A multi-point coordinated temperature control system for precision injection molds

Abstract

A method and system for temperature control of injection molds is disclosed. This method introduces the melt flow front position as a dynamic disturbance input into a model predictive controller, adjusting the power of the corresponding temperature control actuator 0.5 to 2 seconds in advance. A thin-film thermocouple array is used to acquire the temperature distribution on the cavity surface. Virtual temperature control zones are dynamically divided by software, and adjacent zones share boundary region sensor data for weighted fusion calculation. This solution predicts the melt position using a screw displacement and cavity pressure mapping model, eliminating the need for additional visual inspection devices. It is suitable for precision injection molding of parts with wall thicknesses of 0.3 to 5 mm, reducing the product warpage defect rate from 5% to below 1%.

Description

Technical Field

[0001] This invention relates to the field of injection molding equipment technology, specifically a multi-point collaborative temperature control system and method for precision injection molds based on melt position feedforward and model predictive control. This technology is applicable to the production of precision injection molded parts with wall thicknesses of 0.3 to 5 mm, such as optical lenses, precision gears, and medical device housings—plastic products with stringent requirements for dimensional stability and surface quality. Background Technology

[0002] During injection molding, the uniformity and stability of the temperature field on the mold cavity surface directly determine the crystallinity, internal stress distribution, and final dimensional accuracy of the product. Traditional processes use thermocouples for single-point temperature measurement in conjunction with PID controllers to regulate cooling water flow, with each temperature control loop operating independently and lacking a coordination mechanism. This architecture has three fundamental flaws.

[0003] The first drawback is the response lag. After the melt enters the mold cavity, it releases a large amount of latent heat of crystallization and shear heat. Cooling is only initiated after the temperature sensor detects the abnormality, and the lag time is usually 5 to 8 seconds. At this time, a temperature gradient has already formed and locked on the surface of the product, making it difficult to eliminate internal stress through subsequent adjustments.

[0004] The second drawback is spatial incoordination. The surface temperature difference of large precision mold cavities needs to be controlled within ±1℃, but the existing zone control uses fixed boundaries, and the temperature control loops of adjacent areas operate independently. When the melt flow front crosses the zone boundary, the abrupt change in the control values ​​on both sides causes weld lines or stress concentration lines to appear on the product at the boundary.

[0005] The third defect is the disconnect between the process and temperature control. The injection speed and holding pressure curves set by the injection molding process engineer and the cooling curves set by the temperature control engineer belong to different systems and are not aware of each other. During high-speed injection, the melt shear heat increases significantly, but the cooling system still operates according to the preset curve and cannot intervene in time before the melt arrives.

[0006] To address the aforementioned issues, existing technologies have attempted to improve uniformity by increasing sensor density or introducing fuzzy control algorithms. However, these improvements have not addressed the root causes of response lag and spatial inconsistency. Model predictive control has been applied in chemical temperature control, but its standard form cannot directly handle the highly dynamic disturbances of melt flow during injection molding. Furthermore, simply increasing the number of sensors leads to data redundancy and a surge in system complexity, failing to meet the reliability and ease-of-maintenance requirements of industrial environments. Therefore, a new method is needed to fundamentally solve the problems of response lag and spatial inconsistency without significantly increasing hardware costs. Summary of the Invention

[0007] The technical problem to be solved by this invention is to provide a method and system for controlling the temperature of injection molds. This method incorporates the melt flow front position as a predictable dynamic disturbance input into a model predictive control framework. It actively adjusts the power of the temperature control actuator in a specific cavity region 0.5 to 2 seconds before the melt reaches that region, thereby reducing the control response lag from over 5 seconds to less than 0.5 seconds. Simultaneously, by dynamically dividing virtual temperature control zones through software and enabling adjacent zones to share boundary sensor data, it eliminates abrupt changes in the control signal at the zone boundaries, achieving precise control of the cavity surface temperature difference within ±0.5℃.

[0008] To achieve the above objectives, the technical solution adopted in this invention comprises five core elements.

[0009] Step 1: Establishing a lightweight temperature sensing network: A thin-film thermocouple array is arranged on the surface of the mold cavity, with the center-to-center distance between adjacent thermocouples set to 10 to 20 mm. This density is sufficient to capture temperature field gradient changes.

[0010] Each thermocouple junction box has a built-in microprocessor that performs digital filtering on the raw temperature measurement data to eliminate electromagnetic interference noise and only uploads the estimated temperature status value to the main controller.

[0011] The data sampling frequency is set to 10 Hz to 20 Hz, a range that balances response speed and data communication load. The data volume of 128 thermocouples is reduced by 90% after edge computing compression, and the main controller maintains a data refresh rate of 10 Hz, fully meeting the requirement of a 50-millisecond control cycle.

[0012] Step 2: Constructing a soft measurement model for melt position: This invention uses existing screw displacement sensors and cavity pressure sensors in injection molding machines to predict melt position. The screw displacement signal reflects the melt injection volume, and the cavity pressure signal reflects whether the melt front has reached a specific position. By combining the two, the coordinates of the melt flow front can be calculated.

[0013] Specifically, a recursive least squares algorithm is used to establish a mapping model. The model inputs are the current screw displacement, screw speed, and current cavity pressure value. The model output is the predicted melt position within a time range of 0.5 to 2 seconds. The prediction time step is synchronized with the control cycle at 50 milliseconds, forming a rolling prediction mechanism. The prediction error is controlled within 3% of the total cavity length during the melt filling stage, which meets the temperature control feedforward accuracy requirements. In practical applications, for molds with particularly complex structures, the number of pressure sensor measuring points near the gate can be appropriately increased to improve prediction accuracy.

[0014] Step 3: Designing a process-coupled model predictive controller: The state vector in the state space form of the controller is composed of the temperature values ​​of each thin-film thermocouple obtained in step one, the control vector is the heating or cooling power of each virtual temperature control zone, and the disturbance vector is composed of the melt position prediction value obtained in step two.

[0015] The system matrix A, control matrix B, and disturbance matrix E were obtained through order reduction identification using the finite element model of mold heat conduction. The identification process was completed during the first mold trial. Specifically, after the mold was heated to a steady state, pulsed power disturbances were applied sequentially to each temperature control zone, the temperature response curves of each thermocouple were recorded, and the system matrix was extracted using a subspace identification algorithm.

[0016] This identification method does not rely on precise mold material parameters, can adapt to actual machining tolerances, and takes less than 30 minutes to identify.

[0017] The objective function of the optimization problem solved by the model predictive controller in each control cycle is set as the weighted sum of the squares of temperature deviation, the squares of power adjustment, and the squares of temperature difference between adjacent zones.

[0018] The constraints include four items: upper and lower limits of heating power, upper and lower limits of cooling power, upper limit of temperature change rate, and upper limit of temperature difference between adjacent zones. Among them, the upper limit of temperature difference between adjacent zones is set at 5℃. This value is determined based on the material thermal stress cracking experiment. When the temperature exceeds 5℃, the internal stress of the product increases sharply.

[0019] The optimization problem is solved using a quadratic programming algorithm, with a computation time of less than 20 milliseconds, and can run stably within a 50-millisecond control cycle.

[0020] The introduction of the perturbation matrix E is a key improvement in this invention. Its physical meaning is the convective heat transfer influence coefficient of melt flow on cavity temperature. When it is predicted that the melt is about to reach a certain region, the perturbation term causes the controller to reduce the cooling power or increase the heating power in that region in advance, thus achieving feedforward compensation.

[0021] Experiments show that adding a disturbance term reduces the temperature prediction error by 40% and reduces the control overshoot from 15% to less than 5%.

[0022] Step 4: Implementing dynamic partitioning and boundary smoothing in software: Based on the geometric characteristics of the mold cavity, the cavity is divided into N virtual temperature control zones at the software level, where N is an integer between 4 and 64. The number of zones is dynamically determined based on the projected area of ​​the injection molded part: N is 8 when the projected area is less than 0.05 square meters, N is 16 when the projected area is between 0.05 and 0.2 square meters, and N is 32 or 64 when the projected area is greater than 0.2 square meters.

[0023] The partition mapping is stored in the controller memory without altering any hardware wiring.

[0024] Two adjacent virtual temperature control zones share thin-film thermocouple data located at their boundary. The width of the shared area is a distance extending on both sides of the boundary, equivalent to 10% of the width of a single virtual temperature control zone.

[0025] When calculating the control quantity, the temperature data in the shared area is weighted and fused. The weight coefficient is dynamically adjusted according to the boundary temperature difference. The calculation formula is w_i equal to 1 divided by 1 plus e ΔT divided by 2, where ΔT is the real-time temperature difference between two adjacent virtual temperature control zones at the boundary, w_i represents the weight assigned to the current calculation partition, and the weight of the other partition is 1 minus w_i.

[0026] This weighting method ensures that when the boundary temperature difference is large, the control quantity is dominated by the side with the smaller temperature difference, thus avoiding control signal oscillation.

[0027] The partition mapping relationship is refreshed every time the mold is changed or the product material grade is changed, and remains unchanged during normal continuous production to ensure control stability.

[0028] Step 5: Execute rapid response temperature control actions: The heating actuator uses a ceramic heating rod, while the cooling actuator uses a high-speed solenoid valve to regulate the cooling water flow. The response time of both is required to be less than 100 milliseconds.

[0029] The power setpoint output by the model predictive controller is sent to the actuator via analog signal or fieldbus, and the control cycle is strictly set to 50 milliseconds.

[0030] The temperature difference between any two points on the cavity surface is maintained within ±0.8℃, and for optical grade products, it can be further tightened to ±0.5℃.

[0031] The cooling water circuit can be selectively connected to the phase change material auxiliary cooling module, which is an external heat exchanger structure filled with microcapsule phase change material with a phase change temperature of 60 to 80 degrees Celsius and a phase change enthalpy greater than 180 joules per gram.

[0032] The cooling water flows through the heat exchanger before entering the mold cooling channel. The cooling water flow is controlled by a pulse, and the pulse duty cycle is dynamically adjusted by the model prediction controller based on the predicted melt position.

[0033] When it is predicted that the melt is about to reach a certain virtual temperature control zone, the pulse duty cycle of the corresponding cooling water in that zone is reduced to below 20%, and the cooling level is restored to normal after the melt passes through.

[0034] The phase change material absorbs heat from the mold during the cooling pause, and the temperature plateau during the phase change process ensures that the mold temperature does not overshoot. This module can be disassembled and maintained independently, and the main system can still operate without this module, lowering the barrier to entry for users. Attached Figure Description

[0035] Figure 1 System overall control logic flowchart Figure 2 Model predictive control calculation flowchart Figure 3 Virtual Temperature Control Zone Dynamic Division Management Flowchart Detailed Implementation

[0036] Example 1: Application of the present invention in precision injection molding of automotive lamp covers The lampshade is 850 mm long and 450 mm wide, with a wall thickness that gradually decreases from 2 mm at the edge to 4 mm at the center. It is made of polycarbonate, and the optical distortion rate is required to be less than 1%. 128 thin-film thermocouples are arranged in a rectangular grid with a spacing of 15 mm on the mold cavity surface. The thermocouple leads are collected into 16 junction boxes, each with a built-in microprocessor performing Kalman filtering, reducing variance by 80%. The thermocouple mounting slots are 0.3 mm deep and 2 mm wide, and after vacuum brazing, the cavity surface is polished to Ra 0.05 micrometers, without affecting the product's appearance.

[0037] The input signals for the melt position soft measurement model were taken from the screw displacement sensor built into the injection molding machine (0.1 mm resolution, 50 Hz acquisition frequency) and three cavity pressure sensors located at the ends of the main runner and two branch runners (range 0-200 MPa). The model employed recursive least squares method with a forgetting factor set to 0.95, and parameter convergence was achieved through 10 consecutive injections during the initial trial molding. After convergence, the model's prediction error for the melt front position was 25 mm, equivalent to 3% of the total cavity length, fully meeting the temperature control feedforward requirements.

[0038] The model predictive controller runs within the main controller, which is a Beckhoff CX5140 industrial PC using the TwinCAT3 software platform. The state vector has a dimension of 32, composed of 32 equally spaced thermocouples on the cavity surface. The control vector has a dimension of 8, corresponding to 8 independent cooling loops. The disturbance matrix E is obtained through finite element method (FEM) reduction identification. During identification, a 5-second pulse power is applied to each cooling loop, and the response curves of the 32 thermocouples are recorded. The matrix parameters are extracted using the system identification toolbox. In the objective function, the weight for temperature deviation is set to 100, the weight for power adjustment is set to 1, and the weight for temperature difference between adjacent zones is set to 50. The constraints include a maximum heating power of 2000 watts, a maximum cooling power of 3000 watts, a maximum temperature change rate of 5 degrees Celsius per minute, and a maximum temperature difference between adjacent zones of 5 degrees Celsius. The quadratic programming solver used is qpOASES, with an average computation time of 18 milliseconds.

[0039] The virtual temperature control zone is divided into 8 main zones, each corresponding to a cooling circuit. Adjacent main zones share thermocouple data within a 15mm width on both sides of the boundary, with each shared area containing 2 to 3 thermocouples. The weighted fusion weights are dynamically calculated based on the boundary temperature difference. When the boundary temperature difference is less than 2℃, the weights are 0.5 for each zone; as the temperature difference increases, the weights shift towards the lower temperature side to ensure smooth control. The control cycle is strictly set to 50 milliseconds, synchronized with the injection molding machine's action sequence.

[0040] The cooling circuit employs a pulse-type high-speed solenoid valve with a switching frequency of 1 Hz and an adjustable duty cycle from 0 to 100%. When the melt front is predicted to be 50 mm from the boundary of a main zone, the cooling duty cycle of that main zone drops from 70% to 15% for 4 seconds before recovering. During this period, the phase change material (PCM) auxiliary cooling module is activated, filled with 30 kg of microcapsule paraffin at a phase change temperature of 65 degrees Celsius. The mold heat is absorbed by the PCM, and the temperature difference between the module's inlet and outlet water is maintained at 8 to 10 degrees Celsius to ensure the mold temperature does not exceed the limit.

[0041] Application results show that the lampshade molding cycle has been shortened from 75 seconds to 65 seconds, the temperature difference on the cavity surface has decreased from ±2.5℃ to ±0.4℃, the optical distortion rate has decreased from 8% to 0.6%, and the product qualification rate has increased from 82% to 99%. For model changeover debugging, simply input the projected area of ​​the new product on the controller interface, and the system automatically completes the virtual temperature control zone division and mapping relationship refresh, reducing debugging time from 4 hours to 25 minutes.

[0042] Example 2: Application of the present invention in injection molding of medical blood collection tubes.

[0043] The blood collection tube is 100 mm long, 16 mm in diameter, and 0.8 mm thick, made of medical-grade polypropylene, with extremely high requirements for dimensional tolerances and biocompatibility. Thirty-two thin-film thermocouples are arranged on the surface of the mold cavity, spaced 10 mm apart. Due to the product's projected area of ​​only 0.0016 square meters, the virtual temperature control zone is divided into four areas, each corresponding to a ring-shaped cooling circuit.

[0044] The input to the soft-sensor model for melt position is simplified to screw displacement and single-point pressure in the main flow channel, reducing the model prediction time to 0.5 seconds. The model predicts the controller state vector dimension to 8 and the control vector dimension to 4, with the temperature difference constraint between adjacent zones tightened to 3 degrees Celsius. The cooling actuator uses a proportional flow valve instead of a pulse solenoid valve, resulting in higher control accuracy.

[0045] This embodiment does not employ a phase change material-assisted cooling module because the thin-walled product cools quickly, eliminating the need for additional heat storage. The system operates independently, without relying on a PCM module, demonstrating the universality of the core architecture of this invention. Application results show that the roundness error of blood collection tubes decreased from 0.05 mm to 0.02 mm, the pass rate increased from 92% to 99.6%, and due to precise temperature control, material degradation was reduced, improving the product's biocompatibility.

[0046] Example 3: This example illustrates the application of the present invention in the injection molding of precision electronic connectors.

[0047] The connector has external dimensions of 35 mm x 20 mm x 8 mm and internally contains 12 precision pin holes with a diameter of 0.3 mm. The positional tolerance is ±0.02 mm, and the material is PBT with 30% glass fiber. Forty-eight thin-film thermocouples are arranged on the mold cavity surface, spaced 10 mm apart, with a denser arrangement around the pin holes within a 5 mm radius. The thermocouple mounting method is the same as in Example 1, and the leads are aggregated into six junction boxes for edge calculation.

[0048] The soft measurement model for melt position uses a combination of screw displacement and cavity pressure as input. Due to the complex product structure, a pressure sensor is placed at the end of the main runner and near each of the two gates, for a total of three measurement points.

[0049] The forgetting factor of the recursive least squares model was adjusted to 0.9 to accommodate the viscosity fluctuation characteristics of glass fiber materials.

[0050] The prediction time domain is set to 1.0 second, and the control time domain is set to 0.3 seconds to ensure that the temperature of the pinhole filling is stable at the moment of filling.

[0051] The virtual temperature control zone is divided into 12 areas, with each pinhole corresponding to an independent zone. The remaining areas are merged into 4 zones, for a total of 16 zones.

[0052] The model predicts the controller's state vector dimension is 16, and the control vector dimension is 16, corresponding to the independent heating and cooling loops for each zone.

[0053] The pulse power amplitude applied during system matrix identification is reduced to 30% of the normal power to avoid deformation of tiny precision cavities due to thermal shock.

[0054] The weight of temperature difference between adjacent zones in the objective function is increased to 80 to ensure a smooth temperature transition between the pinhole area and the surrounding area, and to avoid uneven fiber orientation.

[0055] The upper limit of the temperature change rate in the optimization solution constraints is tightened to 3 degrees Celsius per minute to prevent internal stress deformation of precision structural parts due to temperature gradient.

[0056] The heating actuator uses a miniature ceramic heating element, which is directly integrated into the core of each pin hole, with a response time of 50 milliseconds.

[0057] The cooling actuator uses a miniature pneumatic needle valve cooling water circuit with a flow rate adjustment accuracy of 0.1 liters per minute.

[0058] The control cycle is maintained at 50 milliseconds, and the temperature difference on the cavity surface is controlled within ±0.6℃.

[0059] The phase change material-assisted cooling module was not used because the injection cycle of the precision connector is only 8 seconds and the cooling time is less than 2 seconds, so the response speed of the PCM module cannot be matched.

[0060] Application results show that the pin hole position error decreased from 0.05 mm to 0.018 mm, the uniformity of glass fiber distribution improved, the product warpage decreased from 0.08 mm to 0.02 mm, and the pass rate increased from 88% to 99.2%. During model changeover debugging, only the new pin hole coordinate data needs to be input into the controller, and the system automatically generates a partition mapping, reducing the debugging time from 3 hours to 40 minutes.

[0061] Example 4: This illustrates the application of the present invention in the injection molding of large automotive dashboard panels.

[0062] The instrument panel is 1.2 meters long, 0.4 meters wide, and has a uniform wall thickness of 2.5 millimeters. The surface requires a leather-textured finish, and the flatness requirement is 0.5 millimeters per meter. The material is ABS / PC alloy. 192 thin-film thermocouples are arranged on the mold cavity surface, spaced 15 millimeters apart, in six staggered rows to cover the narrow cavity. The thermocouple leads are aggregated into 24 junction boxes, with edge calculations implemented.

[0063] In addition to screw displacement and cavity pressure, the injection speed is added as an auxiliary variable to the soft measurement model of melt position. Because the instrument panel adopts a multi-point needle valve hot runner system, the fluctuation of injection speed has a significant impact on the shape of the melt front.

[0064] The forgetting factor for the recursive least squares model was set to 0.93, and the parameters converged after 10 trial injections.

[0065] The prediction time domain is set to 2.0 seconds, covering the time it takes for the melt to flow from the central main gate to the farthest end.

[0066] The virtual temperature control zone is divided into 24 zones, with each zone consisting of a main zone every 200 millimeters along the length. Each main zone is further subdivided into two sub-zones, for a total of 48 zones.

[0067] The model predicts that the controller's state vector has a dimension of 48 and the control vector has a dimension of 24. Each main zone corresponds to a cooling loop, while sub-zones share a cooling loop but have independent heating loops.

[0068] During system matrix identification, the pulse power is applied sequentially from the gate near to the gate far to avoid simultaneous disturbance of multiple zones, which could lead to overall thermal deformation of the cavity.

[0069] The weight of the power adjustment in the objective function is increased to 5 to suppress the overall warping of the narrow cavity caused by uneven cooling.

[0070] The temperature difference constraint between adjacent zones is set at 8℃. Since a slightly wider temperature range is allowed for large-area products, overall flatness is given priority.

[0071] The heating actuator uses a cast aluminum heating plate, which is laid on the back plate of the mold, with a power density of 2 watts per square centimeter.

[0072] The cooling actuator uses a proportional valve group, with each main zone independently adjusting the flow rate from 20 to 80 liters per minute.

[0073] The control cycle is 50 milliseconds, the temperature difference on the cavity surface is controlled within ±0.7℃, and the product flatness reaches 0.4 mm per meter.

[0074] A phase change material-assisted cooling module is adopted, with 50 kg of PCM with a phase change temperature of 70 degrees Celsius filled in the heat exchanger. It is installed at the cooling water return main pipe, and the pulse duty cycle is dynamically adjusted from 0 to 100% so that the mold surface temperature only rises by 3 to 5 degrees Celsius during the filling stage, effectively protecting the surface quality of the leather texture decoration.

[0075] Application results show that the flatness of the dashboard improved from 1.2 mm / m to 0.4 mm / m, the depth of shrinkage marks on the textured surface decreased from 0.03 mm to less than 0.01 mm, and the product qualification rate increased from 75% to 96%. Due to the slow response of the cast aluminum heating plate, in this embodiment, the heating control advance is automatically compensated to 1.5 seconds by the prediction module to ensure precise temperature adjustment. During model changeover debugging, the length and gate position of the new product are input, and the system completes the remapping of 48 zones within 30 seconds, reducing the debugging time from 6 hours to 1 hour.

[0076] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Those skilled in the art can adjust parameters such as thermocouple density, number of partitions, and matrix dimension without departing from the concept of the present invention. If these adjustments still adopt the core mechanism of melt position feedforward perturbation and dynamic software partitioning, they are still within the protection scope of the present invention.

Claims

1. A method for controlling the temperature of an injection mold, characterized in that... Includes the following steps: Step 1: Arrange a thin-film thermocouple array on the surface of the mold cavity, with the spacing between adjacent thermocouples set to 10 to 20 mm, and obtain real-time temperature distribution data of the cavity surface. The data sampling frequency is not less than 20 Hz. Step 2: Real-time acquisition of screw displacement signal and cavity pressure signal of injection molding machine, establishment of mapping relationship model between screw displacement and cavity pressure, calculation of physical position coordinates of melt flow front in cavity at current moment based on the mapping relationship model, and rolling prediction of melt position coordinates at subsequent moments; Step 3: The predicted melt position obtained in Step 2 is used as a disturbance input and input together with the real-time temperature data obtained in Step 1 into the model predictive controller. The model predictive controller solves the constrained optimization problem in each control cycle and generates the heating or cooling power setpoint for each temperature control loop. Step 4: Divide the mold cavity into N virtual temperature control zones according to the geometric features of the mold cavity. N is an integer between 4 and 64. Each virtual temperature control zone corresponds to an independent heating or cooling actuator. Two adjacent virtual temperature control zones share the thin film thermocouple data located at their boundary. The width of the shared area is a distance that extends on both sides of the boundary, equivalent to 10% of the width of a single virtual temperature control zone. When calculating the control quantity, the temperature data in the shared area is weighted and fused. Step 5: Output the power setting value generated in Step 3 to the heating or cooling actuator. The actuator response time is less than 100 milliseconds, and the control cycle is set to 50 milliseconds to keep the temperature difference between any two points on the cavity surface within ±0.8℃.

2. The method according to claim 1, characterized in that: In step two, the mapping relationship model uses a recursive least squares algorithm to update the parameters. The model inputs are the current screw displacement, screw speed, and current cavity pressure value. The model output is the predicted melt position within the time range of 0.5 seconds to 2 seconds in the future. The prediction time step is synchronized with the control cycle of 50 milliseconds.

3. The method according to claim 1, characterized in that: The model predictive controller in step three adopts a state-space form. The state vector is composed of the temperature values ​​of each thin-film thermocouple obtained in step one. The control vector is the heating or cooling power of each virtual temperature control zone. The disturbance vector is composed of the melt position prediction value obtained in step two. The system matrix A, control matrix B, and disturbance matrix E are obtained by order reduction identification of the mold heat conduction finite element model. The identification process is completed in the air shot stage of the first trial mold or in the special test mold and solidified before mass production.

4. The method according to claim 1, characterized in that: In step four, the number N of virtual temperature control zones is dynamically determined based on the projected area of ​​the injection molded part. When the projected area is less than 0.05 square meters, N is 8; when the projected area is between 0.05 and 0.2 square meters, N is 16; and when the projected area is greater than 0.2 square meters, N is 32 or 64. The zone refresh is triggered each time the mold is changed or the product material grade is changed.

5. The method according to claim 1, characterized in that: In step four, the weighting coefficients of the weighted fusion calculation are dynamically adjusted according to the boundary temperature difference. The formula for calculating the weighting coefficients is w_i = 1 divided by 1 plus e (ΔT minus 2) divided by 2, where ΔT is the real-time temperature difference between two adjacent virtual temperature control zones at the boundary, w_i represents the weight assigned to the current calculation partition, and the weight of the other partition is 1 minus w_i.

6. The method according to claim 1, characterized in that: The thin-film thermocouple adopts a patch structure with a thickness of about 0.15 mm. It is fixed in a reserved groove 0.5 mm to 1 mm below the surface of the mold cavity by vacuum brazing process. The groove depth is about 0.3 mm and the groove width is about 2 mm. The thermocouple leads are led out to the terminal through the sealing hole of the mold back plate.

7. A temperature control system for injection molds, used to implement the method according to any one of claims 1 to 6, characterized in that... include: A thin-film thermocouple array is used to acquire real-time temperature distribution data on the cavity surface; a screw displacement sensor and a cavity pressure sensor are used to acquire the basic signals required for melt position calculation. The melt position prediction module, embedded in the injection molding machine controller, performs model calculations to determine the mapping relationship between screw displacement and cavity pressure. The model prediction controller, with a control cycle of 50 milliseconds, solves constrained optimization problems and outputs heating or cooling power settings. The dynamic partition management module divides the cavity into N virtual temperature control zones based on the mold cavity geometry and manages boundary data sharing. The heating and cooling actuator array has at least one actuator for each virtual temperature control zone, with an actuator response time of less than 100 milliseconds.

8. The system according to claim 7, characterized in that: The system further includes a data compression module, which is integrated inside the thin-film thermocouple junction box. It uses a digital filtering algorithm to perform edge calculations on the raw temperature data and only uploads the temperature state estimate to the model predictive controller. The data compression rate is greater than 90%, and the refresh frequency of the data received by the model predictive controller is kept in the range of 10 Hz to 20 Hz.

9. The system according to claim 7, characterized in that: The system further includes a phase change material auxiliary cooling module, which is an external heat exchanger structure filled with microcapsule phase change material with a phase change temperature of 60 to 80 degrees Celsius and a phase change enthalpy greater than 180 joules per gram. The cooling water circuit flows through the heat exchanger and then enters the mold cooling channel. The cooling water flow adopts pulse control, and the pulse duty cycle is dynamically adjusted by the model prediction controller according to the predicted value of the melt position. When the predicted melt is about to reach a certain virtual temperature control zone, the pulse duty cycle of the corresponding cooling water in that zone is reduced to below 20%, and the melt returns to the normal cooling level after passing through.

Images