Injection molding energy consumption optimization control method based on digital twinning

Abstract

The application relates to the technical field of digital twinning, and discloses an injection molding energy consumption optimization control method based on digital twinning, which comprises the following steps: obtaining a sensor thermal response time constant and a baseline disturbance feature by a digital twinning body, calculating an inversion real temperature value by using the thermal response time constant, and arbitrating the calculated output heat flow in combination with the baseline disturbance feature; detecting a physical feature point associated with a gate freezing, and executing a termination pressure maintaining instruction; a digital model of a sensor self dynamic characteristic is established in the digital twinning body, real-time reverse compensation is carried out on measurement delay and signal distortion caused by physical inertia, and environmental disturbance is arbitrated, so that the real transient characteristic of a physical process can be captured with high fidelity, and control deviation caused by measurement lag and false signals is avoided.

Description

Technical Field

[0001] This invention relates to a method for optimizing and controlling energy consumption in injection molding based on digital twins, belonging to the field of digital twin technology. Background Technology

[0002] Digital twin technology, as a means of achieving real-time mapping and interaction between the physical world and digital space, has been widely applied to the monitoring, prediction, and optimization of complex processes. In injection molding, a typical process, the mainstream technology direction in the industry attempts to deduce and predict the melt flow, heat transfer, and solidification state inside the physical black box of the mold cavity by constructing high-fidelity computational fluid dynamics models, finite element models, or artificial intelligence proxy models, hoping to guide the setting and optimization of process parameters based on this. However, when such digital twin systems that rely on complex mathematical models for indirect deduction are deployed in real industrial production sites, their inherent limitations begin to emerge. The injection molding site is a highly dynamic and volatile environment. Frequent mold changes, batch-to-batch differences in the physical properties of recycled materials, and even changes in environmental temperature and humidity can all lead to changes in the physical processes inside the mold cavity. At this time, the high-fidelity mathematical model originally established based on specific assumptions will quickly become distorted, and the reliability of its prediction results will drop significantly.

[0003] To maintain the fidelity of this digital twin system, on-site engineers with specialized knowledge must spend a significant amount of time recalibrating, calibrating, and verifying the model. This transforms the digital twin system from a problem-solving tool into another costly, complex, and highly sensitive digital black box to fluctuations in operating conditions. When the system's predictions deviate from actual production, engineers often struggle to determine whether the deviation is due to a genuine shift in the physical process or errors in the digital model's deductive logic or fundamental assumptions. This is unacceptable in mass production where high stability is paramount. Meanwhile, other digital twin systems in this field, while architecturally different, also have limitations in addressing core real-time control issues. For example, Chinese invention patent CN118544548A discloses a digital twin-based... The in-mold electronic injection molding digital molding system's main idea is to construct a system-level digital model covering multiple processes and analyze the collected real-time data through artificial intelligence algorithms such as neural networks to predict whether the product has defects. However, this method is still essentially still within the scope of indirect inference, relying on the predictive accuracy of the AI ​​model. This model itself is trained based on historical datasets. When faced with working conditions that are not fully reflected in the training set, such as batch fluctuations of recycled materials or environmental disturbances, its predictive reliability will decrease. This makes it difficult to capture the real state of critical physical events that occur instantaneously, such as gate freezing, which determine molding quality and energy consumption. Therefore, its control logic is not based on the real-time, high-fidelity capture of the physical process, but on a predictive result of inference, which still contradicts the on-site demand for highly stable mass production.

[0004] Therefore, the technical problem to be solved by this invention is how to get rid of the heavy reliance on complex mathematical models and indirect deductions in the construction of digital twin systems, and instead seek a simpler digital mapping method that can directly reflect the real state of physical processes, so as to solve the inherent limitations of existing technologies in terms of high fidelity and low cost. Summary of the Invention

[0005] This invention provides a digital twin-based method for optimizing and controlling energy consumption in injection molding. Its main purpose is to solve the problems of high cost and difficulty in adapting to industrial site fluctuations caused by the reliance on complex models for indirect inference in existing digital twin technology.

[0006] To achieve the above objectives, this invention provides a digital twin-based energy consumption optimization and control method for injection molding. The method is executed by the digital twin and includes: During the non-heat exchange critical stage of the injection molding cycle, the inflow temperature of the cooling medium, the outflow temperature of the cooling medium, and the flow meter of the cooling medium are monitored and the measured temperature value of the outflow temperature sensor and the flow rate value of the cooling medium flow meter are obtained. Based on the obtained inflow temperature, measured temperature value and flow rate value of the cooling medium, the baseline disturbance characteristics used to characterize the disturbance of the cooling medium are calculated; the pre-stored thermal response time constant of the cooling medium outflow temperature sensor is obtained, and then the measured temperature value of the cooling medium outflow temperature sensor is obtained in real time during the filling or holding pressure stage of the injection molding cycle. Based on the acquired measured temperature value and thermal response time constant, the digital twin calculates the inverted real temperature value in real time through a sensor dynamic compensation algorithm; using the inverted real temperature value, the inflow temperature and flow rate of the cooling medium, it calculates the output heat flow removed from the mold cavity in real time and generates a real-time time-series curve of the output heat flow. Based on the real-time timing curve of the output heat flux, and using baseline perturbation features to arbitrate the real-time timing curve, a physical feature point that characterizes the abrupt change in the shape of the output heat flux timing curve and is associated with the melt gate freezing event is detected. In response to the detection of the physical feature point, a termination pressure holding command is generated and executed.

[0007] Preferably, the baseline perturbation feature includes the statistical maximum amplitude of the output heat flux calculated during non-critical heat exchange phases; arbitration includes comparing the features detected on the real-time time-series curve of the output heat flux with the statistical maximum amplitude, and the physical feature points include the peak points or inflection points of the real-time time-series curve of the output heat flux.

[0008] Preferably, the method further includes: acquiring process parameters for calculating the input enthalpy flow; during the stage from the completion of plasticization to the start of injection in the injection molding cycle, actively applying a heat detection pulse with known energy using the plasticization unit of the injection molding machine, and monitoring the temperature response characteristics caused by the heat detection pulse; reversely calculating a calibration factor related to the melt thermophysical parameters; calculating the input enthalpy flow using the process parameters and the calibration factor; and based on the input enthalpy flow and the output heat flow, constructing and updating a digital twin mapping of the net enthalpy state inside the mold cavity in real time through integral calculation.

[0009] Preferably, the step of reverse-calculating a calibration factor related to the melt's thermophysical properties includes: obtaining the estimated mass of the melt inside the barrel ( ); Extract a peak temperature from the temperature response characteristics ( ); the specific heat capacity associated with the calibration factor ( The determination of ) is based on having known energy ( The heat detection pulse and the estimated mass () ) and peak temperature ( The physical relationship between them, where .

[0010] Preferably, the method further includes: after executing the termination pressure holding command, continuing to calculate the output heat flow in real time, integrating the output heat flow to obtain the cumulative removed heat, and generating and executing the mold opening and ejection command when the cumulative removed heat reaches the preset ejection heat threshold.

[0011] Preferably, the method further includes: obtaining the actual cooling time determined by the mold opening and ejection command during multiple injection molding cycles. Time-series data; actual cooling time ( Long-term statistical analysis is performed on the time series data to monitor its trend changes. The trend changes are compared with the preset benchmark cooling time that represents the benchmark health status of the mold. Based on the comparison results, diagnostic information related to the overall heat transfer efficiency of the mold or the health status of the mold cooling circuit is generated.

[0012] Preferably, the mold cooling circuit includes at least a first branch and a second branch; the step of calculating the output heat flow removed from the mold cavity in real time further includes: measuring the first output heat flow of the first branch in real time respectively. ) and the second output heat flow of the second branch ( The digital twin is also used to: acquire the first output heat flux (); ) and the second output heat flow ( The time-series curve or its integral value; based on the obtained first output heat flux ( ) and the second output heat flow ( The time-series curve or its integral value is used to calculate the warpage surrogate index characterizing the asymmetry of mold cooling. ); and based on the warp proxy metric ( ), asymmetrically adjust the process setting parameters for the next one or more injection molding cycles.

[0013] Preferably, the method further includes: at the end of the current cycle, using the accumulated heat removed as the total heat removed for that cycle. ), will remove total heat ( ) and the preset target calorie value ( The parameters are compared, and based on the comparison results, one or more process setting parameters are adjusted for the next cycle.

[0014] Preferably, the digital twin is also used to: obtain the total removed heat ( The time-series data over multiple periods is used to determine the total heat removed. The process of decoupling the time-series data in the time domain to obtain the mean index reflecting slow drift and the variance index reflecting high-frequency fluctuations, and adjusting one or more process setting parameters for the next cycle, further includes: a first control loop: adjusting a first process setting parameter based on a comparison of the mean index and a preset target heat value; and a second control loop: adjusting a second process setting parameter related to melt homogenization based on a comparison of the variance index and a preset target stability threshold.

[0015] Preferably, the sensor dynamic compensation algorithm includes: obtaining the time change rate of the measured temperature value, multiplying the time change rate by the thermal response time constant to obtain the compensation amount, and adding the compensation amount to the measured temperature value to calculate the inverted true temperature value.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. By monitoring the inflow temperature, outflow temperature, and flow rate of the cooling medium in the mold cooling circuit, the output heat flow removed from the mold cavity is measured in real time. Based on this output heat flow and the input melt enthalpy flow, a digital mapping reflecting the net enthalpy state inside the mold cavity is constructed. Furthermore, this method uses morphological analysis of the output heat flow time-series curve itself to detect a feature point associated with the physical event of melt gate freezing. In response to the detection of this feature point, control commands for optimizing energy consumption are generated and executed. The digital mapping established in this way is based on the physical principle of heat conservation, avoiding the uncertainty and high maintenance costs caused by relying on complex mathematical models for indirect deduction. At the same time, the data flow used to construct the digital mapping is unified with the feature detection used to identify key physical events, so that the control logic directly depends on the real-time capture of the physical process state, rather than on static, preset time or empirical parameters, thus constructing a simpler, more robust, and more responsive control closed loop.

[0017] 2. After a molding cycle, the total heat removed in that cycle is calculated and compared with a target heat value. Based on the comparison result, the process setting parameters for the next cycle are adjusted. Furthermore, this method decouples the time-series data of the total heat removed in the time domain to obtain a mean index reflecting slow drift and a variance index reflecting high-frequency fluctuations. Two parallel control loops are established: the first process parameter (such as barrel temperature) is adjusted based on the mean index, and a second process parameter (such as plasticizing back pressure) related to melt homogenization is adjusted based on the variance index. This design decomposes a single digital index, total heat, which contains multiple pieces of information, into two physically orthogonal and independently controllable digital state variables: mean and variance. This allows the method to use a single set of sensor information to address two different technical problems simultaneously: long-term average energy consumption optimization and cycle-by-cycle high-frequency stability control. This achieves multidimensional control objectives and refined execution methods.

[0018] 3. Based on determining the cumulative removed heat by integrating the output heat flow and executing the mold opening and ejection command accordingly, this method further acquires the actual cooling time data determined by the command over multiple cycles. This actual cooling time data is used as a digital signal reflecting the physical asset status. Long-term statistical analysis is performed on this data to monitor its trend changes, and it is compared with the benchmark cooling time representing the mold's baseline health status to generate diagnostic information related to the overall heat transfer efficiency of the mold or the health status of the cooling pipeline. This approach reuses the output information of the control loop and transforms it into the input information of the diagnostic loop. This allows the digital system to perform closed-loop control to compensate for the performance drift of the physical system, while also quantifying and diagnosing the drift itself. In addition to ensuring real-time production quality, this expands the ability to predictively maintain the health status of the physical asset throughout its entire life cycle. Attached Figure Description

[0019] Figure 1 This is a flowchart of the digital twin control and optimization logic of the present invention; Figure 2 This is a comparison curve of the correction effect of the sensor dynamic compensation algorithm of the present invention; Figure 3 This is a diagram showing the physical digital architecture and interaction of the digital twin system of this invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the scope of protection of the invention.

[0021] This invention discloses a digital twin-based energy consumption optimization control method for injection molding. The method is executed by a digital twin deployed in an industrial control system. The digital twin constructs a denoised and dynamically compensated physical process mapping by utilizing information from specific stages of the injection molding cycle and the inherent physical characteristics of sensors. Then, the twin captures key physical feature points in this mapping data in real time and executes energy consumption optimization commands in a closed-loop manner. Furthermore, it extends this method to achieve online calibration of material fluctuations, dual optimization of multi-cycle energy consumption and stability, coordinated control of cooling asymmetry, and other related functions. The system includes functions such as health diagnosis of management assets. During the injection molding process, physical signals, especially heat flow signals, are subject to interference from two typical physical limitations. One is that the sensor used to measure the outflow temperature of the cooling medium inevitably suffers signal delay and distortion due to its inherent thermal inertia. The other is that pressure or inflow temperature fluctuations in the factory's public cooling water network introduce false disturbance peaks into the measurement signal. These two types of interference make it difficult for the digital twin to accurately capture real physical events that occur rapidly, such as gate freezing. Therefore, the digital twin in this method has two signal fidelity reconstruction mechanisms.

[0022] The first mechanism is dynamic compensation of the sensor, where a thermal response time constant characterizing the temperature of the cooling medium outflow sensor is pre-stored in the digital twin. ,Should The value can be obtained through a one-time offline calibration. A feasible calibration procedure is to apply an approximately step-like switching between hot and cold fluids in the water circuit and measure the time required for the temperature response curve to reach 63.2% of its final change value, and then determine this value as [value]. Value; During the filling or holding pressure phase of the injection molding cycle, the digital twin acquires the measured temperature values ​​from the sensors in real time. Then, the sensor dynamic compensation algorithm is immediately executed to calculate the rate of change of the measured temperature value over time. In the digital implementation, this is achieved through... To approximate, where The sampling time interval is used; the rate of change over this time is multiplied by the thermal response time constant. To obtain a compensation amount; finally, this compensation amount is compared with the measured temperature value. Add them together to calculate a retrieved true temperature value. ,Right now ,Should The first mechanism is to more realistically reproduce the transient characteristics of the physical process; the second mechanism is the autonomous learning of baseline perturbation characteristics. During the non-heat exchange critical stages of the injection molding cycle, a typical stage is the mold opening, ejection, or mold closing stage. In this stage, there is no heat release from the high-temperature melt in the mold cavity, and the theoretical heat flow should be zero. The digital twin reuses its sensors to continuously monitor the inflow temperature of the cooling medium, the measured temperature value, and the flow rate value, and calculates any non-zero heat flow signal during this period, calibrating it as the baseline perturbation signal. Digital twins for these The data is statistically analyzed to calculate and store the maximum statistical amplitude during this phase in multiple consecutive cycles. This serves as a dynamic baseline perturbation feature for subsequent signal arbitration.

[0023] After obtaining the signal foundation, the digital twin begins to execute closed-loop control logic; during the filling or holding pressure stage, the digital twin utilizes the inverted true temperature value obtained in the previous step. The inflow temperature and flow rate of the cooling medium are determined using thermal calculation formulas, i.e. Real-time calculation of the output heat flow removed from the mold cavity It generates real-time time-series curves; the digital twin continuously analyzes the data. The shape of the curve, and utilizes stored baseline perturbation features. Arbitration of this curve involves, in a specific logic, comparing the amplitude of the features detected on the real-time time-series curve of the output heat flux with the statistical maximum amplitude. Comparison; if and only if the detected physical feature point, which characterizes a sudden change in the output heat flux time-series curve and is associated with a melt gate freezing event, has a characteristic amplitude higher than Only then is it confirmed as a credible physical event; this physical characteristic point typically manifests in engineering practice as... The timing curve shows a peak point or a clear inflection point; the digital twin immediately responds to the detection of this credible physical feature point, generating and executing a termination pressure holding command; after executing the termination pressure holding command, the system enters the cooling phase, and the digital twin continues to calculate the output heat flux in real time. The output heat flux is then integrated over time to obtain the cumulative removed heat; the digital twin then compares this integrated value with a preset ejection heat threshold. Real-time comparisons are performed; this threshold represents the total heat removed required for the part to reach a level of curing sufficient for safe ejection, and can be obtained through a single process calibration test; when the cumulative heat removed reaches this threshold... When the threshold is reached, the digital twin is immediately generated and the mold opening and ejection command is executed.

[0024] To improve the accuracy of integral control, especially when using recycled materials with fluctuating physical properties, the digital twin also needs to input its energy conservation model, namely, the enthalpy flow. Online calibration is performed; this method adds a calibration module for this purpose, which is activated during the stage from the completion of plasticization to the start of injection in the injection cycle. It utilizes the plasticization unit standard in the injection molding machine, namely the barrel heating coil and thermocouple, and reuses them as an in-situ calorimeter; during this stage, the digital twin actively applies a known energy— —A heat detection pulse is generated, and the temperature response characteristics induced by the pulse are monitored at high frequency; the digital twin extracts a peak temperature from the temperature response characteristics. And obtain the estimated mass of the melt in the heated area inside the barrel. ;in accordance with This physical relationship allows for the reverse calculation of a specific heat capacity related to the melt. The relevant calibration factor; the digital twin then uses this calibration factor to adjust the calculations. The process parameters are corrected in real time; finally, the digital twin is based on this online calibrated... With the precisely measured Through integration Real-time construction and updating of the net enthalpy state inside the mold cavity The digital twin is mapped; during periodic operation, the digital twin also performs cycle-by-cycle optimization and diagnostic functions; at the end of a current cycle, the digital twin uses the cumulative heat removed in the previous step as the total heat removed in that cycle. Digital twin acquisition Time series data over multiple periods is used, and this time series data is decoupled in the time domain to obtain two orthogonal numerical indicators: one is a mean indicator reflecting slow drift. A variance index reflecting high-frequency fluctuations can be obtained through low-pass filtering or by calculating a moving average of 50 periods. It can be achieved through high-pass filtering or calculation. relatively The standard deviation is obtained; the digital twin activates two parallel control loops: the first control loop is used for energy consumption mean optimization, and the mean index is... With a preset target calorie value A comparison is made, and based on the comparison results, a first process setting parameter is adjusted, preferably the barrel temperature that affects the energy baseline; a second control loop is used for process stability control, which adjusts the variance index. The process is compared with a preset target stability threshold, and based on the comparison result, a second process setting parameter related to melt homogenization is adjusted. This parameter is preferably plasticizing back pressure or screw speed.

[0025] For specific quality issues, such as warpage, the digital twin can be further expanded. In this case, the mold cooling circuit needs to include at least a first branch (A) and a second branch (B), such as corresponding to the core and cavity sides respectively. The sensor configuration is also adjusted accordingly to monitor the outlet temperature of the two branches separately. Accordingly, the digital twin further includes, in the step of real-time calculation of output heat flow, the real-time measurement of the first output heat flow of the first branch. Second output heat flow of the second branch Digital twin acquisition and The time-series curve or its integral value is used to calculate a warpage surrogate index (WPI) characterizing the mold cooling asymmetry. WPI is calculated as the ratio of the total heat removed from the two pathways. The cycle-by-cycle optimization logic of the digital twin adds a new control objective: the WPI approaches a target value (e.g., 1.0). When a deviation of the WPI from the target value is detected, the digital twin asymmetrically adjusts the process setting parameters for the next one or more cycles based on this WPI index. The parameters to be adjusted are preferably the zone temperature of the barrel or the zone temperature of the mold. Finally, the digital twin of this method also has the ability to diagnose the health of physical assets. This functional module reuses the output information of the control loop. The digital twin obtains the actual cooling time determined by the mold opening and ejection command in multiple injection molding cycles. The time series data, The value itself is a reverse proxy index of the overall heat transfer coefficient of the mold; the digital twin of this value is a reverse proxy index of the overall heat transfer coefficient of the mold. Long-term statistical analysis is performed on time-series data to monitor its trend changes, such as calculating its long-term moving average; then this trend change is compared with a preset benchmark cooling time that represents the baseline health state of the mold. For comparison, this baseline duration can be calibrated after mold installation or cleaning; when the comparison results are as follows... The moving average compared to When the drift index exceeds a preset engineering threshold, such as 1.10, the digital twin automatically generates diagnostic information related to the overall heat transfer efficiency of the mold or the health status of the cooling pipes, such as water blockage, based on the comparison result.

[0026] Example 1: In an injection molding application scenario, the raw materials contain a high proportion of recycled materials, whose melt flow index and enthalpy fluctuate between batches. Simultaneously, the mold cooling circuit is connected to a common pipeline network of multiple parallel devices, causing the temperature and flow rate of the cooling medium to be subject to instantaneous, high-frequency disturbances from the start-up and shutdown of other devices. In the initial deployment phase of this scenario, the digital twin performs self-calibration and learning; through a one-time step signal calibration, it measures and stores the thermal response time constant of the cooling medium outflow temperature sensor. The time is 0.75 seconds; simultaneously, during the non-heat exchange critical stages of the injection molding cycle, namely mold opening and closing, the digital twin continuously monitors and calculates the output heat flow caused by disturbances in the public pipeline network. In the statistics of 100 consecutive periods, this The statistical maximum amplitude was determined to be one The value is stored as a baseline perturbation feature for subsequent arbitration.

[0027] During a production cycle, this batch of recycled material has a high enthalpy; the digital twin acquires the measured temperature values ​​from the sensors in real time during the filling and pressurization stages. And immediately activate the sensor dynamic compensation algorithm, based on and its rate of change over time and the stored A real-time inversion temperature value is calculated with a time interval of 0.75 seconds. ;Should It compensates for the sensor's response delay in the time domain; the digital twin then uses this... Real-time calculation of output heat flux and generate The real-time timing curve shows that during the pressure holding phase of this cycle, the startup of adjacent equipment causes pressure fluctuations in the public pipeline network. This disturbance occurs in... A momentary characteristic point was generated on the curve, with an amplitude lower than The arbitration logic of the digital twin compares the amplitude of this feature point with the stored baseline perturbation feature. Comparison, because it did not exceed that The threshold signal was identified as a pipeline disturbance and filtered out, without triggering any control action. Due to the high enthalpy of the melt during this cycle, a gate freezing event occurred within the mold cavity. A value higher than 1 was generated on the time series curve. The physical characteristic points; the arbitration logic of the digital twin confirms that the signal amplitude is higher than The system identifies this as a gate freezing event and immediately responds by generating and executing a termination of pressure holding command upon detecting the physical characteristic point. In the next production cycle, the batch enthalpy of the recycled material used is lower than normal. The digital twin repeats the compensation and arbitration process, but in this cycle, the gate freezing event is delayed due to the cooler melt. The digital twin accordingly... The timeline curve showed a higher value only at a later point in time. The physical characteristic point of the threshold also delays the termination pressure holding command, extending the pressure holding time to adapt to material fluctuations. This operation mode coordinates the two mechanisms of sensor dynamic compensation algorithm and baseline disturbance characteristic arbitration, so that control based on physical characteristic point, such as termination pressure holding, can respond to the fluctuation of material properties in each cycle and avoid interference from pipeline disturbance signals.

[0028] Example 2: This example objectively verifies the effectiveness of the method of the present invention in addressing the physical limitations of sensors and disturbances in industrial environments. The test platform uses a 250-ton standard injection molding machine, on which a data acquisition system is installed on the main cooling circuit pipe. This system includes an ultrasonic flow meter with a measurement accuracy of ±0.5%, and a Pt100 platinum resistance temperature sensor with a measurement accuracy of 0.1 ohms installed at both the main inlet and outlet of the cooling medium. The sampling frequency was 10Hz; the digital twin was deployed on an industrial programmable controller to execute the control method of this invention; during the experiment, by intermittently starting and stopping another device connected to the same public cooling network, reproducible transient disturbances were introduced into the temperature and flow rate of the cooling medium flowing into this test platform; the experiment was set up with three operating groups, each of which used the same equipment for 500 continuous production cycles, and tested with two materials: standard virgin PP and fluctuating recycled PCR; control group A adopted traditional open-loop control, i.e., a fixed holding time (5.0 seconds) and a fixed cooling time (20.0 seconds) based on experience; control group B enabled the digital twin, based on The real-time timing curve detects physical feature points to trigger the termination of the holding pressure, but the digital twin does not enable the sensor dynamic compensation algorithm and baseline disturbance feature arbitration function; the sample group of this invention adopts the complete method of this invention, and the digital twin has been completed according to the specific implementation method before the experiment. Calibration (measured) =0.75 seconds) and Learning (measured) =1.2kW), during the test, the average holding time, average cooling time, total energy consumption per cycle, and defect rate caused by insufficient or excessive holding pressure (leading to shrinkage marks or flash) of each group under different material conditions were recorded in real time. The statistical results are shown in Table 1.

[0029] Table 1: Performance Comparison of Different Control Methods under Two Material Conditions Data analysis showed that control group A performed reasonably well when using standard virgin PP, but after switching to fluctuating recycled PCR, its fixed 5.0-second holding time could not adapt to the changes in physical properties, leading to an increase in the defect rate to 12.5%; control group B attempted to... The curve employs closed-loop control, but due to the lack of arbitration, its control logic is frequently triggered by spurious feature points introduced by pipeline disturbances, leading to premature termination of the average holding time (e.g., 4.12 seconds) and a higher defect rate than control group A. Simultaneously, the lack of dynamic sensor compensation results in a fixed delay in detecting the physical feature point of gate freezing even during periods free from noise interference, exacerbating control instability. The baseline disturbance characteristics of this invention's sample group... The arbitration function filtered out all false heat flow fluctuations with amplitudes below 1.2kW introduced by the start-up and shutdown of adjacent machines during the test. At the same time, its sensor dynamic compensation algorithm corrected the measurement delay of 0.75 seconds. Therefore, the sample group of this invention not only achieved accurate capture of physical feature points in 4.82 seconds and optimized cooling time in 18.25 seconds under standard virgin PP material, but also, after switching to volatile recycled PCR material, its digital twin accurately captured physical feature points that were delayed due to changes in physical properties in 5.41 seconds, and adjusted the cooling time accordingly in 18.83 seconds. It maintained the lowest total energy consumption and defect rate per cycle under both materials.

[0030] Example 3: This Example 3 uses the same test platform, materials, working conditions, and data acquisition system as Example 2, and sets up the following two control groups to isolate the key technical features of the present invention: Control Group C (lacking dynamic compensation): This group uses the same control logic as the sample group of the present invention in Example 2, but the sensor dynamic compensation algorithm is disabled; the digital twin uses the original, uncompensated measured temperature value. calculate But it was retained =1.2kW baseline disturbance arbitration function; Control group D (missing disturbance arbitration): This group adopts the same sensor dynamic compensation algorithm as the sample group of the present invention in Example 2 ( =0.75 seconds), but the baseline perturbation feature arbitration function was disabled; the digital twin used the compensated inverted true temperature value. calculate However, its control logic is set to respond to the first detected heat flow characteristic point where the amplitude exceeds the static low threshold (e.g., 0.1kW), and control group C and control group D are continuously run for 500 cycles, and their key performance indicators are compared with the data of the sample group of the present invention in Example 2. The results are shown in Table 2.

[0031] Table 2: Performance Comparison Table When Key Technical Features Are Missing Data analysis shows that control group C, due to the lack of dynamic sensor compensation, Curve calculation benchmark There is a physical delay of 0.75 seconds, therefore, despite its arbitration function While it can filter out pipe network disturbances, its detection of the physical characteristic points of gate freezing is consistently delayed by 0.75 seconds. This delay results in an average holding time (e.g., 5.58 seconds for PP material) that is longer than the required 4.82 seconds, causing ineffective over-holding, leading to an increase in the defect rate (4.7%) and a single-cycle total energy consumption increase (0.316 kWh). Control group D, lacking baseline disturbance arbitration, has a digital twin that cannot distinguish between real physical characteristic points and false characteristic points introduced by pipe network disturbances. In the experiment, pipe network disturbances with an amplitude below 1.2 kW caused by the start-up and shutdown of adjacent machines were incorrectly identified by the digital twin as gate freezing disturbances. The freezing event caused the holding pressure to be prematurely and randomly terminated, such as 3.21 seconds for PP material, resulting in a large number of shrinkage defects and a defect rate of 37.9%. The comparative data of this embodiment shows that the control group C, which lacks the sensor dynamic compensation algorithm, will cause a lag in the detection of physical feature points, resulting in over-holding pressure and increased energy consumption. On the other hand, the control group D, which lacks the baseline disturbance feature arbitration mechanism, will cause the control logic to be erroneously triggered by false disturbance signals, resulting in insufficient holding pressure and a significant increase in the defect rate. In contrast, the sample of the present invention in Example 2, which has both mechanisms, avoids the above two failure modes.

[0032] Example 4: This example combines Figures 1 to 3 This section describes a method for optimizing and controlling energy consumption in injection molding based on digital twins. Figure 1As shown, the process begins with physical signal acquisition, which involves monitoring the inflow / outflow temperature and flow rate of the cooling medium. The acquired data flows to the sensor dynamic compensation module. This module combines the sensor thermal response time constant obtained from the sensor dynamic characteristic modeling step to calculate the inverted true temperature value. This inverted true temperature and flow rate value, combined with the melt thermophysical property online calibration step (applying a probe pulse), inversely resolves the calibration factor to correct the input enthalpy calibration information provided by the input enthalpy. Together, they are used for real-time output heat flow calculation to generate a heat flow time series curve. Before entering the physical feature point detection, this heat flow time series curve undergoes heat flow signal arbitration. This arbitration logic, based on the baseline disturbance feature learning module (i.e., in the non-heat exchange stage), calibrates the maximum amplitude of the environmental disturbance to determine whether the current heat flow signal is greater than the baseline disturbance amplitude. If it is a reliable signal, it enters the physical feature point detection module to detect the peak or inflection point associated with gate freezing. If not, it is determined to be a disturbance. The system filters and processes dynamic signals. When a physical feature point is detected, the system executes a termination pressure holding command and optimizes the pressure holding stage and cycle, entering the cumulative heat removal integration stage, continuously integrating the output heat flow after the termination of pressure holding. The integration result enters the ejection threshold judgment module to determine whether the cumulative heat removal has reached the preset ejection threshold. If not, integration continues; if so, a mold opening ejection command is executed to ensure the curing quality of the part, and the next cycle begins. After the mold opening ejection command is triggered, cycle data acquisition is performed, that is, the total heat removal and actual cooling timing data are obtained. This data is used for multi-cycle dual-loop control, adjusting process parameters by decoupling the mean slow / variance fast in the time domain, and feeding back the adjusted process parameters for the next cycle to the sensor dynamic compensation module and the real-time output heat flow calculation module, respectively. On the other hand, it is used for physical asset health diagnosis, diagnosing mold heat transfer efficiency by analyzing the actual cooling time trend and generating diagnostic information, such as cooling pipe blockage warning.

[0033] like Figure 2 As shown, the horizontal axis represents time in seconds, and the vertical axis represents temperature. The dashed line in the figure represents the original measured temperature. The solid line represents the actual temperature after compensation. As can be seen from the curve shape, compared to the original measurement data , The curve exhibits a faster response rate during the temperature rise phase, with a higher peak value appearing earlier. This indicates that the sensor dynamic compensation algorithm effectively corrects the measurement delay and signal distortion caused by the sensor's physical inertia. Figure 3As shown, the left side represents the physical layer, including the factory's public piping network, which acts as an external disturbance source. Pressure and temperature disturbances from this network affect the physical sensor array. The physical layer also includes the physical sensor array containing inflow and outflow temperature sensors, flow meters, and the physical assets of the injection molding unit. These assets consist of the plasticizing unit of the injection molding machine, the barrel, and the cooling circuit of the mold, which may include branch lines A and B. The physical sensor array transmits data to the industrial control system on the right side via arrows labeled with measured temperature and flow values. For example, an industrial programmable logic controller (PLC). The core of the industrial control system is the digital twin core software, which includes: a signal processing and mapping module responsible for sensor dynamic compensation, baseline disturbance arbitration, and net enthalpy state mapping; and a real-time closed-loop control module. The system is responsible for physical feature point detection, instruction generation such as termination of holding pressure / mold opening and ejection, and multi-cycle optimization and diagnosis modules responsible for multi-cycle dual-loop control, online calibration of melt thermophysical properties, physical asset health diagnosis, and warpage proxy index control. The system also relies on a core database to store sensor thermal response time constants, maximum amplitude of baseline disturbance characteristics, ejection heat threshold, mold reference cooling time, and process timing data. The digital twin core software interacts with the core database through read / write operations. At the same time, the industrial control system sends instructions to the injection molding unit through arrows marked with control instructions such as termination of holding pressure, mold opening and ejection, process parameter adjustment temperature, back pressure, and heat detection pulses to achieve closed-loop control and optimization.

[0034] Example 5: In this example, the digital twin executes a standardized calibration procedure for its internal key control parameters after the physical asset (a newly installed mold) is first installed or replaced. In this scenario, a newly installed mold for producing specific automotive lamp covers is connected to an injection molding machine. The physical characteristics of this mold, such as its heat capacity and heat transfer coefficient, are unknown. Therefore, the core parameter related to energy consumption control in the digital twin, namely the ejection heat threshold, is crucial. Reference cooling time and input enthalpy flow The calibration factors must be determined before mass production begins. The first step in the calibration process is to perform calibration using the input enthalpy flow. The procedure involves the digital twin actively applying a known energy source through the heating coil of the plasticizing unit during the stage between the completion of plasticizing and the start of injection in the injection cycle. The heat detection pulse, a specific parameter set, involves applying a 500W energy pulse lasting 0.2 seconds; the digital twin is based on... This physical relationship, when reversed, yields a calibration factor related to the specific heat capacity of the melt, in which... For specific heat capacity related calibration factors, The applied heat probe pulse with known energy is 100 J in this calibration. The estimated mass of the melt in the heated zone inside the barrel, obtained from the CAD model of the mold, is 25g in this calibration. It was captured by high-frequency monitoring using thermocouples. The peak temperature rise caused by the pulse is 1.12 in this calibration. The digital twin then uses this calibration factor to perform calculations. The process parameters are corrected in real time; the second step of the calibration process is to perform actions to determine the outlet heat threshold. The process calibration test was conducted; the system ran for several calibration cycles. In each cycle, the digital twin only calculated and integrated the cumulative removed heat in real time, but did not execute the mold opening command based on it; the operator manually executed the mold opening and ejection command at different cooling times (e.g., 15 seconds, 18 seconds, 21 seconds) and inspected the quality of the ejected parts; by inspecting the dimensional stability, surface quality, and whether ejection deformation occurred, 18 seconds was determined to be the minimum acceptable cooling time; the digital twin then retrieved the cumulative removed heat value corresponding to the 18-second cycle, which was measured to be 152 kJ in this calibration, and set this value as the preset ejection heat threshold for the mold. .

[0035] The third step in the calibration process is to determine the baseline parameters used for long-term diagnosis and control; After setting the J=152kJ, the system enters a 1000-cycle baseline operation phase; in this phase, the digital twin operates according to... The threshold automatically executes the mold opening and ejection command, and records the actual cooling time for each cycle. Total heat removed The time-series data; after the operation is completed, the digital twin analyzes these 1000... The timing data was statistically analyzed, and its average value was calculated. In this calibration, it was measured to be 18.25 seconds, and this value was stored as the baseline cooling time representing the baseline health state of the mold. Meanwhile, the digital twin is for these 1000 Time-domain decoupling of time-series data is used for mean indicators. The calculation was set to a 50-period moving average window, the size of which was chosen to balance sensitivity to slow drift with smoothing ability for high-frequency fluctuations, while the variance index... This is obtained by calculating the standard deviation within the 50-period window. After this calibration process is completed, the digital twin will have... , , The calibration factor and time-domain decoupling parameters have been initialized for the physical characteristics of the new mold, and the system has entered the closed-loop optimization operation state.

[0036] Example 6: This example demonstrates the synergistic control effect of the method of the present invention in addressing the dual challenges of warpage and long-term process fluctuations. The test scenario involves using an injection molding machine calibrated according to Example 4 to produce a large-area, thin-walled (1.2mm thick) PC material automotive interior panel. This part has high requirements for flatness, and warpage is its main defect. The cooling circuit of the mold is designed as two independent branches, namely the first branch (A branch, corresponding to the core side) and the second branch (B branch, corresponding to the cavity side). The digital twin can measure the output heat flow of each branch separately. and During the initial 100 cycles of production, the digital twin only executes the pressure holding and cooling closed-loop control as described in Example 2, but does not activate asymmetric adjustment and dual-loop control; during this period, the digital twin continuously monitors and records... and The integral value is used to calculate the warpage proxy index. The time-series data; simultaneously, the total heat removed is recorded. (Right now The time-series data is used, and time-domain decoupling is performed on it to calculate the mean index. and variance index Monitoring data at this stage shows that The fluctuation between 1.25 and 1.35 indicates that the heat dissipation rate on the core side (A path) is faster than that on the cavity side (B path). It stabilized at 175 kJ, which is higher than the established target calorific value. (160kJ); Then Gaoda The target stability threshold is far exceeded. The corresponding physical component inspection showed that the warpage defect rate at this stage was as high as 18.6%, and the energy consumption per cycle was also high.

[0037] At the start of the 101st cycle, the digital twin activates the full functionality of asymmetric adjustment and dual-loop control; the digital twin detects... If the mean value of 1.30 deviates from the target value of 1.0, the asymmetric adjustment logic is immediately initiated, based on this... To address the deviation, before the start of the next cycle (cycle 102), the process settings will be adjusted asymmetrically. Specifically, the mold temperature controller setting on the core side (A-path) will be increased by 0.5 degrees Celsius. At the same time, the set temperature of the mold temperature controller on the B path of the cavity side is lowered by 0.3 degrees Celsius. Simultaneously, the first control loop detected the mean index. (175kJ) higher than (160kJ), energy consumption optimization adjustment is performed, reducing the barrel temperature, the first process setting parameter, by 0.4. The second control loop detected the variance index. ( ) higher than the threshold ( The system performs stability control adjustments, increasing the plasticizing back pressure (the second process setting parameter) by 0.3 MPa. During continuous operation from cycle 102 to 500, the digital twin repeats the multi-objective optimization calculations and parameter adjustments at the end of each cycle. At cycle 500, system status data shows that the warpage surrogate index... It has been steadily driven to 1.02; mean indicator It has converged to 160.5 kJ, close to... Variance index It has been suppressed to The value was below the stability threshold; corresponding physical component inspections during this period showed that the warpage defect rate dropped to below 0.5%, while the total energy consumption per cycle also reached the optimized stable level.

[0038] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0039] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for optimizing and controlling energy consumption in injection molding based on digital twins, characterized in that, The method is executed by the digital twin and includes: During the non-heat exchange critical stage of the injection molding cycle, the inflow temperature sensor, outflow temperature sensor and flow meter of the cooling medium installed on the mold cooling circuit are monitored, and the measured temperature values ​​of the inflow temperature sensor and the outflow temperature sensor and the flow rate value of the flow meter are obtained. Based on the obtained inflow temperature, measured temperature value and flow rate value of the cooling medium, the baseline disturbance characteristics used to characterize the disturbance of the cooling medium are calculated; the pre-stored thermal response time constant of the cooling medium outflow temperature sensor is obtained, and then the measured temperature value of the cooling medium outflow temperature sensor is obtained in real time during the filling or holding pressure stage of the injection molding cycle. Based on the acquired measured temperature value and thermal response time constant, the digital twin calculates the inverted real temperature value in real time through a sensor dynamic compensation algorithm; using the inverted real temperature value, the inflow temperature and flow rate of the cooling medium, it calculates the output heat flow removed from the mold cavity in real time and generates a real-time time-series curve of the output heat flow. Based on the real-time timing curve of the output heat flux, and using baseline perturbation features to arbitrate the real-time timing curve, a physical feature point that characterizes the abrupt change in the shape of the output heat flux timing curve and is associated with the melt gate freezing event is detected. In response to the detection of the physical feature point, a termination pressure holding command is generated and executed.

2. The energy consumption optimization and control method for injection molding based on digital twins according to claim 1, characterized in that, Baseline perturbation characteristics include the statistical maximum amplitude of the output heat flux calculated during non-critical heat exchange phases; Arbitration includes comparing the features detected on the real-time time-series curve of the output heat flux with the statistical maximum amplitude. The physical feature points include the peak points or inflection points of the real-time time-series curve of the output heat flux.

3. The energy consumption optimization and control method for injection molding based on digital twins according to claim 1, characterized in that, The method also includes: acquiring process parameters for calculating the input enthalpy flow; actively applying a heat detection pulse with known energy using the plasticizing unit of the injection molding machine during the stage from the completion of plasticizing to the start of injection during the injection molding cycle; monitoring the temperature response characteristics caused by the heat detection pulse; reverse-calculating a calibration factor related to the melt thermophysical parameters; calculating the input enthalpy flow using the process parameters and calibration factor; and constructing and updating a digital twin mapping of the net enthalpy state inside the mold cavity in real time through integral calculation based on the input enthalpy flow and the output heat flow.

4. The energy consumption optimization and control method for injection molding based on digital twins according to claim 3, characterized in that, The steps involved in reverse engineering to derive a calibration factor related to the melt's thermophysical properties include: obtaining the estimated mass of the melt inside the barrel. Extract a peak temperature from the temperature response characteristics. ; Specific heat capacity associated with calibration factor The determination is based on having known energy. Thermal detection pulse, mass estimation With peak temperature The physical relationship between them, among which .

5. The energy consumption optimization and control method for injection molding based on digital twins according to claim 1, characterized in that, The method also includes: after executing the termination pressure holding command, continuing to calculate the output heat flow in real time, integrating the output heat flow to obtain the cumulative removed heat, and generating and executing the mold opening and ejection command when the cumulative removed heat reaches the preset ejection heat threshold.

6. The energy consumption optimization and control method for injection molding based on digital twins according to claim 5, characterized in that, The method also includes: obtaining the actual cooling time determined by the mold opening and ejection command during multiple injection molding cycles. Timing data; actual cooling time Long-term statistical analysis is performed on the time-series data to monitor its trend changes. The trend changes are compared with the preset benchmark cooling time that represents the baseline health status of the mold. Based on the comparison results, diagnostic information related to the overall heat transfer efficiency of the mold or the health status of the mold cooling circuit is generated.

7. The energy consumption optimization and control method for injection molding based on digital twins according to claim 1, characterized in that, The mold cooling circuit includes at least a first branch and a second branch; the step of calculating the output heat flow removed from the mold cavity in real time further includes: measuring the first output heat flow of the first branch in real time respectively. With the second output heat flow of the second branch The digital twin is also used to: acquire the first output heat flux. With the second output heat flow The time-series curve or its integral value; based on the obtained first output heat flux. With the second output heat flow The time-series curve or its integral value is used to calculate the warpage surrogate index characterizing the asymmetry of mold cooling. ; and based on warpage proxy metrics The process settings parameters for the next one or more injection molding cycles are adjusted asymmetrically.

8. The energy consumption optimization and control method for injection molding based on digital twins according to claim 5, characterized in that, The method also includes: at the end of the current cycle, using the cumulative heat removed as the total heat removed for that cycle. Total heat removed Compared with the preset target calorie value The comparison is made, and based on the comparison results, one or more process setting parameters are adjusted for the next cycle.

9. The energy consumption optimization and control method for injection molding based on digital twins according to claim 8, characterized in that, Digital twins are also used to: obtain total removed heat Time-series data over multiple periods were used to obtain the total heat removed. The steps of decoupling the time-series data in the time domain to obtain the mean index reflecting slow drift and the variance index reflecting high-frequency fluctuations, and adjusting one or more process setting parameters for the next cycle, further include: a first control loop: adjusting a first process setting parameter based on a comparison of the mean index and a preset target heat value; and a second control loop: adjusting a second process setting parameter related to melt homogenization based on a comparison of the variance index and a preset target stability threshold.

10. The energy consumption optimization and control method for injection molding based on digital twins according to claim 1, characterized in that, The sensor dynamic compensation algorithm includes: obtaining the time change rate of the measured temperature value, multiplying the time change rate by the thermal response time constant to obtain the compensation amount, and adding the compensation amount to the measured temperature value to calculate the inverted true temperature value.

Images