforward parameter and outputs a vacuum suction command to the microporous breathable steel array on the moving mold side to establish a negative pressure environment in the mold cavity.
[0051] After entering the low-speed pressure build-up stage, the injection mechanism propels the polymer melt at an initial velocity. The dynamic impedance sealing module 20 collects the transient injection pressure and transient injection speed of the injection mechanism in real time. It calculates the transient rheological impedance in the cavity based on the ratio of the derivative of the transient injection pressure with respect to time to the transient injection speed. Combined with the transient temperature distribution matrix, it generates an adaptive impedance change judgment threshold. When the calculated transient rheological impedance is greater than the impedance change judgment threshold, the dynamic impedance sealing module 20 determines that the micro gap in the mold cavity has been physically sealed and issues a trigger command to enter the high-speed filling stage.
[0052] After receiving the trigger command to enter the high-speed filling stage, the proxy prediction and interception module 30 is deployed on the external edge computing node. It inputs the real-time collected main pressure of the mold cavity and injection speed into the preset neural network model and outputs the predicted remaining time for the melt wavefront to reach the servo exhaust valve. Combining the mechanical action delay constant of the servo exhaust valve and the control safety redundancy time, the proxy prediction and interception module 30 issues the servo exhaust valve closing command at the feedforward time node.
[0053] During the high-speed filling stage, the gas resistance feedback compensation module 40 operates synchronously and monitors the gas back pressure at the front end of the servo exhaust valve in real time. It calculates the partial derivative of the pressure difference between the main pressure of the mold cavity and the gas back pressure over time, generates a speed compensation signal based on the pressure difference partial derivative, and superimposes it into the instructions of the injection molding machine controller to adjust the transient injection speed.
[0054] During the filling and cooling stages, the dual-modal relaxation module 50 receives control and calculates the integral of the total enthalpy of the mold cavity, including heat transfer from fluid convection and work done by gas compression. When the integral of the total enthalpy of the mold cavity reaches a set critical enthalpy threshold, the dual-modal relaxation module 50 issues a command to cut off the variable mold temperature medium and introduce cooling water. After the cooling water is introduced, the dual-modal relaxation module 50 simultaneously monitors the attenuation derivative of the cavity holding pressure and the decrease derivative of the mold cavity surface temperature. When the attenuation derivative of the cavity holding pressure is less than the preset pressure attenuation threshold and the mold cavity surface temperature... When the decreasing derivative is less than the preset temperature decrease threshold, the dual-modal relaxation module 50 generates a dual-modal trigger signal. In response to the dual-modal trigger signal, it sends an instruction to the injection molding machine controller to control the injection screw to perform a negative pressure withdrawal action, simultaneously cutting off the vacuum suction air path of the microporous permeable steel array and controlling the pneumatic reversing valve group to introduce positive pressure gas into the microporous permeable steel array. The dual-modal relaxation module 50 controls the positive pressure of the gas introduced into the microporous permeable steel array according to the set dynamic gas pressure function containing the solidification layer evolution weight until the injection molded part is cured and shaped.
[0055] See attached document Figure 2 , Figure 2 This is a flowchart of a method according to an embodiment of the present invention. The present invention provides an intelligent control method for injection molding of ultra-thin parts based on variable mold temperature and coordinated venting, comprising the following steps:
[0056] S10 controls the variable mold temperature medium to enter the mold conformal water channel during the injection molding preheating stage. Temperature data is collected by temperature sensor array to construct transient temperature distribution matrix. When the global minimum value in the transient temperature distribution matrix crosses the glass transition temperature threshold of the polymer material, thermal field data is extracted as initial feedforward parameters and a vacuum suction command is output to the microporous breathable steel array.
[0057] S20: In the low-speed pressure building stage, the transient injection pressure and transient injection speed of the injection mechanism are collected in real time. The transient rheological impedance is obtained by calculating the ratio of the transient injection pressure time derivative to the transient injection speed. The impedance change judgment threshold is generated by combining the transient temperature distribution matrix. When the transient rheological impedance is greater than the impedance change judgment threshold, the micro gap is judged to achieve physical sealing and a trigger command to enter the high-speed filling stage is issued.
[0058] S30, after entering the high-speed filling stage, the real-time collected main pressure of the mold cavity and injection speed are input into the preset neural network model, and the remaining time of the melt wavefront reaching the servo exhaust valve is predicted. Combined with the mechanical motion delay constant and the control safety redundancy time, the servo exhaust valve closing command is issued at the feedforward time node.
[0059] S40 synchronously monitors the gas back pressure at the front end of the servo exhaust valve during the high-speed filling stage, calculates the partial derivative of the pressure difference between the main pressure of the mold cavity and the gas back pressure over time, and generates a speed compensation signal based on the pressure difference partial derivative, which is then superimposed on the injection molding machine controller instruction to adjust the transient injection speed.
[0060] S50 calculates the integral of the total enthalpy of the mold cavity, including heat transfer from fluid convection and work done by gas compression, at the end of filling and during the cooling stage. When the integral of the total enthalpy of the mold cavity reaches the critical enthalpy threshold, it issues a command to cut off the variable mold temperature medium and introduce cooling water. Simultaneously, it monitors the derivative of the cavity holding pressure decay and the derivative of the mold cavity surface temperature drop. When the preset pressure decay threshold and temperature drop threshold are met, it controls the injection screw to perform a negative pressure withdrawal action, and at the same time cuts off the vacuum suction air path of the microporous ventilated steel array and introduces positive pressure gas. It controls the positive pressure gas pressure according to the dynamic gas pressure function until the injection molded part is cured and shaped.
[0061] In this embodiment, the intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting acquires the thermodynamic boundary conditions of the mold surface through the thermal field sensing module 10 during the preheating stage to establish an initial coordinated working environment. The thermal field sensing module 10 is further divided into a data acquisition unit, a matrix construction unit, a readiness determination unit, and a coordinated triggering unit.
[0062] During the preheating stage of the injection molding cycle, the main controller directs high-temperature steam or high-temperature heat transfer oil, or other variable mold temperature media, into the conformal water channels on the fixed mold side. From the general principles of heat transfer, the heterogeneity of the mold steel and the differences in the local structure of the water channels inevitably lead to a non-linear spatial distribution of the temperature rise rate across different areas of the mold cavity surface. Based on these thermodynamic characteristics, temperature measurements at a single node cannot accurately reflect the global state of the mold cavity. Therefore, the data acquisition unit acquires discrete temperature data in real time through a temperature sensor array pre-deployed on the mold cavity surface.
[0063] As a preferred approach, the temperature sensor array employs a thin-film thermocouple array, which is fitted and embedded in the non-visible surface area of the mold cavity to avoid leaving physical marks on the injection molded part surface. To address the time axis drift and asynchrony issues that are prone to occur during parallel data acquisition from multiple sensors, the data acquisition unit is equipped with a hardware clock synchronization latch circuit at its underlying layer. This circuit is controlled by the system's global high-frequency clock, ensuring that thousands of discrete temperature data points acquired within the same sampling period are strictly aligned at the microsecond level in terms of timestamps, thereby eliminating reconstruction distortion of the transient temperature field. Regarding the specific selection of thin-film thermocouples, the setting of their operating temperature range, and the wiring and networking method, those skilled in the art can perform conventional designs based on the mold structure dimensions. The principle of acquiring local node temperatures is well-known technology in this field and will not be elaborated upon here.
[0064] The matrix construction unit receives discrete temperature data from the temperature sensor array. Since the temperature sensors are distributed as points in physical space, they cannot directly represent the seamless thermodynamic state of the entire mold cavity. Therefore, the matrix construction unit uses a spatial interpolation algorithm to process the discrete data as continuous data. Specific spatial interpolation algorithms can include inverse distance weighted interpolation or Kriging interpolation. Before performing the inverse calculation of the interpolation matrix, the matrix construction unit actively performs collinearity or coincidence feature detection on the input discrete coordinate point set. If the determinant of the spatial coordinate matrix approaches zero, the system automatically removes redundant physical coincidence points and introduces a preset regularization perturbation term into the matrix diagonal. This completely avoids mathematical errors such as divergence in the solution process or the denominator approaching zero in the calculation of interpolation weights. After spatial interpolation processing, the matrix construction unit generates a transient temperature distribution matrix of the mold cavity surface. This transient temperature distribution matrix can characterize the global temperature gradient information of the three-dimensional forming surface of the mold cavity at a specific time point, denoted as […]. .
[0065] After acquiring continuous thermodynamic distribution information, the system needs to perform quantitative evaluation to determine whether to proceed with the process flow. Due to the complex environment of the injection molding industry, relying solely on the absolute minimum value of the matrix is highly susceptible to interference from poor thermal contact of individual sensors or noise from localized cooling dead zones, potentially leading to a long-term deadlock state. The readiness determination unit reads the transient temperature distribution matrix at a set system clock frequency. The effective high-temperature area ratio of the matrix at the current moment is extracted using a multi-dimensional statistical algorithm.
[0066] The readiness determination unit synchronously compares the effective high-temperature area ratio and the local minimum temperature after filtering with the system's preset glass transition temperature threshold for polymer materials in real time. The glass transition temperature threshold is the physical critical temperature at which a polymer material transitions from a glassy state to a rubbery state. This parameter is determined by the physical properties of the polymer material used in the injection molding process, and its value is typically set between 80℃ and 150℃ depending on the different plastic substrates. The preheating readiness determination equation of the readiness determination unit is expressed as follows:
[0067] ;
[0068] In the formula: This indicates that the temperature is greater than The ratio of the area of the matrix elements to the total surface area of the mold cavity is a function of the matrix element area. This indicates the glass transition temperature threshold of polymer materials; This indicates the set system release area ratio threshold, which is preferably in the range of 0.92 to 0.98. The basis for determining this value is that a small hysteresis temperature difference is allowed at the end of the flow channel or in the non-filled core area. This represents the truncated minimum function after removing the bottom 3% of extremely low outlier data. This represents the local tolerance temperature constant of the system, with a preferred range of 2°C to 5°C.
[0069] The above formula constitutes a logical intersection condition based on the dual constraints of area ratio and truncation minimum value: while ensuring that most of the core filling area of the mold cavity has the conditions of high elasticity and low resistance flow, it tolerates the existence of limited low temperature deviations in non-critical areas that will not cause global freezing flow, thereby improving the robustness of industrial field judgment.
[0070] When all the conditions of the above preheating readiness determination equation are met simultaneously, it indicates that the temperature of the core area of the mold cavity has exceeded the hardening and solidification point of the material, and the mold cavity as a whole has the thermodynamic basis to eliminate the large-area cold material resistance of the melt. The readiness determination unit generates a preheating readiness signal and transmits it to the cooperative triggering unit.
[0071] Upon receiving the preheating ready signal, the collaborative triggering unit synchronously executes collaborative preparation actions across physical domains. Based on the closed-loop connection principle of feedforward control, the collaborative triggering unit extracts the transient temperature distribution matrix data at the current triggering moment and transmits it as the initial feedforward parameter to the dynamic impedance sealing module 20, providing a thermodynamic reference input for the subsequent calculation of the adaptive impedance change judgment threshold.
[0072] Synchronously, the coordinated triggering unit outputs a full-area vacuum suction command to the moving mold side. In this embodiment, a microporous permeable steel array is embedded in the non-appearance surface area of the moving mold side. The microporous permeable steel array has through-hole micron-sized pores, with the equivalent pore size range set from 3μm to 10μm. This pore size allows air molecules to pass through freely while effectively blocking the physical leakage of large molecular chain polymer melt. The full-area vacuum suction command is transmitted to the pneumatic reversing valve assembly connected to the air path on the back of the microporous permeable steel array, controlling the valve core to open the negative pressure channel. Through the full-area suction effect of the microporous permeable steel array, free air inside the mold cavity is rapidly discharged, establishing a uniform initial negative pressure environment before the melt wavefront arrives, eliminating potential air-trapped blind spots during the high-speed filling process of ultra-thin parts.
[0073] The dynamic impedance sealing module 20 replaces traditional mechanical position control with a fluid dynamics and thermodynamics coupling algorithm during the low-speed pressure build-up phase to achieve precise leakage prevention. In this embodiment, the dynamic impedance sealing module 20 is specifically divided into an impedance calculation unit, a threshold adaptive unit, and a phase switching determination unit.
[0074] After entering the low-speed pressure build-up phase of the injection molding cycle, the injection mechanism propels the polymer melt at a set low initial velocity. During this phase, close monitoring of the physical resistance encountered by the melt wavefront is necessary to prevent overflow due to high-speed impacts. The impedance calculation unit uses pressure sensors and displacement encoders at the bottom layer of the injection molding machine to collect the transient injection pressure and transient injection speed of the injection mechanism at high frequency. Due to the unavoidable transmission delay difference between the analog pressure signal and the digital encoder signal in the underlying bus, the impedance calculation unit embeds a data alignment queue based on timestamp difference compensation. This queue uses the encoder clock edge as a reference to perform local linear interpolation of the pressure signal, ensuring strict alignment of pressure and speed conditions within the same control cycle. For the specific model selection and signal conversion circuit of the pressure sensor and displacement encoder, those skilled in the art can perform conventional adaptation according to the hydraulic or servo system architecture of the injection molding machine. The hardware acquisition principle is well-known in the field and will not be elaborated here.
[0075] Based on the general principles of polymer rheology, when a polymer fluid propagates within a confined cavity, the derivative of its pressure over time reflects the intensity of pressure buildup, while the transient velocity represents the current driving boundary conditions. Based on these principles, the impedance calculation unit constructs a transient rheological impedance equation to quantify the physical constraint state of the wavefront. Considering that directly differentiating the discrete pressure signal would cause a sharp amplification of high-frequency noise, the impedance calculation unit pre-smooths the transient injection pressure using a five-point moving average algorithm before performing the derivative calculation. The transient rheological impedance equation is specifically expressed as follows:
[0076] ;
[0077] In the formula: Indicates time Transient rheological impedance within the cavity; This represents the partial derivative of the transient injection pressure with respect to time after real-time acquisition and filtering. Transient injection pressure; This represents the transient injection velocity acquired synchronously. To ensure the integrity of the underlying algorithm and avoid system control failures caused by division by zero, a small positive velocity tolerance constant is introduced into the denominator of the transient rheological impedance equation. This speed tolerance constant The preferred value range is 0.01 mm / s to 0.05 mm / s. This is used to avoid the division-by-zero divergence fault of the arithmetic unit when the injection mechanism is affected by mechanical friction or control blind zone, causing the speed to briefly approach zero, thereby improving the system robustness in the industrial control field.
[0078] After obtaining the transient rheological impedance, the system needs to establish a reasonable judgment criterion to identify the sealing state. Traditional injection molding control usually uses a constant impedance threshold or a fixed position for stage switching. The apparent viscosity of the polymer melt is highly dependent on the contact temperature of the mold interface. When the local mold temperature is too high due to the aforementioned preheating stage, the melt viscosity drops sharply. Even if the wavefront has reached the physical sealing position, the impedance rise it exhibits fails to reach the preset constant trigger dead zone.
[0079] To overcome the aforementioned physical limitations, the threshold adaptive unit combines the transient temperature distribution matrix data transmitted by the aforementioned thermal field sensing module 10 to perform dynamic reconstruction and deduction of the threshold. The relaxation time of the polymer macromolecular chain exhibits a clear exponential dependence on absolute temperature. The threshold adaptive unit constructs a mathematical model based on a variant of the Arrhenius equation. The adaptive impedance mutation determination threshold equation generated by the threshold adaptive unit is expressed as follows:
[0080] ;
[0081] In the formula: This represents the threshold for determining adaptive impedance change at the current moment; This represents the reference impedance constant of the polymer material under reference conditions. This reference impedance constant can be obtained by testing and fitting a capillary rheometer under standard laboratory conditions for a specific polymer material. This represents the viscosity-temperature sensitivity coefficient of this type of material, and its value typically ranges from 0.01K. -1 up to 0.05K -1 Between these values, the coefficient is used to quantify the relative weakening effect of interfacial temperature fluctuations on rheoviscous resistance. This represents the representative interface contact temperature extracted from the transient temperature distribution matrix. As a preferred method, to avoid interference from low-temperature points in non-contact regions, this representative interface contact temperature... Instead of taking the arithmetic mean of the entire domain, the average of the top 30% of data nodes with the highest temperature in the matrix is calculated by area weighting, so as to truly reflect the high temperature environment characteristics of the melt wavefront. This indicates the glass transition temperature of the polymer material.
[0082] The stage switching decision unit is responsible for executing the final control decision. The stage switching decision unit uses a microsecond-level control clock frequency to convert the real-time transient rheological impedance output by the impedance calculation unit... Adaptive impedance change determination threshold with the output of the threshold adaptive unit Continuous comparison is performed. Due to the unavoidable presence of high-frequency electromagnetic noise from hydraulic pulses or servo motors in industrial environments, and the fact that a false pressure surge is easily generated at the moment the screw starts to overcome static friction, the stage switching determination unit is connected in series with basic back pressure constraints and anti-jitter confirmation logic based on time windows.
[0083] When the current transient injection pressure A voltage build-up pressure greater than the set 5MPa to 10MPa threshold is used to filter the startup no-load phase, and the transient rheological impedance is calculated in real time. Greater than the adaptive impedance change judgment threshold When this state is maintained continuously for 33 to 88 consecutive control cycles within the set observation time window, the stage switching determination unit obtains the final determination result. This result confirms that the polymer melt wavefront has completely filled the area around the insert or the preset microscopic physical sealing gap, and the flow front has been blocked by the physical structure of the mold. At the instant this leak-proof sealing state is established, the stage switching determination unit seamlessly sends a trigger command to the injection molding machine main controller via the bus to enter the high-speed filling stage. After receiving the trigger command, the injection molding machine main controller ends the low-speed pressure build-up mode and immediately releases the speed limit to enter the next process control domain.
[0084] The proxy prediction and interception module 30 is activated upon receiving a high-speed filling stage trigger command. During the high-speed filling process of ultra-thin parts, the wavefront propagation of the polymer melt exhibits highly nonlinear and transient characteristics. Traditional feedback control logic typically outputs a valve-closing signal only when the melt reaches the exhaust port. This is often limited by the inherent physical inertia of the electromagnetic coil excitation and valve core mechanical displacement inside the servo exhaust valve, resulting in melt overflow before the valve is fully closed. Based on the aforementioned physical hysteresis characteristics, the proxy prediction and interception module 30 intervenes in advance by introducing a feedforward prediction mechanism. In this embodiment, the proxy prediction and interception module 30 includes a feature construction unit, a timing prediction unit, and an interception execution unit.
[0085] The feature construction unit is responsible for capturing basic operating condition data for model inference from the industrial bus in real time. In complex flow channel systems, transient data at a single moment is insufficient to reflect the acceleration and resistance accumulation trends of melt propulsion. Therefore, the feature construction unit synchronously collects the main cavity pressure and injection speed at the current moment and several historical cycles at a set sampling frequency. To solve the timing misalignment problem caused by the inconsistent sampling frequencies of the main pressure sensor and the injection displacement encoder, the feature construction unit uses a linear interpolation resampling algorithm based on the least common multiple sampling period to strictly align the two heterogeneous signals on the time axis.
[0086] After time alignment, to eliminate weight shifts caused by differences in the dimensions and orders of magnitude of different sensors in subsequent neural network operations, the feature construction unit uses the Z-Score normalization algorithm to preprocess the data in real time. To avoid division-by-zero anomalies caused by the data variance approaching zero during the holding pressure or standby steady state of the injection molding machine, a small regularization constant is introduced into the preprocessing algorithm. The specific normalization equation is as follows:
[0087] ;
[0088] In the formula: This represents the normalized feature data at the current time. This represents the raw data collected at the current moment; This represents the arithmetic mean of the physical quantity within the set sliding time window; This represents the standard deviation of the data within the sliding time window; This represents the regularization tolerance constant, and its preferred value is 1×10. -5 It is used to maintain arithmetic stability when the denominator approaches zero.
[0089] After the above processing, the feature construction unit constructs a two-dimensional feature vector sequence matrix with time sliding window characteristics. As a preferred method, the step size of the sliding time window is set. Feature Dimension This generates a two-dimensional sequence tensor of dimension [50,2], which is then used as the standard input to the downstream neural network proxy model.
[0090] In this embodiment, the timing prediction unit is deployed on a separate external edge computing node, which is directly connected to the main controller via Gigabit Industrial Ethernet and carries a dedicated tensor processing acceleration chip. The timing prediction unit incorporates a lightweight Long Short-Term Memory (LSTM) neural network model. Unlike traditional static feedforward models, the unique forget and update gate structures within the LSTM network can effectively capture the nonlinear dependencies of fluid dynamic parameters in the time dimension.
[0091] For the specific business scenario of this embodiment, the internal hierarchical structure of this lightweight LSTM neural network model is designed as follows: the input layer receives the feature vector sequence matrix with dimensions [50,2]; the hidden layer contains two cascaded LSTM units, where the first layer contains 32 hidden nodes and the second layer contains 16 hidden nodes, and a random deactivation layer with a dropout rate of 0.2 is inserted between the two layers to prevent overfitting caused by industrial noise, thereby achieving layer-by-layer dimensionality reduction extraction of temporal features; the output layer adopts a linear fully connected layer to output a continuous scalar value. This scalar value is used to predict the remaining time required for the polymer melt wavefront to advance from the current position to the servo exhaust valve orifice under the current system boundary conditions.
[0092] To ensure the neural network model possesses practical engineering guidance capabilities, it requires offline training before formal deployment. Those skilled in the art can collect historical trial molding cycle data of the same mold model under different process parameters as a training sample set. The label definition originates from the time difference between the absolute time point of the step change in temperature sensor at the mold's end exhaust port and the data extraction time point in the historical data. During the training phase, the smoothed L1 loss function is preferred. This loss function exhibits mean squared error when the prediction error is small, and absolute error when the error is large, effectively avoiding the gradient explosion problem caused by abnormal jumps in individual sensors. An adaptive moment estimation optimizer is used during training, with an initial learning rate set to 0.001. A cosine annealing learning rate decay strategy is employed in the later stages of training to ensure that the model weights converge to the global optimum.
[0093] The interception execution unit receives the predicted remaining time from the timing prediction unit in real time and performs feedforward decision-making based on the system's physical execution constraints. The process from receiving the command to the valve core completely sealing the exhaust gap includes signal transmission delay and mechanical motion delay. This combined time difference is stored as a constant calibration parameter in the system. The interception execution unit constructs a feedforward triggering interception equation based on the time axis:
[0094] ;
[0095] In the formula: This indicates that the proxy model is at the current moment. The predicted remaining time for the output melt wavefront to reach the servo exhaust valve; This represents the mechanical action delay constant of the servo exhaust valve. Its value can be obtained through valve factory calibration or offline testing with a high-speed camera, and is usually in the range of 10ms to 30ms. This represents the control safety redundancy time set by the system. Its value is preferably between 5ms and 15ms. The setting is based on the residual band predicted by the absorption neural network and the non-Newtonian fluid acceleration jet phenomenon that may occur at the end of the melt, to ensure the arrival time of the absolute envelope wavefront of the interception action.
[0096] The aforementioned feedforward triggering interception equation is used to pre-deduct the response time that the mechanical valve must consume on the time axis, ensuring that the end point of the valve action coincides exactly with the node where the melt arrives. To prevent false triggering caused by small fluctuations in network inference and to mitigate the risk of uninterpretability of the pure data-driven model under extreme conditions, the interception execution unit incorporates multi-dimensional cross-validation logic. The interception execution unit synchronously calculates the real-time first-order time derivative of the current cavity main pressure. Only when the main pressure shows a positive increasing trend, i.e., the first-order derivative is greater than zero, is it proven that the system is in a continuous filling and work state, and the remaining time is predicted. When the above feedforward triggering interception equation is satisfied for three consecutive communication cycles, the interception execution unit confirms it as a valid trigger condition. At the moment the condition is met, the interception execution unit immediately issues a closing command for the servo exhaust valve at the feedforward time node. Through this non-causal system intervention method that trades computing power for time, the system maximizes the discharge of gas from the cavity while physically cutting off the overflow channel at the end of the ultra-thin part filling.
[0097] See attached document Figure 3 The gas resistance feedback compensation module 40 is activated synchronously during the high-speed filling stage. During the molding of ultra-thin parts, at the end of complex flow channels or in areas of abrupt geometric changes, the entrained trace amounts of gas are compressed by the high-speed melt wavefront, easily forming microscopic high-pressure gas resistance. This transient gas resistance acts in reverse on the fluid front, disrupting its constant-speed propulsion state, which in turn leads to water ripples, hysteresis marks, and even localized scorching defects on the surface of the injection molded part. In this embodiment, the gas resistance feedback compensation module 40 maintains wavefront stability by monitoring the fluid dynamic impedance in real time and making dynamic interventions. Its internal components are specifically divided into a data acquisition and filtering unit, a pressure difference trend analysis unit, and a compensation signal generation unit.
[0098] To quantify the hydrodynamic resistance encountered by the wavefront, the data acquisition and filtering unit acquires the microscopic gas back pressure through a miniature high-frequency air pressure sensor deployed in a tiny blind hole at the front end of the servo exhaust valve, and simultaneously acquires data from the main pressure sensor at the bottom of the mold cavity. Considering the differences in electrical response delay and inherent sampling rate of different physical quantity sensors in the underlying bus, the hardware acquisition layer adopts a timestamp latching mechanism based on a global high-frequency clock. As a preferred approach, the data acquisition and filtering unit introduces a first-order linear interpolation alignment algorithm at its underlying layer. This algorithm uses the control cycle of the injection molding machine's main controller as a rigid time grid to force discretely arriving asynchronous air pressure signals to the same time node, thereby eliminating phase lead or lag errors in subsequent differential calculations.
[0099] In industrial high-pressure injection environments, the miniature pressure sensor at the front end of the exhaust valve is susceptible to broadband aerodynamic noise interference caused by high-frequency fluid turbulence and valve outlet shock waves. The data acquisition and filtering unit cleans the acquired gas back pressure time series data. The cleaning algorithm employs a first-order discrete Kalman filter, based on the assumption of continuous pressure change in fluid dynamics. The state equation at the bottom of the filtering algorithm is constructed as a first-order random walk model. This algorithm filters out random high-frequency white noise in the time domain by fusing the state prediction value from the previous moment with the sensor observation value at the current moment, thereby outputting a smooth and effective back pressure signal that reflects the true gas compression state. The conventional settings for the state transition matrix and observation noise covariance matrix in the Kalman filter can be adaptively tuned by those skilled in the art based on the hardware noise floor characteristics of the selected pressure sensor. The principle is well-known in the field and will not be elaborated further here.
[0100] After acquiring smooth and time-aligned time-series data, the differential pressure trend analysis unit performs in-depth analysis of the fluid dynamics. From a kinetic work perspective, the effective work capacity driving the polymer melt to overcome frictional resistance is not determined by the absolute principal pressure of the mold cavity, but by the net pressure difference between the wavefront principal pressure and the microscopic gas back pressure. The differential pressure trend analysis unit calculates the physical difference between the two in real time. To predict the dynamic restraining effect of gas resistance on wavefront propulsion, the system further calculates the partial derivative of this net pressure difference over time.
[0101] Since the control system operates in the discrete time domain, the differential pressure trend analysis unit employs an approximate derivation algorithm based on finite difference to avoid control cycle fluctuations caused by underlying clock jitter. Occasionally, the value approaches zero, causing division to crash and overflow. The partial derivative algorithm introduces a very small time constraint constant into the time denominator term. The partial derivative equation for pressure difference change is expressed as follows:
[0102] ;
[0103] In the formula, Indicates the current time Discrete difference partial derivatives of pressure difference over time; and These represent the transient main pressure of the cavity at the current moment and the previous sampling period, respectively; and These represent the effective gas back pressure at the current moment and the previous sampling period, respectively, after Kalman filtering. Indicates the system's control sampling period; This represents the set time constraint constant, and its preferred value is 1×10. -6 s, ensuring the completeness of mathematical operations. When this partial derivative... When the value is negative, it indicates that the rate at which the gas is compressed and pressurized exceeds the rate at which the main pressure of the screw at the rear end is built up, proving that the effective net driving force of the melt is rapidly decaying and that micromechanical stagnation is about to occur at the wavefront.
[0104] In response to the aforementioned driving force attenuation phenomenon, the compensation signal generation unit performs active feedforward reverse intervention. The compensation signal generation unit generates a dynamic speed compensation signal based on the aforementioned pressure difference partial derivative, and then superimposes the compensation signal into the servo motion control command at the bottom layer of the injection molding machine.
[0105] To prevent a sudden drop in differential pressure due to an unexpected sensor disconnection, which could lead to the injection molding machine misinterpreting a surge in air resistance and executing a dangerous acceleration action, the compensation signal generation unit incorporates multi-dimensional cross-validation logic. The system requires the absolute net differential pressure at the current moment before triggering speed compensation. It must be greater than the set minimum sustaining work threshold. Under the premise of satisfying the above safety constraints, the compensation signal generation unit constructs the dynamic velocity feedback control equations as follows:
[0106] ;
[0107] In the formula: This indicates the final injection speed command sent to the underlying servo driver after real-time reverse superposition. This indicates the maximum safe physical injection speed allowed by the system, set based on the maximum displacement limit of the machine's hydraulic pump. The function implements a forced safety clamp on the compensated speed; This indicates the preset base injection speed profile in the current molding process. The set value for the time; The speed compensation gain coefficient is set, and its value range is preferably from 0.5 mm / (s·MPa / s) to 2.0 mm / (s·MPa / s). The coefficient is determined based on the comprehensive fluid gain volume ratio of the hydraulic response bandwidth of the injection molding machine to the current screw diameter cross-sectional area. This represents the partial derivative of the pressure difference calculated in real time; This indicates the set differential pressure attenuation tolerance dead zone threshold, which typically ranges from 1.0 MPa / s to 5.0 MPa / s. It is used to filter out minute viscous pressure pulsations during normal fluid propulsion and prevent the system from undergoing frequent and ineffective mechanical compensation wear. The function is used to constrain the unidirectionality of the compensation action, that is, the positive velocity gain is only triggered when the negative attenuation of the differential pressure exceeds the dead zone threshold.
[0108] When the screw position of the injection molding machine's main controller reaches the set speed-pressure switching point, the system officially transitions from the high-speed filling stage to the holding and cooling stage. At this time, the dual-modal relaxation module 50 is simultaneously activated. During the cooling process of ultra-thin parts, due to the extremely high surface area to volume ratio, the parts are prone to irreversible shrinkage and warping deformation due to uneven cooling and residual stress release. In this embodiment, the dual-modal relaxation module 50 implements active stress relaxation and deformation suppression by constructing multi-physics boundary conditions. Its internal components are specifically divided into an enthalpy calculation unit, a dual-modal capture unit, and a pneumatic clamping execution unit.
[0109] The enthalpy calculation unit is responsible for macroscopically evaluating the energy decay state inside the mold cavity. Heat transfer of the polymer melt within the mold cavity is a complex transient process involving multiple phases (solid, liquid, and gas). The enthalpy calculation unit constructs a modified integral equation for the total enthalpy of the mold cavity, incorporating fluid convection heat transfer, exhaust mass flow heat dissipation, and work done by gas compression. The specific expression of this equation is as follows:
[0110] ;
[0111] In the formula, Indicates the current time Total enthalpy within the mold cavity; This represents the initial enthalpy reference value of the melt carried into the mold cavity at the moment the filling is completed. (Integral term) Representing the transient fluid convective heat transfer rate, as a preferred method, this term is discretely calculated within the controller using Newton's law of cooling. ,in The calibrated overall convective heat transfer coefficient, The effective contact area between the fluid and the cavity. The transient melt equivalent center temperature is estimated in real time by a first-order heat transfer state observer, based on the initial injection temperature set by the injection molding machine barrel, combined with the material's thermal diffusivity and historical cooling time, since it is difficult to directly install temperature probes inside the cavity. This refers to the mold wall temperature.
[0112] For the exhaust heat dissipation and work terms in the integral equation This represents the instantaneous gas mass flow rate monitored at the exhaust valve. Because the preceding agent prediction and interception module 30 has already issued a command to close the servo exhaust valve at the end of the filling process, the large-scale exhaust mass flow rate is [not specified]. It rapidly decays to zero; This represents the isobaric specific heat capacity constant of the mixed gas within the mold cavity; Indicates the transient temperature of the exhaust gas; Indicates transient gas pressure; It represents the rate of volume change of a gas when it is compressed; This represents an infinitesimally small time element. Since the volume of a microscopic gas cannot be directly measured, the controller utilizes the real-time acquired gas back pressure. This term, obtained indirectly through equivalent substitution using the differential form of the ideal gas law, primarily represents the microscopic energy conversion of the residual trace gas trapped within the microporous permeable steel array and micro-gap of the mold under pressure-holding compression, ensuring energy conservation throughout the entire cycle. In the discrete execution environment of the industrial controller, the above integral operation is approximated using the Euler method to a discrete accumulation and summation algorithm, and the discrete control cycle of the system is used as the basis for this calculation. As a fixed integration step size.
[0113] This modified integral equation overcomes the limitation of relying solely on the surface temperature of the mold steel to assess the cooling state. It incorporates the heat expelled from the mold cavity with the high-temperature gas and the process of the gas being compressed and its internal energy being converted into heat, thus deriving a true cavity energy decay curve. The enthalpy calculation unit calculates the enthalpy in real time. Continuously compare with the set solidification phase change enthalpy threshold, when When the temperature drops and exceeds the threshold for the first time, the enthalpy calculation unit issues a feedforward water-cutting command in advance via the industrial bus to cut off the circulation of the mold cooling water circuit.
[0114] As heat continues to dissipate, the polymer transitions from a highly elastic state to a glassy state, and the volume of the part begins to shrink microscopically. A single pressure sensor is prone to signal blind spots due to localized freezing of the melt, while a single temperature sensor suffers from thermal inertia hysteresis due to probe encapsulation and heat transfer through the steel. The dual-modal capture unit internally performs discrete differential calculations of pressure and temperature synchronously, and its differential control equation is as follows:
[0115] ;
[0116] ;
[0117] In the formula: and These represent the discrete decay derivatives of the holding pressure and surface temperature at the current moment as a function of time; and The cavity pressure and temperature observations represent the current sampling period; and This represents the historical observation value from the previous period; Represents the sampling time interval; The time constant is extremely small, preferably 1×10⁻⁶. -6 s, used to prevent division-by-zero divergence faults caused by time differences approaching zero due to bus communication jitter.
[0118] As a preferred approach, the system sets a pressure attenuation threshold. Typically, a value between -1.5 MPa / s and -3.0 MPa / s is used, along with the temperature drop threshold. Typically, the threshold is taken as -2.0 K / s to -5.0 K / s. The determination of this threshold is based on the derivative characteristics of the PVT (pressure-volume-temperature) characteristic curve of a specific polymer obtained through offline mold testing, at the point of a sudden volume change during phase transition. This threshold is determined if and only if the real-time calculated... and When both conditions are met simultaneously within the set observation window, the system determines that physical and thermal decoupling has occurred between the workpiece and the mold cavity. A sharp drop in pressure indicates that volume shrinkage has broken away from the pressure-holding transmission path, while a sharp drop in temperature indicates that a microscopic air gap has formed between the workpiece and the mold wall, causing heat conduction to be blocked. Through strong coupling verification of these two factors, the system eliminates false triggering caused by single-sensor noise interference.
[0119] Upon responding to the aforementioned dual-modal trigger signal, the system confirms that the solidification and molding inflection point has been reached, and then enters the micro-deformation suppression stage. At this time, the dual-modal capture unit sends a command to the injection molding machine's main controller, controlling the injection screw to immediately stop the forward holding pressure and execute a negative pressure retraction action with a small distance of 2mm to 5mm at a set low speed of 5mm / s to 10mm / s. The timing of this action is used to establish a micro-negative pressure in the main runner at the instant when the part has just solidified but the gate is still in a semi-molten state, using the screw retraction to release the fluid shear stress accumulated near the gate backward, cutting off the physical path of stress transmission to the ultra-thin part body.
[0120] While the screw retracts to release the flow channel stress, the pneumatic clamping actuator simultaneously intervenes to maintain the flatness of the ultra-thin part. To address the inward shrinkage and warping that easily occurs after the ultra-thin part leaves the mold cavity, the pneumatic clamping actuator controls the bottom pneumatic circuit to simultaneously cut off the pre-vacuuming action of the mold's microporous permeable steel array and reverses the control of the proportional pressure regulating valve to introduce positive pressure gas into the permeable steel array. To prevent high-pressure gas from directly penetrating the locally weak layers on the part's surface that are not yet fully hardened, the pneumatic clamping actuator constructs a dynamic gas pressure function based on the solidification layer evolution index weights:
[0121] ;
[0122] ;
[0123] In the formula: The solidification evolution index represents the overall degree of solidification of the part, and is expressed by the current enthalpy relative to the target enthalpy for complete solidification. The attenuation ratio was calculated and obtained through and Functions are strictly constrained to Within the range; Indicates the initial enthalpy reference value; This indicates the amount of heat melted at the current moment; This represents the absolute solid-state enthalpy reference constant corresponding to the polymer reaching the target demolding temperature. This constant is determined in advance by looking up the specific heat capacity integral curve provided by the material supplier, thereby ensuring the absolute reference for normalization calculations. This indicates the command air pressure value sent to the proportional pressure regulating valve at the current moment; This indicates the initial flexible lifting air pressure set. Indicates the maximum permissible rigid compression air pressure; This represents the nonlinear curvature adjustment coefficient, which controls the steepness of the air pressure rise. Its value is preferably in the range of 2.0 to 5.0. The specific value of this coefficient is adapted and calibrated according to the crystallization rate constant of the polymer material.
[0124] This dynamic pressure function is used in In the initial stage, as the part approaches zero and just leaves the mold, the system outputs low-pressure air to provide flexible air cushion support, preventing gas from penetrating the incompletely cross-linked polymer matrix; as... As the part temperature approaches 1, the rigidity gradually increases, and the dynamic air pressure rises exponentially and smoothly, eventually transforming into a large-area hard clamping force on the part surface. This distributed air pressure load forcibly adheres the ultra-thin part to the mold core surface to complete the final cooling, offsetting the internal stress generated by the asymmetric shrinkage of polymer chain segments from a physical and mechanical perspective, ensuring the dimensional accuracy and surface flatness of the ultra-thin injection molded part after demolding.
[0125] To better understand the technical solution of this invention, the following description is based on a specific application scenario.
[0126] See attached document Figure 4 In this embodiment, the test object is an ultra-thin polycarbonate light guide plate for a certain model of smartphone screen. Its average wall thickness is only 0.3 mm, and its projected area reaches 15,000 square millimeters. It is a typical extreme aspect ratio part that is extremely sensitive to internal stress and micro-deformation. The experiment was carried out on a fully electric servo injection molding machine platform equipped with a high-frequency data acquisition bus and a microporous breathable steel mold array.
[0127] To address the molding process of the light guide plate, two parallel control experiments were set up. The control group adopted the industry-standard constant speed injection and fixed-time pressure holding and cooling process, with a fixed threshold of ejection air applied to the pneumatic circuit during demolding. The experimental group, on the other hand, activated the multi-physics coupling control architecture proposed in this invention throughout the process, including proxy prediction and interception, air resistance feedback compensation based on partial derivatives, and a dual-mode relaxation dynamic air pressure loading module.
[0128] During the high-speed filling stage, data from high-frequency sampling by the main pressure sensor at the bottom of the mold cavity showed that in the control group, when the melt wavefront advanced to the dead corner at the end of the runner, the entrained microscopic gas failed to be discharged in time, causing a high-frequency shock wave oscillation with an amplitude of up to 15.5 MPa to erupt inside the mold cavity in a very short time. This violent hydrodynamic backlash directly caused mechanical stagnation of the fluid wavefront, and ultimately, dense water ripples and hysteresis marks visible to the naked eye appeared at the end of the part.
[0129] With the intervention of the gas resistance feedback compensation module, the experimental group demonstrated smooth suppression capabilities. The system, upon detecting the moment when the partial derivative of the net pressure difference between the main pressure and the gas back pressure in the mold cavity showed a negative decay exceeding the -2.0 MPa / s dead zone threshold, inverted the pre-calculated speed compensation signal to the underlying driver within milliseconds. This dynamic feedforward intervention successfully and precisely utilized the increased transient mechanical kinetic energy of the screw to counteract the recoil force of the microscopic gas resistance. Test data showed that the main pressure curve of the experimental group maintained a smooth, monotonically increasing trend throughout the filling process, completely eliminating destructive high-frequency pulsations, and resulting in a perfect optical-grade flatness on the part surface.
[0130] After entering the holding and cooling stage, the control group, due to uneven heat distribution in the mold volume and a single time control mode, experienced inward shrinkage and warping of the part after demolding. The maximum warping deformation of the control group, as indicated by the coordinate measuring machine, reached 1.25 mm, exceeding the tolerance limit of optical assembly. The experimental group, relying on a dual-modal relaxation module, accurately captured the cross-validation inflection point of the holding pressure decay and surface temperature drop, and immediately triggered a 3 mm screw low-speed negative pressure withdrawal to cut off the gate shear stress. Simultaneously, dynamic air pressure loading, which increased exponentially with the solidification layer evolution index, was initiated inside the permeable steel array. Measurement results confirmed that under adaptive aerodynamic constraints ranging from 0.2 MPa flexible lifting to 1.0 MPa rigid compression, the internal stress of the experimental group's part was fully relaxed, and the maximum warping deformation was compressed to within 0.12 mm.
Claims
1. An intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting, characterized in that, include: The thermal field sensing module is used to construct a transient temperature distribution matrix. When the global minimum value of the transient temperature distribution matrix exceeds the glass transition temperature threshold, a negative pressure environment is established and initial feedforward parameters are output. The dynamic impedance sealing module is used to generate an adaptive impedance change judgment threshold based on the initial feedforward parameters, and outputs a high-speed filling stage trigger command when the transient rheological impedance is greater than the adaptive impedance change judgment threshold. The proxy prediction and interception module is used to respond to the high-speed filling stage trigger command to drive the preset neural network model to output the prediction of the remaining time and issue a servo exhaust valve closing command. The air resistance feedback compensation module is used to synchronously monitor the back pressure of the gas at the front end of the servo exhaust valve, and to adjust the transient injection speed based on the partial derivative of the pressure difference between the main pressure of the mold cavity and the back pressure of the gas at the front end of the servo exhaust valve by generating a speed compensation signal. The dual-mode relaxation module is used to introduce cooling water when the integral of the total enthalpy of the mold cavity reaches the critical enthalpy threshold, to perform negative pressure extraction when the derivative of pressure and temperature drop meets the preset threshold conditions, and to control positive pressure gas according to the dynamic gas pressure function.
2. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting as described in claim 1, characterized in that, The thermal field sensing module includes: The data acquisition unit is used to control the flow of the variable mold temperature medium into the mold conformal water channel, and to collect discrete temperature data through an array of temperature sensors distributed on the surface of the mold cavity. The matrix construction unit is used to drive the spatial interpolation algorithm to perform continuous processing based on the discrete temperature data, and reconstruct and generate a transient temperature distribution matrix representing the global temperature gradient information of the mold cavity surface. The readiness determination unit and the collaborative triggering unit are used to extract the effective high-temperature region area ratio based on the transient temperature distribution matrix. When it is determined that the global minimum value of the transient temperature distribution matrix exceeds the glass transition temperature threshold, the collaborative triggering unit extracts the current thermal field data to construct the initial feedforward parameters and outputs a vacuum suction command to the microporous breathable steel array to drive the establishment of a negative pressure environment.
3. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting as described in claim 1, characterized in that, The dynamic impedance sealing module includes: The impedance calculation unit is used to collect the transient injection pressure and transient injection speed of the injection mechanism. By calculating the ratio of the derivative of the transient injection pressure with respect to time to the transient injection speed, the transient rheological impedance in the generated cavity is constructed. The threshold adaptive unit is used to receive initial feedforward parameters to extract representative interface contact temperatures, and combine the viscosity-temperature sensitivity characteristics of polymer materials with the reference impedance constant to generate an adaptive impedance change judgment threshold representing the dynamic leak-proof boundary. The stage switching determination unit is used to perform high-frequency cross-comparison between the transient rheological impedance and the adaptive impedance change determination threshold. When it is determined that the transient rheological impedance is greater than the adaptive impedance change determination threshold and the set observation time window is maintained, it confirms that the micro gap has achieved physical sealing and outputs a high-speed filling stage trigger command.
4. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting as described in claim 1, characterized in that, The proxy prediction and interception module includes: The feature construction unit is used to respond to the high-speed filling stage trigger command, extract historical cycle data of the main pressure of the mold cavity and the injection speed, perform time axis alignment preprocessing, and construct a two-dimensional feature vector sequence matrix with time sliding window characteristics; The timing prediction unit is used to drive a pre-built neural network model deployed on the edge computing node to perform forward nonlinear inference using the two-dimensional feature vector sequence matrix as the underlying feature input, and output the predicted remaining time for the melt wavefront to reach the servo exhaust valve. The interception execution unit is used to construct a feedforward trigger interception boundary by combining the predicted remaining time, the mechanical action delay constant of the servo exhaust valve, and the control safety redundancy time. At the feedforward time node that meets the interception boundary, the servo exhaust valve closing command is issued.
5. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting as described in claim 1, characterized in that, The air resistance feedback compensation module includes: The data acquisition and filtering unit is used to obtain the back pressure of the gas at the front end of the servo exhaust valve and the main pressure of the mold cavity, and to clean up high-frequency aerodynamic noise through the Kalman filter algorithm to generate a smooth pressure time series with strict time alignment. The differential pressure trend analysis unit is used to calculate the net pressure difference between the main pressure of the mold cavity and the back pressure of the gas at the front end of the servo exhaust valve based on the smoothed pressure time series, and to use the finite difference algorithm to approximately derive the partial derivative of the differential pressure as the net pressure difference decays over time. The compensation signal generation unit is used to generate a speed compensation signal that is limited by the maximum safe physical injection speed when the differential pressure partial derivative exceeds the dead zone threshold and decays negatively, and to superimpose the speed compensation signal into the instruction to dynamically adjust the transient injection speed.
6. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting as described in claim 1, characterized in that, The dual-modal relaxation module includes an enthalpy calculation unit; The enthalpy calculation unit is used to receive the physical blockage state where the exhaust mass flow rate is zero after the servo exhaust valve is closed. Based on the continuous accumulation of the differential substitution between the transient fluid convection heat transfer rate and the work done by the microscopic compression of the gas trapped by the exhaust valve, it generates the integral of the total enthalpy of the mold cavity, which represents the energy decay state of the control system. When the integral of the total enthalpy reaches the critical enthalpy threshold, it issues a command to cut off the variable mold temperature medium and introduce cooling water.
7. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting as described in claim 6, characterized in that, The dual-modal relaxation module also includes: The dual-modal capture unit is used to synchronously execute discrete differential operations in response to the cooling water supply command. When the attenuation derivative of the cavity holding pressure is less than the preset pressure attenuation threshold and the decrease derivative of the cavity surface temperature is less than the preset temperature decrease threshold, a dual-modal trigger signal is generated to send a command to the control system to stop the positive pressure holding and execute the negative pressure withdrawal action to cut off the flow channel shear stress. The pneumatic pressing actuator is used to respond to the dual-mode trigger signal to simultaneously cut off the vacuum suction air path of the microporous breathable steel array and introduce positive pressure gas. It extracts the ratio of the current total enthalpy to the enthalpy reference of the completely solid state to construct the solidification layer evolution index, and drives the dynamic gas pressure function to control the positive pressure gas pressure to be applied exponentially smoothly until the injection molded part is solidified and shaped.
8. The intelligent control system for ultra-thin injection molding based on variable mold temperature and coordinated venting as described in claim 2, characterized in that, Before driving the spatial interpolation algorithm, the matrix construction unit performs overlapping feature detection on the discrete coordinate point set to actively remove redundant physical overlapping points, and introduces a preset regularization perturbation term on the diagonal of the spatial coordinate matrix. The readiness determination unit simultaneously compares the effective high-temperature region area ratio and the local minimum temperature after filtering and truncating abnormally low-value data with the glass transition temperature threshold and the system tolerance temperature constant through an intersection constraint comparison.
9. The intelligent control system for ultra-thin injection molding based on variable mold temperature and coordinated venting as described in claim 4, characterized in that, The feature construction unit uses a normalization preprocessing algorithm that introduces a steady-state prevention and elimination of zero redundancy constant standard deviation to standardize the data in real time, eliminate the offset of physical dimension data of heterogeneous sensors, and generate a standard tensor input to the preset neural network model. The interception execution unit has built-in feedforward verification logic, which synchronously calculates the first-order time derivative of the current main pressure of the mold cavity. Only when the first-order time derivative is greater than zero and the system is in a continuous filling state, and the predicted remaining time meets the interception boundary, is it confirmed as a valid trigger and a servo exhaust valve closing command is issued.
10. The intelligent control system for ultra-thin part injection molding based on variable mold temperature and coordinated venting as described in claim 5, characterized in that, When calculating the differential pressure partial derivative, the differential pressure trend analysis unit introduces a time constraint constant into the time denominator of the differential algorithm to prevent division by zero overflow caused by the jitter of the underlying communication network clock. The compensation signal generation unit has built-in multi-dimensional cross-validation logic. Before outputting the speed compensation signal, it verifies that the absolute net pressure difference between the main pressure of the mold cavity and the back pressure of the gas at the front end of the servo exhaust valve at the current moment is greater than the set minimum maintenance work threshold, so as to ensure that the intervention action is applied to the microscopic gas resistance stage.
