Cooperative control method for hybrid heating assembly of two-way stretching machine

Abstract

The invention discloses a cooperative control method for a hybrid heating assembly of a two-way stretcher, and relates to a thin film stretching heating control technology. Comprising the following steps: arranging a hybrid heating assembly and deploying data monitoring equipment; constructing and training a CNN neural network model to predict an ideal temperature curve with a time sequence in each temperature control area; collecting data in real time to construct a real-time temperature curve; the optimal time phase difference between the real-time thickness and the ideal thickness is calculated through a cross-correlation analysis method, time phase compensation is carried out on an ideal temperature curve, high-frequency or low-frequency change characteristics of temperature deviation are distinguished, a PID control strategy is carried out, and protective gradient power adjustment is carried out on an infrared heater array in a corresponding area. According to the invention, the heating uniformity and stability of the diaphragm are improved through the partition control of the mixed heating assembly and the combination of real-time temperature monitoring and dynamic compensation; based on a time phase compensation mechanism of cross-correlation analysis, heating time sequence dislocation caused by speed deviation of a production line is corrected in real time, and it is ensured that heat accurately acts on a target film section.

Description

Technical Field

[0001] This invention relates to the field of film stretching heating control technology, specifically a collaborative control method for a hybrid heating component of a biaxial stretching machine. Background Technology

[0002] In the production of biaxially oriented films, the preheating process is crucial to the final properties of the film, such as thickness uniformity, mechanical properties, and optical properties. Current technologies primarily employ multiple sets of preheating rollers for conductive heating of the film. However, this approach has two inherent drawbacks: Uneven temperature distribution on the preheating roller: The preheating roller is usually heated by circulating heat transfer oil through a spiral or Z-shaped flow channel. Due to the limitations of the flow channel design and the temperature drop of the heat transfer medium, there are axial and circumferential temperature gradients on the roller surface, resulting in uneven heating of the film and affecting the uniformity of subsequent stretching, which can easily lead to larger film thickness tolerances.

[0003] Impact of surface defects on the roller wall: After long-term use, the surface of the preheating roller inevitably develops minor scratches, wear, or contaminant adhesion. These surface defects can leave indentations or scratches on the soft, preheated film, affecting the film's appearance quality; While improving the machining accuracy of the preheating rollers or optimizing the flow channel design can significantly increase manufacturing costs and maintenance difficulty, it does not fundamentally eliminate the problem of uneven temperature. Therefore, we provide a collaborative control method for the hybrid heating components of a biaxial stretching machine. Summary of the Invention

[0004] The purpose of this invention is to provide a method for the coordinated control of the hybrid heating components of a biaxial stretching machine.

[0005] The technical problem solved by this invention is: This invention can be achieved through the following technical solution: a collaborative control method for the mixing and heating components of a biaxial stretching machine, comprising the following steps: Step 1: Arrange the hybrid heating components and deploy data monitoring equipment. The hybrid heating assembly includes a preheating roller and an infrared heater array. The preheating roller is divided into multiple independent temperature control zones along the width direction. A thermocouple is embedded in each temperature control zone. The infrared heater array is controlled in a corresponding zone and works with an infrared thermal imager and a laser thickness gauge to form a temperature monitoring matrix. Step 2: Based on historical production data, select positive samples, construct and train a CNN neural network model to predict the ideal temperature curve with time sequence in each temperature control zone; Step 3: Real-time acquisition of film surface temperature and film thickness within each temperature monitoring matrix to construct real-time temperature curves; Step 4: Calculate the optimal time phase difference between the real-time thickness and the ideal thickness using cross-correlation analysis, and perform time phase compensation on the ideal temperature curve. Step 5: Based on the high-frequency or low-frequency variation characteristics of the temperature deviation, PID control strategies are used to coordinately adjust the flow rate of the heating medium in the preheating roller and the power of the infrared heater array. Step 6: Based on the visual camera, identify the location of defects on the thin film surface and perform protective gradient power adjustment on the infrared heater array in the corresponding area to avoid scratch defects.

[0006] A further technical improvement of the present invention is that the preheating roller adopts a bidirectional spiral gradual flow channel structure, and the flow channel pitch gradually decreases along the medium flow direction to compensate for the temperature difference in the temperature zone.

[0007] A further technical improvement of the present invention is that the method for obtaining positive samples in step two includes: The quality of finished films from historical production batches is scored, and the historical production data corresponding to production batches whose quality scores exceed a set threshold are used as positive samples. The quality score includes a film thickness uniformity score and a surface quality score, and ; ; ; in, These represent the weights for the influence of thickness uniformity and the influence of surface quality, respectively. ; This represents the average thickness range in the width direction across all scanning cycles within each scanning cycle of the thickness gauge. This represents the maximum allowable lateral thickness range; This represents the average thickness range over the entire length direction within a set distance L in the direction of film winding movement. The maximum allowable longitudinal thickness range; The weights for the influence of horizontal and vertical uniformity are respectively. ; These represent the total defect area and the set allowable defect area for this batch of films, respectively.

[0008] A further technical improvement of the present invention lies in: a method for constructing a CNN neural network model, comprising: Using the membrane material, membrane thickness, longitudinal moving speed, target film thickness, and target film width as model inputs, the membrane material is converted into a categorical variable using unique thermal encoding, and then standardized in combination with the other continuous variables and concatenated into a one-dimensional vector. The ideal average temperature time series within each spatial cell corresponding to the temperature monitoring matrix is ​​used as the output. Based on this series, the ideal temperature curve of each spatial cell is plotted, and the corresponding ideal thickness change curve under the ideal temperature is mapped out.

[0009] A further technical improvement of the present invention is that the method for obtaining the optimal time phase difference in step four includes: The real-time acquired data is divided into continuous time windows according to a certain time interval. Within each time window, the cross-correlation coefficient between the real-time thickness and the ideal thickness of the film is calculated at different offset times. The offset time phase corresponding to the maximum cross-correlation coefficient is taken as the optimal time phase difference.

[0010] A further technical improvement of the present invention is that the method for judging the high-frequency / low-frequency changes of temperature deviation in step five is whether the number of peaks within a unit time window exceeds a preset value.

[0011] A further technical improvement of the present invention is that: global PID control is used to adjust the flow rate of the preheating roller for low-frequency temperature deviation, and zoned PID control is used to adjust the power of the infrared heater for high-frequency temperature deviation. In low-frequency scenarios, the temperature deviations in multiple spatial cells along the width direction within the same time window are obtained, and the average value of multiple temperature deviations at the same moment is calculated. Then, the average value of the temperature deviations at each moment within the entire time window is calculated again to obtain the global average temperature deviation corresponding to the time window. The global temperature deviation is used as input, and the valve opening of the preheating roller heating medium is used as output. In high-frequency scenarios, the average temperature difference of the spatial cell within the time window is used as the output, and the adjusted power of the infrared heater array is used as the output.

[0012] A further technical improvement of this invention lies in: in the protective gradient power adjustment, firstly, the position where the preheating roller produces scratches along the width direction is determined, and then the power of multiple infrared heater arrays upstream of that position is increased in a gradient manner, the magnitude of which is... , where i=1,2,3,...,N; N represents the number of infrared heater arrays from upstream of the preheating roller to the end of the position of another preheating roller; It is 10% of the current power of the first adjacent infrared heater array upstream of the preheating roller.

[0013] Compared with the prior art, the present invention has the following beneficial effects: This invention significantly improves the uniformity and stability of membrane heating by using a hybrid heating assembly consisting of a preheating roller and an infrared heater for zoned control, combined with real-time temperature monitoring and dynamic compensation. It employs five-zone independent temperature control and a bidirectional spiral gradient flow channel to compensate for temperature differences, eliminating axial temperature gradients and ensuring uniform heating of the membrane. In addition, based on a time phase compensation mechanism using cross-correlation analysis, it corrects heating timing misalignments caused by production line speed deviations in real time, ensuring that heat is precisely applied to the target membrane segment.

[0014] This invention selects positive samples from historical data, filters the data using thin film surface quality scores, defines the input with the selected data, and constructs a CNN neural network model to predict the ideal temperature curve under the corresponding production parameters. After time phase compensation, the temperature deviation is obtained by comparing it with the real-time temperature curve. After distinguishing between high-frequency and low-frequency temperature deviations, a PID algorithm is used to optimize control, improving control efficiency and targeting, and also saving energy.

[0015] 3. In the temperature regulation process, the present invention introduces a protective gradient regulation mechanism, which softens the diaphragm by local temperature rise, reduces contact damage with roller surface defects, and significantly reduces the probability of scratches and indentations. Attached Figure Description

[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 This is a schematic diagram of the method execution flow of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0019] Please see Figure 1 As shown, the coordinated control method of the hybrid heating component of the biaxial stretching machine specifically includes the following steps: Step 1: Arrange the hybrid heating components and deploy data monitoring equipment. The number of preheating rollers is reduced, thereby increasing the span between adjacent preheating rollers and reducing the number and probability of contact between the film and the surface of the preheating rollers; infrared heater arrays are installed in the upstream and downstream areas of the preheating rollers, thus forming a hybrid heating assembly of preheating rollers and infrared heater arrays. In the hybrid heating assembly, the entire width of the preheating roller is divided into five independent temperature control zones, corresponding to the central zone, the adjacent zones on both sides, and the edge zones on both sides, respectively; a miniature thermocouple array is embedded in each independent temperature control zone of the preheating roller to obtain the temperature data of each temperature control zone of the preheating roller in real time. The infrared heater array also corresponds to five independent temperature control zones for zoned control. Infrared thermal imagers are continuously installed at certain intervals along the membrane material's movement direction to collect real-time temperature distribution data along the width of the membrane as it passes below. The width of the temperature distribution along the membrane material's movement direction collected by the infrared thermal imagers is set, and then combined with the independent temperature control zones along the width direction to form square spatial cells, which are marked as temperature monitoring matrices. A laser thickness gauge is installed in each temperature monitoring matrix. It should be noted that the heating medium in the preheating roller generally enters from one end and flows out from the other end. As the temperature of the medium decreases during the flow and heat conduction process, there is a large temperature difference between the independent temperature control zones of the preheating roller. A bidirectional spiral gradual flow channel is adopted so that the pitch of the spiral flow channel gradually decreases from one edge zone to the other edge zone. For example, spiral gradual flow channels are embedded at both ends of the preheating roller, and the pitch gradually decreases along the flow direction of the medium. This extends the residence time of the medium at the corresponding position after the temperature drops, thereby compensating for the temperature difference.

[0020] Step 2: Obtain the ideal temperature curve that meets current production needs based on historical production data. S21. Select positive sample data from historical production data that can be used as training data. First, positive data from historical production data for excellent production batches are selected as positive samples; among them, excellent production batches refer to production batches whose quality scores of finished film exceed a set threshold. The quality score includes a film thickness uniformity score and a surface quality score, and ; ; ; in, These represent the weights for the influence of thickness uniformity and the influence of surface quality, respectively. Under normal circumstances , ; This represents the thickness range in the width direction (i.e., the X direction) within each scanning cycle of the thickness gauge, and then the average value is taken over all scanning cycles to reflect the lateral uniformity. The maximum value of the allowable lateral thickness range (maximum thickness - minimum thickness); This represents the thickness range within a set distance L in the film winding direction (i.e., the Y direction), and then the average value is taken over the entire length direction to reflect the longitudinal uniformity. The maximum allowable longitudinal thickness range; The weights for the influence of horizontal and vertical uniformity are respectively. ; These represent the total defect area and the set allowable defect area for this batch of films, respectively.

[0021] It should be noted that the thickness uniformity score and surface quality score are normalized separately, and the resulting quality score is taken between 0 and 100%. Specifically, when the quality score is not less than 96%, the production record data of the corresponding film production batch is considered as positive sample data and used to train the ideal temperature curve prediction model.

[0022] S22. Construct and train an ideal temperature curve prediction model based on a CNN neural network. Define the model inputs: membrane material, membrane thickness, longitudinal movement speed, target film thickness, and target film width; Define the model output as: the ideal average temperature sequence with time sequence in each cell of the target diaphragm temperature control matrix, and the corresponding ideal temperature curve can be plotted based on this sequence; Before inputting each input variable into the model, it is preprocessed and then concatenated into an input feature vector. The preprocessing includes one-thermal encoding of the membrane material and converting it into a sparse variable; standardizing the membrane thickness, longitudinal moving speed, target film thickness and target film width; and concatenating the above input features into a one-dimensional input vector. CNN neural networks encode and non-linearly transform input variables through connection layers to increase feature dimensionality. Convolutional layers expand the increased features in both time and space dimensions. Finally, channel compression is performed in the output layer to obtain the output matrix. In this output matrix, each row represents a spatial cell and each column represents a time point. The output matrix is ​​the ideal average temperature time series of each cell in the predicted temperature control monitoring matrix. Based on the ideal average temperature time series, the ideal temperature prediction curve in each cell can be plotted.

[0023] The selected positive sample data is divided into training and validation sets in a 7:3 ratio. The mean squared error is used as the loss function to calculate the difference between the predicted output matrix and the true target matrix extracted from the historical positive sample data. In order to minimize the mean squared error, the model parameters are iteratively optimized to obtain the ideal temperature curve prediction model. It should be noted that although the ideal temperature prediction curve obtained from the model has a time series, this time series is driven by the film thickness in the corresponding region. That is, the ideal temperature at the corresponding time point should be understood as the ideal temperature at the moment when the film is stretched to the corresponding thickness. Thus, the mapping relationship between temperature and film thickness can be mapped from the ideal temperature prediction curve to a curve showing the ideal thickness change of the film under ideal conditions.

[0024] Step 3: Real-time acquisition of temperature and film thickness parameters to construct real-time temperature curves. The temperature distribution of the membrane surface and the thickness distribution of the corresponding membrane in each spatial cell of the temperature monitoring matrix are collected in real time. A real-time temperature curve is constructed based on the temperature distribution of the membrane surface, and the membrane thickness at the corresponding time is marked.

[0025] Step 4: Analyze the time phase shift and temperature deviation by combining the actual temperature curve with the ideal temperature prediction curve. The ideal temperature prediction curve is based on a stable ideal time axis. In actual production, there will be a slight deviation between the actual speed of the production line and the set value, which will cause the time point when the thin film passes through the corresponding spatial cell to deviate from the time point predicted by the model. If the mixing heating component is controlled according to the ideal time axis, the applied heat will not be applied to the correct film segment, and the film quality will not meet the expected effect. Therefore, for time phase shift, continuous time windows are divided according to a certain time interval (e.g., 60s). Within each time window, the optimal time phase difference between the real-time thickness and the ideal film thickness is calculated using cross-correlation analysis. The specific steps include: Calculate the cross-correlation coefficient between the real-time thickness and the ideal film thickness at different offset times, and take the offset time phase corresponding to the maximum cross-correlation coefficient as the optimal time phase difference.

[0026] The optimal time phase difference is used to shift the ideal thickness variation curve and the ideal temperature prediction curve of the thin film onto the ideal time axis, thereby completing the time phase deviation compensation; the compensation mechanism is triggered once in each time window.

[0027] Subsequently, the temperature deviation between the real-time temperature and the ideal predicted temperature was calculated for each spatial cell. Within a time window, the number of peaks within the window is counted to calculate the number of peaks per unit time. When the number of peaks exceeds the set value, the temperature deviation is considered to be a high-frequency change; otherwise, the temperature deviation is considered to be a low-frequency change.

[0028] Step 5: Differentiate and coordinate control based on high / low frequency variations of temperature deviation. S51. For scenarios with low-frequency temperature deviation changes: Low-frequency temperature deviations are generally large-scale and slowly changing temperature unevenness, which is strongly correlated with the temperature control of the preheating roller. The temperature deviations in multiple spatial cells along the width direction within the same time window are obtained, and the average value of multiple temperature deviations at the same moment is calculated. Since the temperature deviation changes slowly, the average value of the temperature deviation at each moment within the entire time window is further calculated to obtain the global average temperature deviation value corresponding to the time window. A PID algorithm is used, with the global temperature deviation average as input and the control quantity as output. This control quantity is then used as a specific command to adjust the flow rate of the heating medium. Specifically: Let the mean global temperature deviation be The output control quantity is Then we have: ; in, This represents the proportionality coefficient, which determines the strength of the control quantity's response to the current temperature deviation. Indicates the integral coefficient, used to eliminate static error; This represents the differential coefficient, which is used to make advance adjustments based on the trend of temperature deviation changes, thus suppressing overshoot. when When the temperature is too high, it indicates that the flow rate through the preheating roller needs to be reduced. when When the temperature is too low, it indicates that the flow rate through the preheating roller needs to be increased. pass The value directly determines the valve opening, for example, Then the valve is instructed to open to 20%. If so, adjust directly according to the value of the nearest endpoint.

[0029] S52. For scenarios with high-frequency temperature deviations. High-frequency temperature deviations are generally localized and rapidly changing temperature unevenness, which is strongly correlated with the temperature control of the infrared heater array. Since the infrared heater array in each spatial cell is controlled independently, a separate PID control is performed based on each spatial cell. Similarly, the average temperature difference of the spatial cell within the time window is used as the input, and the output variable is used as the adjustment power of the infrared heater array to control the infrared heater array to compensate for the corresponding temperature deviation.

[0030] Step 6: Perform protective gradient adjustment on the tensile quality of the film surface. Visual cameras are deployed in the upstream and downstream areas adjacent to the preheating roll to acquire images of the film surface at the corresponding locations and identify scratches and defects marked on them. The coordinates of the scratches and defects are calculated along the width direction to determine that the scratches and defects are located in the corresponding spatial cell in the width direction of the preheating roll. Multiple infrared heater arrays located at the same X-axis position upstream of the preheating roller, positioned along the moving direction, all increase their power by a certain amount. For example, if the required power increase is set to... If there are N infrared heater arrays along the width from upstream of the corresponding preheating roller to the end of another preheating roller position, then the power of the nearest infrared heater array will be increased. The power of the adjacent second infrared heater array is increased. Then the power of the upstream i-th infrared heater array increases. Where i = 1, 2, 3, ..., N; in this embodiment It is 10% of the current power of the first adjacent infrared heater array; By increasing the power gradient of the infrared heater array at each of the aforementioned locations, the corresponding positions of the film reach a temperature higher than the ideal predicted temperature when contacting the preheating roller. Due to the properties of the film's polymer material, a slightly higher temperature makes the film softer and more elastic, allowing it to better "slide" or "fill" in minor contact defects, thereby greatly reducing or even eliminating scratches. It should be noted that in the protective gradient adjustment mode, the ideal predicted temperature is increased in the corresponding spatial cell, thereby avoiding adjustment conflicts caused by protective adjustment when calculating the temperature deviation of the corresponding spatial cell in steps four and five.

[0031] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method of coordinated control of a heating assembly of a two-way drawing machine, characterized in that: The method comprises the following steps: Step one, arranging a mixed heating assembly and deploying a data monitoring device The mixed heating assembly comprises a preheating roller and an infrared heater array, the preheating roller is divided into multiple independently temperature-controlled zones along the width direction, a thermocouple is embedded in each temperature-controlled zone, the infrared heater array is controlled in correspondence with the zones, and an infrared thermal imager and a laser thickness gauge are used to construct a temperature monitoring matrix; Step two, screening positive samples based on historical production data, constructing and training a CNN neural network model, and predicting an ideal temperature curve with time sequence in each temperature-controlled zone; Step three, collecting the film surface temperature and the film thickness in each temperature monitoring matrix in real time, and constructing a real-time temperature curve; Step four, calculating the optimal time phase difference between the real-time thickness and the ideal thickness by using a cross-correlation analysis method, and compensating the ideal temperature curve in time phase; Step five, according to the high-frequency or low-frequency change characteristics of the temperature deviation, using a PID control strategy to cooperatively adjust the heating medium flow of the preheating roller and the power of the infrared heater array; Step six, identifying the position of the film surface defect based on a visual camera, and performing protective gradient power adjustment on the infrared heater array in the corresponding region to avoid scratch defects.

2. The method of claim 1, wherein the method further comprises: The preheating roller adopts a bidirectional spiral gradually changing flow channel structure, and the flow channel pitch gradually decreases along the medium flow direction to compensate for the temperature difference between the zones.

3. The method of claim 1, wherein the method further comprises: The method for obtaining positive samples in step two comprises: Performing quality scoring on the film products of historical production batches, and taking the historical production data of the production batches with quality scores exceeding a set threshold as positive samples; The quality score includes a film thickness uniformity score and a surface quality score, and ; ； ； wherein, respectively, are the thickness uniformity influence weight and the surface quality influence weight, ; represents the average value of the thickness range in the width direction within each scanning cycle of the thickness gauge in all scanning cycles; is the maximum value of the allowed transverse thickness range; represents the average value of the thickness range in the entire length direction within the set distance L of the film winding moving direction; is the maximum value of the allowed longitudinal thickness range; respectively, are the transverse and longitudinal uniformity influence weights, and ; respectively, represent the total defect area of the batch of films and the set allowed defect area.

4. The method of synergic control of the heating assembly of a bidirectional stretching machine according to claim 1, characterized in that, The method for constructing the CNN neural network model comprises: Taking the film sheet material, the film sheet thickness, the longitudinal moving speed, the target film thickness, and the target film width as model inputs, converting the film sheet material into a classification variable by using one-hot encoding, combining the remaining continuous variables for standardization processing, and splicing them into a one-dimensional vector; Taking the ideal average temperature time sequence in each spatial cell of the temperature monitoring matrix as the output, drawing the ideal temperature curve of each spatial cell according to the sequence, and mapping the ideal thickness change curve corresponding to the ideal temperature.

5. The method of claim 1, wherein the method further comprises: The method for obtaining the optimal time phase difference in step four comprises: Dividing the collected data into continuous time windows according to a certain time interval, calculating the cross-correlation coefficients of the real-time thickness and the ideal film thickness at different offset times in each time window, and taking the offset time phase corresponding to the maximum cross-correlation coefficient as the optimal time phase difference.

6. The method of synergic control of the heating assembly of a bidirectional stretching machine according to claim 1, characterized in that, The judgment method for the high-frequency / low-frequency change of the temperature deviation in step five is whether the number of wave peaks in a unit time window exceeds a preset value.

7. The method of synergistic control of the heating and stretching assembly of the bidirectional drawing machine according to claim 1, characterized in that, For the low-frequency change of the temperature deviation, a global PID control is used to adjust the preheating roller flow, and for the high-frequency change, a partitioned PID control is used to adjust the infrared heater power; In the low-frequency scenario, the temperature deviations in multiple spatial cells along the width direction in the same time window are obtained, the mean values of the multiple temperature deviations at the same time are calculated, the mean values of the temperature deviations at all times in the whole time window are further calculated, and the global temperature deviation mean value corresponding to the time window is obtained, taking the global temperature deviation as the input and the valve opening degree of the preheating roller heating medium as the output. In the high frequency scenario, the average temperature difference of the spatial cells in the time window is taken as the output, and the adjustment power of the infrared heater array is taken as the output.

8. The method of synergic control of the heating assembly of a bidirectional stretching machine according to claim 1, characterized in that, In the protective gradient power regulation, first, the position of the preheating roller producing the scratch along the width direction is determined, and the power of the corresponding multiple infrared heater arrays upstream of the position is raised in a gradient manner, and the size of the raised power is where i = 1, 2, 3, …, N; N represents the number of infrared heater arrays from the upstream of the preheating roller to the end of the position of another preheating roller; 10% of the current power of the adjacent first infrared heater array upstream of the preheating roller.

Images