A method and system for monitoring the injection molding process of a cosmetic plastic mold

Abstract

This invention discloses a method and system for monitoring the injection molding process of cosmetic plastic molds, comprising the following steps: acquiring multi-dimensional monitoring data corresponding to the injection molding process of cosmetic plastic molds; classifying and identifying the data sources of each sub-data within the multi-dimensional monitoring data to generate a multi-source monitoring data coding table; performing injection molding characteristic analysis and injection molding status assessment on the multi-dimensional monitoring data based on the multi-source monitoring data coding table to generate a comprehensive evaluation value of the injection molding status; setting and judging corresponding monitoring and early warning thresholds based on the comprehensive evaluation value of the injection molding status, locating the corresponding abnormal data sources, matching the abnormal data sources with a preset fault type library to output preliminary results of abnormal faults including evaluation values, abnormal location, and fault judgment, and generating an injection molding monitoring abnormal location report by combining the timestamp and location code of the abnormal data sources. This invention can significantly improve the efficiency of fault handling in injection molding monitoring.

Description

Technical Field

[0001] This invention relates to the field of injection molding monitoring technology, specifically, to a method and system for monitoring the injection molding process of cosmetic plastic molds. Background Technology

[0002] Cosmetic plastic molds, due to their complex cavity structures (often containing intricate textures and thin-walled designs), are susceptible to defects during injection molding, such as mold temperature fluctuations, uneven injection pressure, and poor melt flow. These defects can lead to flash, short runs, and surface scratches, directly impacting the product's market competitiveness. Cosmetic plastic mold injection molding monitoring technology, with its unique advantages of real-time multi-dimensional process data collection and dynamic evaluation of processing status, is gradually replacing the traditional "post-production sampling inspection" model, becoming a crucial technical means to ensure the production quality of cosmetic plastic parts and reduce defective product losses. Currently, existing methods mostly rely on collecting basic injection molding machine parameters (such as injection pressure and temperature), combining them with fixed thresholds to determine the presence of anomalies, and then using experience to locate the fault. However, existing methods are hampered by the lack of integration of mold cavity visual data (such as the empty cavity state and real-time molded appearance), neglect of the correspondence between data collection locations (such as the feed inlet and cavity sidewalls) and fault types, and the reliance solely on process parameters, which can easily miss surface defects and cavity matching deviations. Furthermore, they cannot accurately pinpoint the source of abnormal data, thus reducing the efficiency of fault handling in cosmetic plastic mold injection molding monitoring. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides a method for monitoring the injection molding process of cosmetic plastic molds, comprising the following steps: Step S1: Obtain multi-dimensional monitoring data corresponding to the injection molding process of cosmetic plastic molds. The multi-dimensional monitoring data includes mold cavity temperature data, injection pressure data, melt flow rate data, mold opening and closing displacement data, as well as mold cavity no-load visual data and injection part real-time forming visual data collected by industrial cameras. Step S2: Classify and identify the data sources of each sub-data in the multi-dimensional monitoring data, and assign a unique data source code to each sub-data according to the data acquisition equipment type and the corresponding acquisition position including the mold inlet, cavity sidewall and injection molding machine nozzle to generate a multi-source monitoring data coding table; perform injection molding process feature analysis on the multi-dimensional monitoring data based on the multi-source monitoring data coding table to obtain the multi-dimensional feature parameter set of injection molding process monitoring; Step S3: Generate the corresponding injection process stability coefficient and mold cavity matching probability based on the multi-dimensional feature parameter set of injection molding process monitoring, and evaluate the injection molding process status of the cosmetic plastic mold based on the injection process stability coefficient and mold cavity matching probability to generate a comprehensive evaluation value of the injection molding process status. Step S4: Set the corresponding monitoring and early warning threshold based on the comprehensive evaluation value of the injection molding process status and make a judgment. For the injection molding process where the comprehensive evaluation value of the injection molding process status is lower than the monitoring and early warning threshold, locate the corresponding abnormal data source. According to the abnormal data source, match the preset fault type library to output the preliminary abnormal fault results containing the evaluation value, abnormal location and fault judgment. Combine the timestamp and location code of the abnormal data source to generate an injection molding process monitoring abnormal location report.

[0004] Furthermore, the present invention also provides an injection molding process monitoring system for cosmetic plastic molds, including a computer-readable storage medium, a processor, a communication interface, and a computer program stored on the computer-readable storage medium and executable on the processor, for performing the injection molding process monitoring method for cosmetic plastic molds as described above.

[0005] The beneficial effects of this application are as follows: By acquiring multi-dimensional monitoring data, which includes not only process parameters such as mold cavity temperature and injection pressure, but also visual data of the cavity under no-load and real-time molding of the injection molded part, this approach breaks through the limitations of traditional methods that only consider parameters and not the shape. Compared to traditional single-parameter acquisition, this step's multi-dimensional data covers the entire chain of "process parameters - mold status - product shape," providing complete data support for subsequent accurate analysis of injection status and fault location, avoiding omissions due to missing data, and improving the comprehensiveness of monitoring from the source. Secondly, through data classification, identification, and feature analysis, and by assigning unique codes according to equipment type and acquisition location (feed port, cavity sidewall, etc.), each type of data can be traced back to its specific source. The multi-dimensional feature parameter set generated by feature analysis based on the coding table not only includes single parameter values ​​but also integrates the correlation features of "location-parameter-vision," avoiding the one-sidedness of traditional methods that only analyze isolated parameters. Compared to traditional unclassified and chaotic data, the coding table and feature parameter set in this step make the data more readable and relevant, providing a structured basis for subsequent status assessment and anomaly localization. This avoids the low analysis efficiency caused by disordered data and enhances the utilization value of monitoring data. Then, by generating process stability coefficients, cavity matching probabilities, and comprehensive evaluation values, the problem of traditional monitoring relying solely on fixed thresholds and failing to quantify injection molding status is solved. The process stability coefficient quantifies the fluctuation degree of parameters such as pressure and temperature, avoiding the traditional crude judgment that "parameters within the threshold are normal"—for example, even if parameters are within the range, large fluctuations can still lead to product defects. The cavity matching probability combines no-load visual data and mold opening and closing displacement data to assess mold fit; traditional methods completely ignore this type of matching problem. The comprehensive evaluation value integrates two indicators to quantify the overall injection molding status, avoiding the subjectivity of traditional status judgments based solely on experience. Even if a single parameter does not exceed the threshold, it can still provide early warning of potential risks, avoid faults in advance, and improve the predictive capability of monitoring. Finally, by judging early warning thresholds, locating anomalies, and generating reports, the problems of traditional monitoring being unable to accurately locate the source of anomalies and having low efficiency in fault handling are solved. Setting early warning thresholds based on comprehensive evaluation values ​​avoids the one-size-fits-all approach of traditional fixed thresholds. When locating the source of abnormal data, it directly traces to the specific equipment and location using a unique code. Traditional methods lack coding support and require checking each item one by one. The output of preliminary results and location reports for abnormal faults integrates evaluation values, abnormal locations, fault types, and timestamps, providing clear guidance for maintenance personnel and avoiding the blindness of traditional troubleshooting based solely on experience. This allows maintenance personnel to quickly pinpoint the fault point and clarify the fault type, significantly shortening the handling time and thus significantly improving the fault handling efficiency of cosmetic plastic mold injection molding monitoring. Attached Figure Description

[0006] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of the injection molding process monitoring method for cosmetic plastic molds in this embodiment; Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S1. Detailed Implementation

[0007] The following drawings disclose several embodiments of the present invention. For clarity, many practical details will be described in the following description. However, it should be understood that these practical details are not intended to limit the invention. That is, in some embodiments of the invention, these practical details are not essential. Furthermore, for the sake of simplicity, some conventional structures and components will be shown in the drawings in a simple schematic manner.

[0008] To further understand the invention's content, features, and effects, the following embodiments are provided, and detailed descriptions are given below in conjunction with the accompanying drawings: Reference Figure 1 , Figure 1 This is a flowchart illustrating the injection molding process monitoring method for cosmetic plastic molds in this embodiment. The injection molding process monitoring method for cosmetic plastic molds in this embodiment includes the following steps: Step S1: Obtain multi-dimensional monitoring data corresponding to the injection molding process of cosmetic plastic molds. The multi-dimensional monitoring data includes mold cavity temperature data, injection pressure data, melt flow rate data, mold opening and closing displacement data, as well as mold cavity no-load visual data and injection part real-time forming visual data collected by industrial cameras. In this embodiment of the invention, for the injection mold of the cosmetic lipstick tube shell (6 cavities), temperature sensors (28 in total) are installed on the sidewalls of the cavities (4 points per cavity), the feed inlet, and the cooling water path, with a sampling frequency of 12Hz, to obtain mold cavity temperature data (e.g., cavity tube body temperature 88℃, feed inlet temperature 92℃); pressure sensors (15 in total) are installed at the injection molding machine nozzle, cavity inlet, and feed channel, with a sampling frequency of 10Hz, to obtain injection pressure data (e.g., nozzle pressure 55MPa, cavity pressure 48MPa); a melt flow rate detector is installed at the injection molding machine nozzle, with a sampling frequency of 5Hz, to obtain melt flow rate data (e.g., 3.8g / min, 4.2g / min); and two displacement sensors are installed at the guide pillars of the mold opening and closing mechanism, with a sampling frequency of 8Hz, to obtain mold opening and closing displacement data (e.g., mold opening 145mm, mold closing 0.2mm). Industrial cameras (2048×2048 pixels resolution, 30fps) are deployed 30cm in front of each cavity and 25cm to the side. Before injection molding, visual data of the mold cavity under no-load condition is collected (showing that the inner wall of the cavity is smooth and free of impurities). During the injection molding process, real-time visual data of the molded part is collected (showing the forming state of the lipstick tube shell opening and tube body). All data are stored continuously according to the collection time to form multi-dimensional monitoring data.

[0009] Step S2: Classify and identify the data sources of each sub-data in the multi-dimensional monitoring data, and assign a unique data source code to each sub-data according to the data acquisition equipment type and the corresponding acquisition position including the mold inlet, cavity sidewall and injection molding machine nozzle to generate a multi-source monitoring data coding table; perform injection molding process feature analysis on the multi-dimensional monitoring data based on the multi-source monitoring data coding table to obtain the multi-dimensional feature parameter set of injection molding process monitoring; In this embodiment of the invention, by classifying and identifying the data sources of each sub-data within the multi-dimensional monitoring data, a unique data source code is assigned according to the rule of "equipment type-acquisition location-parameter type": the cavity sidewall temperature data acquired by the temperature sensor is coded as "T-001" and the feed port temperature data is coded as "T-002"; the nozzle pressure data acquired by the pressure sensor is coded as "P-001" and the cavity pressure data is coded as "P-002"; the data acquired by the melt flow rate detector is coded as "F-001"; the mold opening displacement data acquired by the displacement sensor is coded as "D-001" and the mold closing displacement data is coded as "D-002"; the no-load visual data acquired by the industrial camera is coded as "V-001" and the real-time molding visual data is coded as "V-002", thus generating a multi-source monitoring data coding table. Based on the encoding table, the following sub-data are extracted: for the “T-001” data, the temperature fluctuation characteristic value of 0.013 is calculated using a sliding window (5 seconds); for the “P-001” and “P-002” data, the pressure peak value and the average holding pressure value are extracted, and the pressure deviation characteristic value of 0.0344 is calculated; for the “F-001” data, the unit time change rate is calculated, and the melt flow rate fluctuation coefficient of 0.25 is generated; for the “D-001” and “D-002” data, the deviation from the preset displacement is calculated, and the mold opening and closing displacement deviation value of 1.1mm is generated; for the “V-001” and “V-002” data, the feature point matching rate of 0.834 and the average contour deviation of 0.52mm are extracted, and integrated to obtain the multi-dimensional feature parameter set for injection molding monitoring [0.013, 0.0344, 0.25, 1.1, 0.834, 0.52].

[0010] Step S3: Generate the corresponding injection process stability coefficient and mold cavity matching probability based on the multi-dimensional feature parameter set of injection molding process monitoring, and evaluate the injection molding process status of the cosmetic plastic mold based on the injection process stability coefficient and mold cavity matching probability to generate a comprehensive evaluation value of the injection molding process status. In this embodiment of the invention, based on a multi-dimensional feature parameter set for injection molding monitoring, the process influence weights of each parameter (temperature 0.25, pressure 0.35, melt rate 0.28, displacement 0.12) are determined through regression analysis of historical fault data. After parameter normalization, the parameters are weighted and fused using Pearson correlation coefficients (temperature and pressure correlation coefficient 0.62, temperature and melt rate correlation coefficient 0.58) to obtain a process fluctuation correlation fusion parameter of 0.1891. The standard deviation of the fusion parameter for 10 consecutive processing cycles is calculated to be 0.0032. Combined with the preset upper limit of the allowable range of process fluctuation of 0.01, a time-series stability evaluation factor of 0.68 is generated. The cumulative effect value of time-series decay is simulated using a Markov chain to calculate 0.21, and the injection molding process stability coefficient is calculated as 0.1891 × (0.68 ÷ 1.21) × 100 ≈ 10.63. For the cavity contour feature vector [0.834, 0.52] in the multidimensional feature parameter set, after aligning it with the standard contour vector space (rotation 0.5°, translation -0.2mm), it is substituted into the Dirichlet distribution model to solve for the probability distribution peak value of 0.92, thus obtaining the mold cavity matching probability of 0.92. With a stability coefficient weight of 0.6 and a matching probability weight of 0.4, the comprehensive evaluation value of the injection molding processing state is calculated as (10.63 ÷ 15 × 0.6) + (0.92 × 0.4) × 100 ≈ 79.3.

[0011] Step S4: Set the corresponding monitoring and early warning threshold based on the comprehensive evaluation value of the injection molding process status and make a judgment. For the injection molding process where the comprehensive evaluation value of the injection molding process status is lower than the monitoring and early warning threshold, locate the corresponding abnormal data source. According to the abnormal data source, match the preset fault type library to output the preliminary abnormal fault results containing the evaluation value, abnormal location and fault judgment. Combine the timestamp and location code of the abnormal data source to generate an injection molding process monitoring abnormal location report.

[0012] In this embodiment of the invention, based on the comprehensive evaluation value of injection molding processing status of 79.3, combined with 3000 sets of historical normal data (evaluation values ​​75-95) and 800 sets of abnormal data (evaluation values ​​40-68), a first warning threshold of 78.75 (lower limit of normal data 75 × 1.05) and a second warning threshold of 70.04 (upper limit of abnormal data 68 × 1.03) are set. If the evaluation value of a certain batch is 75 (between the two thresholds), it is determined to be a first-level warning state; the multi-source monitoring data encoding table is retrieved, and the abnormal data source codes "T-001" (temperature fluctuation characteristic value 0.018 exceeding auxiliary threshold 0.015) and "F-001" (melt rate fluctuation coefficient 0.32 exceeding auxiliary threshold 0.3) are located, and the timestamps T0+120s and T0+180s and the location codes cavity sidewall and nozzle are recorded. A pre-defined fault type library (abnormal temperature fluctuation, excessive pressure deviation, etc.) and a characteristic parameter range table were established. The matching degree between the abnormal characteristic parameter set [0.018, 0.038, 0.32, 1.2] and each fault range was calculated. The overall matching degree for abnormal temperature fluctuation was 0.935 (the highest). Preliminary results of the abnormal fault were generated: "Fault type: abnormal temperature fluctuation; evaluation value: 75; source of abnormality: cavity sidewall temperature sensor (T-001), time T0+120s". Combining the abnormal data trend (temperature fluctuation increased from 0.013 to 0.018) with the lipstick tube defect (slight deformation of the tube opening), an injection molding process monitoring anomaly location report was generated, and handling suggestions were given (adjusting the cooling water flow rate).

[0013] Furthermore, refer to Figure 2 , Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S1 is provided below. In this embodiment, step S1 includes the following steps: Step S11: Deploy multiple types of data acquisition equipment to install temperature sensors and pressure sensors on the inner wall of the cavity, feeding channel and cooling water channel of the cosmetic plastic mold, install a melt flow rate detector at the injection molding machine nozzle, install a displacement sensor at the mold opening and closing mechanism, and deploy industrial vision cameras at the front and side of the mold cavity. In this embodiment of the invention, for a plastic mold (containing 6 cavities) for the outer shell of a cosmetic lipstick tube, one temperature sensor and one pressure sensor are installed at the tube opening, the middle of the tube body, and the bottom of each cavity, for a total of 36 sensors (18 temperature sensors and 18 pressure sensors). The sensor probes are embedded in the inner wall of the cavity and flush with the curved surface of the inner wall. One temperature sensor and one pressure sensor are installed at the main runner inlet, the branch runner, and the gate corresponding to each cavity of the mold feeding channel, for a total of 18 sensors (9 temperature sensors and 9 pressure sensors). One temperature sensor is installed at the main inlet, the main outlet, and the cooling ring corresponding to each cavity of the cooling water circuit, for a total of 8 temperature sensors. A melt flow rate detector is installed at the nozzle outlet of the injection molding machine via a threaded connection. The detector probe extends 2.5mm into the nozzle to ensure direct contact with the molten plastic. Two displacement sensors are installed at the connection between the moving mold guide pillar and the fixed mold guide sleeve of the mold opening and closing mechanism to monitor the axial and radial displacement of the mold opening and closing, respectively. An industrial vision camera (2048×2048 pixels resolution, 30fps) is deployed 28cm directly in front of each mold cavity, and another industrial vision camera (1920×1080 pixels resolution, 25fps) is deployed 22cm to the side. The camera lens is adjusted to ensure complete imaging of the inner wall condition from the cavity opening to the bottom of the tube.

[0014] Step S12: The sensor data acquisition module collects the analog signal output by the temperature sensor during the injection molding process of the cosmetic plastic mold in real time, converts the analog signal into a digital signal and then filters it to generate mold cavity temperature data; at the same time, it receives the output signals from the pressure sensor, melt flow rate detector and displacement sensor, processes them to generate injection pressure data, melt flow rate data and mold opening and closing displacement data respectively. In this embodiment of the invention, the 0-10V analog signals output by each temperature sensor are collected in real time by the sensor data acquisition module (sampling frequency 12Hz). The analog signals are converted into digital signals by a 16-bit A / D converter, and then the digital signals are filtered by a 4th-order Chebyshev low-pass filter (cutoff frequency 4Hz) to filter out vibration noise and power grid interference generated by the injection molding machine, thereby generating mold cavity temperature data for each sensor location (e.g., temperature in the middle of the cavity tube is 88℃, and temperature at the gate is 95℃). Simultaneously, the system acquires 4-20mA analog signals from pressure sensors, performs 12-bit A / D conversion and 3-point moving average filtering to generate injection pressure data (e.g., main runner pressure 55MPa, cavity tube bottom pressure 48MPa); receives frequency signals from melt flow rate detectors, calculates melt flow rate data (e.g., 4.2g / min) based on the conversion relationship of 0.08g / min per 1Hz; and acquires RS485 digital signals from displacement sensors, generates mold opening and closing displacement data (e.g., mold opening displacement 145mm, mold closing displacement 0mm) through displacement-pulse conversion formula (1 pulse corresponds to 0.008mm displacement).

[0015] Step S13: Use an industrial vision camera to collect images of the mold cavity in its unloaded state before injection molding and real-time forming images during the injection molding process. Perform white balance correction and distortion correction on the collected unloaded state images and real-time forming images to eliminate the effects of lighting changes and lens distortion, and generate unloaded visual data of the mold cavity and real-time forming visual data of the injection molded part. In this embodiment of the invention, before injection molding is completed and before injection, an industrial vision camera is controlled to continuously capture 5 images of the mold cavity in an unloaded state (simultaneous capture by front and side cameras); during the injection molding process, there are three stages: injection stage (T0-T0+1.5s), holding pressure stage (T0+1.5s-T0+3s), and cooling stage (T0+3s-T0+8s), with 8 real-time molding images captured in each stage. White balance correction was performed on all acquired images. Using a standard neutral gray card as a reference, the gains of the R, G, and B channels were adjusted (R gain 1.15, G gain 1.0, B gain 1.12) to eliminate color cast caused by differences in light intensity in different areas of the workshop. A checkerboard-based camera calibration method was used to capture 15 checkerboard images in different poses. Lens distortion parameters were calculated (radial distortion k1=-0.028, k2=0.009, tangential distortion p1=0.0018, p2=-0.0009). Based on the parameters, the pixel coordinates of the images were geometrically corrected to eliminate the influence of lens distortion, generating visual data of the mold cavity under no-load condition and real-time visual data of the lipstick tube shell forming.

[0016] Step S14: Integrate the mold cavity temperature data, injection pressure data, melt flow rate data, mold opening and closing displacement data, and mold cavity no-load visual data with the real-time molding visual data of the injection molded part to form multi-dimensional monitoring data corresponding to the injection molding process of cosmetic plastic molds.

[0017] In this embodiment of the invention, the mold cavity temperature data, injection pressure data, melt flow rate data, and mold opening and closing displacement data are bound to the timestamp at a sampling frequency of 12Hz, with the injection valve opening time of the injection molding machine as the reference timestamp (T0). Each data point is marked with a specific acquisition time (e.g., T0+0.083s, T0+0.166s). The no-load visual data is marked as the pre-injection timestamp (T0-3s, T0-2.5s), and the real-time molding visual data is marked with timestamps according to the stages (injection stage T0+0.5s, holding pressure stage T0+2s, cooling stage T0+5s). Through the data integration module, all data are linked and integrated in the format of "timestamp-temperature data list-pressure data list-melt rate value-displacement value-frontal visual image path-side visual image path" to form multi-dimensional monitoring data of cosmetic lipstick tube shell plastic mold injection molding process containing 2400 sets of time-synchronized data. Each set of data covers 27 temperature values, 27 pressure values, 1 melt rate value, 2 displacement values ​​and 2 corresponding visual images.

[0018] Furthermore, the injection molding process feature analysis based on the multi-source monitoring data encoding table described in step S2 includes the following steps: Step S201: Collect the process parameter setting data corresponding to the injection molding machine, including preset injection temperature, preset injection pressure, preset holding pressure and preset mold opening and closing displacement, and generate the preset process parameter table of the injection molding machine; In this embodiment of the invention, process parameter settings for the injection molding of cosmetic lipstick tube shells are collected from the injection molding machine control system. The preset injection temperature is set to 210℃ (205℃ for the front section of the barrel, 210℃ for the middle section, and 200℃ for the rear section), the preset injection pressure is set to 60MPa, the preset holding pressure is set to 45MPa, and the preset mold opening / closing displacement is set to 145mm for mold opening and 0mm for mold closing. These parameters are organized in the format of "parameter type-setting value-corresponding station," such as "preset injection temperature-210℃-middle section of barrel" and "preset mold opening / closing displacement-145mm-mold opening station," generating a preset process parameter table for the injection molding machine. This ensures that each parameter is labeled with its specific application location and value.

[0019] Step S202: Based on the multi-source monitoring data encoding table, retrieve the monitoring data corresponding to each type of code. For the mold cavity temperature data, use a sliding window to extract the corresponding temperature value within a continuous time period. Calculate the ratio between the difference between the maximum and minimum temperature values ​​within the window and the preset injection temperature to generate instantaneous temperature fluctuation values. Then, take the average of the instantaneous temperature fluctuation values ​​corresponding to multiple windows to obtain the temperature fluctuation characteristic value. In this embodiment of the invention, the mold cavity temperature data (sampling frequency 12Hz, containing 1200 temperature values) corresponding to the code "T-001" is retrieved based on the multi-source monitoring data encoding table. A sliding window with a duration of 5 seconds (containing 60 temperature values) is used to capture temperature values ​​within a continuous time period. In the first window (0-5 seconds), the maximum temperature is 89℃ and the minimum temperature is 85℃, with a difference of 4℃. The ratio of this difference to the preset injection molding temperature of 210℃ is calculated as 4÷210≈0.019, generating an instantaneous temperature fluctuation value of 0.019 for this window. In the second window (5-10 seconds), the maximum temperature is 88℃ and the minimum temperature is 86℃, with a difference of 2℃. The ratio of this difference is 2÷210≈0.0095, generating an instantaneous fluctuation value of 0.0095. A total of 20 consecutive windows are captured, and the average value of the 20 instantaneous fluctuation values ​​is calculated as (0.019+0.0095+…+0.012)÷20≈0.013, resulting in a temperature fluctuation characteristic value of 0.013.

[0020] Step S203: Extract the pressure peak value corresponding to the injection stage from the injection pressure data, and calculate the difference between the pressure peak value and the preset injection pressure in the preset process parameter table of the injection molding machine. Then, the ratio of this difference to the preset injection pressure is used as the pressure peak deviation rate. At the same time, calculate the deviation rate between the mean value of the injection pressure data in the holding pressure stage and the preset holding pressure to generate the holding pressure deviation rate. The pressure peak deviation rate and the holding pressure deviation rate are weighted and summed to obtain the pressure deviation characteristic value. In this embodiment of the invention, injection pressure data for the injection molding stage (lasting 1.5 seconds) is extracted from multi-source monitoring data, resulting in 18 pressure values ​​(58MPa, 62MPa, 61MPa…63MPa). The peak pressure of 63MPa is selected. The difference of 3MPa between the peak pressure and the preset injection pressure of 60MPa is calculated. The ratio of this difference to the preset injection pressure is 3÷60=0.05, resulting in a peak pressure deviation rate of 0.05. Injection pressure data for the holding pressure stage (lasting 3 seconds) is extracted, resulting in 36 pressure values ​​(44MPa, 46MPa, 45MPa…47MPa). The average value is calculated as (44+46+45+…+47)÷36≈45.5MPa. The difference of 0.5MPa between the average holding pressure and the preset holding pressure of 45MPa is calculated. The ratio is 0.5÷45≈0.011, resulting in a holding pressure deviation rate of 0.011. With a weight of 0.6 for peak pressure deviation rate and 0.4 for holding pressure deviation rate, the weighted summation is 0.05×0.6+0.011×0.4=0.03+0.0044=0.0344, yielding a pressure deviation characteristic value of 0.0344.

[0021] Step S204: Extract the corresponding visual image feature points and cavity contours from the unloaded visual data of the mold cavity and the real-time molding visual data of the injection molded part, and calculate the feature point matching rate and the average contour deviation of the two images to generate the cavity contour feature vector. In this embodiment of the invention, the unloaded visual data of the mold cavity (2048×2048 pixel grayscale image) and the real-time molding visual data of the injection molded part (grayscale image of the same specification) are processed. The SIFT algorithm is used to extract key feature points respectively. 120 feature points are extracted from the unloaded image and 115 feature points are extracted from the real-time image. After generating feature descriptors, they are matched by the FLANN matching algorithm. 98 feature points are successfully matched. The feature point matching rate is calculated to be 98÷[(120+115)÷2]≈0.834. The Canny edge detection algorithm is used to extract the cavity contour of the two images. The pixel-physical coordinate transformation matrix is ​​established by combining the camera intrinsic parameters (focal length 1500 pixels) and extrinsic parameters (translation vector 300mm). 800 physical coordinates of the unloaded contour and 795 physical coordinates of the real-time contour are extracted. After alignment by the ICP algorithm, the Euclidean distance of each pair of corresponding points is calculated. The average value is 0.52mm. The cavity contour feature vector [0.834, 0.52] is generated.

[0022] Step S205: Calculate the rate of change of melt flow rate per unit time for melt flow rate data to generate melt flow rate fluctuation coefficient; calculate the deviation between actual mold opening and closing displacement and preset mold opening and closing displacement for mold opening and closing displacement data to generate mold opening and closing displacement deviation value. In this embodiment of the invention, melt flow rate data (sampling frequency 12Hz, a total of 120 data points, such as 3.8g / min, 4.2g / min, 4.0g / min…4.1g / min) is extracted from multi-source monitoring data. The rate of change of adjacent data points within a unit time (1 second) is calculated. For example, the rate of change in the first 1-2 seconds is (4.2-3.8)÷1=0.4g / (min·s), and the rate of change in the second 2-3 seconds is (4.0-4.2)÷1=-0.2g / (min·s). The absolute value of all the rates of change is taken and the mean is calculated as (0.4+0.2+…+0.3)÷119≈0.25g / (min·s), generating a melt flow rate fluctuation coefficient of 0.25. Extract the mold opening and closing displacement data. The actual mold opening displacement is 143mm and the actual mold closing displacement is 0.2mm. Calculate the deviation from the preset mold opening and closing displacements (145mm when opening and 0mm when closing). The mold opening deviation is 143-145=-2mm and the mold closing deviation is 0.2-0=0.2mm. Take the absolute value and average it (2+0.2)÷2=1.1mm. The resulting mold opening and closing displacement deviation value is 1.1mm.

[0023] Step S206: Summarize the temperature fluctuation characteristic value, pressure deviation characteristic value, cavity contour characteristic vector, melt flow rate fluctuation coefficient, and mold opening and closing displacement deviation value to obtain a multi-dimensional characteristic parameter set for injection molding monitoring.

[0024] In this embodiment of the invention, the temperature fluctuation characteristic value 0.013, pressure deviation characteristic value 0.0344, cavity contour feature vector [0.834, 0.52], melt flow rate fluctuation coefficient 0.25, and mold opening and closing displacement deviation value 1.1mm are summarized in the order of "temperature-pressure-contour vector-melt rate-displacement" and the numerical format is unified (retaining four decimal places or one decimal place, and the vector follows the original format) to form a multi-dimensional feature parameter set for injection molding monitoring: {temperature fluctuation characteristic value: 0.013, pressure deviation characteristic value: 0.0344, cavity contour feature vector: [0.834, 0.52], melt flow rate fluctuation coefficient: 0.25, mold opening and closing displacement deviation value: 1.1mm}, ensuring that each parameter corresponds to a specific calculation result and unit.

[0025] Furthermore, step S204 includes the following steps: The generated visual data of the mold cavity in no-load and the real-time molding visual data of the injection molded part are processed into grayscale to generate a grayscale image of the cavity in no-load and a grayscale image of the real-time molding of the injection molded part. Gaussian filtering is then used for noise reduction to generate a grayscale image of the filtered base and a grayscale image of the filtered real-time part. In this embodiment of the invention, the empty visual data of the cavity of the cosmetic lipstick tube shell mold (a 2048×2048 pixel color image captured by a front-facing camera) and the real-time molding visual data of the injection molded part (a color image of the same specification) are subjected to grayscale processing using a weighted average method. The grayscale value calculation formula is grayscale value = 0.299×R channel value + 0.587×G channel value + 0.114×B channel value. The red area of ​​the tube opening in the empty image (R=220, G=180, B=150) is converted to a grayscale value of 188, and the white area of ​​the tube body in the real-time molding image (R=240, G=240, B=240) is converted to a grayscale value of 240, thereby generating a cavity empty reference grayscale image and an injection molding real-time molding grayscale image. Noise reduction is performed on the two grayscale images using a 5×5 Gaussian filter template. The pixel values ​​of the images are weighted and averaged through convolution operations. The bright spot pixels (grayscale value 250) caused by lens noise in the unloaded reference grayscale image are smoothly reduced to 235, while the dark spot pixels (grayscale value 50) caused by injection molding reflection in the real-time forming grayscale image are increased to 70. The filtered reference grayscale image and the filtered real-time grayscale image are then generated.

[0026] Furthermore, key feature points are detected in the filtered baseline grayscale image and the filtered real-time grayscale image respectively, and the feature descriptor corresponding to each key feature point is calculated to generate the feature descriptor point set of the baseline image and the feature descriptor point set of the real-time image. In this embodiment of the invention, key feature points are detected in the filtered reference grayscale image and the filtered real-time grayscale image using the SIFT algorithm. 120 key feature points are extracted from the pipe opening edge, pipe bottom rounded corner, and cavity inner wall groove in the filtered reference grayscale image. Each feature point generates a 128-dimensional feature descriptor, which contains gradient information in eight directions within a 4×4 neighborhood around the point. For example, the descriptor for a feature point at the pipe opening edge is [0.12, 0.08, ..., 0.15] (a total of 128 values), generating a reference image feature descriptor point set. In the corresponding areas of the filtered real-time grayscale image (pipe opening edge, pipe bottom rounded corner, and inner wall of the pipe body), 115 key feature points are extracted, also generating 128-dimensional feature descriptors. For example, the descriptor for a feature point on the inner wall of the pipe body is [0.11, 0.09, ..., 0.14], generating a real-time image feature descriptor point set. This ensures that the feature descriptors of the two images have the same dimension and a numerical range of 0-1.

[0027] Furthermore, feature matching is performed on the feature description point set of the reference image and the feature description point set of the real-time image to count the number of successfully matched feature points, and the ratio of this number to the total number of feature points in the two visual images is used as the feature point matching rate. In this embodiment of the invention, the FLANN matching algorithm is used to perform feature matching between the feature descriptor point set of the reference image and the feature descriptor point set of the real-time image. A matching distance threshold of 0.6 is set. A successful match is determined when the Euclidean distance between the descriptor of a feature point in the real-time image and the descriptor of a feature point in the reference image is less than 0.6. The number of successfully matched feature points is counted. For example, out of 120 feature points in the reference image, 98 successfully match with feature points in the real-time image; out of 115 feature points in the real-time image, 98 successfully match. The feature point matching rate is calculated as: Number of successfully matched feature points ÷ (Total number of feature points in the reference image + Total number of feature points in the real-time image) × 2, i.e., 98 ÷ (120 + 115) × 2 ≈ 0.834, resulting in a feature point matching rate of 0.834.

[0028] Furthermore, edge extraction is performed on the filtered reference grayscale image and the filtered real-time grayscale image to obtain the reference cavity contour image and the real-time cavity contour image; the intrinsic and extrinsic parameters corresponding to the industrial vision camera are obtained and the mapping relationship between image pixel coordinates and actual physical coordinates of the mold cavity is established to generate a pixel-physical coordinate transformation matrix; In this embodiment of the invention, the Canny edge detection algorithm is used to extract edges from the filtered baseline grayscale image and the filtered real-time grayscale image. A high threshold of 80 and a low threshold of 40 are set. Pixels in the filtered baseline grayscale image with a grayscale value change rate greater than 80 are marked as strong edges, and pixels with a change rate between 40 and 80 and connected to the strong edges are marked as weak edges. The strong edges and weak edges are merged to obtain the baseline cavity contour image (clearly showing the complete contour of the tube opening, tube body, and tube bottom). Similarly, the filtered real-time grayscale image is processed to obtain the real-time cavity contour image (showing the formed contour of the lipstick tube shell). The intrinsic parameters (focal length f_x=1500 pixels, f_y=1500 pixels, principal point coordinates u_0=1024 pixels, v_0=1024 pixels) and extrinsic parameters (rotation matrix R=[[1,0,0],[0,1,0],[0,0,1]], translation vector T=[0,0,-300]mm) of the industrial vision camera are obtained by the checkerboard calibration method. The mapping relationship between image pixel coordinates (u,v) and actual physical coordinates (X,Y,Z) of the mold cavity is established. The pixel-physical coordinate transformation matrix is ​​generated by the perspective projection formula. The matrix dimension is 3×4, and each element is a fixed value calculated based on the intrinsic and extrinsic parameters.

[0029] Furthermore, the coordinates of the contour pixels in the reference cavity contour map and the real-time cavity contour map are extracted to obtain the reference contour pixel coordinate set and the real-time contour pixel coordinate set; the reference contour pixel coordinate set and the real-time contour pixel coordinate set are substituted into the pixel-physical coordinate transformation matrix to generate the reference contour physical coordinate set and the real-time contour physical coordinate set respectively. In this embodiment of the invention, a contour tracking algorithm is used to extract the coordinates of contour pixels in the baseline cavity contour map and the real-time cavity contour map. In the baseline cavity contour map, edge pixels are traversed in a clockwise direction, and 800 pixel coordinates are recorded, including the contour points at the tube opening (u=500, v=500), the contour points in the middle of the tube body (u=500, v=1000), and the contour points at the bottom of the tube (u=500, v=1500), to generate a baseline contour pixel coordinate set. In the real-time cavity contour map, the corresponding forming contour pixels are extracted, such as the forming contour points at the tube opening (u=502, v=501), the forming contour points in the tube body (u=501, v=1002), and the forming contour points at the bottom of the tube (u=500, v=1501), totaling 795 pixel coordinates, to generate a real-time contour pixel coordinate set. Substitute each pixel coordinate (u, v) in the baseline contour pixel coordinate set into the pixel-to-physical coordinate transformation matrix, and calculate the corresponding actual physical coordinates through matrix multiplication. For example, (u=500, v=500) is transformed into (X=20mm, Y=20mm, Z=0mm), generating the baseline contour physical coordinate set. Similarly, the real-time contour pixel coordinate set is transformed into the real-time contour physical coordinate set. For example, (u=502, v=501) is transformed into (X=20.5mm, Y=20.3mm, Z=0mm).

[0030] Furthermore, the physical coordinate set of the baseline contour and the physical coordinate set of the real-time contour are aligned with point clouds to determine the matching relationship of the corresponding contour points and generate a set of cavity contour point matching pairs. The Euclidean distance between each pair of corresponding points in the set of cavity contour point matching pairs is calculated, and the arithmetic mean of all Euclidean distances is used as the mean of contour deviation. The corresponding feature point matching rate and the mean of contour deviation are combined in sequence to generate the cavity contour feature vector.

[0031] In this embodiment of the invention, the ICP algorithm is used to align the point clouds of the reference contour physical coordinate set and the real-time contour physical coordinate set. The reference contour physical coordinate set is used as the target point cloud, and the real-time contour physical coordinate set is used as the source point cloud. By iteratively calculating the rotation matrix and translation vector, the average distance between the source and target point clouds is minimized. After five iterations, the average distance converges to 0.1 mm. The matching relationship of each pair of corresponding contour points is determined. For example, the reference contour (X=20mm, Y=20mm, Z=0mm) matches the real-time contour (X=20.5mm, Y=20.3mm, Z=0mm), generating a cavity contour point matching pair set containing 790 matching points. The Euclidean distance of each pair of matching points is calculated, such as the Euclidean distance of the above matching points = √[(20.5-20)]. 2 +(20.3-20) 2 +(0-0) 2The feature point matching rate is approximately 0.583 mm. The Euclidean distances of 790 matching points are statistically analyzed, and the arithmetic mean is calculated as (0.583 + 0.452 + ... + 0.611) ÷ 790 ≈ 0.52 mm, generating a mean contour deviation of 0.52 mm. The feature point matching rate of 0.834 and the mean contour deviation of 0.52 mm are combined in the order of "feature point matching rate - mean contour deviation" to generate the cavity contour feature vector [0.834, 0.52] of the cosmetic lipstick tube outer shell mold.

[0032] Furthermore, step S3 includes the following steps: Step S31: Based on the multi-dimensional feature parameter set of injection molding process monitoring and combined with regression analysis of historical fault data, determine the process influence weight coefficient corresponding to each parameter. Each parameter includes temperature fluctuation feature value, pressure deviation feature value, melt flow rate fluctuation coefficient, and mold opening and closing displacement deviation value to generate an injection molding process parameter-weight correspondence table; normalize the multi-dimensional feature parameter set of injection molding process monitoring to generate a normalized process parameter set for injection molding process. In this embodiment of the invention, by injection molding the outer shell of a cosmetic lipstick tube, based on a multi-dimensional feature parameter set for injection molding process monitoring (temperature fluctuation feature value 0.013, pressure deviation feature value 0.0344, melt flow rate fluctuation coefficient 0.25, mold opening and closing displacement deviation value 1.1mm), and combined with 500 sets of historical fault data, regression analysis was performed: the failure rate was statistically analyzed to be 25% due to temperature fluctuation, 35% due to pressure deviation, 28% due to melt flow rate fluctuation, and 12% due to displacement deviation. The process influence weight coefficients were assigned according to the failure rate: temperature 0.25, pressure 0.35, melt flow rate 0.28, and displacement 0.12, generating an injection molding process parameter-weight correspondence table. The multidimensional feature parameter set is normalized. The normalized temperature fluctuation feature value is 0.013 ÷ 0.05 (maximum parameter value) = 0.26, the normalized pressure deviation feature value is 0.0344 ÷ 0.1 = 0.344, the normalized melt rate fluctuation coefficient is 0.25 ÷ 0.5 = 0.5, and the normalized displacement deviation value is 1.1 ÷ 5 = 0.22, generating the normalized process parameter set for injection molding [0.26, 0.344, 0.5, 0.22].

[0033] Step S32: Calculate the Pearson correlation coefficients between parameters based on the normalized process parameter set for injection molding to generate a parameter correlation matrix. Then, based on the parameter correlation matrix, combine the process influence weight coefficients with the normalized process parameter set for injection molding to perform weighted fusion to obtain the process fluctuation correlation fusion parameters. Retrieve the process fluctuation fusion parameters corresponding to multiple consecutive processing cycles and calculate the standard deviation of the parameters in the time dimension to generate the time series standard deviation of process fluctuation. Compare the time series standard deviation of process fluctuation with the preset allowable range of process fluctuation to generate a time series stability evaluation factor. Specifically, the time series stability evaluation factor = 1 - (time series standard deviation of process fluctuation / upper limit of allowable range of process fluctuation). In this embodiment of the invention, based on the normalized process parameter set for injection molding [0.26, 0.344, 0.5, 0.22], the Pearson correlation coefficients between each parameter are calculated: temperature and pressure correlation coefficient 0.62, temperature and melt rate correlation coefficient 0.58, temperature and displacement correlation coefficient 0.31, pressure and melt rate correlation coefficient 0.75, pressure and displacement correlation coefficient 0.43, and melt rate and displacement correlation coefficient 0.28, generating a 4×4 parameter correlation matrix. Combined with the process influence weight coefficients [0.25, 0.35, 0.28, 0.12], the following calculations are performed using the formula: "Process fluctuation correlation fusion parameter = Σ(normalized parameter × weight × mean correlation coefficient of corresponding row)": Temperature term = 0.26 × 0.25 × (0.62 + 0.58 + 0.31) ÷ 3 ≈ 0.0327, Pressure term = 0.344 × 0.35 × (0.62 + 0.75 + ... 0.43)÷3≈0.0722, melt rate term=0.5×0.28×(0.58+0.75+0.28)÷3≈0.0752, displacement term=0.22×0.12×(0.31+0.43+0.28)÷3≈0.009, summing gives the fusion parameter≈0.0327+0.0722+0.0752+0.009≈0.1891. Retrieving the fusion parameters for 10 consecutive processing cycles [0.1891,0.192,0.185,0.195,0.188,0.19,0.186,0.193,0.187,0.191], the standard deviation is calculated to be ≈0.0032. The upper limit of the allowable range of process fluctuation is preset to 0.01. The time stability evaluation factor of 0.68 is generated by "time stability evaluation factor = 1 - (0.0032 ÷ 0.01) = 0.68".

[0034] Step S33: Based on the time-series stability assessment factor, evaluate the time-series impact decay of the process fluctuation correlation fusion parameters to obtain the injection molding process stability coefficient; In this embodiment of the invention, based on the time-series stability evaluation factor of 0.68, combined with the process fluctuation correlation fusion parameter of 0.1891 and the time-series decay cumulative effect value of 0.21 (obtained through Markov chain simulation), the injection molding process stability coefficient is calculated as follows: 0.1891×(0.68÷1.21)×100≈10.63. Thus, the injection molding process stability coefficient for cosmetic lipstick tubes is 10.63 (a coefficient ≤ 15 is considered a stable process).

[0035] Step S34: Estimate the corresponding mold cavity matching probability based on the cavity contour feature vector in the multi-dimensional feature parameter set of injection molding process monitoring; In this embodiment of the invention, the cavity contour feature vector [0.834, 0.52] (feature point matching rate 0.834, mean contour deviation 0.52 mm) is extracted from the multi-dimensional feature parameter set of injection molding process monitoring. This vector is then spatially aligned with the standard cavity contour feature reference vector (rotation 0.5°, translation -0.2 mm). The Euclidean distance after alignment is calculated to be 0.152, and the spatial similarity index is approximately 6.579. The variance of the contour deviation distribution is calculated to be 0.00087, and the contour uniformity index is approximately 1149.43. Combined with the process-contour correlation correction coefficient of 0.0217, the corrected feature matching probability is obtained to be 0.0879. After standardizing the three indices, they are input into the Dirichlet distribution model, and the peak value of the probability distribution is 0.92, thus determining the mold cavity matching probability to be 0.92.

[0036] Step S35: Based on the injection molding process stability coefficient and the mold cavity matching probability, evaluate the injection molding status of the cosmetic plastic mold to generate a comprehensive evaluation value of the injection molding status.

[0037] In this embodiment of the invention, based on the injection molding process stability coefficient of 10.63 (weight 0.6) and the mold cavity matching probability of 0.92 (weight 0.4), the comprehensive evaluation value of injection molding processing status is calculated as follows: (10.63 ÷ 15 × 0.6) ≈ 0.425, (0.92 × 0.4) = 0.368. The sum is 0.425 + 0.368 = 0.793. Multiplying by 100 gives 79.3, and the comprehensive evaluation value of injection molding processing status of cosmetic lipstick tube mold is 79.3 (an evaluation value ≥ 70 is considered as a qualified processing status).

[0038] Furthermore, step S33 includes the following steps: Extract process fluctuation correlation fusion parameters corresponding to N consecutive injection molding cycles to construct a time-series fusion parameter sequence. At the same time, retrieve the equipment running time and mold preheating time corresponding to each injection molding cycle, and calculate the equipment-mold time-series coordination coefficient by the ratio of equipment running time to mold preheating time. In this embodiment of the invention, by injection molding the outer shell of a cosmetic lipstick tube, process fluctuation correlation and fusion parameters are extracted corresponding to 10 consecutive injection molding cycles (each cycle is 45 seconds). These parameters are obtained by weighting temperature fluctuation characteristic values ​​(e.g., 0.013, 0.015), pressure deviation characteristic values ​​(e.g., 0.0344, 0.032), and melt flow rate fluctuation coefficients (e.g., 0.25, 0.23) with weights of 0.3, 0.4, and 0.3, respectively. The fusion parameters for the 10 cycles are obtained as follows: 0.028, 0.029, 0.031, 0.027, 0.030, 0.026, 0.028, 0.032, 0.029, 0.027. The temporal fusion parameter sequence [0.028, 0.029, 0.031, 0.027, 0.030, 0.026, 0.028, 0.032, 0.029, 0.027] is constructed. The equipment runtime (cumulative time from equipment startup to the end of the cycle, in sequence: 45s, 90s, 135s, 180s, 225s, 270s, 315s, 360s, 405s, 450s) and mold preheating time (fixed at 300s) corresponding to each cycle are retrieved. The coordination coefficients for the 10 cycles are calculated as follows: "Equipment-Mold Timing Coordination Coefficient = Equipment Runtime ÷ Mold Preheating Time". The coordination coefficients for the 10 cycles are 0.15, 0.3, 0.45, 0.6, 0.75, 0.9, 1.05, 1.2, 1.35, 1.5, respectively, generating the equipment-mold timing coordination coefficient sequence [0.15, 0.3, 0.45, 0.6, 0.75, 0.9, 1.05, 1.2, 1.35, 1.5].

[0039] Furthermore, based on the time-series fusion parameter sequence, the time-series fluctuation characteristics of the parameters are calculated. A sliding window is used to extract the corresponding parameter subsequences within different time periods, and the coefficient of variation of each parameter subsequence is calculated to generate a set of time-series fluctuation coefficients of variation. The set of time-series fluctuation coefficients of variation is corrected based on the equipment-mold time-series coordination coefficient, so that the time-series decay correlation coefficient of variation can be obtained by the corrected coefficient of variation = coefficient of variation × (1 - equipment-mold time-series coordination coefficient). In this embodiment of the invention, based on the time-series fusion parameter sequence [0.028,0.029,0.031,0.027,0.030,0.026,0.028,0.032,0.029,0.027], a sliding window with a duration of 3 periods is used to extract parameter subsequences, resulting in a total of 8 subsequences: [0.028,0.029,0.031], [0.029,0.027], etc. [0.029,0.031,0.027]、[0.031,0.027,0.030]、[0.027,0.030,0.026]、[0.030,0.026,0.028]、[0.026,0.028,0.032]、[0.028,0.032,0.029]、[0.032,0.029,0.027]. Calculate the coefficient of variation (coefficient of variation = standard deviation ÷ mean) for each subsequence. For example, the first subsequence has a mean of 0.0293 and a standard deviation of 0.0015, so the coefficient of variation is 0.0015 ÷ 0.0293 ≈ 0.0512; the second subsequence has a mean of 0.029 and a standard deviation of 0.002, so the coefficient of variation is 0.002 ÷ 0.029 ≈ 0.069, generating a set of time series fluctuation coefficients of variation [0.0512, 0.069, 0.067, 0.072, 0.068, 0.081, 0.065, 0.083]. Based on the equipment-mold time-series synergy coefficient (taking the average synergy coefficient of each window corresponding to the period, such as the average of 0.3 for the first window), the corrected coefficient of variation is calculated by "corrected coefficient of variation = coefficient of variation × (1 - synergy coefficient)". The corrected coefficient of the first window is 0.0512 × (1 - 0.3) = 0.0358, and the remaining windows are calculated in sequence as 0.0414, 0.0368, 0.0288, 0.0238, 0.0162, 0.0195, and 0.0166, generating a time-series decay-related coefficient of variation set [0.0358, 0.0414, 0.0368, 0.0288, 0.0238, 0.0162, 0.0195, 0.0166].

[0040] Furthermore, the time-series decay correlation coefficient is used as the simulation input, and a process decay factor is introduced. Specifically, the trend characteristics of parameter fluctuation decay over time are extracted based on historical data, and the decay path of parameter fluctuation in the time dimension is simulated through Markov chain state transition to generate a time-series decay path matrix, where each element represents the decay probability of parameter fluctuation at a certain moment to subsequent moments. In this embodiment of the invention, a process attenuation factor is introduced by using the time-series decay correlation coefficient set [0.0358, 0.0414, 0.0368, 0.0288, 0.0238, 0.0162, 0.0195, 0.0166] as the simulation input. Based on parameter fluctuation data from 300 historical lipstick tube injection molding cycles, the trend characteristics of parameter fluctuation decay over time are extracted: 5% decay rate for the first 5 cycles, 8% decay rate for cycles 5-10, and 12% decay rate for cycles 10-15. The decay path of parameter fluctuations in the time dimension is simulated by Markov chain state transitions. The 10 injection molding cycles are divided into 5 time nodes (T1-T5, each node contains 2 cycles), and a 5x5 time decay path matrix is ​​constructed. The matrix elements represent the decay probability of parameter fluctuations at a certain time moment being passed to subsequent time moments. For example, the probability of T1→T2 is 0.95 (corresponding to a decay rate of 5% in the first 5 cycles), the probability of T1→T3 is 0.92 (combined with the decay rate of 5-10 cycles), the probability of T2→T3 is 0.95, the probability of T2→T4 is 0.92, the probability of T3→T4 is 0.88 (corresponding to a decay rate of 12% in 10-15 cycles), the probability of T3→T5 is 0.85, and the probability of T4→T5 is 0.88. Other elements without a transmission path are set to 0, generating a complete time decay path matrix.

[0041] Furthermore, the cumulative effect value of time-series decay is calculated based on the time-series decay path matrix; In this embodiment of the invention, based on the time-series decay path matrix (5 rows and 5 columns, time nodes T1-T5 correspond to lipstick tube injection molding cycles 1-2, 3-4, 5-6, 7-8, 9-10), the in-degree and out-degree of each time node are extracted: T1 in-degree 0, out-degree 2 (T2, T3); T2 in-degree 1 (T1), out-degree 2 (T3, T4); T3 in-degree 2 (T1, T2), out-degree 2 (T4, T5); T4 in-degree 2 (T2, T3), out-degree 1 (T5); T5 in-degree 2 (T3, T4), out-degree 0. The impact of nodes on transmission efficiency is calculated as follows: T1=2.0, T2=2.0, T3=1.0, T4=0.5, T5=0.0. The decay probabilities at each time step in the matrix (T1=0.95, T2=0.95, T3=0.88, T4=0.88, T5=0.85) are coupled with efficiency to obtain the optimized probability set [1.14, 1.14, 0.968, 0.924, 0.85]. A time-decay weighted network is constructed, and the betweenness centrality of nodes [T1=0.1, T2=0.3, T3=0.4, T4=0.2, T5=0.0] is calculated. The cumulative value of a single path is obtained by multiplying the edge weights together, and the contribution value is calculated by combining the betweenness mean. Finally, the cumulative effect value of time-decay is calculated by geometric mean, which is approximately 0.21.

[0042] Furthermore, the process fluctuation correlation fusion parameter is attenuated and calculated based on the ratio between the time-series stability assessment factor and the 1+ time-series decay cumulative effect value to obtain the injection molding process stability coefficient.

[0043] In this embodiment of the invention, the ratio is calculated by using the time-series stability assessment factor based on cosmetic lipstick tube injection molding (calibrated to 0.85 using historical qualified cycle data, representing the benchmark factor during stable production) and "1 + time-series decay cumulative effect value" (1 + 0.21 = 1.21), i.e., 0.85 ÷ 1.21 ≈ 0.702. This ratio is multiplied by the mean of the process fluctuation-related fusion parameters (mean of fusion parameters over 10 cycles, 0.0285), and calculated using "injection molding process stability coefficient = ratio × mean of fusion parameters × 100", 0.702 × 0.0285 × 100 ≈ 1.99, resulting in an injection molding process stability coefficient of 1.99 (a coefficient below 2.5 is considered a stable process, meeting the quality requirements for lipstick tube injection molding).

[0044] Furthermore, the calculation of the cumulative effect value of time-series decay based on the time-series decay path matrix includes the following steps: Extract the injection molding cycle time and parameter fluctuation transmission direction corresponding to each row and column of the time-series decay path matrix, and count the in-degree and out-degree of each node at each time point. The in-degree is the number of influence paths of the previous time point to this time point, and the out-degree is the number of influence paths of this time point to the subsequent time point. Calculate the node influence transmission efficiency by the ratio of in-degree to out-degree, and generate a set of node influence efficiencies at each time point. In this embodiment of the invention, the injection molding process of cosmetic lipstick tube shells is analyzed using a time-series attenuation path matrix (5 rows and 5 columns, where rows represent preceding times and columns represent subsequent times, and the time nodes are divided into T1 (mold closing), T2 (injection), T3 (holding pressure), T4 (cooling), and T5 (mold opening) according to the injection molding cycle). The injection molding cycle times and parameter fluctuation transmission directions corresponding to each row and column are extracted. For example, matrix element (1,2) represents the temperature fluctuation transmission path from T1 to T2, and (2,3) represents the pressure fluctuation transmission path from T2 to T3, for a total of 8 effective attenuation paths. The in-degree and out-degree of each node at each time step are calculated as follows: T1: in-degree 0 (no preceding time step), out-degree 2 (affects T2 and T3); T2: in-degree 1 (affected by T1), out-degree 2 (affects T3 and T4); T3: in-degree 2 (affected by T1 and T2), out-degree 2 (affects T4 and T5); T4: in-degree 2 (affected by T2 and T3), out-degree 1 (affects T5); T5: in-degree 2 (affected by T3 and T4), out-degree 0 (no subsequent time step). The node influence propagation efficiency is calculated using the formula: "Node influence propagation efficiency = out-degree ÷ in-degree" (efficiency is set to 2.0 when in-degree is 0): T1 efficiency 2.0, T2 efficiency 2.0, T3 efficiency 1.0, T4 efficiency 0.5, T5 efficiency 0.0, generating the node influence efficiency set {2.0, 2.0, 1.0, 0.5, 0.0}.

[0045] Furthermore, the decay probability corresponding to each time node in the time-series decay path matrix is ​​retrieved and coupled with the node influence transmission efficiency at the corresponding time in the time node influence efficiency set to generate an optimized node decay probability set. In this embodiment of the invention, by retrieving the attenuation probabilities corresponding to each time node in the time-series attenuation path matrix (T1 attenuation probability 0.92, T2 attenuation probability 0.88, T3 attenuation probability 0.85, T4 attenuation probability 0.90, T5 attenuation probability 0.95), the attenuation probability of each time node is coupled with the node influence transmission efficiency at the corresponding time in the time node influence efficiency set. The coupling formula is "optimized node attenuation probability = attenuation probability × (1 + node influence transmission efficiency × 0.1)". The calculated probabilities are: T1 after optimization: 0.92×(1+2.0×0.1)=1.104; T2 after optimization: 0.88×(1+2.0×0.1)=1.056; T3 after optimization: 0.85×(1+1.0×0.1)=0.935; T4 after optimization: 0.90×(1+0.5×0.1)=0.945; T5 after optimization: 0.95×(1+0.0×0.1)=0.95. The optimized node decay probability set is generated as {1.104, 1.056, 0.935, 0.945, 0.95}.

[0046] Furthermore, each time node is taken as a network node, the decay path between network nodes is taken as a network edge, and the probability value of the optimized node decay probability set is taken as the edge weight to generate a time-series decay weighted network. The betweenness centrality value of each network node is calculated through network topology analysis. In this embodiment of the invention, a time-decay weighted network is generated by using T1-T5 as network nodes and the 8 decay paths in the time-decay path matrix as network edges (such as T1→T2, T1→T3, T2→T3, T2→T4, T3→T4, T3→T5, T4→T5). The probability values ​​of the corresponding path starting nodes in the optimized node decay probability set are used as edge weights (such as T1→T2 edge weight 1.104, T2→T3 edge weight 1.056). The betweenness centrality value of each network node is calculated through network topology analysis (betweenness centrality is the proportion of times a node is located on all shortest paths out of the total number of shortest paths): T1 betweenness centrality 0.1 (located on 2 shortest paths), T2 betweenness centrality 0.3 (located on 6 shortest paths), T3 betweenness centrality 0.4 (located on 8 shortest paths), T4 betweenness centrality 0.2 (located on 4 shortest paths), T5 betweenness centrality 0.0 (no subsequent paths), resulting in betweenness centrality values ​​of {0.1, 0.3, 0.4, 0.2, 0.0} for each node.

[0047] Furthermore, the weights of each edge in the time-series decay weighted network are multiplied together in the order of the path to obtain the cumulative decay value of a single path. The cumulative contribution value of the single path decay is then calculated by combining the mean of the betweenness centrality values ​​of the network nodes traversed by the decay path. In this embodiment of the invention, by multiplying the weights of each edge in the 8 decay paths of the time-series decay weighted network in the order of the paths, the cumulative decay value of a single path is obtained: cumulative value of T1→T2 path 1.104, cumulative value of T1→T3 path 1.104, cumulative value of T2→T3 path 1.056, cumulative value of T2→T4 path 1.056, cumulative value of T3→T4 path 0.935, cumulative value of T3→T5 path 0.935, and cumulative value of T4→T5 path 0.945. The mean values ​​are calculated based on the betweenness centrality values ​​of the network nodes traversed by each path: the mean value for the path T1→T2, which passes through nodes T1 and T2, is (0.1+0.3)÷2=0.2; the mean value for the path T1→T3, which passes through nodes T1 and T3, is (0.1+0.4)÷2=0.25; the mean value for the path T2→T3, which passes through nodes T2 and T3, is (0.3+0.4)÷2=0.35; the mean values ​​for the remaining paths are calculated similarly to be 0.25, 0.3, 0.2, and 0.1 respectively. The contribution values ​​for each path are calculated using the formula: "Cumulative contribution value of single-path attenuation = Cumulative contribution value of single-path attenuation × Mean betweenness of nodes". The contribution values ​​for T1→T2 are 1.104×0.2=0.2208, T1→T3 are 1.104×0.25=0.276, and T2→T3 are 1.056×0.35=0.3696. The contribution values ​​for the remaining paths are 0.264, 0.2805, 0.187, and 0.0945, respectively.

[0048] Furthermore, the geometric mean of the cumulative single-path attenuation contribution values ​​of all effective paths in the statistical time-decay weighted network is calculated to obtain the cumulative time-decay effect value.

[0049] In this embodiment of the invention, the cumulative contribution values ​​of single-path attenuation of the eight effective paths in the time-decay weighted network (0.2208, 0.276, 0.3696, 0.264, 0.2805, 0.187, 0.0945, 0.25) are statistically analyzed. The geometric mean formula "cumulative effect value of time-decay attenuation = (product of all single-path contribution values)^(1 / number of paths)" is used for calculation. First, the product of all contribution values ​​is calculated: 0.2208×0.276×0.3696×0.264×0.2805×0.187×0.0945×0.25≈0.0000186. Then, the eighth root of the product is taken to obtain the cumulative effect value of time-decay attenuation ≈0.205, thus completing the quantitative calculation of the cumulative effect of time-decay attenuation in the injection molding process of cosmetic lipstick tube shells.

[0050] Furthermore, step S34 includes the following steps: Extract standard contour data from mold cavity design drawings, convert standard contour data into standard cavity contour digital images, extract feature points and contour parameters of standard contours to generate standard cavity contour feature reference vectors; In this embodiment of the invention, standard contour data, including physical dimensional parameters such as a tube opening diameter of 8mm, a tube body length of 60mm, and a tube bottom rounded corner radius of 1.5mm, is extracted from the design drawings of the mold cavity of a cosmetic lipstick tube. These parameters are imported into an image generation tool to generate a 2048×2048 pixel standard cavity contour digital image at a ratio of 1 pixel = 0.02mm. The SIFT algorithm is used to extract key feature points from the standard contour image, including the four vertices of the tube opening edge, the endpoints of the two generatrix lines in the middle of the tube body, and the corresponding pixel points of the tube bottom rounded corner center, totaling 120 feature points. Each feature point generates a 128-dimensional feature descriptor. Simultaneously, contour parameters (tube opening circumference 25.12mm, tube body straightness 0.02mm, tube bottom rounded corner curvature 0.67mm) are calculated. -1 The features are integrated according to the format of "feature point descriptor-contour parameter" to generate a standard cavity contour feature reference vector with a vector dimension of 120×128+3=15363.

[0051] Furthermore, the cavity contour feature vector in the multi-dimensional feature parameter set of injection molding process monitoring is spatially aligned with the standard cavity contour feature reference vector to eliminate the vector space offset caused by shooting angle and mold installation deviation, and the Euclidean distance between the aligned vector and the standard reference vector is calculated to obtain the vector space distance value. In this embodiment of the invention, the cavity contour feature vector [0.834, 0.52] (feature point matching rate 0.834, average contour deviation 0.52mm) is retrieved from the multi-dimensional feature parameter set of injection molding monitoring, and spatially aligned with the standard cavity contour feature reference vector. Using the ICP algorithm, with the contour parameters in the standard reference vector as the target, the spatial orientation of the monitoring feature vector is adjusted: the shooting angle deviation is corrected by a rotation matrix (rotating 0.5° around the Z-axis), and the mold installation deviation is corrected by a translation vector (X-direction offset -0.2mm, Y-direction offset -0.1mm), eliminating vector spatial offset. The Euclidean distance between the aligned monitoring feature vector and the standard reference vector is calculated, and the feature point matching rate and average contour deviation are standardized to the range of 0-1 (aligned matching rate 0.85, average deviation 0.026). The Euclidean distance is calculated using the formula "Euclidean distance = √[(0.85-1)". 2 +(0.026-0) 2 The calculation yields √[0.0225+0.000676]≈0.152, resulting in a vector space distance value of 0.152.

[0052] Furthermore, based on the mean of the contour deviation in the aligned cavity contour feature vector, the distribution variance of the contour deviation in different regions of the cavity is calculated, and combined with the optimized feature point matching rate, the probability distribution of feature point matching is fitted by the Poisson distribution model to obtain the feature point matching probability density value. In this embodiment of the invention, based on the mean contour deviation of 0.52mm in the aligned cavity contour feature vector, the lipstick tube cavity is divided into three regions: the tube opening, the tube body, and the tube bottom. The contour deviation values ​​for each region are calculated (0.48mm for the tube opening, 0.55mm for the tube body, and 0.53mm for the tube bottom). The variance is then calculated using the formula "Distribution variance = Σ(Region deviation value - Mean)". 2 The result is calculated as [(0.48-0.52)] ÷ number of regions. 2 +(0.55-0.52) 2 +(0.53-0.52) 2 ] ÷ 3 ≈ 0.00087. The feature point matching rate was optimized from 0.834 to 0.85 (excluding blurred edge feature points). A Poisson distribution model was used to fit the probability distribution of feature point matching. The Poisson distribution parameter λ = 0.85 (probability of successful matching) was set, and the probability density function value was calculated: when the number of successful matches was 102 (120 × 0.85), the probability density value = (λ... 102 ×e -λ )÷102!≈0.086, yielding a feature point matching probability density value of 0.086.

[0053] Furthermore, the melt flow rate fluctuation coefficient and cavity temperature fluctuation characteristic value of the injection molding process are retrieved to analyze the influence of melt flow state and temperature change on cavity contour forming. The correlation between the two parameters and the contour deviation is calculated through covariance analysis to generate the process-contour correlation correction coefficient. The coefficient is coupled with the feature point matching probability density value to obtain the corrected feature matching probability. In this embodiment of the invention, the influence of the melt flow rate fluctuation coefficient (0.25 g / (min·s) and the cavity temperature fluctuation characteristic value (0.013) of the same injection molding process on the cavity contour forming is analyzed: for every 0.1 g / (min·s) increase in melt flow rate fluctuation, the average contour deviation increases by 0.08 mm; for every 0.001 increase in temperature fluctuation characteristic value, the average contour deviation increases by 0.03 mm. Covariance analysis is used to calculate the correlation between the two parameters and the contour deviation. The covariance between the melt flow rate fluctuation coefficient and the contour deviation is 0.25 × 0.52 × 0.32 (correlation coefficient) ≈ 0.0416; the covariance between the temperature fluctuation characteristic value and the contour deviation is 0.013 × 0.52 × 0.28 (correlation coefficient) ≈ 0.00188. Taking the average of the two (0.0416 + 0.00188) ÷ 2 ≈ 0.0217, a process-contour correlation correction coefficient of 0.0217 is generated. The corrected feature matching probability is calculated as follows: "Corrected feature matching probability = feature point matching probability density value × (1 + correction coefficient)". The result is 0.086 × (1 + 0.0217) ≈ 0.0879.

[0054] Furthermore, the reciprocal of the distance value in the vector space is calculated as a spatial similarity index, and the reciprocal of the distribution variance is combined as a contour uniformity index. The spatial similarity index, contour uniformity index and the corrected feature matching probability are substituted into the Dirichlet distribution model for probability fusion. The peak value of the probability distribution output by the model is determined as the mold cavity matching probability.

[0055] In this embodiment of the invention, the reciprocal of the vector space distance value 0.152, 1 ÷ 0.152 ≈ 6.579, is calculated as a spatial similarity index (the larger the value, the higher the spatial matching degree); the reciprocal of the distribution variance 0.00087, 1 ÷ 0.00087 ≈ 1149.43, is calculated as a contour uniformity index (the larger the value, the more uniform the contour deviation distribution). The spatial similarity index, contour uniformity index, and corrected feature matching probability 0.0879 are standardized to the range of 0-1 (after standardization, they are 0.658, 0.892, and 0.088 respectively), substituted into the Dirichlet distribution model, and the model parameters are set to α = [0.658, 0.892, 0.088] to calculate the probability distribution function. Through numerical iteration, when the three indicators are 0.65, 0.88, and 0.085 respectively, the distribution function reaches a peak value of 0.92. This peak value is determined to be the cavity matching probability of the cosmetic lipstick tube mold of 0.92 (a probability ≥ 0.9 indicates that the cavity forming meets the design requirements).

[0056] Furthermore, step S4 includes the following steps: Step S41: Collect normal and abnormal working condition data corresponding to historical cosmetic plastic mold injection molding, and determine the corresponding monitoring and early warning thresholds based on the normal and abnormal working condition data, including a first early warning threshold and a second early warning threshold, wherein the first early warning threshold is higher than the second early warning threshold; In this embodiment of the invention, historical processing data from the injection molding of cosmetic lipstick tube shells over the past 12 months was collected, including 3000 sets of normal operating condition data and 800 sets of abnormal operating condition data. In the normal operating condition data, the comprehensive evaluation value of the injection molding processing status ranged from 75 to 95. 1.05 times the lower limit of this range (75 × 1.05 = 78.75) was taken as the first warning threshold. In the abnormal operating condition data, the evaluation value ranged from 40 to 68. 1.03 times the upper limit of this range (68 × 1.03 = 70.04) was taken as the second warning threshold, clearly indicating that the first warning threshold of 78.75 is higher than the second warning threshold of 70.04. Simultaneously, for each monitoring parameter (temperature fluctuation characteristic value, pressure deviation characteristic value, etc.), the upper limit of the 95% confidence interval for normal data and the lower limit of the 95% confidence interval for abnormal data were statistically calculated as auxiliary warning thresholds for the corresponding parameters, generating a monitoring and warning threshold table containing the comprehensive evaluation value and parameter thresholds.

[0057] Step S42: Compare the comprehensive evaluation value of the injection molding process status with the monitoring and early warning threshold. If the comprehensive evaluation value of the injection molding process status is higher than the first early warning threshold, the processing status is determined to be normal. If the comprehensive evaluation value of the injection molding process status is between the first and second early warning thresholds, it is determined to be a level one early warning status. If the comprehensive evaluation value of the injection molding process status is lower than the second early warning threshold, it is determined to be a level two early warning status. Retrieve the corresponding multi-source monitoring data encoding table to locate the abnormal monitoring data source codes corresponding to the level one and level two early warning statuses, and generate an abnormal data source location table by combining the data collection timestamp and location code. In this embodiment of the invention, by comparing the comprehensive evaluation value of the injection molding processing status of cosmetic lipstick tubes (79.3) with the monitoring and early warning threshold, if 79.3 is higher than the first early warning threshold (78.75), the processing status is determined to be normal; if the evaluation value of a certain batch is 75 (between 78.75 and 70.04), it is determined to be a first-level early warning status; if the evaluation value is 65 (lower than 70.04), it is determined to be a second-level early warning status. Retrieve the multi-source monitoring data coding table (e.g., "T-001" corresponds to cavity temperature data, "P-002" corresponds to injection pressure data). For the first-level warning state, locate the abnormal data source coding as "T-001" (temperature fluctuation characteristic value 0.018 exceeds the auxiliary threshold 0.015) and "F-003" (melt flow rate fluctuation coefficient 0.32 exceeds the auxiliary threshold 0.3). Combine the data acquisition timestamp (T0+120s, T0+180s) and location code (cavity tube temperature sensor, injection molding machine nozzle detector), organize them according to the format of "warning level-code-timestamp-location code", and generate an abnormal data source location table.

[0058] Step S43: Establish a preset fault type library, which contains each fault type and its corresponding characteristic parameters. For each fault type in the preset fault type library, extract the corresponding characteristic parameter range to generate a fault characteristic parameter range table. At the same time, extract the characteristic parameters of the first-level warning state and the second-level warning state from the generated injection molding process monitoring multi-dimensional characteristic parameter set to generate an abnormal characteristic parameter set. In this embodiment of the invention, a preset fault type library is established, including common fault types in lipstick tube injection molding: abnormal temperature fluctuation (causing tube deformation), excessive pressure deviation (causing insufficient filling of the tube body), unstable melt rate (causing material shortage at the bottom of the tube), and mold opening and closing displacement deviation (causing poor cavity fit). For each fault type, feature parameter ranges are extracted: abnormal temperature fluctuation range (0.016-0.03), excessive pressure deviation range (0.04-0.08), unstable melt rate range (0.31-0.5), and mold opening and closing displacement deviation range (1.8-3.5mm), generating a fault feature parameter range table. Feature parameters under the first-level warning state are extracted from the multi-dimensional feature parameter set of injection molding monitoring (temperature fluctuation feature value 0.018, pressure deviation feature value 0.038, melt rate fluctuation coefficient 0.32, displacement deviation value 1.2mm), generating an abnormal feature parameter set [0.018, 0.038, 0.32, 1.2].

[0059] Step S44: Calculate the matching degree between each parameter in the abnormal feature parameter set and the corresponding feature parameter interval in the fault feature parameter interval table. If the parameter falls within the feature parameter interval, the matching degree is counted as 1; if it falls outside the feature parameter interval, calculate the distance between the parameter and the interval boundary to generate the single-parameter fault matching degree; for each fault type, sum all the corresponding single-parameter fault matching degrees by weight to obtain the fault type comprehensive matching degree. In this embodiment of the invention, the degree of matching between the abnormal characteristic parameter set and the fault characteristic parameter interval table is calculated. The temperature fluctuation characteristic value of 0.018 falls within the abnormal temperature fluctuation interval (0.016-0.03), and the matching degree is 1; the pressure deviation characteristic value of 0.038 is lower than the lower limit of the excessive pressure deviation interval (0.04-0.08) of 0.002, and the single parameter fault matching degree = 1 - (0.002 ÷ 0.04) = 0.95; the melt rate fluctuation coefficient of 0.32 falls within the unstable melt rate interval (0.31-0.5), and the matching degree is 1; the displacement deviation value of 1.2mm is lower than the lower limit of the mold opening and closing displacement deviation interval (1.8-3.5) of 0.6mm, and the single parameter fault matching degree = 1 - (0.6 ÷ 1.8) ≈ 0.667. Parameter weights are assigned according to the fault type (temperature 0.3, pressure 0.3, melt rate 0.25, displacement 0.15), and the comprehensive matching degree of the fault type is calculated: abnormal temperature fluctuation = 1×0.3+0.95×0.3+1×0.25+0.667×0.15≈0.3+0.285+0.25+0.1≈0.935; excessive pressure deviation = 0.8×0.3+1×0.3+0.9×0.25+0.667×0.15≈0.24+0.3+0.225+0.1≈0.865. The other fault types are calculated similarly, and the melt rate instability is 0.91 and the mold opening and closing displacement deviation is 0.72.

[0060] Step S45: Compare the overall matching degree of all fault types, select the fault type with the highest matching degree, and combine it with the corresponding evaluation value and anomaly location to generate the corresponding preliminary result of the abnormal fault; if the highest matching degree is lower than the preset matching threshold, it is marked as an unknown fault type and the review process is triggered; otherwise, an injection molding process monitoring anomaly location report is generated by combining the timestamp and location code of the abnormal data source.

[0061] In this embodiment of the invention, by comparing the comprehensive matching degree of all fault types, the abnormal temperature fluctuation (0.935) is the highest. The preset matching threshold is 0.75. 0.935 is higher than the threshold. Combining the first-level warning status assessment value of 75 and the abnormal location (code "T-001", timestamp T0+120s, location code of the cavity tube body), a preliminary result of the abnormal fault is generated: "Fault type: abnormal temperature fluctuation; assessment value: 75; abnormal source: cavity tube body temperature sensor (T-001), acquisition time T0+120s". Based on this result, the abnormal data trend (temperature fluctuation value increased from 0.013 to 0.018) and the corresponding lipstick tube defect description (slight deformation of the tube opening) are supplemented. Combining the timestamp and location code, an injection molding process monitoring abnormal location report is generated, and the fault handling suggestion is clarified (adjust the flow rate of the cavity tube body cooling water circuit). If the comprehensive matching degree of a certain fault is 0.6 (below 0.75), it is marked as an unknown fault type, triggering the review process (re-collecting 3 sets of data and manually inspecting the lipstick tube molding status).

[0062] Furthermore, the present invention also provides an injection molding process monitoring system for cosmetic plastic molds, including a computer-readable storage medium, a processor, a communication interface, and a computer program stored on the computer-readable storage medium and executable on the processor, for performing the injection molding process monitoring method for cosmetic plastic molds as described above.

[0063] The above description is merely an embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of the present invention should be included within the scope of the claims of the present invention.

Claims

1. A method for monitoring the injection molding process of cosmetic plastic molds, characterized in that, Includes the following steps: Step S1: Obtain multi-dimensional monitoring data corresponding to the injection molding process of cosmetic plastic molds. The multi-dimensional monitoring data includes mold cavity temperature data, injection pressure data, melt flow rate data, mold opening and closing displacement data, as well as mold cavity no-load visual data and injection part real-time forming visual data collected by industrial cameras. Step S2: Classify and identify the data sources of each sub-data in the multi-dimensional monitoring data, and assign a unique data source code to each sub-data according to the data acquisition equipment type and the corresponding acquisition position including the mold inlet, cavity sidewall and injection molding machine nozzle to generate a multi-source monitoring data coding table; perform injection molding process feature analysis on the multi-dimensional monitoring data based on the multi-source monitoring data coding table to obtain the multi-dimensional feature parameter set of injection molding process monitoring; Step S3: Generate the corresponding injection process stability coefficient and mold cavity matching probability based on the multi-dimensional feature parameter set of injection molding process monitoring, and evaluate the injection molding process status of the cosmetic plastic mold based on the injection process stability coefficient and mold cavity matching probability to generate a comprehensive evaluation value of the injection molding process status. Step S4: Set the corresponding monitoring and early warning threshold based on the comprehensive evaluation value of the injection molding process status and make a judgment. For the injection molding process where the comprehensive evaluation value of the injection molding process status is lower than the monitoring and early warning threshold, locate the corresponding abnormal data source. According to the abnormal data source, match the preset fault type library to output the preliminary abnormal fault results containing the evaluation value, abnormal location and fault judgment. Combine the timestamp and location code of the abnormal data source to generate an injection molding process monitoring abnormal location report.

2. The injection molding process monitoring method for cosmetic plastic molds according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Deploy multiple types of data acquisition equipment to install temperature sensors and pressure sensors on the inner wall of the cavity, feeding channel and cooling water channel of the cosmetic plastic mold, install a melt flow rate detector at the injection molding machine nozzle, install a displacement sensor at the mold opening and closing mechanism, and deploy industrial vision cameras at the front and side of the mold cavity. Step S12: The sensor data acquisition module collects the analog signal output by the temperature sensor during the injection molding process of the cosmetic plastic mold in real time, converts the analog signal into a digital signal and then filters it to generate mold cavity temperature data; at the same time, it receives the output signals from the pressure sensor, melt flow rate detector and displacement sensor, processes them to generate injection pressure data, melt flow rate data and mold opening and closing displacement data respectively. Step S13: Use an industrial vision camera to collect images of the mold cavity in its unloaded state before injection molding and real-time forming images during the injection molding process. Perform white balance correction and distortion correction on the collected unloaded state images and real-time forming images to eliminate the effects of lighting changes and lens distortion, and generate unloaded visual data of the mold cavity and real-time forming visual data of the injection molded part. Step S14: Integrate the mold cavity temperature data, injection pressure data, melt flow rate data, mold opening and closing displacement data, and mold cavity no-load visual data with the real-time molding visual data of the injection molded part to form multi-dimensional monitoring data corresponding to the injection molding process of cosmetic plastic molds.

3. The injection molding process monitoring method for cosmetic plastic molds according to claim 1, characterized in that, Step S2, which involves analyzing the injection molding process characteristics of multi-dimensional monitoring data based on a multi-source monitoring data encoding table, includes the following steps: Step S201: Collect the process parameter setting data corresponding to the injection molding machine, including preset injection temperature, preset injection pressure, preset holding pressure and preset mold opening and closing displacement, and generate the preset process parameter table of the injection molding machine; Step S202: Based on the multi-source monitoring data encoding table, retrieve the monitoring data corresponding to each type of code. For the mold cavity temperature data, use a sliding window to extract the corresponding temperature value within a continuous time period. Calculate the ratio between the difference between the maximum and minimum temperature values ​​within the window and the preset injection temperature to generate instantaneous temperature fluctuation values. Then, take the average of the instantaneous temperature fluctuation values ​​corresponding to multiple windows to obtain the temperature fluctuation characteristic value. Step S203: Extract the pressure peak value corresponding to the injection stage from the injection pressure data, and calculate the difference between the pressure peak value and the preset injection pressure in the preset process parameter table of the injection molding machine. Then, the ratio of this difference to the preset injection pressure is used as the pressure peak deviation rate. At the same time, calculate the deviation rate between the mean value of the injection pressure data in the holding pressure stage and the preset holding pressure to generate the holding pressure deviation rate. The pressure peak deviation rate and the holding pressure deviation rate are weighted and summed to obtain the pressure deviation characteristic value. Step S204: Extract the corresponding visual image feature points and cavity contours from the unloaded visual data of the mold cavity and the real-time molding visual data of the injection molded part, and calculate the feature point matching rate and the average contour deviation of the two images to generate the cavity contour feature vector. Step S205: Calculate the rate of change of melt flow rate per unit time for melt flow rate data to generate melt flow rate fluctuation coefficient; calculate the deviation between actual mold opening and closing displacement and preset mold opening and closing displacement for mold opening and closing displacement data to generate mold opening and closing displacement deviation value. Step S206: Summarize the temperature fluctuation characteristic value, pressure deviation characteristic value, cavity contour characteristic vector, melt flow rate fluctuation coefficient, and mold opening and closing displacement deviation value to obtain a multi-dimensional characteristic parameter set for injection molding monitoring.

4. The injection molding process monitoring method for cosmetic plastic molds according to claim 3, characterized in that, Step S204 includes the following steps: The generated visual data of the mold cavity in no-load and the real-time molding visual data of the injection molded part are processed into grayscale to generate a grayscale image of the cavity in no-load and a grayscale image of the real-time molding of the injection molded part. Gaussian filtering is then used for noise reduction to generate a grayscale image of the filtered base and a grayscale image of the filtered real-time part. Key feature points are detected in both the filtered baseline grayscale image and the filtered real-time grayscale image, and feature descriptors corresponding to each key feature point are calculated to generate feature descriptor point sets for the baseline image and the real-time image. Feature matching is performed on the feature description point set of the reference image and the feature description point set of the real-time image to count the number of successfully matched feature points, and the ratio of the number of successfully matched feature points to the total number of feature points in the two visual images is used as the feature point matching rate. Edge extraction is performed on the filtered baseline grayscale image and the filtered real-time grayscale image to obtain the baseline cavity contour image and the real-time cavity contour image; the intrinsic and extrinsic parameters corresponding to the industrial vision camera are obtained and the mapping relationship between image pixel coordinates and actual physical coordinates of the mold cavity is established to generate a pixel-physical coordinate transformation matrix; The coordinates of the contour pixels in the reference cavity contour map and the real-time cavity contour map are extracted to obtain the reference contour pixel coordinate set and the real-time contour pixel coordinate set; the reference contour pixel coordinate set and the real-time contour pixel coordinate set are substituted into the pixel-physical coordinate transformation matrix to generate the reference contour physical coordinate set and the real-time physical coordinate set respectively. Align the physical coordinate set of the baseline contour with the physical coordinate set of the real-time contour to determine the matching relationship of the corresponding contour points and generate a set of cavity contour point matching pairs. Calculate the Euclidean distance between each pair of corresponding points in the set of cavity contour point matching pairs and calculate the arithmetic mean of all Euclidean distances as the mean of the contour deviation. Combine the corresponding feature point matching rate and the mean of the contour deviation in sequence to generate the cavity contour feature vector.

5. The injection molding process monitoring method for cosmetic plastic molds according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Based on the multi-dimensional feature parameter set of injection molding process monitoring and combined with regression analysis of historical fault data, determine the process influence weight coefficient corresponding to each parameter. Each parameter includes temperature fluctuation feature value, pressure deviation feature value, melt flow rate fluctuation coefficient, and mold opening and closing displacement deviation value to generate an injection molding process parameter-weight correspondence table; normalize the multi-dimensional feature parameter set of injection molding process monitoring to generate a normalized process parameter set for injection molding process. Step S32: Calculate the Pearson correlation coefficients between parameters based on the normalized process parameter set for injection molding to generate a parameter correlation matrix. Then, based on the parameter correlation matrix, combine the process influence weight coefficients with the normalized process parameter set for injection molding to perform weighted fusion to obtain the process fluctuation correlation fusion parameters. Retrieve the process fluctuation fusion parameters corresponding to multiple consecutive processing cycles and calculate the standard deviation of the parameters in the time dimension to generate the time series standard deviation of process fluctuation. Compare the time series standard deviation of process fluctuation with the preset allowable range of process fluctuation to generate a time series stability evaluation factor. Specifically, the time series stability evaluation factor = 1 - (time series standard deviation of process fluctuation / upper limit of allowable range of process fluctuation). Step S33: Based on the time-series stability assessment factor, evaluate the time-series impact decay of the process fluctuation correlation fusion parameters to obtain the injection molding process stability coefficient; Step S34: Estimate the corresponding mold cavity matching probability based on the cavity contour feature vector in the multi-dimensional feature parameter set of injection molding process monitoring; Step S35: Based on the injection molding process stability coefficient and the mold cavity matching probability, evaluate the injection molding status of the cosmetic plastic mold to generate a comprehensive evaluation value of the injection molding status.

6. The injection molding process monitoring method for cosmetic plastic molds according to claim 5, characterized in that, Step S33 includes the following steps: Extract process fluctuation correlation fusion parameters corresponding to N consecutive injection molding cycles to construct a time-series fusion parameter sequence. At the same time, retrieve the equipment running time and mold preheating time corresponding to each injection molding cycle, and calculate the equipment-mold time-series coordination coefficient by the ratio of equipment running time to mold preheating time. The time-series fluctuation characteristics of parameters are calculated based on the time-series fusion parameter sequence. A sliding window is used to extract the corresponding parameter subsequences in different time periods, and the coefficient of variation of each parameter subsequence is calculated to generate a set of time-series fluctuation coefficients. The set of time-series fluctuation coefficients is corrected based on the equipment-mold time-series coordination coefficient, so that the time-series decay correlation coefficient can be obtained by the corrected coefficient of variation = coefficient of variation × (1 - equipment-mold time-series coordination coefficient). The time-series decay correlation coefficient is used as the simulation input, and the process decay factor is introduced. Specifically, the trend characteristics of parameter fluctuation decay over time are extracted based on historical data, and the decay path of parameter fluctuation in the time dimension is simulated through Markov chain state transition to generate a time-series decay path matrix, where each element represents the decay probability of parameter fluctuation at a certain moment to subsequent moments. Calculate the cumulative effect value of time-series decay based on the time-series decay path matrix; The degradation assessment of the process fluctuation correlation fusion parameter is calculated based on the ratio between the time-series stability assessment factor and the 1+ time-series decay cumulative effect value to obtain the injection molding process stability coefficient.

7. The injection molding process monitoring method for cosmetic plastic molds according to claim 6, characterized in that, The calculation of the time-decayed cumulative effect value based on the time-decayed path matrix includes the following steps: Extract the injection molding cycle time and parameter fluctuation transmission direction corresponding to each row and column of the time-series decay path matrix, and count the in-degree and out-degree of each node at each time point. The in-degree is the number of influence paths of the previous time point to this time point, and the out-degree is the number of influence paths of this time point to the subsequent time point. Calculate the node influence transmission efficiency by the ratio of in-degree to out-degree, and generate a set of node influence efficiencies at each time point. The decay probability corresponding to each node at each time step in the temporal decay path matrix is ​​retrieved and coupled with the node influence propagation efficiency at the corresponding time step in the time step node influence efficiency set to generate an optimized node decay probability set. Each time node is taken as a network node, the decay path between network nodes is taken as a network edge, and the probability value of the optimized node decay probability set is taken as the edge weight to generate a time-series decay weighted network. The betweenness centrality value of each network node is calculated through network topology analysis. For each decay path in the time-decay weighted network, the weights of each edge are multiplied together in the path order to obtain the cumulative decay value of a single path. The cumulative contribution value of the decay of a single path is then calculated by combining the mean of the betweenness centrality values ​​of the network nodes traversed by the decay path. The cumulative contribution of single-path attenuation in all effective paths of the statistical time-decay weighted network is calculated by geometric mean to obtain the cumulative effect value of time-decay.

8. The injection molding process monitoring method for cosmetic plastic molds according to claim 5, characterized in that, Step S34 includes the following steps: Extract standard contour data from mold cavity design drawings, convert standard contour data into standard cavity contour digital images, extract feature points and contour parameters of standard contours to generate standard cavity contour feature reference vectors; Spatial alignment processing is performed on the cavity contour feature vector in the multi-dimensional feature parameter set of injection molding process monitoring and the standard cavity contour feature reference vector to eliminate the vector space offset caused by shooting angle and mold installation deviation. The Euclidean distance between the aligned vector and the standard reference vector is calculated to obtain the vector space distance value. Based on the mean of the contour deviation in the feature vector of the aligned cavity contour, the distribution variance of the contour deviation in different regions of the cavity is calculated. Combined with the optimized feature point matching rate, the probability distribution of feature point matching is fitted by the Poisson distribution model to obtain the feature point matching probability density value. The melt flow rate fluctuation coefficient and cavity temperature fluctuation characteristic value of the injection molding process are retrieved to analyze the influence of melt flow state and temperature change on cavity contour forming. The correlation between the two parameters and the contour deviation is calculated by covariance analysis to generate process-contour correlation correction coefficient. The coefficient is coupled with the feature point matching probability density value to obtain the corrected feature matching probability. The spatial similarity index is calculated based on the reciprocal of the distance value in the vector space, and the contour uniformity index is combined with the reciprocal of the distribution variance. The spatial similarity index, the contour uniformity index and the corrected feature matching probability are substituted into the Dirichlet distribution model for probability fusion. The peak value of the probability distribution output by the model is used to determine the mold cavity matching probability.

9. The injection molding process monitoring method for cosmetic plastic molds according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Collect normal and abnormal working condition data corresponding to historical cosmetic plastic mold injection molding, and determine the corresponding monitoring and early warning thresholds based on the normal and abnormal working condition data, including a first early warning threshold and a second early warning threshold, wherein the first early warning threshold is higher than the second early warning threshold; Step S42: Compare the comprehensive evaluation value of the injection molding process status with the monitoring and early warning threshold. If the comprehensive evaluation value of the injection molding process status is higher than the first early warning threshold, the processing status is determined to be normal. If the comprehensive evaluation value of the injection molding process status is between the first and second early warning thresholds, it is determined to be a level one early warning status. If the comprehensive evaluation value of the injection molding process status is lower than the second early warning threshold, it is determined to be a level two early warning status. Retrieve the corresponding multi-source monitoring data encoding table to locate the abnormal monitoring data source codes corresponding to the level one and level two early warning statuses, and generate an abnormal data source location table by combining the data collection timestamp and location code. Step S43: Establish a preset fault type library, which contains each fault type and its corresponding characteristic parameters. For each fault type in the preset fault type library, extract the corresponding characteristic parameter range to generate a fault characteristic parameter range table. At the same time, extract the characteristic parameters of the first-level warning state and the second-level warning state from the generated injection molding process monitoring multi-dimensional characteristic parameter set to generate an abnormal characteristic parameter set. Step S44: Calculate the matching degree between each parameter in the abnormal feature parameter set and the corresponding feature parameter interval in the fault feature parameter interval table. If the parameter falls within the feature parameter interval, the matching degree is counted as 1; if it falls outside the feature parameter interval, calculate the distance between the parameter and the interval boundary to generate the single-parameter fault matching degree; for each fault type, sum all the corresponding single-parameter fault matching degrees by weight to obtain the fault type comprehensive matching degree. Step S45: Compare the overall matching degree of all fault types, select the fault type with the highest matching degree, and combine it with the corresponding evaluation value and anomaly location to generate the corresponding preliminary result of the abnormal fault; if the highest matching degree is lower than the preset matching threshold, it is marked as an unknown fault type and the review process is triggered; otherwise, an injection molding process monitoring anomaly location report is generated by combining the timestamp and location code of the abnormal data source.

10. A monitoring system for injection molding of cosmetic plastic molds, characterized in that, The method includes a computer-readable storage medium, a processor, a communication interface, and a computer program stored on the computer-readable storage medium and executable on the processor, for performing the injection molding monitoring method for cosmetic plastic molds as described in any one of claims 1-9.

Images