Master batch mixing strength optimization method for image granularity recognition

Abstract

The invention relates to the technical field of data quality monitoring, and provides a master batch mixing strength optimization method for image granularity recognition. The method comprises the following steps: randomly sampling master batch produced by mixed extrusion according to an initial sampling area to obtain a first image and identify granularity to obtain first master batch granularity distribution; calculating a change parameter according to the particle size distribution, adjusting the sampling area and continuing acquisition and analysis until a parameter difference value is smaller than a threshold value or reaches the maximum sampling area, and obtaining a particle size change parameter; setting a component granularity adjustment step length according to the parameter, and optimizing the granularity of the master batch in a granularity space to obtain a maximum combination of mixing strength; the granularity is adjusted according to the combination, and the optimal mixing strength is combined to serve as a mixing strength optimization result. The technical problem of unstable mixing strength caused by uneven particle size distribution in the master batch mixing process is solved, and the technical effects of optimizing master batch components based on particle size identification and adjustment and improving mixing uniformity and production stability are achieved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, specifically to the field of data quality monitoring technology, and in particular to a method for optimizing masterbatch mixing strength by image particle size recognition. Background Art

[0002] In the production and processing of masterbatch, the uniformity of particle size distribution is crucial to the quality and stability of the product. However, in the existing production process, the particle size distribution of the masterbatch is often difficult to be effectively controlled, resulting in poor mixing uniformity, which in turn affects the performance of the final product. Especially in the extrusion production process, the uneven particle size distribution will cause fluctuations in mixing intensity, which will significantly affect the physical properties and quality stability of the product. The existing technology mainly relies on physical stirring, empirical setting and other methods to control masterbatch mixing, but these methods are difficult to ensure the accuracy of particle size distribution. Due to the lack of real-time monitoring of the masterbatch particle size and distribution, it is difficult to detect and adjust the fluctuation of particle size in time, resulting in unstable masterbatch mixing effect between batches or within the same batch. In addition, traditional particle size detection methods are mostly offline detection, which cannot meet the rapid feedback requirements in the production process, which further increases the difficulty of particle size control in the production process. Therefore, how to achieve accurate monitoring and optimization control of particle size distribution in the production process has become a major technical problem in the current masterbatch mixing process. Summary of the invention

[0003] The present application provides a method for optimizing masterbatch mixing strength by image particle size recognition, aiming to solve the technical problem of unstable mixing strength caused by uneven particle size distribution during masterbatch mixing.

[0004] In view of the above problems, the present application provides a method for optimizing masterbatch mixing strength by image particle size recognition.

[0005] The present application provides a method for optimizing masterbatch mixing intensity by image particle size recognition, the method comprising: performing random image sampling on a masterbatch to be mixed and extruded according to an initial sampling area to obtain a first sampling image, performing image particle size recognition, and obtaining a first masterbatch particle size distribution; analyzing and calculating a first particle size change parameter according to the first masterbatch particle size distribution, adjusting the initial sampling area according to the first particle size change parameter, and continuing random image acquisition and analysis until the difference in the particle size change parameter is less than a preset particle size change threshold or reaches a maximum sampling area to obtain a particle size change parameter; setting an adjustment step for adjusting the component particle sizes of a plurality of masterbatch components in the masterbatch according to the particle size change parameter, optimizing the masterbatch component particle size adjustment within the masterbatch particle size space of the plurality of masterbatch components, and obtaining an optimal masterbatch component particle size combination with the maximum mixing intensity; adjusting the particle size of the masterbatch according to the optimal masterbatch component particle size combination, and combining the optimal mixing intensity as the masterbatch mixing intensity optimization result.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: The above-mentioned method for optimizing the mixing intensity of a masterbatch by image particle size recognition first performs random image sampling on the masterbatch to be processed, obtains the first batch of images based on the sampling area initially set, and then performs particle size recognition on these images to obtain the initial particle size distribution of the masterbatch. By analyzing the particle size distribution of the first masterbatch, the first particle size change parameter is calculated. If the particle size change parameter shows a significant change, the sampling area is appropriately adjusted, and random image acquisition and analysis are continued to gradually narrow the gap in the particle size change parameter until the difference in the particle size change parameter is less than the set threshold, or the sampling area reaches the maximum value. This step ensures the stability and accuracy of the particle size distribution and the reliability of subsequent adjustments. Subsequently, according to the obtained particle size change parameter, a suitable adjustment step is set to optimize the particle size of each component in the masterbatch. At this stage, the component particle size combination that can achieve the maximum mixing intensity is found by fine-tuning the component particle size within the masterbatch particle size space. This optimization process makes the particle size more uniform, thereby improving the mixing intensity of the masterbatch. Afterwards, the optimal masterbatch component size combination is applied to the actual masterbatch size adjustment to ensure that the mixed masterbatch has the best strength and uniformity. The entire process gradually optimizes sampling and adjusts the step length to continuously improve the accuracy of particle size distribution and the mixing strength of the masterbatch, achieving stable and high-quality extrusion production.

[0007] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0009] Figure 1 A schematic flow chart of a method for optimizing masterbatch mixing strength by image particle size recognition in one embodiment; Figure 2 The figure is a schematic flow chart of obtaining the particle size distribution of the first masterbatch in a masterbatch mixing strength optimization method based on image particle size recognition in one embodiment. DETAILED DESCRIPTION

[0010] The embodiment of the present application solves the technical problem of unstable mixing strength caused by uneven particle size distribution during masterbatch mixing by providing a method for optimizing masterbatch mixing strength through image particle size recognition.

[0011] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0012] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0013] Examples, such as Figure 1 As shown, the present application provides a method for optimizing masterbatch mixing intensity by image particle size recognition, the method comprising: The masterbatch to be mixed and extruded is randomly sampled according to the initial sampling area to obtain a first sampling image, and image particle size recognition is performed to obtain a first masterbatch particle size distribution.

[0014] In an embodiment of the present application, before starting the mixed extrusion production, a sample is first taken from the masterbatch to be processed in order to analyze the particle size distribution of its particles. The specific method is to randomly select a certain number of areas from the masterbatch for image acquisition according to the preset initial sampling area to obtain a first sampling image. These images provide the particle distribution of the masterbatch within the current sampling area. Subsequently, the collected images are analyzed using a pre-trained masterbatch particle size identifier to identify and record the particle size of the masterbatch particles, and the particle size distribution of the first masterbatch is obtained in combination with the position information of the masterbatch particles in the first sampling image. This distribution data includes the size, quantity and distribution characteristics of the particles, which can serve as the basis for subsequent analysis and adjustment. Through this preliminary image sampling and recognition, the particle state of the masterbatch can be intuitively understood, providing basic information for the subsequent calculation of particle size change parameters.

[0015] Further, if Figure 2 As shown, the present application provides a method of performing random image sampling on a masterbatch to be produced by mixed extrusion according to an initial sampling area, obtaining a first sampling image, performing image particle size recognition, and obtaining a first masterbatch particle size distribution, including: The masterbatch to be mixed and extruded is randomly sampled according to the initial sampling area to obtain a first sampling image; a convolutional neural network is used to train a masterbatch particle size identifier; the first sampling image is input into the masterbatch particle size identifier for identification to obtain multiple masterbatch particle sizes of the masterbatch particles in the first sampling image, and a first masterbatch particle size distribution is constructed in combination with the position information of the masterbatch particles in the first sampling image.

[0016] Preferably, in the masterbatch to be mixed and extruded, an area is randomly selected according to a preset initial sampling area, and image sampling is performed to obtain a first sampling image, which contains information such as the arrangement, size and position of the masterbatch particles in the area. This step is intended to obtain the basic particle size characteristics of the raw material, laying the foundation for subsequent identification and analysis. In order to accurately identify the particle size of the masterbatch particles, a masterbatch particle size identifier based on a convolutional neural network (CNN) is constructed and trained. During the training process, the masterbatch particle images with annotations in the historical data are used as the training set, and these images have been annotated with the particle size. Through the layer-by-layer feature extraction of the convolutional neural network, the masterbatch particle size identifier learns how to accurately identify particles of different sizes and shapes, and gradually forms the ability to identify the masterbatch particle size. After the training is completed, this identifier has the ability to process actual sampling images and identify the particles therein. The collected first sampling image is input into the trained masterbatch particle size identifier for image analysis and particle identification. The identifier automatically identifies the particle size (such as diameter, area, etc.) of each particle in the image, and generates multiple masterbatch particle sizes of the masterbatch particles in the first sampling image. Then, based on the recognition results, the particle size of each particle and the corresponding position information in the first sampling image are combined to generate the first masterbatch particle size distribution. This distribution information can reflect the size, spacing and arrangement of the masterbatch particles in the current sampling area, providing basic data for the subsequent analysis of particle size change parameters.

[0017] Furthermore, the present application provides a method for training a masterbatch particle size identifier using a convolutional neural network, including: According to the masterbatch particle size detection data in historical time, sampling images under different sampling areas are collected to obtain a sample sampling image set, and the particle size of the masterbatch particles in the sample sampling image is marked to obtain multiple sample masterbatch particle size sets; the sample sampling image set and the multiple sample masterbatch particle size sets are used as supervised training data, and a masterbatch particle size identifier is trained based on a convolutional neural network.

[0018] Optionally, the masterbatch particle size detection data at different times and batches are collected, focusing on the image information under different sampling areas. These historical data can provide the particle size distribution of masterbatch particles at different times or under different conditions. According to the collected historical data, multiple sampling areas (such as 5 square centimeters, 10 square centimeters, etc.) are selected to sample the masterbatch under these different sampling areas, so as to obtain a set of sample sampling images. These images contain various states of different particle size distributions and masterbatch particles, so as to cover more sample features. In the obtained sample sampling images, the masterbatch particles are annotated, and the particle size information of each particle (such as diameter, area, etc.) is recorded, so as to obtain multiple sample masterbatch particle size sets, providing accurate labels for subsequent supervised training. Subsequently, all the annotated sample sampling images and their corresponding masterbatch particle size information are combined to form a supervised training data set. This data set contains image information and label information (i.e., the particle size and position of the particles in each image), providing a complete input and output pair for the training of the convolutional neural network. After obtaining the supervised training data set, it is divided into 80% training set and 20% validation set to prevent the model from overfitting. The architecture of the convolutional neural network is redesigned, including convolutional layers, pooling layers, and fully connected layers. The convolutional layers are used to extract local features such as the edge and texture of the masterbatch particles. The pooling layers reduce the computational complexity by downsampling and retain significant features. The fully connected layers are used to integrate feature information and output the predicted particle size value. After that, the training hyperparameters are set, including the learning rate, batch size, and number of iterations. The learning rate controls the step size of parameter updates, the batch size determines the amount of data at each update, and the number of iterations determines the number of times the entire data set passes through the model. During the model training process, the training set data will be input into the network for forward propagation, the image features will be extracted through the convolutional layer, the pooling layer will perform downsampling, and the fully connected layer will finally output the prediction results. The model calculates the loss between the prediction result and the labeled value through the mean square error, measures the error size of the prediction, and calculates the gradient of the weights of each layer through back propagation. The Adam optimizer is used to optimize the model parameters and gradually adjust the weights to minimize the loss value. The process of forward propagation, loss calculation, back propagation, and parameter update is repeated until the maximum number of iterations is reached. At the end of each iteration, the validation set is used to test the performance of the model, calculate the validation loss and accuracy, and evaluate the generalization ability of the model. If the validation loss stops decreasing, the training is stopped in advance to prevent overfitting. After the training is completed, the model with the best performance is saved to obtain the masterbatch particle size identifier, which is used for particle size identification tasks in the production process.

[0019] According to the particle size distribution of the first masterbatch, the first particle size change parameter is analyzed and calculated, and according to the first particle size change parameter, the initial sampling area is adjusted, and random image acquisition and analysis are continued until the difference in the particle size change parameter is less than the preset particle size change threshold or the maximum sampling area is reached, thereby obtaining the particle size change parameter.

[0020] In one embodiment, according to the obtained first masterbatch particle size distribution, the first particle size variation parameter is calculated by averaging the particle size variation coefficient. This parameter reflects the uniformity of the masterbatch particle size distribution in the current sampling area. The larger the variation parameter, the more uneven the particle size distribution is, and the greater the impact on the mixing intensity. If this variation parameter is large, it means that the image currently sampled may not be representative. In order to obtain more accurate particle size distribution data, the initial sampling area is adjusted according to the first particle size variation parameter, and the sampling area is expanded to obtain a more representative sampling area. New random images are collected and analyzed in the expanded sampling area to obtain a second particle size variation parameter. This process is continued, and the sampling area is adjusted according to the previous particle size variation parameter each time, and the image is collected and analyzed again until the difference between the two adjacent particle size variation parameters is less than the preset particle size variation threshold (such as 0.05) or reaches the preset maximum sampling area, and finally a relatively stable particle size variation parameter is obtained. In order to more accurately represent the particle size distribution under different sampling areas, the particle size variation parameters obtained under different sampling areas are weighted. According to the size of each sampling area, weights are assigned, and all particle size change parameters are weighted and summed to obtain a comprehensive particle size change parameter. The larger the comprehensive particle size change parameter, the more uneven the overall particle size distribution is, which provides a key basis for the subsequent optimization of mixing intensity.

[0021] Further, the present application provides a method of obtaining a first particle size change parameter by analyzing and calculating the particle size distribution of the first masterbatch, adjusting the initial sampling area according to the first particle size change parameter, and continuing to collect and analyze random images until the difference in the particle size change parameter is less than a preset particle size change threshold or reaches a maximum sampling area, and obtaining the particle size change parameter, including: A first masterbatch particle size is extracted within the first masterbatch particle size distribution, and a first random masterbatch particle size is randomly extracted from other masterbatch particle sizes within the first masterbatch particle size distribution; the ratio of the difference between the first masterbatch particle size and the first random masterbatch particle size to the first random masterbatch particle size is calculated to obtain a first particle size variation coefficient; multiple particle size variation coefficients are continuously calculated to obtain an average value to obtain a first particle size variation parameter; according to the sum of the first particle size variation parameter and 1, the initial sampling area is calculated and expanded to obtain an adjusted sampling area, and random image acquisition and analysis are continued until the difference in the particle size variation parameter is less than a preset particle size variation threshold or the maximum sampling area is reached to obtain the particle size variation parameter.

[0022] Preferably, in the first masterbatch particle size distribution, a masterbatch particle size as a reference is first randomly selected and recorded as the first masterbatch particle size. A comparison particle size is randomly selected from other masterbatch particle sizes in the distribution and recorded as the first random masterbatch particle size. Subsequently, the absolute value of the difference between the first masterbatch particle size and the first random masterbatch particle size is calculated, and the difference is divided by the first random masterbatch particle size to obtain the first particle size variation coefficient. The larger this ratio is, the greater the difference in particle size between the two is, and the worse the uniformity of the masterbatch particle size distribution is. Repeat this process, extract different random masterbatch particle sizes and reference masterbatch particle sizes for calculation multiple times, so as to obtain multiple particle size variation coefficients. Afterwards, take the average of these particle size variation coefficients to obtain the first particle size variation parameter. This parameter can reflect the uniformity of the particle size distribution in the current sampling area. The higher the value, the more uneven the particle size distribution is. If the particle size variation parameter is large, it may indicate that the current sampling area is not representative enough, so the sampling area needs to be expanded. At this time, the particle size change parameter and 1 are directly multiplied by the initial sampling area, and then rounded to get a new adjusted sampling area. The adjusted sampling area is used to continue random image acquisition and analysis, and the new particle size change parameter is recalculated. This cycle is repeated until the difference between two adjacent particle size change parameters is less than the preset particle size change threshold (such as 0.05), or the sampling area has reached the maximum setting value, thereby finally obtaining a stable particle size change parameter.

[0023] Further, the present application provides that according to the particle size distribution of the first masterbatch, random image acquisition and analysis are continued until the difference of the particle size change parameter is less than a preset particle size change threshold or the maximum sampling area is reached, and the particle size change parameter is obtained, including: According to the adjusted sampling area, random image sampling is performed on the masterbatch to be mixed and extruded to obtain a second sampling image and analyze and calculate it to obtain a second particle size change parameter; the difference between the second particle size change parameter and the first particle size change parameter is calculated to determine whether it is less than a preset particle size change threshold value, if so, the image sampling analysis is stopped, and the first particle size change parameter and the second particle size change parameter are weightedly calculated according to the size of the adjusted sampling area and the initial sampling area to obtain the particle size change parameter; if not, the adjusted sampling area is continued to be calculated and expanded according to the sum of the second particle size change parameter and 1, and random image acquisition and analysis are continued until the difference between the particle size change parameters is less than the preset particle size change threshold value or the maximum sampling area is reached, and the particle size change parameter is obtained by weighted calculation.

[0024] Optionally, random image sampling is performed on the masterbatch to be mixed and extruded according to the adjusted sampling area to obtain a new sampling image as the second sampling image. The same particle size analysis is performed on the second sampling image to calculate the second particle size change parameter of the masterbatch, which is used to evaluate the uniformity of the particle size distribution of the masterbatch under the current sampling area. Subsequently, the difference between the first particle size change parameter and the second particle size change parameter is calculated to determine whether the difference is less than a preset particle size change threshold (e.g., 0.05). If the difference is less than the threshold, it means that the sampling area is sufficient to represent the particle size distribution characteristics of the current masterbatch, and the image sampling and analysis process can be stopped. Afterwards, according to the size of the adjusted sampling area and the initial sampling area, weights are assigned to perform weighted calculations on the first and second particle size change parameters, and finally a comprehensive particle size change parameter is obtained. If the difference is greater than or equal to the preset particle size change threshold, it means that the current sampling area is still insufficient to obtain stable particle size distribution data. At this time, according to the sum of the second particle size change parameter and 1, the adjusted sampling area is continued to be expanded, and a new sampling area is obtained by multiplying the adjusted sampling area by the sum and rounding. Then, random image sampling and analysis are performed again according to the new sampling area to calculate the new particle size change parameters. This process will continue. After each sampling and calculation to obtain new particle size change parameters, check whether the difference between two adjacent particle size change parameters is less than the preset threshold. If it is less than the threshold, stop sampling and analysis; otherwise, continue to expand the sampling area. Until the difference in particle size change parameters is less than the preset particle size change threshold, or the maximum set sampling area is reached. Finally, the particle size change parameters obtained multiple times are weighted and calculated according to the weight of the sampling area to obtain the final particle size change parameters, which can accurately reflect the uniformity of the masterbatch particle size and provide reliable data support for the subsequent optimization of mixing intensity.

[0025] According to the particle size change parameter, an adjustment step is set for adjusting the component particle sizes of the various masterbatch components in the masterbatch, and within the masterbatch particle size space of the various masterbatch components, the masterbatch component particle size adjustment optimization is performed to obtain an optimal masterbatch component particle size combination with the greatest mixing intensity.

[0026] In one embodiment, the step size of the masterbatch component particle size adjustment is set according to the calculated particle size change parameter. The larger the particle size change parameter, the more uneven the particle size distribution is, so a larger step size is set to speed up the optimization efficiency; when the particle size change parameter is small, a smaller step size is set to improve the optimization accuracy. On this basis, different step sizes are used to gradually adjust the particle sizes of multiple components in the masterbatch. Optimization operations are performed within the particle size space of multiple masterbatch components. By gradually adjusting the particle size of each component, the particle size combination of each masterbatch component is predicted and evaluated by the mixing strength predictor to find the optimal particle size combination that can improve the mixing strength. Finally, after multiple adjustments and optimizations, the optimal masterbatch component particle size combination with the greatest mixing strength is obtained, providing a component configuration scheme with optimal shear strength for subsequent mixed extrusion production.

[0027] Further, the present application provides that according to the particle size distribution of the first masterbatch, random image acquisition and analysis are continued until the difference of the particle size change parameter is less than a preset particle size change threshold or the maximum sampling area is reached, and the particle size change parameter is obtained, including: Obtain a preset particle size adjustment step for adjusting the component particle size; multiply the particle size change parameter by the preset particle size adjustment step to obtain the adjustment step; obtain the masterbatch particle size space of the component particle sizes of the various masterbatch components in the masterbatch; within the masterbatch particle size space, use the adjustment step to adjust and optimize the masterbatch component particle sizes to obtain an optimal masterbatch component particle size combination with the maximum mixing intensity.

[0028] Preferably, a preset particle size adjustment step is set for the masterbatch component particle size adjustment. The preset step is usually set according to experimental experience or historical data, and will be used as a basic step for subsequent particle size adjustment processes, such as 0.05 mm, 0.2 mm, etc. Then multiply the particle size change parameter by the preset particle size adjustment step to obtain an adjustment step that adapts to the current particle size distribution. The larger the particle size change parameter, the larger the adjustment step, thereby accelerating the optimization efficiency; and when the particle size change parameter is small, the adjustment step is reduced to improve the adjustment accuracy. Subsequently, the spatial range of the particle sizes of various masterbatch components in the masterbatch is obtained, that is, the masterbatch particle size space. This space defines the adjustable range of the particle size of each component to ensure that the adjustment process is carried out within a reasonable particle size range. In the masterbatch particle size space, the adjustment step is gradually applied to optimize the particle size of each component. After each adjustment, the mixing intensity predictor is used to calculate the mixing intensity under the current particle size combination to understand the influence of the component on the mixed extrusion intensity. Through multiple iterative adjustments, the maximum value of the mixing intensity is gradually approached. Finally, after repeated adjustments and tests, an optimal masterbatch component particle size combination with the greatest mixing strength was obtained, which provided the best particle size configuration for subsequent extrusion production and improved the uniformity and shear strength of the product.

[0029] Furthermore, the present application provides a method of adjusting and optimizing the particle sizes of masterbatch components by using the adjustment step size within the masterbatch particle size space to obtain an optimal masterbatch component particle size combination with the maximum mixing intensity, including: The test obtains the current multiple average masterbatch particle sizes of the multiple masterbatch components as the first masterbatch component particle size combination; the adjustment step is used to adjust the first masterbatch component particle size combination within the masterbatch particle size space to obtain the second masterbatch component particle size combination; according to the second masterbatch component particle size combination, the mixed extrusion strength is predicted to obtain the second mixing strength; the masterbatch component particle size combination is adjusted and the mixed extrusion strength is predicted until the optimization converges, and the masterbatch component particle size combination with the largest mixing strength is output to obtain the optimal masterbatch component particle size combination.

[0030] Optionally, the current particle size of multiple components in the masterbatch is tested, that is, the particle size of multiple components is detected, and the mean of multiple particle sizes in each component is calculated to obtain multiple average masterbatch particle sizes, so as to obtain the first masterbatch component particle size combination in the initial state. This combination reflects the particle size distribution state of each component under the current conditions, which serves as the starting point for subsequent optimization and adjustment. The pre-built mixing intensity predictor is then used to predict the mixing intensity of the combination as the first mixing intensity. Subsequently, based on the calculated adjustment step, the first masterbatch component particle size combination is adjusted in the masterbatch particle size space. By gradually increasing or decreasing the particle size of each component (according to the adjustment step), a new particle size distribution combination is obtained as the second masterbatch component particle size combination. The mixing intensity predictor is then used to predict the mixed extrusion intensity of the second masterbatch component particle size combination to obtain the second mixing intensity. If the second mixing intensity is higher than the previous first mixing intensity, it means that the adjustment direction is effective; if it is lower than the previous first mixing intensity, the direction needs to be adjusted. On this basis, the same method is used to optimize and adjust the masterbatch component particle size combination. After each update of the particle size combination, the mixed extrusion strength is predicted, and the mixed strength value obtained by each prediction is recorded. The optimization process is continuously iterated, gradually approaching the maximum value of the mixing strength. This process will continue until the mixing strength converges, that is, further adjustment of the particle size combination will no longer significantly improve the mixing strength, or the preset number of iterations is reached. Finally, the masterbatch component particle size combination with the largest mixing strength is output as the optimal masterbatch component particle size combination, which provides the best configuration for mixed extrusion in the actual production process and helps to improve the uniformity and physical strength of the product.

[0031] Further, the present application provides a method for predicting the mixed extrusion strength according to the particle size combination of the second masterbatch components to obtain a second mixed strength, including: According to the extrusion production detection data of the masterbatch, a set of sample masterbatch component particle size combinations is collected, and the shear strength of different sample masterbatch component particle size combinations after mixed extrusion is tested to obtain a sample mixing strength set; the sample masterbatch component particle size combination set and the sample mixing strength set are used to train a mixing strength predictor; the second masterbatch component particle size combination is input into the mixing strength predictor, and the second mixing strength is obtained by prediction output.

[0032] Optionally, according to the extrusion production test data of the masterbatch, a sample set of different masterbatch component particle size combinations is collected. For each sample masterbatch component particle size combination, the specific particle size information of each component is recorded to form a sample masterbatch component particle size combination set. Then, each sample masterbatch component particle size combination is actually mixed and extruded to test the shear strength of the combination after extrusion. By measuring and recording the shear strength of each combination, a sample mixing strength set is obtained. This set and the sample masterbatch component particle size combination set together constitute training data, providing data support for the training of the subsequent mixing strength predictor. Subsequently, the sample masterbatch component particle size combination set and the sample mixing strength set are used to train the mixing strength predictor, which can be constructed based on neural networks, decision trees, etc. The training process is similar to the aforementioned training masterbatch particle size identifier, and is carried out through steps such as forward propagation, loss calculation, back propagation, and parameter updating. After the training is completed, the target second masterbatch component particle size combination is input into the trained mixing strength predictor for prediction. The predictor outputs the mixing strength value of the combination according to the previously learned pattern to obtain the second mixing strength. This intensity value is used to evaluate the effect of the current particle size combination and provide a basis for decision-making during the particle size optimization process.

[0033] According to the optimal masterbatch component particle size combination, the masterbatch particle size is adjusted, combined with the optimal mixing strength, as the masterbatch mixing strength optimization result.

[0034] In one embodiment, according to the optimal masterbatch component particle size combination obtained by the optimization process, the particle sizes of each component of the masterbatch are adjusted to meet the particle size distribution requirements of the optimal combination. This adjustment process involves precise control of the size and distribution of the masterbatch particles to ensure that the particle size of each component is consistent with the optimal combination. After the particle size adjustment is completed, the masterbatch is put into mixed extrusion production, and an actual shear strength test is carried out to verify its mixing effect. The mixing strength of the actual test should be close to or reach the optimal mixing strength predicted by the previous optimization, thereby proving the effectiveness of the obtained particle size combination in improving mixing uniformity and strength. Finally, the masterbatch configuration that meets the optimal particle size combination and achieves the best mixing strength is used as the final result of the masterbatch mixing strength optimization to guide the masterbatch mixing process in production to ensure product consistency and physical property optimization.

[0035] In summary, the embodiments of the present application have at least the following technical effects: The embodiment of the present application obtains the particle size change parameter by sampling the image of the masterbatch and identifying the particle size distribution, and adjusts the sampling area based on the parameter for further analysis to obtain a stable particle size change parameter. The adjustment step size is set according to this parameter, and the particle sizes of the various components in the masterbatch are optimized and adjusted to obtain the particle size combination with the maximum mixing intensity. Subsequently, the intensity of different particle size combinations is predicted using a trained mixing intensity predictor, and finally the optimal masterbatch component particle size combination that maximizes the mixing intensity is determined. These technical effects jointly solve the technical problem of unstable mixing intensity caused by uneven particle size distribution during masterbatch mixing, and realize the optimization of masterbatch components based on particle size identification and adjustment, and improve the technical effect of mixing uniformity and production stability.

[0036] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0037] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0038] This specification and drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, a person skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A masterbatch mixing strength optimization method based on image particle size recognition, characterized in that: The method comprises: Performing random image sampling on the masterbatch to be mixed and extruded according to the initial sampling area to obtain a first sampling image, performing image particle size recognition to obtain a first masterbatch particle size distribution; According to the particle size distribution of the first masterbatch, a first particle size change parameter is obtained by analysis and calculation, and according to the first particle size change parameter, the initial sampling area is adjusted, and random image acquisition and analysis are continued until the difference of the particle size change parameter is less than a preset particle size change threshold or the maximum sampling area is reached, thereby obtaining the particle size change parameter; According to the particle size variation parameter, an adjustment step for adjusting the particle sizes of the various masterbatch components in the masterbatch is set, and within the masterbatch particle size space of the various masterbatch components, the masterbatch component particle size adjustment optimization is performed to obtain an optimal masterbatch component particle size combination with the greatest mixing intensity; According to the optimal masterbatch component particle size combination, the masterbatch particle size is adjusted, combined with the optimal mixing strength, as the masterbatch mixing strength optimization result.

2. The masterbatch mixing strength optimization method based on image particle size recognition according to claim 1, characterized in that: The masterbatch to be mixed and extruded is subjected to random image sampling according to the initial sampling area to obtain a first sampling image, and image particle size recognition is performed to obtain a first masterbatch particle size distribution, including: Performing random image sampling on the masterbatch to be mixed and extruded according to the initial sampling area to obtain a first sampling image; Use convolutional neural network to train masterbatch particle size identifier; The first sampling image is input into the masterbatch particle size identifier for identification, a plurality of masterbatch particle sizes of the masterbatch particles in the first sampling image are obtained, and a first masterbatch particle size distribution is constructed in combination with the position information of the masterbatch particles in the first sampling image.

3. The masterbatch mixing strength optimization method based on image particle size recognition according to claim 2, characterized in that: A convolutional neural network is used to train the masterbatch particle size identifier, including: According to the masterbatch particle size detection data in the historical time, sampling images under different sampling areas are collected to obtain a sample sampling image set, and the particle size of the masterbatch particles in the sample sampling image is marked to obtain multiple sample masterbatch particle size sets; The sample sampling image set and multiple sample masterbatch particle size sets are used as supervised training data, and a masterbatch particle size identifier is trained based on a convolutional neural network.

4. The masterbatch mixing strength optimization method based on image particle size recognition according to claim 1, characterized in that: According to the particle size distribution of the first masterbatch, a first particle size change parameter is obtained by analysis and calculation, and according to the first particle size change parameter, the initial sampling area is adjusted, and random image acquisition and analysis are continued until the difference of the particle size change parameter is less than a preset particle size change threshold or the maximum sampling area is reached, and the particle size change parameter is obtained, including: Selecting a first masterbatch particle size within the first masterbatch particle size distribution, and randomly selecting a first random masterbatch particle size from other masterbatch particle sizes within the first masterbatch particle size distribution; Calculate the ratio of the difference between the first masterbatch particle size and the first random masterbatch particle size to the first random masterbatch particle size to obtain a first particle size variation coefficient; Continue to calculate to obtain multiple particle size variation coefficients, and calculate the average to obtain a first particle size variation parameter; According to the sum of the first particle size change parameter and 1, the initial sampling area is calculated and expanded to obtain the adjusted sampling area, and random image acquisition and analysis are continued until the difference in the particle size change parameter is less than the preset particle size change threshold or the maximum sampling area is reached to obtain the particle size change parameter.

5. The method for optimizing masterbatch mixing strength by image particle size recognition according to claim 4, characterized in that: Continue to collect and analyze random images until the difference in the particle size change parameter is less than the preset particle size change threshold or reaches the maximum sampling area, and obtain the particle size change parameter, including: According to the sampling area adjustment, random image sampling is performed on the masterbatch to be mixed and extruded, a second sampling image is obtained, and analysis and calculation are performed to obtain a second particle size change parameter; Calculate the difference between the second particle size change parameter and the first particle size change parameter to determine whether it is less than a preset particle size change threshold, if so, stop image sampling analysis, and perform weighted calculation on the first particle size change parameter and the second particle size change parameter according to the size of the adjusted sampling area and the initial sampling area to obtain the particle size change parameter; If not, the adjusted sampling area is continued to be calculated and expanded according to the sum of the second particle size change parameter and 1, and random image acquisition and analysis are continued until the difference in the particle size change parameter is less than the preset particle size change threshold or the maximum sampling area is reached, and the particle size change parameter is obtained by weighted calculation.

6. The masterbatch mixing strength optimization method based on image particle size recognition according to claim 1, characterized in that: According to the particle size variation parameter, an adjustment step for adjusting the particle sizes of the various masterbatch components in the masterbatch is set, and within the masterbatch particle size space of the various masterbatch components, the masterbatch component particle size adjustment optimization is performed, including: Obtaining a preset particle size adjustment step for adjusting the particle size of the component; The particle size change parameter is multiplied by the preset particle size adjustment step to obtain the adjustment step; Obtaining a masterbatch particle size space of component particle sizes of a plurality of masterbatch components in the masterbatch; In the masterbatch particle size space, the adjustment step length is adopted to adjust and optimize the particle sizes of the masterbatch components to obtain the optimal masterbatch component particle size combination with the maximum mixing intensity.

7. The method for optimizing masterbatch mixing strength by image particle size recognition according to claim 6, characterized in that: In the masterbatch particle size space, the masterbatch component particle sizes are adjusted and optimized by using the adjustment step size to obtain an optimal masterbatch component particle size combination with the maximum mixing intensity, including: Testing and obtaining a plurality of current average masterbatch particle sizes of the plurality of masterbatch components as a first masterbatch component particle size combination; Using the adjustment step, within the masterbatch particle size space, the first masterbatch component particle size combination is adjusted to obtain a second masterbatch component particle size combination; According to the particle size combination of the second masterbatch components, a mixed extrusion strength prediction is performed to obtain a second mixed strength; Continue to adjust the masterbatch component size combination and predict the mixed extrusion strength until the optimization converges, output the masterbatch component size combination with the largest mixing strength, and obtain the optimal masterbatch component size combination.

8. The method for optimizing masterbatch mixing strength by image particle size recognition according to claim 7, characterized in that: According to the particle size combination of the second masterbatch components, a mixed extrusion strength prediction is performed to obtain a second mixed strength, including: According to the extrusion production test data of the masterbatch, a set of sample masterbatch component particle size combinations is collected, and the shear strength of different sample masterbatch component particle size combinations after mixed extrusion is tested to obtain a sample mixing strength set; Using the sample masterbatch component particle size combination set and the sample mixing intensity set, training a mixing intensity predictor; The second masterbatch component particle size combination is input into the mixing intensity predictor, and the prediction output is used to obtain the second mixing intensity.

Images