Comprehensive control method and system for extrusion molding process of PVC (polyvinyl chloride) skinning and foaming profile

Abstract

The invention discloses a comprehensive control method and system for the extrusion molding process of a PVC skinning and foaming sectional material, and relates to the technical field of plastic sectional material machining. The control method comprises the steps that real-time operation data of all links of the whole process of the PVC skinning and foaming sectional material are collected; a fusion type prediction model is built by using historical production operation data, and the fusion type prediction model combines two data processing modes. According to the comprehensive control method and system for the extrusion molding process of the PVC skinning and foaming profile, the parameter deviation trend is captured in advance by fusing the prediction model and the cooperative control technology, that is, when the environment temperature changes abnormally, the model can pre-judge the fluctuation direction of the decomposition rate of the foaming agent, and the decomposition rate of the foaming agent is determined. A regulation and control instruction is sent to the extruder and the ultrasonic vibration device in advance, the temperature and vibration parameters of the foaming area are synchronously and finely adjusted, and foam hole form abnormity is avoided; for different scene requirements, the dynamic quality evaluation system can flexibly allocate index weights, and it is ensured that each batch of profiles can be accurately matched with the requirements.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of plastic profile processing, in particular to a comprehensive control method and system for the extrusion molding process of PVC skin foamed profile. BACKGROUND

[0002] PVC skin foamed profile is widely used in fields of building doors and windows, decorative lines, furniture frames, etc. due to its light weight, high strength, excellent heat and sound insulation performance. Its extrusion molding process involves multiple links such as raw material mixing, extrusion plasticizing, foaming reaction, mold forming, cooling and shaping, etc. The parameters of each link are mutually coupled and significantly affect each other. The traditional control method has the following technical pain points: The traditional method mainly adjusts single link independently, such as the temperature of the extruder and the amount of foaming agent added. The collaborative correlation of parameters in each link is not established, which leads to the whole body being affected by a single factor. For example, when the uniformity of raw material mixing is insufficient, adjusting the speed of the extruder cannot solve the problem of uneven plasticization, which easily causes uneven thickness of the profile skin and disorderly distribution of internal cells. Moreover, it is particularly dependent on the experience or real-time feedback of the operator for passive adjustment, and lacks early prediction of parameter deviation trend. For example, when the decomposition rate of the foaming agent starts to decline due to temperature fluctuations, the traditional method needs to adjust the parameters after the cell density detection is unqualified, which leads to the production of a large number of unqualified products and reduces the production efficiency. In addition, the quality detection of PVC skin foamed profile mainly focuses on such intuitive indicators as size accuracy and surface finish, ignoring factors such as bending strength and aging resistance that affect long-term performance. Moreover, it is unable to dynamically adjust the index weight according to product use, which leads to the evaluation standard of such unified production being unable to meet the differentiated needs. Therefore, we propose a comprehensive control method and system for the extrusion molding process of PVC skin foamed profile. SUMMARY

[0003] (I) Technical problems solved In view of the deficiencies of the prior art, the present application provides a comprehensive control method and system for the extrusion molding process of PVC skin foamed profile, which solves the technical problems of coordination loss and passive adjustment lag in the production of existing PVC skin foamed profile.

[0004] (II) Technical solutions To achieve the above purpose, the present application is implemented by the following technical solutions: The comprehensive control method for the extrusion molding process of PVC skin foamed profile comprises: Collecting real-time running data of each link from raw material processing to cooling and shaping of PVC skin foamed profile, the real-time running data covering core parameters of key links such as raw material mixing, extruder operation, foaming reaction, mold forming and cooling treatment, and the data acquisition frequency being 1 time per second; A fusion prediction model is built using historical production and operation data. The fusion prediction model combines two data processing methods: one is used to mine the complex correlation features between parameters, and the other is used to track the time series change patterns of parameters. Real-time operation data is input to continuously optimize the model, enabling early prediction of the deviation trend of parameters in each link within the next 5 to 10 minutes. Based on the prediction results and real-time feedback of production operation data, formulate a collaborative control scheme that integrates prediction, execution, verification, and correction. The coordinated control scheme is used to adjust all aspects of the extrusion molding process in a coordinated manner. The full range of quality inspection data of the formed profiles is obtained through the online inspection device on the production line; The weights of each quality indicator are dynamically allocated according to product usage requirements, and the comprehensive quality score is calculated using the analytic hierarchy process. If the overall quality score meets the preset qualification standard, the current collaborative control scheme is maintained; if it does not meet the preset qualification standard, the key parameters of quality deviation are located by analyzing the correspondence between quality defects and real-time operating data, and the collaborative control scheme is iteratively optimized until the overall quality score meets the preset qualification standard.

[0005] Preferably, the step of collecting real-time operational data adopts a dual-mode acquisition method combining distributed sensing and online analysis: Data from the raw material processing stage was collected using a near-infrared spectroscopy analyzer and a particle size analyzer. Specifically, the data included the mixing uniformity of PVC resin and auxiliary materials, moisture content, particle size distribution, proportion of auxiliary materials, and dispersion uniformity. Extruder operating data is collected through the extruder's built-in sensor array, including real-time temperature of each section of the extruder barrel, screw speed, feeding speed, melt pressure, material plasticity, and screw torque; Data from the foaming reaction process is collected through an integrated detection device, including foaming zone temperature, foaming zone pressure, amount of foaming agent added, foaming agent addition accuracy, foaming agent decomposition rate, ultrasonic vibration frequency, and ultrasonic amplitude. Data from the mold forming process is collected through built-in sensors in the mold, including mold zone temperature, sizing sleeve vacuum, sizing time, mold cavity pressure, and profile extrusion speed. Data from the cooling process is collected by intelligent flow sensors and temperature monitoring instruments, including cooling water temperature, cooling time, cooling air volume, cooling medium flow rate, and profile surface temperature gradient.

[0006] Preferably, the steps for building and optimizing a fusion prediction model include: Collect real-time operating data, raw material batch information, and corresponding profile quality inspection data for nearly one year to establish a standardized database. The raw material batch information includes resin molecular weight distribution, auxiliary material purity, stabilizer content, and lubricant ratio. The wavelet transform method is used to carry out noise reduction processing on the database data, the key parameters are screened through the Pearson correlation coefficient calculation, and only the parameters with a correlation degree of ≥85% with the profile quality are retained; A fusion prediction model is built, the screened key parameters are used as input, and the future 5 to 10 minute parameter deviation trend is used as output, the parameter characteristics are extracted and the time sequence rules are captured by using two data processing methods, and the model training is completed; A model self-updating mechanism is set, and every 100 batches of production are included in the new data for incremental training to ensure that the model prediction accuracy is stable≥98%; An abnormal model checking function is added, when the prediction deviation is >3%, the backup gradient boosting tree model is automatically switched to, and the continuity of the prediction process is ensured.

[0007] Preferably, the raw material processing and extruder operation cooperative control in the cooperative control scheme comprises: The influence of raw material mixing uniformity on plasticizing efficiency is predicted through the fusion prediction model, and a coupling adjustment rule combining mixing uniformity, screw speed and feeding speed is established; If the predicted plasticizing efficiency is <90%, the mixing machine stirring speed is increased by 10% to 15% and the stirring time is extended by 2 to 3 minutes in the raw material processing stage, and at the same time, a pre-adjustment instruction is sent to the extruder control system to increase the screw speed by 6% to 9%; Based on the auxiliary material ratio obtained by the near-infrared spectrum detector, the material melting index correction value is calculated, the temperature of the feeding section, compression section, melting section and homogenization section of the extruder barrel is dynamically adjusted by the PID controller, the temperature adjustment precision is controlled within ±0.5℃, and the material plasticizing degree is≥95%, wherein the material melting index correction value is a corrected value calculated by a specific formula or model, which is more suitable for the actual processing environment, in combination with the melting index measured under standard conditions and the process parameter deviation in actual production, such as temperature, pressure, raw material characteristic fluctuation, etc.; Melt flow rate (MFI) is a key indicator to measure the flowability of polymer melt. It is usually measured under standard test conditions, such as a specific temperature and load. For example, PVC is usually tested at 190℃ / 2.16kg. MFI value is used to represent the flowability of material in the molten state. The higher the MFI value, the better the flowability. However, in actual production, such as PVC extrusion molding, the process conditions often deviate from the standard test conditions. For example, the actual extruder barrel temperature may be different from the standard test temperature, such as 180℃ vs. 190℃. The raw materials may have batch differences, such as molecular weight distribution and additive content fluctuation. The production environment pressure, humidity, etc. may also affect the melt flowability. At this time, directly using the MFI value under standard conditions to guide production will cause deviation. For example, the actual flowability may be worse / better than the standard MFI. Therefore, it is necessary to calculate the correction value to match the actual working condition and ensure the accuracy of the process parameter setting.

[0008] The core of calculating the material melt index correction value is to establish a correlation model between the standard MFI and the actual process parameters. The common formula or idea is as follows: 1. Temperature correction Polymer melt viscosity is sensitive to temperature, and temperature deviation will significantly affect the flowability; correction is usually based on the simplified form of Arrhenius equation: Where, MFI 修正 is the melt index correction value, MFI 标准 is the set melt index standard value, T 标准 is the standard test temperature, such as 190℃, which needs to be converted to Kelvin temperature; T 实际 is the actual melt temperature in production, such as the temperature of the melt section of the extruder; k is a material property constant related to the molecular weight of PVC, plasticizer content, etc., which is fitted by experiment; for example: if the standard MFI is 10 g / 10 min (190℃), the actual temperature is 180℃, and the corrected MFI may be reduced to 7 g / 10 min through formula calculation, indicating that the actual flowability is worse, and the screw speed needs to be adjusted to improve the shear to improve the flow.

[0009] Raw material property correction If the batch difference of raw materials leads to the change of molecular weight distribution (MWD), the molecular weight correction coefficient needs to be introduced: Where, MWD 实际 is the molecular weight distribution index of the actual raw material, MWD 标准 is the set standard molecular weight distribution index; n is an empirical index, usually -3 to -2 for PVC, because the flowability decreases when the molecular weight distribution widens.

[0010] Pressure correction In the extrusion process, the die pressure will affect the melt flow, and the pressure correction term may need to be introduced under high pressure: Where, ΔP is the difference between the actual pressure and the standard test pressure; c is the pressure sensitivity coefficient, and the PVC melt usually has low pressure sensitivity, so the value of c is small.

[0011] In the production of PVC structural foamed profiles, the core role of melt index correction value is: as the input of the fusion prediction model, such as the random forest and LSTM models mentioned earlier, to more accurately predict the plasticizing efficiency and guide the coordinated adjustment of screw speed and feeding speed; to avoid defects caused by flowability misjudgment, such as lowering the melt section temperature to improve the flow when the MFI correction value is low to prevent the profile from being under-filled or having a rough surface; to eliminate the influence of raw material fluctuations on the process through correction to ensure the consistency of the quality of profiles produced by different batches of raw materials; in short, calculating the material melt index correction value is a bridge connecting standard test data and actual production conditions, which quantifies the influence of process deviation on flowability and provides data support for precise control; Synchronous monitoring of the temperature of each section of the extruder barrel and the torque of the screw, when the torque fluctuation is >5%, fine-tune the feeding speed to maintain the melt glue pressure stable.

[0012] Preferably, the precise control of ultrasonic-assisted foaming in the synergistic control scheme includes: Setting a partitioned ultrasonic vibration device in the foaming reaction zone, divided into a foaming start-up zone, a cell growth zone, and a cell shaping zone, with independent adjustment of ultrasonic parameters in each zone; Foaming start-up zone: when the decomposition rate of the foaming agent is <60%, high-frequency (35-45 kHz) and small-amplitude (5-8 μm) ultrasonic vibration is used to accelerate the decomposition of the foaming agent through ultrasonic cavitation effect, and the parameters are switched when the decomposition rate is ≥80%; Cell growth zone: real-time monitoring of cell morphology is performed through a high-speed camera, and if the cell diameter deviation is >0.05 mm, the vibration is adjusted to medium frequency (25-35 kHz) and medium amplitude (8-12 μm) to promote uniform cell growth; Cell shaping zone: low-frequency (15-25 kHz) and large-amplitude (12-15 μm) vibration is used to enhance cell stability; Synchronous monitoring of pressure changes in the foaming zone throughout the process, when the pressure fluctuation is >0.02 MPa, fine-tune the amount of foaming agent added to maintain stable pressure, ultimately achieving a 30-40% increase in cell density and ≥90% uniformity of cell uniformity.

[0013] Preferably, the mold forming and cooling process in the synergistic control scheme includes: Predicting the size deviation trend of the profile through a fusion prediction model, and establishing a synergistic adjustment rule for mold temperature, vacuum degree of the shaping sleeve, and cooling rate; If the predicted positive deviation of the profile width is >0.2 mm, 10-15 seconds before the material enters the mold, increase the vacuum degree of the shaping sleeve by 15-20%, reduce the outlet temperature of the mold by 3-5°C through a mold temperature machine, and simultaneously increase the initial air volume of the cooling system; Using an infrared thermal imager to detect the surface temperature distribution of the formed profile, dividing it into high-temperature, medium-temperature, and low-temperature zones, and using zoned cooling control: high-temperature zone uses 20-25°C cooling water and 15-20 m 3 / h air volume; medium-temperature zone uses 25-30°C cooling water and 10-15 m 3 / h air volume; low-temperature zone uses 30-35°C cooling water and 5-10 m 3 / h air volume; 3 / h air volume; 3 / h air volume; 3 / h air volume; During the cooling process, the profile warpage is monitored in real time. When the warpage is greater than 0.3%, the cooling airflow in the corresponding area is adjusted, ultimately achieving a cooling rate uniformity improvement of more than 20% and a profile warpage of ≤0.5%.

[0014] Preferably, the steps for calculating the overall quality score include: The quality inspection indicators are defined as profile density, dimensional accuracy, surface skin thickness, bending strength, impact strength, heat distortion temperature, aging resistance and surface finish. The weights of the indicators are dynamically allocated according to product type: for structural profiles, bending strength is assigned a weight of 30%, dimensional accuracy a weight of 25%, and the remaining indicators each a weight of 10% to 15%; for decorative profiles, surface finish is assigned a weight of 30%, surface skin thickness a weight of 20%, and the remaining indicators each a weight of 10% to 15%. The test values ​​of each indicator are compared with the production standards of foamed materials and converted into a standardized score of 0 to 100: 100 points are awarded for an indicator value that is more than 10% better than the standard, and 6 points are deducted for each 1% lower than the standard. The weighted total score is used as the comprehensive quality score, with a preset passing score of 92 points and no individual indicator score lower than 70 points.

[0015] Preferably, the key parameters for quality deviation location and solution iteration steps include: If the overall quality score fails to meet the standard, the mapping relationship between the quality inspection data and the real-time operation data is analyzed by association rule mining algorithm to locate the key parameters of quality deviation. When the critical deviation is insufficient dimensional accuracy, prioritize adjusting the vacuum gradient of the shaping sleeve, adjusting it by 2kPa to 3kPa per level, and adjusting it in conjunction with the mold cavity pressure, with an adjustment range of 0.1MPa to 0.2MPa, while simultaneously correcting the cooling zone parameters; When the critical deviation is insufficient flexural strength, the focus should be on optimizing the extruder melt zone temperature, screw speed, and cell density. Specifically, the extruder melt zone temperature should be adjusted by 1°C to 2°C; the screw speed by 5 r / min to 10 r / min; and the cell density by 5% to 8% through ultrasonic parameter adjustment. After each adjustment, the optimization effect is predicted by the fusion prediction model. If the predicted overall quality score improves by ≥3 points, the adjustment is executed; otherwise, the deviation source is traced again. Quality can be achieved with ≤3 iterations.

[0016] A comprehensive control system for the extrusion molding process of PVC skin foam profiles, the control system comprising: The dual-mode data acquisition unit consists of a distributed sensor array, online analytical instruments, and a data transmission module. The online analytical instruments include a near-infrared spectrometer and a particle size analyzer, which enable real-time acquisition and preprocessing of data throughout the entire process, with an acquisition frequency of 1 time per second. The fusion prediction modeling unit integrates a fusion prediction model, a data noise reduction module, and a model self-updating module to complete parameter deviation trend prediction and model dynamic optimization. The collaborative control unit includes a parameter coupling rule base, a PID adjustment module, and a linkage control module to formulate and execute a collaborative control scheme. The intelligent quality evaluation unit integrates a quality index weight distribution module and an analytic hierarchy calculation module to calculate a comprehensive quality score based on online detection data and determine whether the quality meets the standards. The deviation traceability iteration unit locates key parameters of quality deviation through an association rule mining module to realize automatic iterative optimization of the collaborative control scheme. The multi-device execution unit connects and controls the collaborative operation of raw material mixing equipment, an extruder, a partitioned ultrasonic vibration device, an intelligent molding die, and a partitioned cooling device. The energy consumption optimization unit monitors real-time energy consumption data of each device and reduces unit product energy consumption by 5% to 10% through parameter optimization under the premise of meeting quality requirements. Each unit realizes data interaction through an industrial Ethernet to ensure that the response delay of control instructions is less than or equal to 0.5 seconds.

[0017] Preferably, the control system further comprises: A digital twin unit: a virtual simulation model corresponding to the physical production line 1:1 is constructed to real-time map the device operation state, support virtual verification of the collaborative control scheme, and the simulation accuracy is greater than or equal to 95%. An intelligent auxiliary decision-making unit: based on the historical database and real-time working conditions, parameter adjustment suggestions are generated, and three sets of alternative solutions are automatically output when the system is abnormal. A blockchain storage unit: raw material information, real-time operation data, quality detection data, and scheme adjustment records are uploaded to the blockchain system to realize tamper-proof storage and traceability of the whole process data. A remote operation and maintenance unit: through a 5G network, device remote monitoring, parameter debugging, and fault diagnosis are realized, the response delay is less than or equal to 2 seconds, and multi-factory centralized management is supported. The above units are linked with core units such as the dual-mode data acquisition unit and the fusion prediction modeling unit to form a full-chain control system from data acquisition, prediction, control, to evaluation and optimization.

[0018] (Three) beneficial effects 1. By fusing the prediction model with the cooperative control technology, the parameter deviation trend is captured in advance, that is, when the environmental temperature changes abnormally, the model can predict the fluctuation direction of the foaming agent decomposition rate, and send control instructions to the extruder and ultrasonic vibration device in advance, so as to synchronize the fine adjustment of the foaming zone temperature and the vibration parameters, and avoid abnormal cell morphology; for different scene requirements, the dynamic quality evaluation system can flexibly allocate index weight, to ensure that each batch of profile can accurately match the demand; secondly, by capturing the data of raw material mixing, extrusion plasticization and foaming shaping in real time, the abnormal parameters are ensured to be not missed, so that when the problem occurs, the root cause of the quality problem can be quickly located, and the corresponding adjustment scheme can be generated according to the problem root cause, without long time shutdown for troubleshooting; 2. Under the premise of meeting the quality requirements, the parameters are dynamically optimized according to the real-time working condition, such as adjusting the cooling air volume and water temperature according to the surface temperature gradient of the profile in summer high temperature, to avoid excessive cooling; when producing thin-walled decorative profiles, the screw speed of the extruder and the power of the mixer are reduced to reduce the invalid energy consumption; a virtual model with a 1:1 ratio with the physical production line is reconstructed, the control scheme can be verified in advance to reduce the physical trial and error cost, and the raw material information, production parameters and quality data are stored throughout the process, so that the root cause of any problem can be quickly traced, the responsibility is avoided to be fuzzy, and the corresponding solution scheme is given according to the problem, so that the technical personnel can complete the parameter adjustment and fault diagnosis without being on site, which greatly reduces the cross-factory operation and maintenance cost; 3. By automatically strengthening the control of mechanical performance related parameters, when switching to the production of decorative lines, the quality evaluation weight is dynamically adjusted, and the surface finish and skin uniformity are focused on, so that the scene switching can be completed without replacing the core equipment, which provides strong support for responding to market demand changes. BRIEF DESCRIPTION OF DRAWINGS

[0019] The above description is only a summary of the technical scheme of the present application. In order to more clearly understand the technical means of the present application and can be implemented according to the content of the description, the following will be described in detail with the preferred embodiments of the present application and with the help of the drawings.

[0020] Figure 1 The above description is only a summary of the technical scheme of the present application. In order to more clearly understand the technical means of the present application and can be implemented according to the content of the description, the following will be described in detail with the preferred embodiments of the present application and with the help of the drawings. Figure 2 The above description is only a summary of the technical scheme of the present application. In order to more clearly understand the technical means of the present application and can be implemented according to the content of the description, the following will be described in detail with the preferred embodiments of the present application and with the help of the drawings. DETAILED DESCRIPTION

[0021] The embodiment of the application provides a comprehensive control method and system for a PVC skin foamed profile extrusion molding process, solves the technical problem of the lack of synergy and passive lagging adjustment of the existing PVC skin foamed profile during production, and through the fusion of a prediction model and a synergy control technology, the parameter deviation trend is captured in advance when the environmental temperature abnormally changes, the model can predict the fluctuation direction of the foaming agent decomposition rate, the control instruction is sent to the extruder and the ultrasonic vibration device in advance, the foaming zone temperature and the vibration parameter are simultaneously fine-tuned, and the bubble hole morphology is prevented from being abnormal; for different scene requirements, the dynamic quality evaluation system can flexibly distribute index weights, and ensures that each batch of profiles can accurately match the requirements; secondly, the data of each link of raw material mixing, extrusion plasticization and foaming shaping are captured in real time, and abnormal parameters are ensured to be not missed, so that when a problem occurs, the quality problem root cause can be quickly located, and the corresponding adjustment scheme is generated according to the problem root cause, without long-time shutdown troubleshooting.

[0022] Embodiment: The technical scheme in the embodiment of the application solves the technical problem of the lack of synergy and passive lagging adjustment of the existing PVC skin foamed profile during production, and the general idea is as follows: The PVC skin foamed profile is the core material of building doors and windows, furniture frames and home decoration lines, and occupies an important position in the building material and furniture markets due to the characteristics of low density, high bending strength, excellent heat and sound insulation performance and the like; in 2023, the annual output of PVC skin foamed profiles in China broke through 5 million tons, among which, the structural profiles such as building door and window frames and furniture load-bearing frames accounted for about 60%, and the decorative profiles such as skirting lines and ceiling lines accounted for about 40%, and with the development of green buildings and custom furniture industry, the market demand is still growing at an average annual rate of 10%.

[0023] The extrusion molding of the PVC skin foamed profile is sequentially composed of five links of raw material mixing, extrusion plasticization, foaming reaction, shaping and setting, and cooling treatment, that is, the raw material mixing needs to ensure that the resin and auxiliary materials are uniformly dispersed, otherwise the uneven decomposition of the foaming agent will cause local density deviation of the profile; the extrusion plasticization, the temperature gradient of the barrel and the screw speed directly affect the plasticization degree, and insufficient plasticization will reduce the mechanical properties of the profile; the foaming reaction, the temperature, the pressure and the ultrasonic parameter cooperatively control the bubble hole morphology, and uneven bubble holes will cause strength fluctuation; the shaping and setting, the matching of the vacuum degree and the extrusion speed determine the dimensional accuracy, and the excessive deviation will affect the assembly; the cooling treatment, the temperature gradient control avoids warping, and uneven cooling will cause the profile to deform; the five links are tightly coupled and complex processes, and the parameters of each link are strongly associated, and any small deviation of any link may cause a chain of quality defects; for example, in the production of PVC skin foamed window frame profiles, when the uniformity of the raw material mixing is lower than the standard threshold, only by increasing the screw speed of the extruder to try to make up for it, the residence time of the material in the barrel will be shortened far below the standard time, at this time, the plasticization degree will be reduced, and finally the bending strength of the profile will not meet the required standard.

[0024] As can be seen from the above, the limitations of the traditional production control method cannot meet the industry demand, which is specifically manifested in that the parameter collection is one-sided, only measuring the temperature and pressure of the extruder, the prediction ability is missing, relying on passive adjustment of artificial experience, the evaluation standard is rough, using a one-size-fits-all weight, the energy consumption is redundant, and the long-term high cooling air volume operation is significant, the data tracing is difficult, and a large amount of time is spent on troubleshooting; Based on this, this paper takes data-driven combined with scenario-based precise control as the core, builds a comprehensive control method and system covering the whole process, and realizes the transformation from experience-driven to data-precise control through multi-dimensional technical innovation.

[0025] In view of the problems existing in the prior art, the present application provides a comprehensive control method for the PVC skin foaming profile extrusion molding process, and the specific content of the comprehensive control method is as follows: 1. Architecture: The overall technical scheme is designed to solve the pain points of the traditional method, and a whole-chain comprehensive control technical system is constructed from data collection, prediction modeling, collaborative control, quality evaluation to deviation tracing. Taking the production of PVC skin foaming window frame profile as an example, the whole process parameters are controllable, the trend is predictable, and the problem is traceable. The overall process of the method is as follows: Firstly, the double-mode method of distributed sensing combined with online analysis is adopted, and the key parameters of the five links of raw material mixing, extrusion, foaming, molding and cooling, such as mixing uniformity, melting section temperature and foaming agent decomposition rate, are collected at a frequency of 1 time per second. After wavelet transform denoising and normalization preprocessing, these data are transmitted to the subsequent module; secondly, based on the production data of 500 batches of window frame profiles, a random forest and long short-term memory network fusion model is constructed to predict the parameter deviation trend in the next 5 to 10 minutes by inputting real-time parameters; then, combined with the prediction result and real-time feedback, the control strategy from prediction, execution, verification to correction is formulated, and the raw material mixer, extruder, ultrasonic device, molding die and cooling equipment are synchronously executed and adjusted to quickly correct and recover the original production state; next, according to the product use structure, the quality index weight is allocated for structural use or decorative use, the comprehensive quality score of window frame profile is calculated, and whether it meets the standard is judged; finally, if the score does not meet the standard, the key deviation parameters are located through association rule mining, such as insufficient bending strength, and whether it is due to low melting section temperature is detected, so that an optimization scheme can be generated, and the iteration adjustment is not more than 3 times to meet the standard. Among them, the algorithm details of association rule mining, algorithm selection and parameters: Apriori association rule algorithm is adopted, the minimum support is set to 0.2 to ensure that the rule covers more than 20% of the production batches, and the minimum confidence is set to 0.8 to ensure the rule reliability; taking the surface roughness Ra>0.8μm as the target item, the strong association rule {mold temperature>170℃}∧{cooling water temperature>25℃}→{surface roughness insufficient} is mined, with a confidence of 0.85 and a support of 0.22; association analysis example: when the surface roughness Ra=1.0μm of a batch, the system automatically matches the association rule; the mold temperature is 172℃, and the cooling water temperature is 28℃; both parameters meet the rule premise, with a matching degree of 100%; it is directly determined that the high mold temperature and high cooling water temperature are the root cause, without the need to check other parameters one by one.

[0026] The overall architecture is composed of 6 core units and 4 auxiliary units, each unit realizes data interaction and control instruction response delay through industrial Ethernet; the core units include a dual-mode data acquisition unit responsible for parameter acquisition and preprocessing, a fusion prediction modeling unit responsible for deviation trend prediction, a collaborative control unit responsible for generating linkage control instructions, an intelligent quality evaluation unit responsible for comprehensive score calculation, a deviation tracing iteration unit responsible for problem positioning and scheme optimization, and a multi-device execution unit responsible for production equipment control; The auxiliary unit includes an energy consumption optimization unit responsible for reducing energy consumption under the premise of meeting quality standards, a digital twin unit responsible for virtual simulation and verification of the scheme, a blockchain storage unit responsible for non-tamperable data throughout the process, and a remote operation and maintenance unit responsible for remote management of multiple factories. Next, the OPCUA protocol is adopted to support cross-vendor device interfacing, such as Siemens extruders and Schneider frequency converters. An OPCUA server is first built, with a sampling period of 1 second and data caching time of 5 minutes to prevent data loss due to network disconnection. The extruders, ultrasonic devices, etc. are connected to the server through Ethernet, and the node addresses are named according to the device type, link, and parameters, such as extruder-melt section-temperature. The production workshop nodes can only read and write parameters, and the management nodes can configure parameters to ensure data security.

[0027] 2. Real-time data acquisition throughout the process: To address the problems of one-sidedness, lag, and low precision in traditional data acquisition, this module designs a dual-mode acquisition architecture based on full-link coverage combined with high-precision preprocessing logic, with distributed sensing arrays and online analytical instruments working together. For window profile production needs, a near-infrared spectrum detector model NIRS6500 is installed on the distributed sensing array next to the raw material mixer to detect real-time mixing uniformity. PT100 temperature sensors are installed on the 4 sections of the extruder barrel, and HBMT40B torque sensors are installed on the screw end. OMEGATQ934 temperature sensors and PG735 pressure sensors are installed in the foaming area. BDSENSOR SDMP331 vacuum degree sensors are installed in the sizing sleeve. The cooling system is equipped with a FLIRE60 infrared thermal imager to detect profile temperature gradient errors. Online analytical instruments are set up with high-speed cameras with a frame rate of 1000 frames / second next to the foaming area to observe bubble morphology in real time. Every 5 minutes, the foaming agent decomposition rate is analyzed by sampling detection devices to avoid manual detection lag.

[0028] Emergency plan for abnormal situations: (1) If the near-infrared spectrometer fails and there is no mixing uniformity data, switch to manual sampling mode and prompt the operator to take samples every 5 minutes to detect mixing uniformity with an offline particle size analyzer. The mixer speed is increased by 5% to prevent further degradation of uniformity. After the fault is repaired, the system automatically compares offline and online data, calibrates the sensor, and restores the automatic mode. (2) If the blockchain is disconnected and data cannot be uploaded, cache the data to the local hard drive of the industrial computer and upload it after the connection is restored. A temporary traceability code is generated locally, containing raw material batch and key parameters, and is printed and attached to the product packaging to avoid missing responsibility definition. (3) If the fusion model fails and the prediction deviation is greater than 5%, the system automatically activates the backup gradient boosting tree model and triggers an alarm, with audible and visual prompts and SMS notifications to the technician. The technician checks the model log on the remote operation and maintenance platform, manually uploads new raw material data if the failure is caused by abnormal raw material batch, and starts incremental training.

[0029] The data preprocessing uses wavelet transform to separate effective signals from high-frequency interference, with the formula wherein, W f (a, b) is the wavelet coefficient, reflecting the characteristics of the original data at scale a and translation b, used to distinguish between effective signals and interference signals, a is the scale parameter controlling the wavelet scaling, the smaller the value of a, the higher the frequency, which is used to capture the transient interference in the data, and the larger the value of a, the lower the frequency, which is used to extract the trend signal, b is the translation parameter controlling the wavelet positioning, which is used to locate the time point of the interference; f(t) is the original data signal, specifically the real-time parameter signal collected, such as the temperature curve of the melting section changing with time; ψ(·) is the wavelet basis function, db4 wavelet basis is selected due to its excellent time-frequency localization characteristics in non-stationary signal processing; is the integral operation, which calculates the inner product of the original signal and the wavelet basis function to realize signal decomposition and reconstruction; is the normalization coefficient, which ensures the energy conservation at different scales; through this formula, the electromagnetic fluctuation of ±0.3℃ of the temperature sensor can be eliminated, and after normalization processing, the parameters are unified to the range of 0 to 1, avoiding the influence of the order of magnitude difference on the subsequent model prediction.

[0030] In the raw material processing link of window frame profile production, the fiber probe of the near-infrared spectrometer is fixed 10 cm above the discharge port of the mixer, and the near-infrared absorption spectrum of the material is collected in real time. It is assumed that the initial uniformity is 88% calculated by the built-in quantitative model, and at the same time, the laser particle size analyzer detects the particle size distribution of 0.2 to 0.4 mm, and the moisture meter displays the moisture content of 0.15%. After wavelet transform noise reduction, the uniformity is confirmed to be 88.2%, and after normalization processing, it is transmitted to the fusion prediction modeling unit, triggering the prediction prompt that the plasticizing efficiency may drop to 92%, 92% being the preset standard threshold, realizing early detection of raw material problems and reducing subsequent invalid consumption. Among them, the error is excluded after the data is confirmed by wavelet transform noise reduction; the data is transmitted to the fusion prediction modeling unit in real time through industrial Ethernet, providing a basis for subsequent prediction of plasticizing efficiency; the temperature sensors of each section of the extrusion cylinder collect data in real time; after data preprocessing, the cooperative control unit immediately captures the risk of insufficient plasticizing degree due to low melting section temperature, triggering the subsequent prediction and adjustment process. The effect brought by this design is that the data coverage realizes real-time collection of various parameters in the five links, and eight core parameters such as mixing uniformity, blowing agent decomposition rate, and profile temperature gradient are added compared with traditional methods, solving the problem of raw material problems caused by only measuring extrusion parameters.

[0031] 3. Fusion prediction model: To address the limitations of traditional models in simultaneously considering parameter correlations and temporal patterns, this module constructs a fusion model of random forest and long short-term memory network, achieving complementary advantages through the collaboration of the two modules. The random forest module consists of 100 decision trees. It randomly samples 70% of the window frame profile production samples using Bootstrap sampling and randomly selects 50% of the features for each node to split, extracting nonlinear correlations between parameters. For example, a 1% decrease in mixing uniformity requires a 2% increase in screw speed to maintain plasticity. The mathematical expression is: Where H(x) represents the final prediction result of the random forest, and the correlation between output parameters; x is the model input vector, composed of key parameters such as mixing uniformity and melting section temperature; argmax y∈Y To find the variable value corresponding to the maximum value, a voting mechanism is used to select the optimal result; y represents the prediction result of a single decision tree, such as acceptable plasticity or parameter influence coefficient; Y represents the set of possible prediction results, including all possible judgments; K represents the number of decision trees, which is set to 100 in this paper to balance accuracy and efficiency; h k (x) represents the prediction result of the k-th decision tree, which is independently trained based on random samples; I(h) k (x)=y) is an indicator function, which takes the value 1 if the condition is met and 0 otherwise. It is used to count the number of votes. The number of decision trees is set to 100 through 5-fold cross-validation. The prediction accuracy of 50 / 80 / 100 / 120 decision trees is tested respectively. The feature random selection ratio is 50%, based on the profile production parameter dimension, to avoid a single feature dominating the decision, such as relying solely on the melting section temperature.

[0032] LSTM parameters: The Long Short-Term Memory (LSTM) network module controls information transmission through input gates, forget gates, and output gates, capturing the temporal changes in parameters. For example, if the temperature in the foaming zone increases by 1°C, the decomposition rate increases by 5% after 3 minutes. The core cell state update formula is: , where C t Stores temporal information for the current cell state; f t The output of the forget gate takes a value from 0 to 1, retaining or forgetting historical information to filter out irrelevant fluctuations; ⊙ represents element-wise multiplication, regulating cell state dimension by dimension; C t-1 This transmits historical memory of the cell's state at the previous moment; t The input gate output takes a value from 0 to 1, filtering the current key information; The candidate cell state is represented by a value from -1 to 1, containing the latest information. The input sequence length for this module is set to 60, corresponding to 60 seconds of historical data. The hidden layer has 64 nodes to accommodate the temporal variation cycle of production parameters. Weighted fusion is performed using a random forest (0.4) and a long short-term memory network (0.6) weighting. Cross-validation of the temporal patterns in window frame profile production data has a more significant impact on prediction, resulting in the final prediction result. ; Dynamic optimization aspect every production 100 batches of window frame profiles, automatically include new data for incremental training random forest add 20% decision tree, long and short term memory network freeze 70% bottom weight to avoid forgetting history rule; When the prediction deviation is more than 3%, switch to gradient boosting tree backup model, train based on 500 batches of data to improve accuracy and ensure long-term accuracy, wherein the input sequence length is 60, corresponding to 60 seconds of historical data, because the previous statistics show that the significant fluctuation period of production parameters is about 50 to 60 seconds, such as the influence of environmental temperature change on foaming rate lagging about 40 seconds; The number of hidden layer nodes is 64, which is obtained by grid search verification.

[0033] During the training process, nearly 500 batches of data in one year were divided into 350 batches of training set, 100 batches of validation set and 50 batches of test set according to the ratio of 7:2:1, to ensure uniform distribution of raw material batches and environmental conditions; Then the mean square error (MSE) is used to calculate the prediction deviation, the optimizer is Adam, the learning rate is 0.001, β1=0.9, β2=0.999, the training iteration is 200 times, and the validation set loss is stopped after stabilization; After every 100 batches of new data, freeze 70% of the original decision trees of random forest, add 30% of the new decision trees, and train based on the new data; LSTM freezes 60% of the bottom weight, only updates the top fully connected layer, to avoid model forgetting historical rules.

[0034] In the foaming link of window frame profile production, the real-time data input into the model includes foaming zone temperature 180℃, pressure 1.0MPa, foaming agent addition amount 5g / min, ultrasonic frequency 30kHz; The random forest module extracts the associated features of current temperature and pressure, and obtains the decomposition rate of 80%; The long and short term memory network module captures the timing rule that the future 5 minutes of environmental temperature will rise by 2℃, and the decomposition rate will rise to 88% exceeding the standard 80% to 85%; After weighted fusion, the model output predicts that the foaming agent decomposition rate will exceed 85% after 5 minutes, and the cell diameter may reach 0.35mm exceeding the standard interval 0.1 to 0.3mm; The prediction result is transmitted to the collaborative control unit to trigger the advance adjustment instruction; Solve the problem that a single model cannot consider both correlation and timing; Capture the deviation 5 to 10 minutes in advance, such as predicting that the melting section temperature will drop by 2℃, adjust the heating power in advance to avoid insufficient plasticization, and solve the problem of accuracy decline caused by the lack of model update mechanism.

[0035] 4. Collaborative control strategy: Aiming at the problem of traditional regulation isolated adjustment, no pre-judgment guide, the module design from prediction, execution, verification to correction control strategy, linkage multi-link parameter; for the process characteristics of window frame profile production, raw materials and extrusion collaborative control if model prediction plasticizing efficiency < 90%, for example, the mixing uniformity 88% < standard 90%, the synchronous adjustment of the mixer speed from 300 r / min to 340 r / min, the stirring time from 5 minutes to 7 minutes, the mixing uniformity is improved to 92%; the screw speed of the extruder is pre-adjusted to 64 r / min, and the PID controller formula is: Wherein, u(t) is the controller output, that is, the heating power adjustment amount; K p =2.5 is the proportional coefficient, which responds to the deviation quickly; e(t) is the deviation, that is, the difference between the target temperature and the current temperature, such as 180℃-178℃=2℃; K i =0.1 is the integral coefficient, which eliminates static error; is the integral operation, which accumulates historical deviation; K d =0.05 is the differential coefficient, which predicts the deviation trend; is the differential operation, which calculates the deviation rate; through the formula, the melting section temperature is raised from 178℃ to 180℃, the precision is ±0.5℃, and the plasticizing degree is ≥95%; when the screw torque fluctuation is more than 5%, the feeding speed is adjusted by ±0.5 r / min, the molten glue pressure fluctuation is maintained ≤0.02 MPa, wherein the initial parameter setting, first set K p =3.0, K i =0, K d =0, the temperature overshoots to 185℃ after the heating power output, the target is 180℃, the overshoot is 5℃, it is judged that Kp is too large; reduce Kp to 2.5, K i =0.1, eliminate static error, K d =0.05, suppress overshoot, at this time the temperature rises from 178℃ to 180℃ in 1 minute, the overshoot is ≤1℃, the fluctuation range is ±0.5℃, which meets the plasticizing demand; verify under different environment temperature, the parameters can maintain temperature accuracy, finally solidify K p =2.5, K i =0.1, K d =0.05.

[0036] The partitioned ultrasonic wave assisted foaming divides the foaming zone into three zones, and the cell requirements of 0.1 to 0.3 mm of the window frame profile are independently controlled. When the decomposition rate in the start-up zone is less than 60%, high frequency of 35 to 45 kHz plus small amplitude of 5 to 8 μm is used to accelerate the decomposition of the foaming agent; when the cell diameter deviation in the growth zone is greater than 0.05 mm, medium frequency of 25 to 35 kHz plus medium amplitude of 8 to 12 μm is used to promote uniform growth; in the setting zone, low frequency of 15 to 25 kHz plus large amplitude of 12 to 15 μm is used to enhance the cell wall strength; when the pressure fluctuation is ±0.02 MPa, the foaming agent addition amount is fine-tuned by ±0.1 g / min, such as the addition amount is reduced from 5 g / min to 4.8 g / min when the predicted decomposition rate is more than 85%; the molding and cooling are optimized, and when the width deviation of the window frame profile is more than 0.2 mm, the vacuum degree of the setting sleeve is adjusted from-80 kPa to-95 kPa 10 to 15 seconds in advance to enhance the adsorption force, the mold outlet temperature is reduced from 170 °C to 166 °C to reduce the material flowability, the cooling system infrared thermal imager is used for partitioned cooling, the high temperature zone is cooled by 22 °C water and 18 m 3 / h wind, the medium temperature zone is cooled by 25 °C water and 15 m 3 / h wind, and the low temperature zone is cooled by 30 °C water and 10 m 3 / h wind; when the warping degree is more than 0.3%, the air volume of the corresponding area is adjusted by ±0.5 m 3 / h, such as the left air volume is increased from 18 m 3 / h to 18.5 m 3 / h.

[0037] Among them, for the parameters related to each other and influencing each other in each production link, such as raw material mixing and extrusion plasticization, foaming reaction and molding and setting, the coordinated adjustment logic and operation criteria are formulated, and the core is to solve the problem of abnormal other parameters caused by single parameter adjustment, and to realize the overall optimization of multiple links and multiple parameters: in the production process, the parameters of each link have strong correlation, that is, coupling: for example, the uniformity of the raw materials of the mixing machine will directly affect the plasticizing efficiency of the extruder, and the extrusion temperature will change the decomposition rate of the foaming agent, and then affect the dimensional accuracy in the molding stage; if a parameter is adjusted alone, such as only increasing the extrusion temperature, it may cause excessive foaming, and thus reduce the product quality; therefore, the core of the coupled adjustment rule is: when a parameter deviates from the standard, not only the parameter itself is adjusted, but also the associated parameters are adjusted, the side effects of single adjustment are offset through multiple parameter coordination, and the overall production state is stable, and the core coupled adjustment rule can be divided into three categories: (1) Coupling regulation rules of raw material processing and extrusion plasticization: associated parameters, mixing uniformity A, screw speed B, extruder barrel temperature C, feeding speed D; when the mixing uniformity A decreases by 1%, for example, from 92% to 91%, the screw speed B increases from 300 r / min to 306 r / min to enhance the shearing mixing effect; the melting section temperature C increases from 180°C to 181°C to compensate for the plasticization delay caused by insufficient dispersion of raw materials; the feeding speed D decreases from 50 kg / h to 49.5 kg / h to avoid extrusion fluctuations caused by uneven raw materials; when A returns to more than 92%, parameters B, C, and D are inversely adjusted according to the original proportion to avoid over-adjustment; in this way, the interlocking problems of uneven raw material mixing, insufficient plasticization, and decreased bending strength are solved, ensuring stable plasticization degree.

[0038] (2) Coupling regulation rules of ultrasonic foaming and pressure control: associated parameters, ultrasonic amplitude E, foaming agent addition amount F, foaming zone pressure G, cell density H; when the cell density H is lower than the standard value of 5%, for example, from 500 cells / cm 3 to 475 cells / cm 3 , the ultrasonic amplitude E increases from 10 μm to 11 μm to enhance the cavitation effect and promote cell generation; the foaming agent addition amount F increases from 5 g / min to 5.15 g / min to supplement the foaming power; the foaming zone pressure G decreases from 100 kPa to 98 kPa to provide space for cell growth; if H exceeds the standard value by 3%, then E decreases by 8%, F decreases by 2%, and G increases by 1 kPa to avoid strength decrease caused by excessive cell size; balance foaming efficiency and cell stability to ensure that the foaming rate fluctuates ≤±3%.

[0039] (3) Coupling regulation rules of molding and cooling system: associated parameters, setting sleeve vacuum degree I, mold temperature J, cooling water temperature K, profile width deviation L; when the width deviation L is +0.2 mm, then the upper limit is increased by 0.1 mm, the setting sleeve vacuum degree I is increased from -90 kPa to -95 kPa to enhance the adsorption and setting effect; the mold temperature J is decreased from 170°C to 168°C to reduce material expansion; the cooling water temperature K is decreased from 28°C to 27°C to accelerate surface solidification and lock the size; when L is -0.15 mm, which is lower than the lower limit, I decreases by 3 kPa, J increases by 1°C, and K increases by 0.5°C to avoid excessive narrow profile; control the size precision within ±0.1 mm to solve the size deviation problem caused by asynchronous molding and cooling.

[0040] Built-in 500+ batch production parameter correlation data, such as the correlation coefficient of A and B is 0.89, and the correlation coefficient of E and H is 0.92, to provide data support for the rules; when multiple parameters deviate at the same time, order by quality impact weight, such as bending strength related parameters are preferred to appearance parameters; after every 100 batches of production, update the rule parameters based on new data, such as adjust the linkage ratio of amplitude and blowing agent, to ensure adaptability; coupling adjustment rule is the core strategy of PVC foamed profile production control, its essence is to replace local adjustment with systematic thinking, to solve the problem of pulling one hair and moving the whole body in the production process by defining the linkage relationship and adjustment ratio between parameters, and finally achieve the goal of stable product quality and reduced energy consumption.

[0041] In the forming and cooling links of window frame profile production, the model predicts that the profile width will deviate by 0.3mm in the next 10 minutes, the coordinated control unit sends instructions 12 seconds in advance, the vacuum degree of the sizing sleeve is increased from-80kPa to-95kPa, the mold outlet temperature is reduced from 170℃ to 166℃, and the cooling system high temperature zone air volume is increased from 15m 3 / h to 18m 3 / h; the multi-device execution unit controls the sizing sleeve vacuum pump, mold temperature machine, and cooling fan to execute synchronously, and completes parameter adjustment within 10 seconds; the infrared thermal imager detects the profile temperature gradient in real time, and the high temperature zone temperature is reduced from 45℃ to 42℃; the width detection initial value is 60.08mm deviation 0.08mm still close to the upper limit of the standard, the vacuum degree is fine-tuned to-98kPa, and the final width is stabilized at 60.05mm deviation 0.05mm, and the warpage is 0.3% which meets the ≤0.5% standard.

[0042] 5. Dynamic quality evaluation and deviation tracing: To solve the problem of one-size-fits-all and difficult to trace of traditional evaluation standard, this module constructs a dynamic weight, standardized scoring and correlation analysis system; according to the difference between window frame profile for structure and skirting line for decoration, the dynamic weight allocation is 30% for bending strength of window frame profile for structure, 25% for dimensional accuracy, 15% for heat distortion temperature, 10% for impact strength, 5% for density, 5% for surface skin thickness, 5% for aging resistance, and 5% for surface finish, with mechanics priority; for skirting line for decoration, the surface finish is 30%, the surface skin thickness is 20%, the dimensional accuracy is 15%, the aging resistance is 10%, the bending strength is 10%, the impact strength is 5%, the heat distortion temperature is 5%, and the density is 5%, with appearance priority; the standardized score is executed according to industry specifications, and the detection value is 10% better than the standard to get 100 points, and 6 points are deducted for every 1% lower, the formula is: Wherein, Si is the standardized score of the i-th quality index, 0-100 points; Vi is the actual detection value, such as bending strength 62 MPa; Vi0 is the standard value, such as bending strength 60 MPa; 1.1Vi0 is the excellent threshold, such as 66 MPa, 100 points; 0.65Vi0 is the serious unqualified threshold, such as 39 MPa, 0 points; the score of the intermediate interval is calculated by linear interpolation or deduction, realizing the unified quantification of different indexes; the comprehensive quality score formula is: Wherein, Q is the comprehensive quality score, the window frame profile qualified standard is 92 points; n is the total number of quality indexes, that is, 8; wi is the weight of the i-th index, such as bending strength 0.3; Si is the standardized score; the window frame profile qualified standard is 92 points, the skirting line qualified standard is 90 points, and the single index score shall not be less than 70 points; the deviation traceability uses Pearson correlation coefficient r≥80% to screen the correlation parameters, and the formula is: Wherein, r is the correlation coefficient, -1 to 1, ≥0.8 is strong correlation; n is the sample quantity, that is, 500 batches; x i is the operating parameter value, such as mold temperature; y i is the quality index value, such as surface finish; the numerator and denominator are eliminated by product sum, square sum and other operations, which accurately identify the key influence parameters of quality defects, such as the correlation parameters of low positioning melting section temperature, high bubble density and the like when the bending strength of window frame profile is insufficient, and the adjustment scheme is generated in 15 minutes.

[0043] In the quality evaluation link of window frame profile production, the detection data includes bending strength 62 MPa, dimensional accuracy 60.05 mm, thermal deformation temperature 72℃, impact strength 11 kJ / m 2 , density 0.62 g / cm 3 , surface skin thickness 0.8 mm, aging resistance 82%, surface finish Ra=0.7 μm; in the standardized score, the bending strength 62 MPa>60 MPa scores 93.3 points, the dimensional accuracy 60.05 mm≤60+0.1 mm scores 95 points, and the rest of the indexes all score≥90 points; the comprehensive score Q=0.3×93.3+0.25×95+0.15×90+0.1×90+4×0.05×90=93.5 points, which is higher than the qualified standard 92 points to determine that it meets the standard.

[0044] In the deviation traceability link of window frame profile bending strength only 58 MPa, the score is 78 points, the correlation analysis calculates the Pearson correlation coefficient, the bending strength and the melting section temperature r=0.85, the bubble density r=-0.82, the correlation degree is more than 80%; parameter comparison shows that the melting section temperature is actually 176℃, which is lower than the standard 180℃, and the bubble density is actually 450 pieces / cm 3 , which is higher than the standard 400 pieces / cm 3; the generation adjustment scheme will melt the segment temperature to 180℃, the foaming agent addition amount is reduced from 5g / min to 4.9g / min, and the expected bending strength is increased to 62MPa; after the adjustment, the bending strength reaches 63MPa, scoring 95 points, and the comprehensive score is 94 points, reaching the standard.

[0045] 6. Comprehensive control system architecture: In terms of core unit function cooperation, the dual-mode data acquisition unit collects various parameters for window frame profile production at a frequency of 1 time / second, and transmits the preprocessed data to other units; the fusion prediction modeling unit outputs 5 to 10 minute deviation trends, such as predicting that the foaming agent decomposition rate is out of tolerance, with an accuracy of ≥98%; the cooperative control unit generates linkage instructions, such as adjusting the parameters of the mixer and extruder, and controls multiple devices to execute synchronously; the intelligent quality evaluation unit calculates the comprehensive score to determine whether the window frame profile meets the standard; the deviation tracing iteration unit locates the key deviations such as insufficient bending strength and generates optimization schemes; the multi-device execution unit connects devices such as mixers, extruders, and ultrasonic devices to execute control instructions in response; in terms of auxiliary unit value, the energy optimization unit reduces the cooling air volume from 25m 3 / h to 20m 3 / h, the cooling water temperature from 20℃ to 28℃, and the unit energy consumption from 85kWh / ton to 78kWh / ton, reducing by 8.2%; the digital twin unit constructs a 1:1 virtual model of window frame profile production, and the new scheme is verified in a virtual environment first; the blockchain storage unit uploads the raw material batch resin SG-5 type, operating parameter melt segment temperature 180℃, and quality data bending strength 65MPa to the blockchain, and the fault tracing is ≤30 minutes; the remote operation and maintenance unit remotely monitors the window frame profile production of 3 factory areas through 5G network.

[0046] Implementation case: Taking batch production of PVC skinned foamed window frame profiles for example, a profile factory produces 10,000 tons of structural window frame profiles per year, with a total of 2 production lines; the product requires to execute industry standards, with bending strength ≥60MPa, dimensional accuracy ±0.1mm, heat distortion temperature ≥70℃, comprehensive score ≥92 points, and unit energy consumption ≤80kWh / ton; equipment and system deployment includes raw material mixer with NIRS6500, SJZ65 / 132 extruder, US-4000 partition ultrasonic device, intelligent forming die, FLIRE60 infrared thermal imager, and the comprehensive control system designed in this research; model initialization imports nearly 1 year of 500 batches of window frame profile production data, and the initial accuracy of the fusion model is 98.2%.

[0047] Raw material processing 0-10 minutes, data acquisition shows that the mixing uniformity is 88%, the moisture content is 0.15%, and the particle size distribution is 0.2-0.4 mm; the fusion model predicts that the plasticizing efficiency will decrease to 88%, the mixing machine speed is adjusted to 340 r / min, the time is 7 minutes, and the extruder screw speed is predicted to 64 r / min; the result is that the mixing uniformity is 92%, and the data is transmitted to the extrusion link.

[0048] Extrusion plasticization link 10-30 minutes, data acquisition shows that the melting section temperature is 178℃, the screw torque is 120 N・m, and the feeding speed is 20 r / min; the model predicts that the temperature is low by 2℃→ plasticizing degree 92%, the PID controller raises the temperature to 180℃, and the feeding speed is reduced to 19.5 r / min; the result is that the plasticizing degree is 96%, the glue pressure is 1.1 MPa, and it is transmitted to the foaming link.

[0049] Foaming reaction link 30-50 minutes, data acquisition shows that the foaming zone temperature is 180℃, the pressure is 1.0 MPa, the decomposition rate is 78%, and the ultrasonic wave is 30 kHz; the model predicts that the decomposition rate is 88% after 5 minutes, the addition amount is reduced from 5 g / min to 4.8 g / min, and the ultrasonic wave is reduced to 28 kHz; the result is that the decomposition rate is 82%, and the cell density is 400 cells / cm 3 , which is transmitted to the forming link.

[0050] Forming and shaping link 50-70 minutes, data acquisition shows that the shaping sleeve vacuum degree is-80 kPa, the mold temperature is 170℃, and the extrusion speed is 70 mm / min; the model predicts that the width deviation is 0.3 mm, the vacuum degree is increased to-95 kPa, and the mold temperature is reduced to 166℃; the result is that the width is 60.05 mm, which is transmitted to the cooling link.

[0051] Cooling treatment link 70-90 minutes, data acquisition shows that the profile temperature gradient is 45℃ high temperature zone, 35℃ medium temperature zone, and 30℃ low temperature zone; adjust the high temperature zone 22℃ water and 18m 3 / h wind, the medium temperature zone 25℃ water and 15m 3 / h wind, and the low temperature zone 30℃ water and 10m 3 / h wind; the result is that the warpage is 0.3%, and the comprehensive score after cooling detection is 93.5 points.

[0052] Suppose that the surface finish of a batch of products is Ra=1.0 μm, the traceability correlation analysis locates the mold temperature at 172℃ which is 2℃ higher than the standard, and the cooling water temperature at 28℃ which is 3℃ higher than the standard; the mold temperature is adjusted to 170℃ and the cooling water temperature is adjusted to 25℃; the result is that the surface finish is Ra=0.8 μm with a score of 84 and a comprehensive score of 91, which meets the standard; the long-term effect of continuous production for 3 months shows that the quality index fluctuates by ±0.5 MPa in bending strength and ±0.1 mm in size deviation, the quality meets the standard at a rate of 99%; the energy consumption index is 78 kWh / ton, which is 8.2% lower than the traditional one, and the annual electricity cost is more than 80,000 yuan; the efficiency index is 0.3 hours per day, which is reduced by 80%, the changeover time is 1.5 hours, which is reduced by 62.5%, and the daily output is 23 tons, which is increased by 15%.

[0053] Finally, it should be noted that: obviously, the above embodiments are only examples for clearly illustrating the present application, and are not limitations on the embodiments. Based on the above description, other different forms of changes or variations can also be made by those of ordinary skill in the art. Here, it is not necessary and impossible to exhaust all the embodiments. The obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. A comprehensive control method for the extrusion molding process of PVC skin-foamed profiles, characterized in that, The control method includes: Collect real-time operational data for the entire process of PVC skin foaming profiles. The real-time operational data covers parameters for each stage, including raw material mixing, extruder operation, foaming reaction, mold forming, and cooling treatment. A fusion-based predictive model is built using historical production and operation data. This model combines two data processing methods: one for mining complex correlation features between parameters and the other for tracking the time-series changes of parameters. Real-time operation data is input to continuously optimize the model and analyze the deviation trends of parameters in each stage. Based on the analysis results and real-time feedback of production operation data, formulate a control scheme that coordinates verification and correction. The entire extrusion molding process is coordinated and adjusted according to the collaborative control scheme. Full quality inspection data of the formed profiles are obtained through online testing; The weights of each quality indicator are dynamically allocated according to product usage requirements, and a comprehensive quality score is calculated. If the overall quality score meets the preset pass standard, the current collaborative control scheme is maintained; if it does not meet the preset pass standard, the key parameters of quality deviation are located by analyzing the correspondence between quality defects and real-time operation data, and the collaborative control scheme is iteratively optimized until the overall quality score meets the preset pass standard.

2. The comprehensive control method for the extrusion molding process of PVC skin-forming foamed profiles according to claim 1, characterized in that, Real-time operational data is collected using a dual-mode acquisition method combining distributed sensing and online analysis. The process includes: Data from the raw material processing stage was collected through a combination of near-infrared spectroscopy and particle size analysis. The collected data included the mixing uniformity of PVC resin and auxiliary materials, moisture content, particle size distribution, proportion of auxiliary materials, and dispersion uniformity. Extruder operating data is collected through the extruder's built-in sensor array. The collected data includes real-time temperature of each section of the extruder barrel, screw speed, feeding speed, melt pressure, material plasticity, and screw torque. Data from the foaming reaction process is collected using an integrated detection device. The collected data includes foaming zone temperature, foaming zone pressure, amount of foaming agent added, foaming agent addition accuracy, foaming agent decomposition rate, ultrasonic vibration frequency, and ultrasonic amplitude. Data during the mold forming process is collected through sensors built into the mold. The collected data includes mold zone temperature, sizing sleeve vacuum, sizing time, mold cavity pressure, and profile extrusion speed. Data from the cooling process is collected using intelligent flow sensors and temperature monitoring instruments. The collected data includes cooling water temperature, cooling time, cooling air volume, cooling medium flow rate, and temperature gradient on the profile surface.

3. The comprehensive control method for the extrusion molding process of PVC skin-forming foamed profiles according to claim 1, characterized in that, The steps for building a fusion prediction model are as follows: Collect real-time operating data, raw material batch information, and corresponding profile quality inspection data for nearly a year to establish a standardized database. The raw material batch information includes resin molecular weight distribution, auxiliary material purity, stabilizer content, and lubricant ratio. Wavelet transform was used to denoise the database data, and key parameters were selected by calculating the Pearson correlation coefficient, retaining only those parameters with a correlation of ≥85% with the quality of the profiles. A fusion prediction model was built, with the selected key parameters as input and the parameter deviation trend as output. Two data processing methods were used to extract parameter correlation features and capture time series change patterns, respectively, to complete the model training. Set up a model self-updating mechanism, and incorporate new data for incremental training after each preset batch production is completed; A model anomaly verification function has been added. When the prediction deviation exceeds the preset deviation ratio, the system switches to the backup gradient boosting tree model for continuous prediction.

4. The comprehensive control method for the extrusion molding process of PVC skin-forming foamed profiles according to claim 1, characterized in that, The collaborative control scheme includes collaborative control of raw material handling and extruder operation, specifically including: The impact of raw material mixing uniformity on plasticizing efficiency is predicted by a fusion prediction model, and a coupled adjustment rule is established between mixing uniformity, screw speed and feeding speed. If the predicted plasticizing efficiency is less than the preset plasticizing ratio, the stirring speed and stirring time in the raw material processing stage will be adjusted, and a pre-adjustment command will be sent to the extruder control system to control the screw speed. Based on the proportion of auxiliary materials obtained by near-infrared spectroscopy detection, the corrected value of the material melt index is calculated, and the temperature of the extruder barrel feeding section, compression section, melting section and homogenization section is dynamically adjusted by a PID controller; Simultaneously monitor the temperature and screw torque changes of each section of the extruder barrel. When the torque fluctuation is greater than 5%, adjust the feeding speed to maintain the pressure of the melt.

5. The comprehensive control method for the extrusion molding process of PVC skin-forming foamed profiles according to claim 1, characterized in that, The collaborative control scheme also includes ultrasonic-assisted foaming control, specifically including: A zoned ultrasonic vibration device is set up in the foaming reaction zone, which is divided into a foaming start-up zone, a cell growth zone, and a cell shaping zone. The ultrasonic parameters of each zone can be adjusted independently. A. Foaming start-up zone: When the decomposition rate of the foaming agent is <60%, high-frequency, small-amplitude ultrasonic vibration is used to accelerate the decomposition of the foaming agent by utilizing the ultrasonic cavitation effect. When the decomposition rate is ≥80%, the parameters are switched. B. In the bubble growth zone, the bubble morphology is monitored in real time using a high-speed camera. If the bubble diameter deviation is >0.05mm, the vibration is adjusted to medium frequency and medium amplitude. C. The bubble-forming zone uses low-frequency, large-amplitude vibration; The pressure change in the foaming zone is monitored synchronously throughout the process. When the pressure fluctuation is greater than 0.02 MPa, the amount of foaming agent added is adjusted to maintain a constant pressure.

6. The comprehensive control method for the extrusion molding process of PVC skin-forming foamed profiles according to claim 1, characterized in that, The collaborative control scheme also includes the coordination of mold forming and cooling processes, specifically including: By using a fusion prediction model to predict the trend of profile dimensional deviation, a coordinated adjustment rule for mold temperature, sizing sleeve vacuum degree and cooling rate is established. If the predicted positive deviation of the profile width is >0.2mm, increase the vacuum degree of the shaping sleeve before the material enters the mold, and adjust the mold outlet temperature and the initial cooling air volume. Infrared thermal imagers are used to detect the surface temperature distribution of the formed profile, which is divided into high-temperature zone, medium-temperature zone and low-temperature zone, and zoned cooling control is adopted.

7. The comprehensive control method for the extrusion molding process of PVC skin-forming foamed profiles according to claim 1, characterized in that, The steps for calculating the overall quality score include: The quality inspection indicators are profile density, dimensional accuracy, surface skin thickness, bending strength, impact strength, heat distortion temperature, aging resistance and surface finish; Based on the quality inspection indicators, the weights of the indicators are assigned, the test values ​​of the indicators are calculated, and the test values ​​of each indicator are compared with the quality standard indicators to convert them into standardized scores from 0 to 100. The weighted total score is calculated as the comprehensive quality score and compared with the preset pass standard. When the comprehensive quality score is greater than the preset pass standard, a pass instruction is issued; otherwise, a fail instruction is issued.

8. The comprehensive control method for the extrusion molding process of PVC skin-forming foamed profiles according to claim 1, characterized in that, The steps for locating key parameters of quality deviation and iterating solutions are as follows: If the overall quality score fails to meet the standard, the mapping relationship between the quality inspection data and the real-time operation data is analyzed by association rule mining algorithm to locate the key parameters of quality deviation. When the critical deviation is insufficient dimensional accuracy, prioritize adjusting the vacuum gradient of the shaping sleeve and the mold cavity pressure, and simultaneously correct the cooling zone parameters. When the critical deviation is insufficient bending strength, adjust the extruder melt zone temperature, screw speed, and cell density. After each adjustment, the optimization effect is predicted by the fusion prediction model. If the predicted overall quality score improves by ≥3 points, the adjustment is executed; otherwise, the deviation source is traced again. The quality is considered to be met after ≤3 iterations.

9. A comprehensive control system for the extrusion molding process of PVC skin-foamed profiles, characterized in that, The control system includes: The dual-mode data acquisition unit consists of a distributed sensor array, an online analysis instrument, and a data transmission module, which detects the acquisition and preprocessing of real-time running data throughout the entire process. The fusion prediction modeling unit integrates a fusion prediction model, a data denoising module, and a model self-updating module to complete the prediction of parameter deviation trends and dynamic optimization of the model. The collaborative control unit, which includes a parameter coupling rule base, a PID adjustment module, and a linkage control module, formulates and executes collaborative control schemes. The intelligent quality assessment unit integrates a quality indicator weight allocation module and a hierarchical analysis calculation module, and calculates a comprehensive quality score by combining online detection data to determine whether the standard is met. The deviation tracing and iteration unit locates key parameters of quality deviation through the association rule mining module, thereby realizing the automatic iterative optimization of the collaborative control scheme; A multi-device execution unit connects and controls the coordinated operation of raw material mixing equipment, extruder, zoned ultrasonic vibration device, intelligent molding die and zoned cooling equipment; Each unit communicates with other units via industrial Ethernet.

10. The integrated control system for the extrusion molding process of PVC skin foam profiles according to claim 9, characterized in that, The control system also includes: The energy consumption optimization unit monitors the energy consumption data of each device in real time and reduces the energy consumption per unit product by adjusting parameters while meeting quality requirements. Digital twin units construct virtual simulation models that correspond 1:1 to the physical production line, mapping the equipment operating status in real time, and enabling virtual verification through collaborative control schemes; The intelligent auxiliary decision-making unit generates parameter adjustment suggestions based on historical databases and real-time operating conditions, and automatically outputs alternative solutions when the system malfunctions. The blockchain-based evidence storage unit uploads raw material information, real-time operational data, quality testing data, and scheme adjustment records to the blockchain system. The remote operation and maintenance unit enables remote monitoring, parameter debugging, and fault diagnosis of equipment via a 5G network.

Images