Efficient film bag feeding control system and method

Abstract

The invention relates to the field of co-extrusion production control, and particularly discloses an efficient film bag feeding control system, which is applied to a multi-raw-material synchronous conveying device of a co-extrusion process, and comprises a full-link data acquisition module, a full-link data processing module and a full-link data processing module, the raw material characteristic parameters comprise a melting point T, a stacking density rho and a magnetic sensitive particle loading amount M; the instantaneous load pressure P of each raw material conveying channel, the temperature Tm of a co-extrusion die head melting area, the real-time thickness H of a co-extrusion film layer, stock bin liquid level data and motor rotating speed data are collected and integrated to obtain a full-link original data set; the single-drive linkage conveying mechanism comprises a main servo motor, a plurality of sets of electromagnetic clutches and a plurality of sets of spiral conveying assemblies, and the main servo motor drives a transmission shaft through a synchronous belt. According to the technical scheme, for the co-extrusion process, the multi-raw-material synchronous conveying precision can be matched, and the problems that parameter adjustment lags behind, adaptability is poor and a downstream process feedback closed loop is lacked are solved.

Description

Technical Field

[0001] This invention relates to the field of co-extrusion production control, and in particular to a high-efficiency feeding control system and method for film bags. Background Technology

[0002] In the field of film bag co-extrusion production, the co-extrusion process, due to its ability to achieve multi-layer material composites to improve the barrier properties, strength, and functional characteristics (such as magnetic adhesion) of film bags, has been widely used in food vacuum packaging, medical protective bags, and other scenarios. The core structure of the existing co-extrusion feeding system includes multiple sets of spiral conveyor components (corresponding to different raw material channels), independent drive motors (each conveyor component is equipped with a servo motor), a basic parameter monitoring module (monitoring only motor speed and raw material hopper level), and an execution module (motor speed regulator, hopper replenishment valve). The data acquisition points are concentrated at the front end of the feeding process (such as hopper level sensor and motor speed encoder), and the execution points only cover motor drive and replenishment valve control. The entire system relies on preset parameters and open-loop conveying logic for operation.

[0003] However, this system suffers from insufficient accuracy in synchronous conveying of multiple raw materials in practical applications. This is because the system uses multiple motors to independently drive each spiral conveyor component, based on the "motor synchronous belt drive" concept disclosed in Chinese Patent Publication No. CN108107830A. However, it lacks a linkage mechanism designed to address the differences in the characteristics of the co-extruded raw materials. When conveying raw materials with different physical properties (such as magnetically sensitive particles and conventional polyolefin particles), fluctuations in conveying resistance due to differences in material bulk density and flowability can cause a disconnect between the motor output speed and the actual material conveying volume. This results in deviations in the conveying ratio of each raw material exceeding preset values, directly causing uneven co-extruded film thickness (unevenness). This necessitates frequent corrections by the subsequent amplitude adjustment compensation mechanism, reducing production efficiency.

[0004] There are also problems with parameter adjustment lag and poor adaptability. The above system can only preset feeding parameters based on a single raw material type. When changing raw materials (such as switching from ordinary polyolefin to vacuum bag film materials with different melting points) or adjusting the formula (such as changing the loading amount of magnetically sensitive particles), it is necessary to manually recalibrate parameters such as the speed of the screw conveyor assembly and the torque of the motor. The operation is time-consuming and cannot match the dynamic balance of the melting rate and feeding rate at the co-extrusion die head in real time, resulting in raw material waste.

[0005] Furthermore, there is a lack of downstream process feedback loops. The existing data acquisition points do not cover key downstream nodes of co-extrusion, and can only monitor the status of the feeding front end, unable to receive feedback signals such as the temperature of the die melting zone and the quality of the film. For example, when the high-melting-point raw material at the co-extrusion die is not fully melted, the above system cannot adjust the feeding rate based on the die temperature sensor signal, and the problem needs to be discovered through manual inspection. This lag leads to batch products being unqualified.

[0006] Therefore, there is an urgent need for a high-efficiency film bag feeding control system and method that can match the synchronous conveying accuracy of multiple raw materials in co-extrusion processes, solve the problems of parameter adjustment lag and poor adaptability, and address the lack of downstream process feedback closed loop. Summary of the Invention

[0007] This invention provides a high-efficiency film bag feeding control system that addresses the issues of synchronous conveying accuracy of multiple raw materials in co-extrusion processes, the problem of lag in parameter adjustment and poor adaptability, and the lack of a closed-loop feedback mechanism from downstream processes.

[0008] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0009] A high-efficiency film bag feeding control system, applied to multi-raw material synchronous conveying devices in co-extrusion processes, including:

[0010] The end-to-end data acquisition module is used to acquire the raw material characteristic parameters of the raw material to be transported, including melting point T, bulk density ρ and magnetically sensitive particle loading M; it also collects the instantaneous load pressure P of each raw material conveying channel, the melting zone temperature Tm of the co-extrusion die, the real-time thickness H of the co-extrusion film, the silo liquid level data and the motor speed data, and integrates them to obtain the end-to-end raw dataset.

[0011] A single-drive linkage conveying mechanism includes a main servo motor, multiple sets of electromagnetic clutches, and multiple sets of screw conveyor components. The main servo motor drives the transmission shaft through a synchronous belt. The transmission shaft is connected to the screw conveyor components of the corresponding raw material channels through different sets of electromagnetic clutches. The electromagnetic clutches are used to independently control the power transmission on / off and transmission ratio of each raw material channel.

[0012] The intelligent control unit is equipped with a semantic encoding algorithm, a timing matching algorithm, and a fuzzy PID closed-loop algorithm. The intelligent control unit is used to obtain the optimal feeding rate benchmark value based on the raw material characteristic parameters and the semantic encoding algorithm and timing matching algorithm. The fuzzy PID closed-loop algorithm extracts the instantaneous load pressure P, melting zone temperature Tm, and film thickness H from the original dataset of the entire chain, calculates the deviation value of the above parameters from the preset threshold, and dynamically generates adjustment instructions in combination with the optimal feeding rate benchmark value.

[0013] The feedback execution module receives adjustment commands and adjusts the power transmission ratio of the corresponding raw material channel by controlling the disengagement time of each electromagnetic clutch. At the same time, it adjusts the output speed of the main servo motor to match the actual conveying rate of each raw material with the melting rate of the co-extrusion process.

[0014] The basic principle and beneficial effects of the solution are as follows: This invention focuses on the core requirement of precise and coordinated transportation of multiple raw materials in the co-extrusion process, and constructs a closed-loop control logic of full-link data acquisition, intelligent algorithm processing, and linkage execution feedback.

[0015] The end-to-end data acquisition principle achieves comprehensive data acquisition by covering multiple data collection points across raw material characteristics, the feeding process, and downstream co-extrusion. Raw material characteristic parameters (melting point T, bulk density ρ, and magnetically sensitive particle loading M) directly relate to the conveying and melting characteristics of the raw material (e.g., high-melting-point T raw materials require a slower feeding rate to ensure sufficient melting, and high-ρ raw materials are prone to load fluctuations due to high conveying resistance); instantaneous load pressure P reflects real-time resistance changes in the raw material conveying channel (e.g., particle agglomeration causes a sudden increase in P); and the co-extrusion die melting zone temperature Tm and real-time film thickness H directly relate to the final quality of the co-extrusion process (e.g., a co-extrusion die melting zone temperature Tm below a threshold indicates insufficient raw material melting, and H deviation reflects an imbalance in the raw material conveying ratio). By integrating the above data to form an end-to-end raw dataset, complete data support is provided for subsequent algorithm processing, avoiding the information bias caused by traditional systems that only collect data from the feeding end.

[0016] Abandoning the traditional multi-motor independent drive mode, this design adopts a linkage structure of a single main servo motor and multiple electromagnetic clutches. The main servo motor provides a unified power source by driving the transmission shaft via a synchronous belt, ensuring consistent power across all material channels. Each set of electromagnetic clutches corresponds to a different material channel, and by independently controlling their on / off states and disengagement times, the power transmission ratio of each screw conveyor component along the transmission shaft is adjusted. For example, when conveying materials with high bulk density ρ, the closing time of the electromagnetic clutch is extended to increase the power share of that channel and compensate for conveying resistance; when conveying materials with low melting point T, the closing time is shortened to reduce the conveying rate and prevent excessive accumulation of materials in front of the die. This structure solves the synchronization problem caused by the response delay of multiple motors at the power source level, laying the hardware foundation for precise proportioning of multiple materials.

[0017] The data-to-instruction conversion is achieved through a three-level algorithm: semantic encoding, temporal matching, and fuzzy PID closed-loop. First, the semantic encoding algorithm transforms unstructured raw material characteristic parameters (T, ρ, M) into quantifiable feature vectors, establishing a foundation for the correlation between raw material characteristics and feeding parameters. Second, the temporal matching algorithm performs 1D-CNN temporal encoding on historical motor speed data, extracting short-term fluctuations and long-term trends in the feeding rate. Combined with the raw material characteristic vector, mutual information is calculated to determine the optimal feeding rate baseline value for the current raw material (e.g., when the loading amount M of magnetically sensitive particles is high, the baseline value needs to match the core layer's magnetic performance requirements to avoid overloading or underloading). Finally, the fuzzy PID closed-loop algorithm dynamically corrects the baseline value by combining the deviation between real-time parameters (P, Tm, H) in the entire raw data set and a preset threshold. For example, when Tm is below the threshold, the corresponding raw material feeding rate is reduced to allow for a longer melting time; when H is too thick, the conveying power of high-percentage raw materials is reduced. This three-level algorithm progressively achieves both pre-adaptation of raw material characteristics and feeding parameters and dynamic response to process fluctuations through real-time feedback.

[0018] The feedback execution module translates the adjustment commands from the intelligent control unit into hardware actions. On one hand, it adjusts the power ratio of each channel by regulating the disengagement time of the electromagnetic clutch, achieving differentiated control of the conveying rate of different raw materials. On the other hand, it synchronously adjusts the output speed of the main servo motor to adapt to the overall rhythm of the co-extrusion process (e.g., when the die melting efficiency is improved, the main motor speed is increased to increase the overall material feeding). The two work together to ensure that the actual conveying rate of each raw material matches the co-extrusion melting rate in real time, forming a complete closed loop.

[0019] Traditional multi-motor independent drives are prone to large deviations in conveying ratios due to response delays. This invention addresses this by using a single main motor + electromagnetic clutch linkage structure, combined with an optimal reference value determined by a timing matching algorithm, to control the conveying rate deviation of each raw material within a low range. For example, in the production of magnetically sensitive film bags, the synchronous conveying deviation between magnetic particles (M=50%) and polyolefin raw materials is reduced, the magnetic properties of the co-extruded film core layer are more uniform, and the problem of uneven magnetic particle distribution leading to localized failure of magnets to adhere to the film is avoided.

[0020] Traditional systems require manual calibration for over 30 minutes when changing raw materials, and suffer from poor adaptability. This invention automatically identifies raw material characteristics (T, ρ, M) using a semantic coding algorithm, quickly generates baseline values ​​using a time-series matching algorithm, and corrects them in real time using a fuzzy PID closed-loop algorithm. The entire parameter adaptation process requires no manual intervention, significantly reducing time consumption. For example, changing raw materials from ordinary polyolefins (T = 130℃, ρ = 0.92 g / cm³)... 3 Switching to magnetically sensitive material (T=150℃, ρ=1.2g / cm) 3 When M=40%, the present invention automatically adjusts the feed rate benchmark value to the appropriate target parameter and the electromagnetic clutch closing time to the appropriate target parameter, thereby improving the adaptation efficiency and avoiding the waste of raw materials caused by errors in manual calibration.

[0021] Traditional systems only monitor data at the feeding end and cannot cope with fluctuations in downstream processes. This invention forms a downstream feedback loop by collecting the temperature Tm of the die head melting zone and the film thickness H. When Tm is below a preset threshold of 5°C, the fuzzy PID algorithm immediately reduces the corresponding raw material feeding rate to allow for additional melting time. When H is too thick, it reduces the power transmission to high-percentage raw materials, correcting the thickness deviation in real time. This closed loop transforms the problem of uneven co-extruded film thickness from traditional batch defects to real-time correction, improving product yield and avoiding batch scrap due to downstream process fluctuations.

[0022] The intelligent control unit's three-level algorithm adopts a lightweight model design. The calculation time for semantic encoding and timing matching is ≤10ms, the generation time for fuzzy PID closed-loop control instructions is ≤5ms, the hardware response delay of the feedback execution module is ≤20ms, and the total response time of the entire system from data acquisition to execution is ≤35ms. It can quickly respond to sudden situations such as raw material agglomeration (sudden increase in P) and die head temperature fluctuation (sudden change in Tm) to avoid the expansion of process deviation in a short period of time.

[0023] The semantic coding algorithm constructs feature vectors using multi-dimensional raw material parameters (T, ρ, M) and combines them with a classification model trained on historical data. It has a high accuracy rate in identifying different types of raw materials (such as polyolefins, magnetically sensitive composite materials, and high-barrier materials). It can accurately distinguish raw materials with similar characteristics (such as two polyolefins with similar T values ​​but a 5% difference in ρ values), avoid deviations in feeding parameter adaptation caused by incorrect raw material identification, and ensure the production stability of film bags with different formulations.

[0024] This invention is adaptable to various co-extruded film bag production scenarios. Whether it is a high-barrier film for food packaging (which requires precise control of the delivery ratio of barrier agent and substrate), a medical protective film (which requires matching the melting characteristics of low-melting-point raw materials), or a magnetically sensitive film bag (which requires control of the magnetic particle loading amount M and the core layer thickness), the control strategy can be automatically adjusted by inputting raw material characteristic parameters. There is no need to change hardware or reconstruct the algorithm, reducing process switching costs and meeting the needs of multi-variety, small-batch production.

[0025] On the one hand, a single master servo motor reduces standby power consumption compared to multiple motors driving independently; on the other hand, through precise synchronous delivery and real-time feedback correction, the waste rate of raw materials is reduced, significantly lowering production costs.

[0026] This invention eliminates the need for manual parameter calibration. Operators only need to input parameters such as T, ρ, and M through the raw material characteristic input interface, and the subsequent process runs automatically. At the same time, the intelligent control unit can automatically record the feeding parameter templates for different raw materials, which can be directly called up the next time they are used, reducing reliance on experienced operators, shortening the training cycle for new employees, and reducing the company's labor costs and the risk of operational errors.

[0027] Through closed-loop control across the entire supply chain, the thickness deviation of co-extruded films is reduced, the magnetic adhesion intensity fluctuation of magnetically sensitive film bags is reduced, and the oxygen permeability deviation of high-barrier films is reduced. This significantly improves product quality consistency, meets the stringent quality stability requirements of high-end fields such as food packaging and medical protection, and enhances product market competitiveness.

[0028] In summary, this invention addresses the issues of matching the synchronous conveying accuracy of multiple raw materials, the problem of lag in parameter adjustment and poor adaptability, and the lack of a closed-loop feedback mechanism from downstream processes in the co-extrusion process.

[0029] Furthermore, the intelligent control unit is used to obtain the optimal feeding rate benchmark value based on the semantic encoding algorithm and the temporal matching algorithm, including: the semantic encoding algorithm is used to call a preset thermophysical property embedding matrix to perform semantic embedding encoding on the raw material characteristic parameters in the full-link original dataset to obtain the raw material characteristic semantic vector Vmat; the temporal matching algorithm constructs the motor speed data in the full-link original dataset into a historical feeding rate time queue, performs temporal implicit encoding on the time queue through a preset 1D-CNN model to obtain the rate time sequence vector Vspeed, calculates the mutual information MI between the raw material characteristic semantic vector Vmat and the rate time sequence vector Vspeed, and determines the optimal feeding rate benchmark value for each raw material based on the mutual information MI.

[0030] Furthermore, the mutual information MI between the raw material characteristic semantic vector Vmat and the rate time series vector Vspeed is calculated using the following formula:

[0031] MI=H(Vmat)+H(Vspeed)-H(Vmat,Vspeed)

[0032] Where H(Vmat) represents the information entropy of Vmat, H(Vspeed) represents the information entropy of Vspeed, and H(Vmat,Vspeed) represents the joint entropy of Vmat and Vspeed.

[0033] Furthermore, the end-to-end data acquisition module includes load pressure sensors and raw material characteristic input interfaces installed at the outlets of each spiral conveyor component at the feeding end, as well as temperature sensors and film thickness detectors installed in the die melting zone downstream of co-extrusion. It also retains traditional silo level sensors and motor speed encoders. The raw material characteristic parameters, including melting point T, bulk density ρ, and magnetically sensitive particle loading M, are acquired through the raw material characteristic input interface. The instantaneous load pressure P of each raw material conveying channel is acquired through the load pressure sensors. The temperature Tm of the co-extrusion die melting zone is acquired through the die melting zone temperature sensor. The real-time thickness H of the co-extruded film is acquired through the film thickness detector. The silo level sensor and motor speed encoder respectively acquire silo level data and motor speed data, integrating them to obtain the end-to-end raw dataset.

[0034] Furthermore, in the semantic encoding algorithm, the thermophysical property embedding matrix is ​​obtained using the formula: E = α·E T +β·E ρ +γ·E M Construct, where: E is the final embedding matrix; E T The melting point characteristic matrix is ​​obtained by normalizing the deviation of T from the standard melting point; E ρE is the bulk density characteristic matrix, calculated by the ratio of ρ to the reference density; M The magnetic particle loading feature matrix is ​​obtained by converting the magnetically sensitive particle loading amount M by weight percentage; α, β, and γ are weighting coefficients, and α+β+γ=1.

[0035] Furthermore, the 1D-CNN model includes three convolutional layers and two pooling layers. The kernel size of the first convolutional layer is set to 5×1 with a stride of 1, which is used to extract short-term fluctuation features of the feeding rate. The kernel size of the second convolutional layer is set to 10×1 with a stride of 2, which is used to extract medium-term trend features. The kernel size of the third convolutional layer is set to 20×1 with a stride of 4, which is used to extract long-term periodic features. All pooling layers use max pooling.

[0036] Furthermore, in the fuzzy PID closed-loop algorithm, the deviation value is calculated using a dynamic weighting method, through the formula D = k P ·|P-P0|+k T ·|T m -T0|+k H • |H-H0| is achieved, where: D is the overall deviation value; k P k T k H For dynamic weighting coefficients; when P-P0>10%P0, k P Automatically adjusted to 0.5, k T and k H Each is 0.25, prioritizing the correction of load fluctuations; when Tm-T0>5℃, k T Automatically adjusted to 0.5, k P and k H Each is 0.25, prioritizing the melting effect; the preset threshold for instantaneous load pressure P is P0, the preset threshold for melting point T is T0, and the preset threshold for real-time thickness H of the co-extruded film layer is H0.

[0037] Furthermore, the intelligent control unit also includes a parameter self-optimization module, which optimizes parameters using formulas. The proportional coefficient, integral coefficient, and derivative coefficient of the fuzzy PID controller are iteratively optimized, where: K n+1 The parameter values ​​for the (n+1)th iteration, K n Here, λ is the parameter value for the nth iteration, λ is the learning rate (0.01-0.1), θ is the decay coefficient (0.05-0.2), and n is the number of iterations.

[0038] Furthermore, in the time-series matching algorithm, the calculation of mutual information MI incorporates a temperature correction factor, and the corrected formula is MI. ′ =MI·(1+δ·|T) m-T0| / T0), where: MI ′ δ represents the corrected mutual information; δ is the temperature sensitivity coefficient, which is 0.8 for raw materials with a melting point above 150℃ and 0.3 for raw materials with a melting point below 150℃. Attached Figure Description

[0039] Figure 1 This is a logic block diagram of an embodiment of a film bag efficient feeding control system. Detailed Implementation

[0040] The following detailed description illustrates the specific implementation method:

[0041] High-efficiency film bag feeding control system (such as Figure 1 As shown), a multi-raw material synchronous conveying device applied to the co-extrusion process includes:

[0042] The end-to-end data acquisition module is used to acquire the raw material characteristic parameters of the raw material to be transported, including melting point T, bulk density ρ and magnetically sensitive particle loading M; it also collects the instantaneous load pressure P of each raw material conveying channel, the melting zone temperature Tm of the co-extrusion die, the real-time thickness H of the co-extrusion film, the silo liquid level data and the motor speed data, and integrates them to obtain the end-to-end raw dataset.

[0043] A single-drive linkage conveying mechanism includes a main servo motor, multiple sets of electromagnetic clutches, and multiple sets of screw conveyor components. The main servo motor drives the transmission shaft through a synchronous belt. The transmission shaft is connected to the screw conveyor components of the corresponding raw material channels through different sets of electromagnetic clutches. The electromagnetic clutches are used to independently control the power transmission on / off and transmission ratio of each raw material channel.

[0044] The intelligent control unit is equipped with a semantic encoding algorithm, a timing matching algorithm, and a fuzzy PID closed-loop algorithm. The intelligent control unit is used to obtain the optimal feeding rate benchmark value based on the raw material characteristic parameters and the semantic encoding algorithm and timing matching algorithm. The fuzzy PID closed-loop algorithm extracts the instantaneous load pressure P, melting zone temperature Tm, and film thickness H from the original dataset of the entire chain, calculates the deviation value of the above parameters from the preset threshold, and dynamically generates adjustment instructions in combination with the optimal feeding rate benchmark value.

[0045] The feedback execution module receives adjustment commands and adjusts the power transmission ratio of the corresponding raw material channel by controlling the disengagement time of each electromagnetic clutch. At the same time, it adjusts the output speed of the main servo motor to match the actual conveying rate of each raw material with the melting rate of the co-extrusion process.

[0046] In practical use: This embodiment takes the production scenario of a three-layer co-extruded magnetically sensitive film bag as an example to explain in detail the specific structure, workflow and parameter calculation of the film bag high-efficiency feeding control system. The core layer of the film bag is a magnetically sensitive composite layer (containing magnetically sensitive particles), and the inner and outer layers are ordinary polyolefin protective layers. It is necessary to achieve synchronous and accurate feeding of three raw materials: magnetically sensitive particle raw material (material A), inner polyolefin raw material (material B), and outer polyolefin raw material (material C) to ensure that the magnetic properties of the co-extruded film core layer meet the standards and the layer thickness is uniform.

[0047] The raw material characteristic input interface uses an industrial touch screen (model: Weintek TK6071IP), supporting manual input or import of raw material characteristic parameters from an Excel spreadsheet. In this embodiment, the parameters for material A (magnetically sensitive particle raw material, the magnetic particles are magnetite, purity 99%) are: melting point T... A =160℃, bulk density ρ A =1.3g / cm 3 Magnetic-sensitive particle loading amount M A =50%; Component B (inner layer polyolefin, homopolymer polypropylene) parameters are: T B =130℃, ρ B =0.92g / cm 3 M B =0% (non-magnetic particles); C material (outer layer polyolefin, modified polyethylene) parameters are: T C =135℃, ρ C =0.94g / cm 3 M C =0%.

[0048] The instantaneous load pressure P was collected by installing a high-precision diffused silicon pressure sensor (model: PT124G-111, measurement range 0-500kPa, accuracy ±0.2%FS) at the outlet of each screw conveyor component corresponding to materials A, B, and C (10cm from the die head inlet) to collect the pipeline resistance pressure during material conveying in real time, with a sampling frequency of 100Hz.

[0049] Co-extrusion die melting zone temperature T m Before data collection, three K-type thermocouples (model: OMRON E52-CA1D, measurement range 0-300℃, accuracy ±0.5℃) were installed at three evenly distributed points around the circumference of the co-extrusion die head melt cavity (5cm from the die lip). The average value of the three points was taken as the real-time T value via a PLC. m Value, in this embodiment, T is preset. m Threshold temperature T0 = 155℃ (adapts to the melt balance requirements of three raw materials).

[0050] The real-time thickness H of the co-extruded film was acquired by installing a laser thickness gauge (model: Keyence LK-G80, measurement range 0-5mm, accuracy ±1μm) 2m downstream of the die exit. One measurement point was taken every 20mm along the film width (a total of 50 points), and the average value was taken as the real-time H value. In this embodiment, the target total thickness H0 of the film bag is 80μm, where the core layer A material accounts for 40% (32μm), the inner layer B material accounts for 30% (24μm), and the outer layer C material accounts for 30% (24μm).

[0051] The liquid level and motor speed data were collected from silos A, B, and C (1 m³ / s). 3 An ultrasonic level sensor (model: Banner T30UXIA, measuring range 0.1-5m, accuracy ±1%) is installed at the bottom to monitor the remaining material in the hopper in real time (a replenishment alarm is triggered when the level is below 10%); a rotary encoder (model: Omron E6B2-CWZ6C, resolution 1000p / r) is installed at the end of the main servo motor shaft to collect the real-time speed of the motor for calculating the feeding rate.

[0052] The data integration method is as follows: data from each sensor is transmitted to a data acquisition card (model: NI-9234, 16-bit resolution) via Profinet industrial Ethernet (transmission rate 100Mbps, latency ≤10ms). The acquisition card timestamps the data at a sampling frequency of 100Hz and integrates them to form a complete raw dataset of raw material characteristics, feeding process, and co-extrusion quality. The dataset is then stored in a local MySQL database (data retention period of 30 days, supporting historical traceability).

[0053] The hardware composition and parameters of the single-drive linkage conveyor mechanism are as follows: the main servo motor is a Panasonic A6 series servo motor (model: MSME152G1U, rated power 1.5kW, rated speed 3000r / min, position control accuracy ±1 pulse), which drives the transmission shaft (material 45# steel, diameter 50mm, length 1200mm, chrome-plated for rust prevention) through an HTD-5M type synchronous belt (circumference 1500mm, number of teeth 300, transmission ratio 1:1).

[0054] Three sets of single-plate dry electromagnetic clutches (model: DLM9-5, rated torque 5N·m, response time ≤20ms, voltage DC24V) are selected and installed on the drive shaft by connecting them with flat keys. They correspond to the screw conveyor components of materials A, B and C (screw diameter 80mm, pitch 50mm, material 304 stainless steel, inner wall polishing treatment to reduce material adhesion).

[0055] The power transmission logic is as follows: the main servo motor drives the drive shaft to rotate at a constant speed (initial reference speed 1500 r / min). The electromagnetic clutch controls the power connection between the drive shaft and the screw conveyor assembly by energizing and disengaging—when closed, 100% of the power from the drive shaft is transmitted to the screw assembly, and the conveying speed is at its maximum; when disengaged, the power is cut off, and the conveying speed drops to 0. By adjusting the proportion of disengagement time within one working cycle (100ms) of the electromagnetic clutch, the actual power transmission ratio of each channel is changed: for example, material A has a high bulk density (ρ... A =1.3g / cm 3 To improve conveying power and compensate for pipeline resistance, the clutch engagement time needs to be extended (reducing the proportion of disengagement time); material B has a low melting point (T). B =130℃), the closing time needs to be shortened (the proportion of the opening time needs to be increased) to reduce the conveying rate and avoid excessive melting and accumulation of raw materials in front of the die head.

[0056] The intelligent control unit uses a Siemens S7-1500 PLC (model: 1511C-1PN, 1.2GHz main frequency, 1MB memory, supports Python script execution) as its hardware carrier. It incorporates semantic coding algorithms, timing matching algorithms, and fuzzy PID closed-loop algorithms. The specific implementation process is as follows:

[0057] Semantic encoding algorithm implementation (constructing raw material characteristic semantic vector V) mat )

[0058] Core logic: Transform unstructured raw material characteristic parameters (T, ρ, M) into quantifiable feature vectors to establish the correlation between raw material characteristics and feeding parameters.

[0059] Formula for constructing the thermophysical property embedding matrix E:

[0060] E = α·E T +β·E ρ +γ·E M

[0061] in:

[0062] E T The melting point characteristic matrix is ​​obtained by normalizing the deviation between the raw material melting point and the standard polyolefin melting point (130℃), and the formula is E. T = (T-130) / 30 (The denominator 30 represents the common melting point deviation range of polyolefins, ensuring E T (Values ​​range from 0 to 1);

[0063] E ρ Bulk density characteristic matrix, obtained by comparing the raw material bulk density with the reference density (0.9 g / cm³). 3 The ratio of the average density of ordinary polyolefins is obtained by the formula E.ρ =ρ / 0.9;

[0064] E M The magnetic particle loading feature matrix is ​​directly converted from the percentage of magnetic particle loading, using the formula E. M =M / 100;

[0065] α, β, γ: Weighting coefficients. Since the production of magnetically sensitive thin film bags requires prioritizing the magnetic properties of the core layer and the melting effect of the raw materials, α = 0.3 (melting point matching weight), β = 0.3 (bulk density matching weight), and γ = 0.4 (magnetic particle loading weight) are chosen, and α + β + γ = 1.

[0066] Calculation of embedding matrices for each raw material:

[0067] Assume that material A has a value of V. mat-A =0.933; B material V mat-B =0.307; C material V mat-C =0.363.

[0068] Then, the core logic of the time-series matching algorithm (determining the optimal feed rate baseline value) is to extract the time-series features of historical feed rates, combine them with the raw material characteristic vector to calculate mutual information, and determine the feed rate baseline value suitable for the current raw material, avoiding blind adjustment. The specific steps are as follows:

[0069] Step 1: Construct a historical feeding rate time queue. Extract motor speed data (sampling frequency 100Hz) for the same raw material combination over the past 24 hours. Convert the speed into feeding rate using the formula "Feeding rate Q = n × S × p × ρ × η" (n is the motor speed, in r / min; S is the screw cross-sectional area; p is the screw pitch; ρ is the raw material bulk density; η is the transmission efficiency, taken as 0.9), thus forming a time queue.

[0070] Step 2: 1D-CNN Temporal Encoding. A 3-layer convolutional layer + 2-layer max-pooling layer structure is used to extract features from the temporal sequence.

[0071] First convolutional layer: kernel size 5×1, stride 1, activation function ReLU, extracting short-term fluctuation features of the feeding rate (such as instantaneous rate decrease caused by raw material agglomeration);

[0072] First pooling layer: pooling kernel size 2×1, max pooling, preserving key fluctuation characteristics;

[0073] Second convolutional layer: kernel size 10×1, stride 2, extracts mid-term trend features (such as the slow decrease in rate caused by the drop in silo liquid level);

[0074] Second pooling layer: pooling kernel size 2×1, max pooling, compressing data dimensionality;

[0075] The third convolutional layer has a kernel size of 20×1 and a stride of 4, and extracts long-term periodic features (such as the periodic changes in rate caused by equipment temperature rise).

[0076] Output: Obtain the time-series vector V of the rate of each raw material. speed (Assuming material A is V) speed-A =0.85, B material V speed-B =1.2, C material V speed-C =1.15, the larger the vector value, the higher the historical adaptation rate).

[0077] Step 3: Calculate mutual information (MI). Using the formula...

[0078] MI=H(V mat )+H(V speed )-H(V mat V speed )

[0079] Calculate the correlation between raw material characteristics and feeding rate, where the information entropy H(X) is calculated using the following formula:

[0080]

[0081] Where p(x) i ) represents the probability distribution of eigenvalues, obtained based on historical data statistics.

[0082] Assume material A: H(V) mat-A H(V)≈0.228 speed-A H(V)≈0.61 mat-A V speed-A If )≈0.883, then MI A =0.228+0.61-0.883≈-0.045 (Negative mutual information indicates that the current historical rate is too high, and the baseline value needs to be reduced to adapt to the high characteristic requirements of material A);

[0083] Material B: MI B ≈0.15 (Positive mutual information indicates historical rate adaptation, which can slightly improve the baseline value);

[0084] Material C: MI C ≈0.12 (Positive mutual information indicates that the historical rate is basically adapted, and the baseline value can be finely adjusted).

[0085] Step 4: Determine the optimal feed rate baseline value. Combining mutual information and film thickness requirements, calculate the initial baseline value using the formula: Q0 = (H0 × k × v) / (ρ × W × η) (where k is the raw material layer thickness ratio; v is the co-extrusion linear speed, assumed to be 5 m / min; W is the film width, assumed to be 1 m). Then, correct according to MI. Assuming the result is:

[0086] Material A: Initial baseline value Q 0-A ≈0.178 kg / min, combined with MI A =-0.045 corrected to Q 0-A =0.17 kg / min;

[0087] Material B: Initial baseline value Q 0-B ≈0.24 kg / min, combined with MI B =0.15 corrected to Q 0-B =0.25 kg / min;

[0088] Material C: Initial baseline value Q 0-C ≈0.23 kg / min, combined with MI C =0.12 corrected to Q 0-C =0.24 kg / min.

[0089] The core logic of the fuzzy PID closed-loop algorithm (generating control commands) is to combine the real-time parameters (P, T) in the original dataset of the entire link. m The deviation between H) and the preset threshold is used to dynamically correct the optimal feeding rate baseline value and generate hardware execution instructions. The specific steps are as follows:

[0090] Step 1: Deviation value calculation. A dynamic weighted formula is used.

[0091] D = k P ·|P-P0|+k T ·|T m -T0|+k H ·|H-H0|

[0092] in:

[0093] P0 is the load pressure threshold (assuming material A has a load pressure threshold). 0-A =300kPa, B material P 0-B =200kPa, C material P 0-C =220kPa (determined based on raw material conveying resistance test);

[0094] k P k T k H The dynamic weighting coefficients are initially 1 / 3; when |P-P0|>10%P0 (e.g., material A, P>330kPa), k P =0.5, k T =k H =0.25 (prioritizing correction of transport resistance); when |T m When -T0|>5℃ (e.g., T) m <150℃), k T =0.5, kP =k H =0.25 (prioritizing melting effect);

[0095] Then, real-time deviation calculation is performed, assuming that material A has a pressure of P = 320 kPa (|320-300| = 20 ≤ 30), and T... m =152℃(|152-155|=3≤5), H=82μm(|82-80|=2), then D=(1 / 3)×20+(1 / 3)×3+(1 / 3)×2≈8.33 (the deviation level is "slight").

[0096] Step 2: PID parameter adjustment. The fuzzy PID controller automatically adjusts the proportional coefficient K according to the deviation level (slight / moderate / severe). p Integral coefficient K i Differential coefficient K d :

[0097] Slight deviation (D<10): K p =2.0, K i =0.5, K d =0.2 (Small adjustment to avoid overshoot);

[0098] Moderate deviation (10≤D<20): K p =3.0, K i =1.0, K d =0.5 (moderate adjustment, accelerating deviation correction);

[0099] Severe deviation (D≥20): K p =4.0, K i =1.5, K d =0.8 (Significant adjustment to address process fluctuations).

[0100] Step 3: Generate adjustment commands. Combining the corrected PID parameters with the optimal feeding rate baseline, calculate the electromagnetic clutch disengagement time percentage and the main servo motor speed adjustment value:

[0101] Electromagnetic clutch disengagement time percentage: Calculated using the formula Disengagement Time Percentage = 1 - (Q0 × ρ × η) / (n × S × p). In this embodiment, the disengagement time percentage for material A is adjusted from the initial 30% to 32% (reducing the conveying speed to adapt to MI). A (Negative correlation), B material was adjusted from 25% to 24% (increasing the rate to adapt to MI). B (Positive correlation), C material remains at 26%;

[0102] Main servo motor speed: Based on the full-link deviation D=8.33, the speed is finely adjusted from 1500r / min to 1480r / min to avoid the overall material feeding amount exceeding the melting load capacity of the die head.

[0103] The electromagnetic clutch is controlled by three Omron G6B-4BND relay modules, which receive PWM (Pulse Width Modulation) signals output from the PLC. The PWM duty cycle of clutch A is adjusted from 70% (70% of the engagement time) to 68% (32% of the disengagement time), clutch B is adjusted from 75% to 76%, and clutch C remains at 74%. The relay response time is ≤10ms to ensure that the clutch action is synchronized with the algorithm instructions.

[0104] The main servo motor is controlled by the PLC sending speed commands to the Panasonic A6 servo driver via the Profinet protocol. The driver uses position control mode to reduce the main motor speed from 1500 r / min to 1480 r / min, with a speed adjustment accuracy of ±1 r / min, which meets the requirements of the co-extrusion process for speed stability.

[0105] The logic of closed-loop verification and iteration is as follows: after the feedback execution module takes action, the end-to-end data acquisition module collects P and T data in real time. m The H data is then transmitted back to the intelligent control unit. If material A's pressure (P) decreases from 320 kPa to 310 kPa (approaching P)... 0-A =300kPa), T m As the temperature rises from 152℃ to 153℃ (approaching T0 = 155℃), H decreases from 82μm to 81μm (approaching H0 = 80μm), and the current adjustment command is maintained. If the deviation still exists (e.g., H is still > 81μm), the intelligent control unit repeats the deviation calculation and command generation process every 50ms until all parameter deviations are ≤ preset thresholds (P deviation ≤ 5kPa, T0 deviation ≤ 80μm). m With deviation ≤2℃ and H deviation ≤1μm, a stable closed loop is formed.

[0106] In practical use, operators input the T, ρ, and M parameters of materials A, B, and C through the raw material characteristic input interface. The system automatically loads preset thresholds (P0, T0, H0) and historical process database (optimal parameter templates for the same raw material combination in the past 30 days).

[0107] The intelligent control unit runs semantic coding algorithm and timing matching algorithm to calculate the initial optimal feeding rate baseline value (0.17 kg / min for material A, 0.25 kg / min for material B, and 0.24 kg / min for material C), and sends the initial command to the feedback execution module: the electromagnetic clutch disengagement time accounts for 30% of material A, 25% of material B, and 26% of material C, and the main motor speed is 1500 r / min.

[0108] Check whether the signals of each sensor are normal (e.g., no zeroing error of the pressure sensor, no open circuit of the thermocouple), whether the electromagnetic clutch operates smoothly, and whether there is no overload alarm of the main motor. After the self-test is passed, enter the production standby state.

[0109] During the production operation phase (continuous production process), the end-to-end data acquisition module collects P and T data for materials A, B, and C at a frequency of 100Hz. m Data such as H, silo liquid level, and motor speed are synchronously transmitted to the data acquisition card via industrial Ethernet and integrated to form a complete raw dataset (each data entry includes a timestamp, accurate to milliseconds).

[0110] The intelligent control unit calls the fuzzy PID closed-loop algorithm every 50ms to calculate the deviation value D based on the real-time dataset, dynamically correct the optimal feeding rate benchmark value, and generate adjustment commands for the electromagnetic clutch and the main motor.

[0111] The feedback execution module receives instructions and drives hardware actions, while simultaneously transmitting the execution status (such as clutch on / off status and actual motor speed) back to the PLC to ensure that the instructions are executed properly.

[0112] If the silo level is below 10%, an audible and visual alarm (buzzer + red LED) will be triggered, and a replenishment prompt will be displayed on the touchscreen. Normal adjustment will automatically resume after replenishment is complete. m If the temperature is below 145℃ (significantly deviating from the threshold), the system will automatically reduce the main motor speed by 20% and extend the disengagement time of the A material clutch to avoid film defects caused by insufficient melting of raw materials.

[0113] After production is completed, when the operator triggers the shutdown command, the intelligent control unit first controls the electromagnetic clutch to disengage completely (cutting off the material conveying), and then reduces the speed of the main motor to 0 to prevent material residue in the screw assembly when the machine stops.

[0114] Generate a production report including: raw material consumption (approximately 8.5 kg / h for material A, 12.5 kg / h for material B, and 12 kg / h for material C), historical adjustment command records (key parameters are stored every 10 minutes), and product quality data (average H value 80.2 μm, pass rate 99.5%). The report can be exported to Excel format for production traceability. Finally, remind operators to clean the screw conveyor assembly and die head inlet to prevent residual raw materials from clumping and affecting the next production run.

[0115] After 8 hours of continuous production, 100 film bag samples were randomly selected for testing.

[0116] Regarding the uniformity of layer thickness, the average thickness of core layer A material is 32.1 μm with a deviation of ±0.8 μm; the average thickness of inner layer B material is 24.2 μm with a deviation of ±0.7 μm; and the average thickness of outer layer C material is 23.9 μm with a deviation of ±0.6 μm, all meeting the accuracy requirement of ±1 μm.

[0117] The magnetically sensitive particles in the core layer are uniformly distributed, and the average magnetic adhesion strength measured by a magnetometer is 0.85 N / cm. 2 Deviation ±0.05 N / cm 2 This meets the requirement of "being able to stably attach magnets";

[0118] In terms of appearance quality, the membrane surface is free of unevenness and bubbles, and the edge material width is ≤2mm (the edge material width of the traditional system is about 5mm), improving the raw material utilization rate by 6%.

[0119] This embodiment achieves the following technical advantages through an integrated design of full-link data acquisition, single-drive linkage conveying, intelligent algorithm closed loop, and feedback execution: reduced deviation in the conveying rate of multiple raw materials, and layer thickness uniformity meeting the production requirements of high-end film bags; no manual calibration is required when changing raw materials, reducing parameter adaptation time and improving adaptation efficiency; through downstream process feedback closed loop, the product qualification rate is improved while avoiding batch scrap; the single main motor reduces energy consumption compared to multi-motor solutions, and the reduction of edge material improves raw material utilization.

[0120] This embodiment can be directly extended to other co-extrusion processes such as ordinary polyolefin film bags and high-barrier film bags. Only the raw material characteristic parameters and preset thresholds need to be adjusted to achieve efficient feeding control, which has strong versatility and practicality.

[0121] The above are merely embodiments of the present invention. The invention is not limited to the fields covered by these embodiments. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are able to access all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A high-efficiency film bag feeding control system, applied to a multi-raw material synchronous conveying device in a co-extrusion process, characterized in that: include: The end-to-end data acquisition module is used to acquire the raw material characteristic parameters of the raw material to be transported, including melting point T, bulk density ρ and magnetically sensitive particle loading M; it also collects the instantaneous load pressure P of each raw material conveying channel, the melting zone temperature Tm of the co-extrusion die, the real-time thickness H of the co-extrusion film, the silo liquid level data and the motor speed data, and integrates them to obtain the end-to-end raw dataset. A single-drive linkage conveying mechanism includes a main servo motor, multiple sets of electromagnetic clutches, and multiple sets of screw conveyor components. The main servo motor drives the transmission shaft through a synchronous belt. The transmission shaft is connected to the screw conveyor components of the corresponding raw material channels through different sets of electromagnetic clutches. The electromagnetic clutches are used to independently control the power transmission on / off and transmission ratio of each raw material channel. The intelligent control unit is equipped with a semantic encoding algorithm, a timing matching algorithm, and a fuzzy PID closed-loop algorithm. The intelligent control unit is used to obtain the optimal feeding rate benchmark value based on the raw material characteristic parameters and the semantic encoding algorithm and timing matching algorithm. The fuzzy PID closed-loop algorithm extracts the instantaneous load pressure P, melting zone temperature Tm, and film thickness H from the original dataset of the entire chain, calculates the deviation value of the above parameters from the preset threshold, and dynamically generates adjustment instructions in combination with the optimal feeding rate benchmark value. The feedback execution module receives adjustment commands and adjusts the power transmission ratio of the corresponding raw material channel by controlling the disengagement time of each electromagnetic clutch. At the same time, it adjusts the output speed of the main servo motor to match the actual conveying rate of each raw material with the melting rate of the co-extrusion process.

2. The high-efficiency film bag feeding control system according to claim 1, characterized in that, The intelligent control unit is used to obtain the optimal feeding rate benchmark value according to the semantic encoding algorithm and the temporal matching algorithm, including: the semantic encoding algorithm is used to call the preset thermophysical property embedding matrix to perform semantic embedding encoding on the raw material characteristic parameters in the full-link original dataset to obtain the raw material characteristic semantic vector Vmat; the temporal matching algorithm constructs the motor speed data in the full-link original dataset into a historical feeding rate time queue, performs temporal implicit encoding on the time queue through the preset 1D-CNN model to obtain the rate time sequence vector Vspeed, calculates the mutual information MI between the raw material characteristic semantic vector Vmat and the rate time sequence vector Vspeed, and determines the optimal feeding rate benchmark value for each raw material based on the mutual information MI.

3. The high-efficiency film bag feeding control system according to claim 2, characterized in that, The mutual information (MI) between the raw material characteristic semantic vector Vmat and the rate time series vector Vspeed is calculated using the following formula: MI=H(Vmat)+H(Vspeed)-H(Vmat,Vspeed) Where H(Vmat) represents the information entropy of Vmat, H(Vspeed) represents the information entropy of Vspeed, and H(Vmat,Vspeed) represents the joint entropy of Vmat and Vspeed.

4. The high-efficiency film bag feeding control system according to claim 3, characterized in that, The end-to-end data acquisition module includes load pressure sensors and raw material characteristic input interfaces at the outlets of each spiral conveyor component at the feeding end, as well as temperature sensors and film thickness detectors located in the die melting zone downstream of co-extrusion. It also retains traditional silo level sensors and motor speed encoders. The module acquires the raw material characteristic parameters of the material to be conveyed through the raw material characteristic input interface, including melting point T, bulk density ρ, and magnetically sensitive particle loading M. It acquires the instantaneous load pressure P of each raw material conveying channel through the load pressure sensors, the temperature Tm of the co-extrusion die melting zone through the die melting zone temperature sensor, and the real-time thickness H of the co-extruded film through the film thickness detector. Finally, it acquires silo level data and motor speed data through the silo level sensor and motor speed encoder, respectively, and integrates these to obtain the end-to-end raw dataset.

5. The high-efficiency film bag feeding control system according to claim 4, characterized in that, In the semantic encoding algorithm, the thermophysical property embedding matrix is ​​expressed by the formula: E = α·E T +β·E ρ +γ·E M Construct, where: E is the final embedding matrix; E T The melting point characteristic matrix is ​​obtained by normalizing the deviation of T from the standard melting point; E ρ E is the bulk density characteristic matrix, calculated by the ratio of ρ to the reference density; M The magnetic particle loading feature matrix is ​​obtained by converting the magnetically sensitive particle loading amount M by weight percentage; α, β, and γ are weighting coefficients, and α+β+γ=1.

6. The high-efficiency film bag feeding control system according to claim 5, characterized in that, The 1D-CNN model contains three convolutional layers and two pooling layers. The first convolutional layer has a kernel size of 5×1 and a stride of 1, which is used to extract short-term fluctuation features of the feeding rate. The second convolutional layer has a kernel size of 10×1 and a stride of 2, which is used to extract medium-term trend features. The third convolutional layer has a kernel size of 20×1 and a stride of 4, which is used to extract long-term periodic features. All pooling layers use max pooling.

7. The high-efficiency film bag feeding control system according to claim 6, characterized in that, In the fuzzy PID closed-loop algorithm, the deviation value is calculated using a dynamic weighting method, through the formula D = k P ·|P-P0|+k T ·|T m -T0|+k H • |H-H0| is achieved, where: D is the overall deviation value; k P k T k H For dynamic weighting coefficients; when P-P0>10%P0, k P Automatically adjusted to 0.5, k T and k H Each is 0.25, prioritizing the correction of load fluctuations; when Tm-T0>5℃, k T Automatically adjusted to 0.5, k P and k H Each is 0.25, prioritizing the melting effect; the preset threshold for instantaneous load pressure P is P0, the preset threshold for melting point T is T0, and the preset threshold for real-time thickness H of the co-extruded film layer is H0.

8. The high-efficiency film bag feeding control system according to claim 7, characterized in that, The intelligent control unit also includes a parameter self-optimization module, which optimizes parameters using formulas. The proportional coefficient, integral coefficient, and derivative coefficient of the fuzzy PID controller are iteratively optimized, where: K n+1 The parameter values ​​for the (n+1)th iteration, K n Here, λ is the parameter value for the nth iteration, λ is the learning rate (0.01-0.1), θ is the decay coefficient (0.05-0.2), and n is the number of iterations.

9. The high-efficiency film bag feeding control system according to claim 8, characterized in that, In the time-series matching algorithm, the calculation of mutual information (MI) incorporates a temperature correction factor, and the corrected formula is MI. ′ =MI·(1+δ·|T) m -T0| / T0), where: MI ′ δ represents the corrected mutual information; δ is the temperature sensitivity coefficient, which is 0.8 for raw materials with a melting point above 150℃ and 0.3 for raw materials with a melting point below 150℃.

10. A method for efficient feeding control of film bags, characterized in that, The system described in any one of claims 1-9 is employed.

Images