Machine learning based toy injection molding process optimization method and system

Abstract

The application discloses a toy injection molding process optimization method and system based on machine learning, and relates to the technical field of injection molding process data processing. The method performs molding area attribution processing on injection process data generated by a target toy injection molding part in the pressure compensation stage and / or the mold cooling stage to obtain corresponding process data; a process quality correlation sample is constructed based on the process data, and the state change risk of the corresponding stage is predicted based on a trained process optimization model; corresponding control parameters are output based on real-time compensation process data and / or cooling process data, and iterative optimization is performed through convergence judgment. Thus, the overall data processing reduces the masking of local anomalies, realizes the correlation optimization between the local molding state of the complex structure toy injection molding part and the corresponding process control stage, and improves the pertinence of quality prediction and parameter adjustment.

Description

Technical Field

[0001] This invention relates to the field of injection molding process data processing technology, and in particular to a method and system for optimizing toy injection molding processes based on machine learning. Background Technology

[0002] In the field of plastic product processing and manufacturing, injection molding is a commonly used molding process, widely applied to the mass production of plastic shell parts, connectors, decorative parts, functional structural parts, and soft rubber components. Injection molding typically includes the following processes: raw material supply and drying, melting and plasticizing, injection filling, pressure holding and shaping, cooling and molding, mold opening and demolding, and quality inspection. In this production process, injection molding equipment uses heating and plasticizing mechanisms, feeding mechanisms, injection mechanisms, mold closing mechanisms, and control mechanisms to melt, transport, inject, and shape the plastic material. Mold temperature control devices, cooling medium circulation devices, and related auxiliary equipment are used to maintain the temperature and cooling conditions required for product molding. Temperature detection devices, pressure detection devices, flow detection devices, position detection devices, and controllers installed in the production equipment or production line can be used to collect data related to the injection molding operation status. After product molding, quality inspection results can be obtained through visual inspection equipment, dimensional inspection equipment, appearance inspection equipment, or manual inspection.

[0003] Because toys typically have diverse appearances, assembly requirements, and surface quality requirements, existing technologies already utilize production process data and product inspection data for process analysis and optimization. Specifically, data such as temperature, pressure, time, flow rate, displacement, equipment operating status, mold status, raw material status, and product inspection results generated during injection molding can be acquired. This data can then be filtered, cleaned, normalized, labeled, classified, or statistically processed to form a dataset for process analysis or model training. Machine learning models, predictive analytics models, or parameter recommendation models are then used to train and analyze historical production data and quality inspection results, yielding model outputs that characterize product quality changes, the likelihood of defects, or the direction of process adjustments. In subsequent production processes, based on real-time collected production data and model outputs, adjustments are made to relevant process settings in injection molding equipment, mold temperature control equipment, cooling auxiliary equipment, or production control systems to improve the production quality and processing stability of injection-molded plastic toy parts.

[0004] However, plastic toy shells, connectors, decorative parts, functional structural parts, or soft rubber toy components typically feature complex shapes, numerous curved surfaces and corners, uneven wall thickness distribution, dense assembly structures such as ribs and snap-fits, and high requirements for appearance quality. During the injection molding process of these products, various factors, including material condition, mold condition, equipment operating condition, changes in process settings, and environmental or continuous production influences, can affect the final molding quality. Furthermore, the quality variations exhibited by different product structures, different production batches, and different equipment operating conditions may not be consistent. Therefore, when conducting unified analysis based on overall production data and final inspection results, the analysis results regarding the relationship between process state changes and product quality performance are prone to inaccuracies. This makes it difficult for the quality predictions or process adjustment suggestions output by the model to be consistently applied to the actual production process of injection-molded toy components with complex structures.

[0005] Furthermore, the operational data generated during injection molding production is typically characterized by continuous collection, diverse sources, varying update frequencies, and large data volumes. In contrast, product quality inspection results are usually obtained after the product has completed molding, demolding, transportation, or entered the inspection station. Due to the differences in acquisition time, collection method, data granularity, and feedback timing between these production data and quality inspection results, existing methods typically organize and use relevant data on a product, batch, or production cycle basis when training models or performing process analysis on historical production data. When there are short-term changes in operating status, fluctuations in some equipment parameters, localized molding quality anomalies, or multiple factors jointly affecting product quality during production, the relevant anomaly information may be weakened during data processing, feature extraction, or overall evaluation. This results in the trained model's insufficient ability to reflect changes in actual production conditions, leading to a low degree of matching between process adjustment suggestions and actual production situations. Consequently, this affects the quality control effectiveness and process optimization reliability of the plastic toy injection molding production process. Summary of the Invention

[0006] Therefore, embodiments of the present invention provide a method and system for optimizing toy injection molding processes based on machine learning. The technical solution is as follows:

[0007] On the one hand, a machine learning-based method for optimizing toy injection molding processes is provided, which includes:

[0008] S1. Perform molding region attribution processing on the injection molding process data generated during the target process control stage of the target toy injection molded part to obtain process data associated with different molding regions and corresponding target process control stages. The target process control stages include one or both of the pressure holding and shrinkage compensation stages and the mold cooling stages.

[0009] S2, construct process quality correlation samples based on process data, and input them into a preset machine learning model for training to generate a process optimization model; the process optimization model is used to predict the risk of state changes in different forming areas during the pressure holding and feeding stage and / or the mold cooling stage.

[0010] S3: Collect the corresponding shrinkage process data and / or the cooling process data corresponding to the mold cooling stage, synchronously call the process optimization model, and output the corresponding shrinkage control parameters and / or cooling and shaping control parameters after state change risk prediction.

[0011] S4. Determine whether the current process optimization state meets the corresponding convergence condition. If yes, output the corresponding target process optimization result; otherwise, update the current process optimization state and execute the next round of process optimization.

[0012] On the other hand, a machine learning-based toy injection molding process optimization system is provided, which includes:

[0013] The process data attribution processing module is used to process the molding process data generated by the target toy injection molded part in the target process control stage, and to obtain process data associated with different molding areas and corresponding target process control stages. The target process control stages include one or both of the pressure holding and shrinkage compensation stages and the mold cooling stages.

[0014] The process quality modeling module is used to construct process quality correlation samples based on process data, and input them into a preset machine learning model for training to generate a process optimization model. The process optimization model is used to predict the risk of state changes in different forming areas during the pressure holding and feeding stage and / or the mold cooling stage.

[0015] The parameter recommendation module is used to collect corresponding feeding process data and / or cooling process data corresponding to the mold cooling stage, synchronously call the process optimization model, and output the corresponding feeding control parameters and / or cooling shaping control parameters after state change risk prediction.

[0016] The optimization status judgment module is used to determine whether the current process optimization status meets the corresponding convergence condition. If so, the corresponding target process optimization result is output; otherwise, the current process optimization status is updated and the next round of process optimization is executed.

[0017] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0018] 1. This invention assigns molding region data generated during the pressure-holding and shrinkage-compensation stage and / or mold cooling stage of the target toy injection molded part to different structural regions susceptible to shrinkage or cooling conditions by continuously collecting process information such as pressure, time, temperature, and flow rate. Furthermore, it constructs process quality correlation samples by combining the quality inspection data corresponding to the current injection cycle and uses these samples to train a process optimization model, enabling the model to learn the correlation between process state changes in specific molding regions and quality inspection results. In actual production, real-time shrinkage process data and / or real-time cooling process data for the corresponding stage are collected, and the process optimization model is invoked to output shrinkage control parameters and / or cooling and shaping control parameters, allowing parameter adjustments to be determined based on the current process state and changes in the corresponding molding region. Finally, by determining whether the current process optimization state meets the convergence condition, if a stable optimization state is not reached, the process is updated and the next round of process optimization is executed. When a stable optimization state is reached, the target process optimization result is output, thereby achieving a closed-loop process between quality risk prediction, process parameter adjustment, and optimization result confirmation in the toy injection molding process, improving quality control effectiveness and process optimization reliability.

[0019] 2. By extracting data such as pressure, duration, screw stroke, cavity pressure response, and switching status corresponding to the holding and shrinkage compensation stage from the injection molding process data, and dividing the target toy injection molded part into a thick-walled area near the gate, a thin-walled area far from the gate, and an intermediate transition area based on the three-dimensional structure, wall thickness distribution, and positional relationship between the part and the mold gate, the system reduces the possibility of local insufficient material replenishment being masked by the overall process data. Furthermore, by combining the influence of wall thickness and pressure transmission distance in each region, the relationship between screw stroke changes and pressure response attenuation is corrected to obtain holding pressure transmission efficiency parameters. Based on the difference between the regional shrinkage volume and the replenishment volume, shrinkage supply and demand matching parameters are formed, thereby generating regional shrinkage compensation correction amounts, which helps to accurately characterize the differences in shrinkage effects at different structural locations. Simultaneously, data such as temperature, cooling medium flow rate, cooling duration, and execution status corresponding to the mold cooling stage are extracted from the injection molding process data. Cooling state characterization parameters are determined based on the actual mold temperature deviation, the temperature difference deviation between the inlet and outlet of the cooling medium, the flow rate deviation, and the deviation duration. Effective cooling process data that meets the cooling state requirements are then selected to ensure stable characterization of the cooling and shaping state. This improves the reliability of quality risk prediction and parameter adjustment for injection-molded parts of complex-structured toys.

[0020] 3. By acquiring the surface shrinkage location, depression depth, dimensional detection location, and dimensional deviation value related to the holding pressure and shrinkage compensation state during the current injection molding cycle, and based on the coordinates of the detection location in the 3D design model, the detection results are mapped to shrinkage-related areas such as the thick-walled area near the gate, the thin-walled area far from the gate, or the intermediate transition area. This ensures that the local quality performance of the product after demolding accurately corresponds to the previously generated regional shrinkage compensation process data. Furthermore, based on whether the depression depth and dimensional deviation value exceed the upper limit of the allowable deviation, normal or abnormal marks are generated for each shrinkage-related area, and the shrinkage compensation process data, regional shrinkage compensation correction amount, and detection data are integrated. The corresponding labeled combinations form a shrinkage process quality correlation sample, enabling the trained pressure-holding shrinkage optimization model to predict shrinkage anomaly risks for different structural regions, reducing the masking effect of overall inspection results on local shrinkage or dimensional anomalies. Simultaneously, by acquiring warpage height, contour offset distance, and post-molding dimensional deviation values, the cooling and setting state of the current injection molding cycle is marked as normal or abnormal. The cooling-related process data, cooling state characterization parameters, inspection data, and corresponding labels are combined into a cooling process quality correlation sample, allowing the cooling and setting optimization model to learn the relationship between the cooling process state and the overall stability after molding. Therefore, it is possible to establish prediction criteria for regional shrinkage quality risks and cyclic cooling and setting risks, improving the identification of quality risks in complex toy injection molded parts.

[0021] 4. During actual injection molding production, real-time shrinkage process data for the current cycle is collected and processed using the same data processing method as during model training. This allows the pressure-holding shrinkage optimization model to output abnormal risk values ​​corresponding to each shrinkage-related region. When this value exceeds the corresponding threshold, the adjustment direction and magnitude of the holding pressure, holding duration, holding switching time, or holding switching position are determined based on the regional shrinkage compensation correction amount. This helps to implement shrinkage adjustment for locations prone to shrinkage, depressions, or dimensional deviations. Simultaneously, real-time cooling process data is processed to determine cooling state characterization parameters. The cooling shaping optimization model is used to identify abnormal risks in the current cycle's cooling shaping. When the risk exceeds limits, the adjustment direction and magnitude of the mold temperature, cooling medium temperature, cooling medium flow rate, or cooling duration are determined. Furthermore, the recommended adjusted control parameters are applied to the next injection molding cycle, and corresponding process data is re-collected for risk prediction. When the updated control's abnormal risk meets the threshold requirements, the target process optimization result is output; otherwise, the parameters are further corrected, and the next round of optimization is executed, thus forming a closed-loop process optimization process of prediction, adjustment, verification, and convergence output. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart illustrating a machine learning-based toy injection molding process optimization method provided in an embodiment of the present invention;

[0024] Figure 2 A schematic diagram of the structure of a machine learning-based toy injection molding process optimization system provided in an embodiment of the present invention;

[0025] Figure 3 This is a graph showing the change in shrinkage risk value of different molding areas under different holding pressures, provided in an embodiment of the present invention.

[0026] Figure 4 The graphs showing the variation of insufficient shrinkage risk value or dimensional deviation in different molding areas under different holding pressure durations are provided in the embodiments of the present invention.

[0027] Figure 5 The graph shows the change curves of warpage risk value or warpage deformation of different molding areas under different cooling times, as provided in the embodiments of the present invention. Detailed Implementation

[0028] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0029] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0030] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0031] The embodiments of the present invention provide, as follows Figure 1 The flowchart shown is for a machine learning-based optimization method for toy injection molding processes. The process flow of this method may include the following steps:

[0032] S1 includes:

[0033] Scenario 1: When the target toy injection molded part is injection molded by an injection molding machine and a molding die, and the toy injection molded part has assembly pillars, reinforcing ribs, thin-walled shells and connecting transition structures, different structural areas have different material replenishment requirements due to different wall thicknesses and distances from the gate. It is necessary to identify the material replenishment status of the area during the pressure holding and shrinkage compensation stage in order to reduce the risk of shrinkage, depressions and local dimensional deviations.

[0034] The shrinkage compensation process data corresponding to the holding pressure and shrinkage compensation stage is extracted from the injection molding process data. This data includes holding pressure data, holding pressure duration data, screw stroke variation data, cavity pressure response data, and holding pressure switching status data. The injection molding process data includes, but is not limited to, shrinkage compensation process data and cooling process data. The target toy injection molded part is a toy shell, connector, or functional structural part produced by the target injection molding machine and the target molding die. The holding pressure and shrinkage compensation stage is the stage from mold filling completion to cooling and solidification, during which the screw continues to move forward and replenish the cavity with molten material. The holding pressure, holding pressure duration, and holding pressure switching status are read by the injection molding machine controller; screw stroke variation data is collected by the screw position encoder; and cavity pressure response data is collected by pressure sensors located at different molding positions in the mold. The shrinkage compensation process data is obtained by extracting the above data according to the injection cycle number and the start and end times of the holding pressure.

[0035] Based on the 3D design model of the target toy injection molded part and the gate location data of the mold, the wall thickness, volume, and flow path length to the gate are read for each structural location. The thick-walled structural area near the gate and containing assembly pillars, reinforcing ribs, or bosses is defined as the near-gate thick-walled area; the area far from the gate and formed by the end thin-walled shell surface or edge structure is defined as the far-gate thin-walled area; and the area located between the two and serving as the connection between the thick and thin structures is defined as the intermediate transition area.

[0036] The installation positions of each pressure sensor are mapped to the corresponding feeding association area, so that the cavity pressure response data belongs to the corresponding area; the holding pressure, holding duration, and holding switching status are synchronously associated with each area. For screw stroke change data, the total feeding volume is first obtained by multiplying the screw cross-sectional area by the forward movement distance during the holding period; then, the feeding allocation value for each area is obtained by multiplying the area volume ratio, the ratio of the area's average wall thickness to the standard wall thickness, and the feeding reachability coefficient corresponding to the gate transfer distance; the feeding allocation value for each area is divided by the sum of the feeding allocation values ​​for all areas to obtain the area allocation ratio; the total feeding volume is multiplied by the area allocation ratio to obtain the area feeding volume. The area feeding process data consists of area number, area type, wall thickness, volume, gate transfer distance, holding pressure, holding time, area feeding volume, cavity pressure response, and holding switching status.

[0037] For each region, the screw stroke change rate is obtained by dividing the screw forward distance during the holding pressure period by the holding pressure duration; the pressure response decay rate is obtained by dividing the difference between the gate-side reference pressure and the region's cavity pressure by the gate-side reference pressure and the corresponding pressure transmission time. The baseline stroke change rate and baseline pressure decay rate are determined by averaging the data from multiple samples with no shrinkage, no dents, and acceptable dimensions. The current screw stroke change rate is divided by the baseline stroke change rate to obtain the feeding state value; the current pressure response decay rate is divided by the baseline pressure decay rate to obtain the pressure loss state value; the feeding state value is divided by the pressure loss state value to obtain the basic transmission efficiency value. Furthermore, the average wall thickness of the region is divided by the standard wall thickness to obtain the wall thickness correction coefficient, and the distance from the region to the gate is divided by the maximum effective material replenishment distance of the mold to obtain the distance correction coefficient; the basic transmission efficiency value is divided by the sum of the wall thickness correction coefficient and the distance correction coefficient to obtain the holding pressure transmission efficiency parameter, which is used to characterize the effectiveness of the actual material replenishment effect transmitted to the region.

[0038] The estimated zone shrinkage volume is determined by multiplying the zone volume by the molding shrinkage rate of the current material at the corresponding mold temperature. The estimated zone replenishment volume uses the aforementioned zone replenishment volume and is corrected by checking whether the cavity pressure in that zone reaches the pressure range corresponding to the qualified sample. Subtracting the estimated zone shrinkage volume from the estimated zone replenishment volume and then dividing by the estimated zone shrinkage volume yields the replenishment supply-demand matching parameter. A negative result indicates insufficient replenishment, a result within the allowable range of the qualified sample indicates a balanced replenishment, and a result exceeding the allowable upper limit indicates excessive replenishment.

[0039] The preset shrinkage matching state is determined by the range of pressure holding and transfer efficiency parameters and the range of shrinkage supply and demand matching parameters corresponding to multiple qualified samples in appearance and size inspection. The current two types of parameters are compared with their corresponding ranges, and the difference exceeding the range boundary is divided by the corresponding boundary value to obtain dimensionless transfer anomaly state values ​​and supply and demand anomaly state values. The weights of supply and demand influence and transfer influence are not directly set, but are based on historical data of qualified samples and samples with shrinkage, dents, or over-pressure defects. After verifying multiple sets of candidate weights, the set with the highest defect identification accuracy is selected; for example, when the calibration results are 60% and 40%, it indicates that the direct impact of shrinkage supply and demand differences on defect formation is higher than the pressure transfer influence. The two types of anomaly state values ​​are multiplied by their corresponding weights and then summed to obtain the comprehensive shrinkage anomaly degree of the region. Based on the comprehensive anomaly degree and the deviation direction of the shrinkage supply and demand matching parameters, it is determined whether to enhance feeding, maintain parameters, or weaken feeding. The degree exceeding the corresponding judgment threshold is converted into an adjustment amount for pressure holding, pressure holding time, or pressure holding switching time, as the regional shrinkage compensation correction amount. The above data is associated with and stored with status markers to obtain the shrinkage associated process data.

[0040] Scenario 2: After the target toy injection molding part has completed the pressure holding and shrinkage compensation, the injection molding machine mold cooling system cools and shapes the part in the cavity. In cases where curved shells, buckles, thin-walled edges and local reinforcement structures may warp, deform or fluctuate in size due to inconsistent cooling rates, it is necessary to identify whether the mold cooling stage meets the requirements for stable shaping.

[0041] Cooling process data corresponding to the mold cooling stage is extracted from the injection molding process data. This data includes mold temperature data, cooling medium temperature data, cooling medium flow rate data, cooling duration data, and cooling execution status data. The mold cooling stage is the period from the end of the pressure holding and shrinkage compensation phase to the opening and demolding phase, during which the mold temperature controller or cooling medium circulation device removes heat from the mold and the target toy injection molded part through the mold cooling water circuit. Mold temperature data is collected by temperature sensors located on the fixed mold side, moving mold side, or near critical molding areas; cooling medium temperature data is collected by temperature sensors at the inlet and outlet of the cooling pipes; cooling medium flow rate data is collected by flow meters in the cooling circuit; the cooling duration is obtained by subtracting the cooling start and end times recorded by the injection molding machine controller; and the cooling execution status data is read by the mold temperature controller or injection molding machine controller, including the operating status of the cooling pump, valves, and cooling circuit. Data is extracted according to the current injection cycle number and the start and end times of the cooling stage, and then sorted by acquisition time to obtain the cooling process data.

[0042] Based on the continuous changes in cooling process data, cooling state characterization parameters are determined. Specifically, the mold temperature deviation is obtained by subtracting the set temperature from the actual mold temperature at each data collection moment; the actual temperature difference is obtained by subtracting the inlet temperature from the outlet temperature of the cooling medium, and then compared with the preset temperature difference to obtain the heat transfer temperature difference deviation; the flow rate deviation is obtained by subtracting the set flow rate from the actual flow rate; and the duration for which each deviation continuously exceeds the allowable range is also recorded. The set temperature, preset temperature difference, and set flow rate are all determined based on historical cooling data corresponding to qualified samples under the same material and mold conditions.

[0043] To avoid incompatibility due to differences in temperature, flow rate, and time units, the temperature deviation, heat transfer temperature difference deviation, and flow rate deviation were each divided by their respective upper limits of allowable deviation to obtain dimensionless abnormal state values. The duration of each deviation was then divided by the current cooling duration to obtain a persistence influence coefficient. Each abnormal state value was then multiplied by its corresponding persistence influence coefficient to obtain corrected temperature, heat transfer, and flow rate abnormal values. The weights for handling these three types of abnormal values ​​were calibrated using historical data from qualified samples and samples with warping, deformation, or dimensional deviations, selecting weight combinations that could accurately distinguish between normal and abnormal cooling. The corrected three types of abnormal values ​​were then multiplied by their respective weights and summed to obtain the cooling state characterization parameters.

[0044] When the cooling status characterization parameters of the current cooling cycle are within the allowable range, do not exceed the allowable upper limit at the end of the cooling cycle, and the cooling pump, valve, and cooling circuit complete the predetermined execution process, the corresponding cooling process data is determined as valid cooling process data; otherwise, it is marked as abnormal cooling process data. Finally, a mold cooling stage identifier is added to the valid cooling process data, and the current injection cycle number, injection machine number, mold number, stage start and end time, mold temperature data, cooling medium temperature data, cooling medium flow rate data, cooling duration data, cooling execution status data, and cooling status characterization parameters are jointly associated and stored to obtain cooling-related process data associated with the mold cooling stage.

[0045] S2 includes:

[0046] On one hand, to address the potential shrinkage, indentation, and dimensional deviations in the target toy injection molded parts during the pressure-holding and shrinkage compensation stage in Scenario 1 due to insufficient local material replenishment or uneven pressure transfer, pressure-holding and shrinkage compensation detection data corresponding to the current injection molding cycle is obtained. This data includes one or more of the following: surface shrinkage location, indentation depth, dimensional detection location, and dimensional deviation value. Specifically, the surface shrinkage location and indentation depth can be obtained using an industrial camera located at the demolding inspection station in conjunction with a 3D contour detection device. The industrial camera identifies abnormal areas on the part surface that are lower than the standard appearance surface, and the 3D contour detection device measures the depth difference between the lowest point of the abnormal area and the standard surface. The dimensional detection location and dimensional deviation value can be obtained using a dimensional measurement device; the dimensional deviation value is the difference between the actual measured dimension and the corresponding design dimension in the 3D design model.

[0047] Based on the spatial coordinates of the surface shrinkage location and dimensional inspection location in the 3D design model, the corresponding depression depth and dimensional deviation values ​​are assigned to the near-gate thick-walled area, the far-gate thin-walled area, or the intermediate transition area. When the coordinates fall within the boundary of a certain area, the corresponding inspection result is recorded in that area; when the inspection location is at the boundary of an area, the inspection result is assigned to the nearest area according to its distance from the center of each area. When the depression depth of each depression within a certain shrinkage-related area does not exceed the upper limit of the allowable deviation for depression depth, and the dimensional deviation value does not exceed the upper limit of the allowable deviation for dimensional deviation, the area is recorded as a normal shrinkage-holding area, and a normal shrinkage-holding mark is generated; when any inspection value exceeds the corresponding upper limit of the allowable deviation, the area is recorded as an optimized shrinkage-holding area, and an abnormal shrinkage-holding mark is generated. The upper limits of the allowable deviation for depression depth and the upper limits of the allowable deviation for dimensional deviation are predetermined based on the design tolerances, appearance quality requirements, and inspection results of multiple qualified samples of the target toy injection molded part.

[0048] The shrinkage-related process data, regional shrinkage compensation correction amount, holding pressure shrinkage detection data, and corresponding status markers for the same shrinkage-related region are associated and combined to form shrinkage process quality association samples. The shrinkage process data includes holding pressure, holding pressure duration, regional material replenishment volume, regional cavity pressure response, and holding pressure switching status; the regional shrinkage compensation correction amount indicates the degree to which the material replenishment effect needs to be strengthened or weakened relative to the normal shrinkage state; and the status markers serve as training labels. Samples from multiple injection molding cycles are input into a pre-set machine learning model, such as a random forest model or a gradient boosting tree model. Supervised training is performed using the shrinkage process data and regional shrinkage compensation correction amount as input, and the holding pressure shrinkage normal label or holding pressure shrinkage abnormal label as the output label, to obtain a holding pressure shrinkage optimization model for predicting the holding pressure shrinkage abnormality risk corresponding to each shrinkage-related region.

[0049] On the other hand, for the target toy injection molded parts that experience warping, contour deformation, and dimensional deviations during the mold cooling stage corresponding to Scenario 2 due to uneven heat dissipation, insufficient cooling, or abnormal operation of the cooling medium, cooling and shaping inspection data corresponding to the current injection molding cycle is obtained. The cooling and shaping inspection data includes one or more of the following: warpage height, contour offset distance, and post-molding dimensional deviation value. Specifically, the warpage height can be measured by a 3D contour detection device at the demolding inspection station, measuring the height difference between the highest deviation position on the part surface and the standard molding surface; the contour offset distance can be determined by acquiring the part contour image using an industrial camera and comparing it with the corresponding projected view of the 3D design model; and the post-molding dimensional deviation value can be obtained by acquiring the actual dimension using a dimensional measurement device and subtracting it from the design dimension.

[0050] A normal cooling and setting indicator is generated when the warpage height, contour offset distance, and post-molding dimensional deviation values ​​collected during the current injection molding cycle all do not exceed the corresponding upper limit of the allowable deviation. A normal cooling and setting indicator is generated when any of the detected values ​​exceeds the corresponding upper limit of the allowable deviation. The upper limits of each allowable deviation are predetermined based on the design tolerances, assembly requirements, appearance quality requirements, and test results of qualified samples of the target toy injection molded part.

[0051] Cooling-related process data, cooling state characterization parameters, cooling settling detection data, and status markers corresponding to the same injection molding cycle are correlated and combined to obtain cooling process quality-related samples. The cooling-related process data includes mold temperature, cooling medium temperature, cooling medium flow rate, cooling duration, and cooling execution status; the status markers serve as training labels. Samples from multiple injection molding cycles are input into a random forest model or a gradient boosting tree model for supervised training to obtain a cooling settling optimization model used to output cooling settling anomaly risks.

[0052] It is important to understand that the aforementioned preset machine learning models can be random forest classification models or gradient boosting tree classification models, and pressure holding and shrinkage compensation optimization models and cooling and shaping optimization models are constructed respectively. During construction, process data, detection data, and status markers are matched one-to-one according to the injection cycle number and region number. Invalid samples with missing data or abnormal execution status are deleted, and the samples are divided into training sample sets and validation sample sets. For the pressure holding and shrinkage compensation optimization model, the inputs are pressure holding pressure, pressure holding duration, region replenishment volume, cavity pressure response, pressure holding switching status, and region shrinkage compensation correction amount. The output is the abnormal risk value of shrinkage, depression, or dimensional deviation in each shrinkage-related region and the corresponding normal or abnormal category. For the cooling and shaping optimization model, the inputs are mold temperature, cooling medium temperature, cooling medium flow rate, cooling duration, cooling execution status, and cooling status characterization parameters. The output is the abnormal risk value of warpage, contour deformation, or dimensional deviation in the current injection cycle and the corresponding category. During training, normal and abnormal labels are used as supervision labels. The classification correspondence between input parameters and quality status is established through training samples, and the prediction accuracy is tested using validation samples. When the accuracy meets the preset requirements, the model parameters are saved to form the corresponding process optimization model.

[0053] S3 includes:

[0054] On one hand, to address the potential for localized shrinkage, depressions, or dimensional deviations in the holding and compensation stage of the current injection molding cycle for target toy injection molded parts, real-time compensation process data corresponding to the current injection molding cycle is collected. This real-time compensation process data includes the actual holding pressure, holding duration, and holding switching status read by the injection molding machine controller; the screw forward distance collected by the screw position encoder; and the pressure response data at each detection position collected by the mold cavity pressure sensor. Following the data processing method corresponding to the compensation process quality correlation sample, the pressure sensor positions are mapped to the near-gate thick-walled area, the far-gate thin-walled area, or the intermediate transition area. The effective cross-sectional area of ​​the screw is multiplied by the screw forward distance during the holding period to obtain the total compensation volume. The volume of each region is multiplied by the ratio of the region's average wall thickness to the standard wall thickness, and then multiplied by a factor minus the ratio of the gate transmission distance to the maximum effective compensation distance to obtain the region allocation value. The total compensation volume is multiplied by the proportion of this region allocation value to the sum of all region allocation values ​​to obtain the corresponding region compensation volume.

[0055] Further calculations are performed on the pressure holding and transfer efficiency parameters, the supply and demand matching parameters for shrinkage compensation, and the regional shrinkage compensation correction amount for each region. These parameters form the model input data, which is then input into the pressure holding and shrinkage compensation optimization model to obtain the pressure holding and shrinkage compensation anomaly risk value for each related region. The pressure holding and shrinkage compensation anomaly risk value is a value between zero and one from the model output, representing the probability of shrinkage, indentation, or dimensional deviation occurring in the corresponding region. The higher the value, the higher the anomaly risk.

[0056] When the abnormal risk value of the holding pressure compensation corresponding to any compensation-related region exceeds the preset compensation risk threshold, the corresponding region is identified as the current holding pressure compensation optimization region. The preset compensation risk threshold is determined based on the distribution of risk values ​​corresponding to qualified and abnormal regions in the verification sample, and the risk value that can accurately distinguish between the two types of regions is selected as the judgment boundary.

[0057] For the current pressure-holding and shrinkage compensation optimization area, the adjustment direction is determined based on the shrinkage supply and demand matching parameters: when the area's compensation volume is less than the shrinkage volume requirement, it is recommended to enhance the compensation effect; when the area's compensation volume exceeds the allowable compensation range, it is recommended to weaken the compensation effect. The recommended adjustment range is determined based on the difference between the abnormal pressure-holding and shrinkage compensation risk value and the preset shrinkage compensation risk threshold, as well as the correction degree corresponding to the area's shrinkage compensation correction amount. The shrinkage compensation control parameters include holding pressure, holding duration, holding switching time, and / or holding switching position; to enhance the compensation effect, increase the holding pressure, extend the holding duration, or delay the holding switching; to weaken the compensation effect, decrease the holding pressure, shorten the holding duration, or advance the holding switching. Each adjustment amount shall not exceed the maximum allowable adjustment amount determined by the target injection molding machine's allowable adjustment range and qualified trial production data.

[0058] When the abnormal risk values ​​of the holding pressure and shrinkage corresponding to each shrinkage-related area do not exceed the preset shrinkage risk threshold, the current holding pressure and shrinkage state is determined to be normal. The holding pressure, holding duration, and holding switching parameters corresponding to the current injection cycle remain unchanged, and a shrinkage control parameter hold result is generated. The shrinkage control parameter hold result is a parameter control record, including the injection cycle number, holding pressure and shrinkage stage identifier, risk value of each area, preset shrinkage risk threshold, current holding pressure, current holding duration, current holding pressure switching time, current holding pressure switching position, and the "Hold Current Parameters" instruction. This record is used to record that no shrinkage parameter correction is required in this cycle and outputs the control result of continuing to use the current set values ​​to the injection molding machine controller.

[0059] On the other hand, to address situations where target toy injection molded parts may experience warping, contour deformation, or dimensional deviations during the mold cooling phase of the current injection cycle due to insufficient heat dissipation, unstable heat exchange, or abnormal cooling medium supply, real-time cooling process data is collected. This real-time cooling process data includes the actual mold temperature collected by the mold temperature sensor, the cooling medium temperature collected by the inlet and outlet temperature sensors of the cooling pipes, the actual cooling medium flow rate collected by the flow meter, and the cooling duration and cooling execution status read by the injection molding machine controller or mold temperature controller.

[0060] Based on the data processing method corresponding to the cooling process quality-related samples, the real-time cooling state characterization parameters are calculated. Specifically, the absolute value of the difference between the actual mold temperature and the set temperature is divided by the upper limit of the mold temperature allowable deviation to obtain the temperature anomaly value; the actual temperature difference between the outlet and the inlet is obtained by subtracting the actual temperature difference from the preset temperature difference, and the absolute value is then divided by the upper limit of the temperature difference allowable deviation to obtain the heat transfer anomaly value; the absolute value of the difference between the actual flow rate and the set flow rate is divided by the upper limit of the flow rate allowable deviation to obtain the flow rate anomaly value. The continuous over-limit time of each of the three types of deviations is divided by the current cooling duration to obtain the corresponding continuous influence coefficient; then, the temperature anomaly value, heat transfer anomaly value, and flow rate anomaly value are multiplied by their respective continuous influence coefficients to obtain the continuously corrected anomaly values. The three continuously corrected anomaly values ​​are multiplied by the influence weights obtained based on the historical qualified cycle and abnormal cycle calibration, and then the three results are added together to obtain the real-time cooling state characterization parameters. By inputting real-time cooling process data and real-time cooling status characterization parameters into the cooling and shaping optimization model, the abnormal risk value of cooling and shaping corresponding to the current injection molding cycle is obtained. This risk value is used to indicate the degree of possibility that the current part will warp, deflect its contour, or have dimensional deviations after molding.

[0061] When the abnormal cooling and setting risk value exceeds the preset cooling risk threshold, the current injection molding cycle is designated as the cooling and setting optimization cycle. The preset cooling risk threshold is determined based on the verification results of the risk values ​​output by the corresponding models of qualified and abnormal cooling cycles. The recommended adjustment direction is determined based on the abnormal directions of temperature deviation, heat exchange temperature difference deviation, and flow rate deviation in the real-time cooling status characterization parameters, and the recommended adjustment range is determined based on the degree to which the abnormal cooling and setting risk value exceeds the threshold. Cooling and setting control parameters include mold temperature, cooling medium temperature, cooling medium flow rate, and / or cooling duration. When the mold temperature is too high or heat exchange is insufficient, it is recommended to reduce the cooling medium temperature, increase the cooling medium flow rate, or extend the cooling duration. When the cooling effect is too strong or the cooling time exceeds a reasonable range, it is recommended to appropriately increase the cooling medium temperature, reduce the flow rate, or shorten the cooling duration.

[0062] When the cooling and solidification anomaly risk value does not exceed the preset cooling risk threshold, the cooling and solidification state corresponding to the current injection molding cycle is determined to be normal. The current mold temperature, cooling medium temperature, cooling medium flow rate, and cooling duration remain unchanged, and a cooling and solidification control parameter hold result is generated. This result is a data record containing the injection molding cycle number, mold cooling stage identifier, cooling and solidification anomaly risk value, preset cooling risk threshold, current mold temperature setpoint, cooling medium temperature setpoint, cooling medium flow rate setpoint, cooling duration setpoint, and the instruction to hold the current parameters. It is used to indicate that the current cooling process meets the stable solidification requirements, and the mold temperature controller or injection molding machine controller continues to execute the current cooling parameters.

[0063] S4 includes:

[0064] For the process control procedures corresponding to Scenario 1 or Scenario 2, based on the recommended adjustment direction and magnitude, the shrinkage compensation control parameters corresponding to the current pressure holding and shrinkage compensation optimization area, or the cooling and shaping control parameters corresponding to the current cooling and shaping optimization cycle, are updated. Updating the shrinkage compensation control parameters includes increasing, decreasing, advancing, or delaying one or more of the following: pressure holding, pressure holding duration, pressure holding switching time, and pressure holding switching position. Updating the cooling and shaping control parameters includes increasing or decreasing one or more of the following: mold temperature, cooling medium temperature, cooling medium flow rate, and cooling duration. The updated parameter values, adjustment objects, adjustment direction, and adjustment magnitude are used to generate parameter execution instructions, which are then sent to the injection molding machine controller or mold temperature controller in the next injection molding cycle. This prompts and controls the corresponding equipment to execute pressure holding and shrinkage compensation control or cooling and shaping control according to the updated parameters.

[0065] Collect real-time data on the feeding process after implementing pressure holding and feeding control, and form model input data according to the processing methods of regional feeding volume, pressure holding transfer efficiency parameters, and feeding supply and demand matching parameters. Input the data into the pressure holding and feeding optimization model to obtain the pressure holding and feeding anomaly risk value after the update control. Alternatively, collect real-time cooling process data after implementing cooling and shaping control, calculate the corresponding real-time cooling state characterization parameters, and input the data into the cooling and shaping optimization model to obtain the cooling and shaping anomaly risk value after the update control.

[0066] When the updated pressure-holding and feeding anomaly risk value does not exceed the preset feeding risk threshold, or the updated cooling-setting anomaly risk value does not exceed the preset cooling risk threshold, the current process optimization state is determined to meet the convergence condition. The corresponding updated feeding control parameter or cooling-setting control parameter is marked as a converged parameter and output as the target process optimization result. Both the preset feeding risk threshold and the preset cooling risk threshold are predetermined based on the distinction between the risk values ​​corresponding to normal and abnormal states in the validation sample.

[0067] If the abnormal risk value after the control update still exceeds the corresponding preset risk threshold, the current process optimization status is updated to a state requiring further optimization. Specifically, the current optimization round, the current injection molding cycle number, the executed control parameters, the abnormal risk value after the control update, and the change between the abnormal risk value and the previous round are recorded. When the abnormal risk value decreases, the current control parameters are used as the adjustment benchmark for the next round, and adjustments are continued along the original recommended direction based on the degree to which the risk threshold is still exceeded. When the abnormal risk value does not decrease or increases, the current adjustment direction is canceled or the adjustment range is reduced, the recommended adjustment direction and recommended adjustment range for the corresponding control parameters are redefined, the parameter execution instruction for the next round is generated, and the next round of process optimization is executed.

[0068] It should be noted that Scenario 1 and Scenario 2 can be executed as independent embodiments. Scenario 1 is suitable for optimizing shrinkage, depressions, or local dimensional deviations occurring during the pressure holding and shrinkage compensation stage. In this case, only shrinkage process data and pressure holding and shrinkage compensation detection data are collected, and the shrinkage compensation control parameters are output by the pressure holding and shrinkage compensation optimization model. Scenario 2 is suitable for optimizing warpage, contour deformation, or post-molding dimensional deviations occurring during the mold cooling stage. In this case, only cooling process data and cooling and shaping detection data are collected, and the cooling and shaping optimization model outputs cooling and shaping control parameters.

[0069] In another embodiment, the pressure holding and shrinkage compensation stage and the mold cooling stage are sequentially connected within the same injection molding cycle and jointly affect the molding quality of the target toy injection molded part. For example, insufficient pressure holding and material compensation may lead to local shrinkage defects, while uneven cooling may further aggravate dimensional deviations or warping deformation. In this case, shrinkage compensation-related process data and cooling-related process data are generated separately and associated according to the same injection molding cycle number. When the abnormal risk output by either model exceeds the corresponding risk threshold, the parameters of the corresponding stage are adjusted. When abnormal risks exist in both stages, adjustment instructions for shrinkage compensation control parameters and cooling and shaping control parameters are generated separately, and joint optimization is performed in the next injection molding cycle according to the execution order of the pressure holding and shrinkage compensation stage first and the mold cooling stage second, until the abnormal risk values ​​corresponding to both stages do not exceed their respective risk thresholds.

[0070] This invention provides a machine learning-based toy injection molding process optimization system, such as... Figure 2 The diagram shown illustrates the structure of a machine learning-based toy injection molding process optimization system. This system may include:

[0071] The process data attribution and processing module is used to acquire the injection molding process data generated during the holding pressure and shrinkage stage and / or the mold cooling stage of the target toy injection molding part. Combined with the 3D design model, mold gate position and stage start and end marks, the injection molding process data is processed to assign molding areas, and process data associated with different molding areas and corresponding holding pressure and shrinkage stage and / or mold cooling stage are obtained.

[0072] The process quality modeling module is used to acquire quality inspection data such as surface shrinkage, indentation depth, dimensional deviation, warpage height and / or contour offset distance corresponding to the current injection molding cycle, and associate and combine the quality inspection results with the process data of the corresponding stage and the corresponding molding area to generate process quality association samples with normal or abnormal labels; the samples are input into the machine learning model for supervised training to generate pressure holding and shrinkage compensation optimization model and / or cooling and shaping optimization model.

[0073] The parameter recommendation module is used to collect real-time shrinkage process data corresponding to the holding and shrinkage stage and / or real-time cooling process data corresponding to the mold cooling stage within the current injection cycle. It forms model input data according to the data processing method corresponding to the training samples and calls the corresponding process optimization model to output abnormal risk values. When the abnormal risk value exceeds the corresponding risk threshold, it determines the molding area or cooling cycle that needs to be optimized and outputs the recommended adjustment direction and recommended adjustment range of the shrinkage control parameters and / or cooling and shaping control parameters.

[0074] The optimization status judgment module is used to update the holding pressure, holding duration, holding switching status, mold temperature, cooling medium temperature, cooling medium flow rate and / or cooling duration according to the recommended adjustment direction and recommended adjustment range, and control the next injection cycle to execute the updated parameters; it also acquires the corresponding process data again and calculates the abnormal risk value. When the abnormal risk value does not exceed the corresponding risk threshold, it determines that the process optimization status meets the convergence condition and outputs the target process optimization result; otherwise, it records the current optimization round and continues to execute the next round of parameter adjustment.

[0075] Further explanation is needed, such as Figure 3 The figure shows the shrinkage risk value variation curves for different molding regions under different holding pressures. The horizontal axis represents the holding pressure in MPa; the vertical axis represents the shrinkage risk value or volume shrinkage rate in %. The blue line represents the thick-walled connection region, the orange line represents the thin-walled appearance region, and the green line represents the region far from the gate. As the figure shows, as the holding pressure gradually increases from low to high, the shrinkage risk corresponding to all three molding regions shows a continuous downward trend, indicating that increasing the holding pressure can enhance the melt's shrinkage compensation ability and reduce the shrinkage risk caused by material cooling shrinkage. Among them, the thick-walled connection region has the highest initial risk and the most significant decrease, indicating that this region is more sensitive to changes in holding pressure; the thin-walled appearance region has a generally lower risk and a more gradual change; the region far from the gate is in between, indicating that its shrinkage compensation effect is affected by both the holding pressure and the melt transfer distance.

[0076] like Figure 4The figure shows the variation curves of insufficient feeding risk value or dimensional deviation in different molding regions under different holding pressure durations. The horizontal axis represents the holding pressure duration in seconds (s); the vertical axis represents the insufficient feeding risk value or dimensional deviation in millimeters (mm). The blue line represents thick-walled regions, the orange line represents thin-walled regions, and the green line represents corner regions. The figure also uses dashed lines to divide the insufficient feeding segment, the effective feeding convergence segment, and the stable interval. It can be seen that when the holding pressure duration is short, the dimensional deviation in each region is large, especially in the thick-walled region where the deviation is most obvious. As the holding pressure time increases, all three curves decrease rapidly, indicating that extending the holding pressure time can improve the insufficient feeding problem. Once the holding pressure time enters the stable interval, the curves tend to flatten out, and further extending the holding pressure time has limited effect on improving the dimensional deviation. This can be used as a basis for determining a reasonable holding pressure duration range.

[0077] like Figure 5 The figure shows the warpage risk value or warpage deformation of different forming areas under different cooling times. The horizontal axis represents cooling time in seconds (s), and the vertical axis represents the warpage risk value or warpage deformation in millimeters (mm). The blue broken line represents the corner area, the orange broken line represents the long-side thin-walled area, and the green broken line represents the thick-walled assembly area. The dashed lines divide the cooling process into a rapid shaping section, a heat release section, and a stable interval. As can be seen from the figure, the warpage deformation of all three forming areas decreases with increasing cooling time, decreasing rapidly in the early stage and gradually stabilizing later. The corner area initially exhibits the largest warpage, indicating a significant impact from structural transitions and uneven cooling shrinkage; the long-side thin-walled area is next; and the thick-walled assembly area has the lowest overall deformation. This figure reflects the influence of cooling time on the shaping stability of different areas and provides a basis for recommending cooling shaping control parameters.

[0078] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0079] This invention is described with reference to flowchart illustrations and / or block diagrams of systems, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0080] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0081] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

Claims

1. A machine learning-based optimization method for toy injection molding processes, characterized in that, The method includes: S1, perform molding area attribution processing on the injection process data generated by the target toy injection molded part in the target process control stage, and obtain process data associated with different molding areas and corresponding target process control stages. The target process control stage includes one or both of the pressure holding and shrinkage compensation stage and the mold cooling stage. S2, construct process quality correlation samples based on the process data, and input them into a preset machine learning model for training to generate a process optimization model; The process optimization model is used to predict the risk of state changes in different forming areas during the pressure holding and shrinkage compensation stage and / or the mold cooling stage. S3, collect the shrinkage process data corresponding to the pressure holding and shrinkage stage and / or the cooling process data corresponding to the mold cooling stage, synchronously call the process optimization model, and output the corresponding shrinkage control parameters and / or cooling and shaping control parameters after state change risk prediction. S4. Determine whether the current process optimization state meets the corresponding convergence condition. If yes, output the corresponding target process optimization result; otherwise, update the current process optimization state and execute the next round of process optimization.

2. The toy injection molding process optimization method based on machine learning as described in claim 1, characterized in that, S1 includes: Extract the shrinkage compensation process data corresponding to the pressure holding and shrinkage compensation stage from the injection molding process data. The shrinkage compensation process data includes pressure holding data, pressure holding duration data, screw stroke change data, cavity pressure response data, and pressure holding switching status data. Based on the three-dimensional design model of the target toy injection molded part, according to its geometric structural features and the relative positional relationship between each structural position and the mold gate, the target toy injection molded part is divided into multiple shrinkage-related regions, including the thick-walled region near the gate, the thin-walled region far from the gate, and the intermediate transition region. Based on the wall thickness, volume, and gate transfer distance of each feeding associated region, the feeding process data is associated with the corresponding feeding associated region to obtain the regional feeding process data. Each feeding associated region corresponds to one or more molding regions of the target toy injection molded part. For each feeding-related region, the ratio between the screw stroke change rate and the pressure response decay rate is corrected by combining the influence of the wall thickness and the influence of the pressure transmission distance in the corresponding region, so as to obtain the pressure holding and transmission efficiency parameter. Simultaneously, the estimated shrinkage volume of the corresponding shrinkage-related region and the estimated material replenishment volume are processed to obtain the shrinkage supply and demand matching parameters. Based on the comprehensive characterization results of the pressure holding and transmission efficiency parameters and the feeding supply and demand matching parameters, a regional feeding compensation correction amount is generated, and the feeding status is marked on the regional feeding process data to obtain feeding-related process data associated with the pressure holding and feeding stage and different feeding-related regions. The regional shrinkage compensation correction amount is used to characterize the degree of correction of the feeding effect of the corresponding shrinkage-related region under the current pressure-holding shrinkage state relative to the preset shrinkage matching state.

3. The toy injection molding process optimization method based on machine learning as described in claim 1, characterized in that, S1 further includes: Extract cooling process data corresponding to the mold cooling stage from the injection molding process data. The cooling process data includes mold temperature data, cooling medium temperature data, cooling medium flow rate data, cooling duration data, and cooling execution status data. Based on the continuous changes of the cooling process data during the mold cooling stage, cooling state characterization parameters are determined to characterize the stability of the corresponding cooling and shaping process of the target toy injection molded part. The cooling state characterization parameters are determined by weighting the deviation between the actual temperature of the mold and the set temperature, the deviation between the actual temperature difference between the inlet and outlet of the cooling medium and the preset temperature difference, the deviation between the actual flow rate of the cooling medium and the set flow rate, and the duration of each deviation. Based on the cooling state characterization parameters, effective cooling process data that meets the preset cooling state requirements is determined from the cooling process data, and the effective cooling process data is associated with the stage identifier corresponding to the mold cooling stage to obtain cooling associated process data associated with the mold cooling stage.

4. The toy injection molding process optimization method based on machine learning as described in claim 2, characterized in that, S2 includes: Obtain the pressure holding and shrinkage compensation detection data corresponding to the current injection molding cycle. The pressure holding and shrinkage compensation detection data includes at least the surface shrinkage location, indentation depth, size detection location, and size deviation value of the target toy injection molded part. Based on the coordinates of the surface shrinkage location and / or size detection location in the corresponding three-dimensional design model of the target toy injection molded part, the indentation depth and / or size deviation value are mapped to the shrinkage compensation associated area in the thick-walled area near the gate, the thin-walled area far from the gate, or the intermediate transition area. When the indentation depth and the size deviation value do not exceed their respective upper limits of allowable deviation, the corresponding shrinkage-related area is recorded as a normal shrinkage-holding area and a normal shrinkage-holding mark is generated; otherwise, the corresponding shrinkage-related area is recorded as an optimized shrinkage-holding area and an abnormal shrinkage-holding mark is generated. The data on the associated process of the same associated shrinkage area, the regional shrinkage compensation correction amount, the pressure holding shrinkage detection data, and the normal or abnormal pressure holding shrinkage markers are associated and combined to obtain the associated shrinkage process quality samples. These samples are then input into a preset machine learning model for training to obtain an optimized pressure holding shrinkage model for predicting the abnormal risk of pressure holding shrinkage corresponding to each associated shrinkage area.

5. The toy injection molding process optimization method based on machine learning as described in claim 3, characterized in that, S2 further includes: Obtain the cooling and shaping test data corresponding to the current injection molding cycle. The cooling and shaping test data includes the warpage height, contour offset distance, and post-molding dimensional deviation value of the target toy injection molded part. When the warpage height, the contour offset distance and the post-molding dimensional deviation value do not exceed their respective upper limits of allowable deviation, a normal cooling and shaping mark corresponding to the current injection molding cycle is generated; otherwise, an abnormal cooling and shaping mark corresponding to the current injection molding cycle is generated. The cooling-related process data, cooling state characterization parameters, cooling shaping detection data, and cooling shaping normal or abnormal markings corresponding to the same injection molding cycle are associated and combined to obtain cooling process quality associated samples. These samples are then input into a preset machine learning model for training to obtain a cooling shaping optimization model for predicting the cooling shaping abnormal risk corresponding to the current injection molding cycle.

6. The toy injection molding process optimization method based on machine learning as described in claim 4, characterized in that, S3 includes: Collect real-time shrinkage process data corresponding to the current injection cycle, and input the processed real-time shrinkage process data into the pressure holding and shrinkage optimization model according to the data processing method corresponding to the shrinkage process quality related sample, to obtain the pressure holding and shrinkage abnormal risk value corresponding to each shrinkage related area; When the abnormal risk value of pressure holding and pressure holding corresponding to any pressure holding and pressure holding associated region exceeds the preset pressure holding and pressure holding risk threshold, the corresponding pressure holding and pressure holding associated region is determined as the current pressure holding and pressure holding optimized region. Based on the abnormal risk value of pressure holding and pressure holding corresponding to the current pressure holding and pressure holding optimized region and the region pressure holding compensation correction amount, the corresponding pressure holding control parameters and the corresponding recommended adjustment direction and recommended adjustment range are determined. The compensation control parameters include holding pressure, holding duration, holding switching time and / or holding switching position; When the abnormal risk values ​​of pressure holding and shrinkage corresponding to each shrinkage-related area do not exceed the preset shrinkage risk threshold, the current pressure holding and shrinkage state is determined to be normal, the shrinkage control parameters corresponding to the current injection cycle remain unchanged, and the shrinkage control parameter holding result is generated.

7. The toy injection molding process optimization method based on machine learning as described in claim 5, characterized in that, S3 further includes: Collect real-time cooling process data corresponding to the current injection molding cycle, process the real-time cooling process data according to the data processing method corresponding to the cooling process quality related sample, determine the corresponding real-time cooling state characterization parameters, and input them into the cooling shaping optimization model to obtain the cooling shaping abnormal risk value corresponding to the current injection molding cycle. When the abnormal risk value of cooling and shaping exceeds the preset cooling risk threshold, the current injection molding cycle is determined as the cooling and shaping optimization cycle, and based on the abnormal risk value of cooling and shaping and the real-time cooling status characterization parameters, the corresponding cooling and shaping control parameters and the corresponding recommended adjustment direction and recommended adjustment range are determined. The cooling and shaping control parameters include mold temperature, cooling medium temperature, cooling medium flow rate and / or cooling duration; When the abnormal risk value of cooling and shaping does not exceed the preset cooling risk threshold, the cooling and shaping state corresponding to the current injection molding cycle is determined as the normal state, the cooling and shaping control parameters corresponding to the current injection molding cycle remain unchanged, and a cooling and shaping control parameter holding result is generated.

8. The toy injection molding process optimization method based on machine learning as described in claim 6 or 7, characterized in that, S4 includes: Based on the recommended adjustment direction and recommended adjustment range, update the shrinkage compensation control parameters or cooling and shaping control parameters corresponding to the current pressure holding and shrinkage compensation optimization area, and in the next injection molding cycle, control the injection molding machine to execute the corresponding pressure holding and shrinkage compensation control or cooling and shaping control according to the updated shrinkage compensation control parameters or cooling and shaping control parameters; Collect real-time data on the compensation process after the pressure holding and compensation control is executed, or real-time cooling status characterization parameters after the cooling and shaping control is executed, and input the corresponding pressure holding and compensation optimization model or cooling and shaping optimization model to obtain the pressure holding and compensation abnormal risk value or cooling and shaping abnormal risk value after the control is updated.

9. The toy injection molding process optimization method based on machine learning as described in claim 8, characterized in that, S4 further includes: When the updated pressure holding and feeding abnormal risk value or cooling and shaping abnormal risk value does not exceed the corresponding preset risk threshold, the current process optimization state is determined to meet the corresponding convergence condition, and the updated feeding control parameters or cooling and shaping control parameters are output as the target process optimization result. When the abnormal risk value of pressure holding and feeding or the abnormal risk value of cooling and shaping after the update control exceeds the corresponding preset risk threshold, the recommended adjustment direction and recommended adjustment range of the corresponding control parameters are re-determined according to the corresponding abnormal risk value after the update control, the current process optimization status is updated, and the next round of process optimization is executed.

10. A machine learning-based toy injection molding process optimization system, characterized in that, The system includes: The process data attribution processing module is used to perform molding area attribution processing on the injection process data generated by the target toy injection molded part in the target process control stage, and obtain process data associated with different molding areas and corresponding target process control stages. The target process control stage includes one or both of the pressure holding and shrinkage compensation stage and the mold cooling stage. The process quality modeling module is used to construct process quality correlation samples based on process data, and input them into a preset machine learning model for training to generate a process optimization model; The process optimization model is used to predict the risk of state changes in different forming areas during the pressure holding and shrinkage compensation stage and / or the mold cooling stage. The parameter recommendation module is used to collect the corresponding shrinkage process data and / or the cooling process data corresponding to the mold cooling stage, and synchronously call the process optimization model. After the state change risk prediction, it outputs the corresponding shrinkage control parameters and / or cooling shaping control parameters. The optimization status judgment module is used to determine whether the current process optimization status meets the corresponding convergence condition. If so, the corresponding target process optimization result is output; otherwise, the current process optimization status is updated and the next round of process optimization is executed.

Images