A method for precisely adding PVC plastic tile additive ingredients

Abstract

The present application relates to the technical field of raw material automatic batching, and specifically discloses a method for precise feeding of PVC plastic tile additive batching, which comprises the following steps: obtaining the target feeding amount of each additive after order switching; comparing the target feeding amount with the collected feeding weight to determine the error accumulation value of the additive; predicting the flow difference feature vector of the material in the mixing stage according to the error accumulation, the material density value and the shear rate; generating the device control instruction according to the flow difference feature vector and adjusting the stirring speed of the stirring device to obtain the collected image of the mixed material; performing pixel-level feature extraction on the collected image and determining the defect grade of the collected image; and generating the corresponding compensation feeding weight according to the defect grade. In the method for precise feeding of PVC plastic tile additive batching, the problem of difficulty in quickly adjusting the formula parameters after order switching is solved.

Description

Technical Field

[0001] This invention relates to the field of automatic raw material batching technology, and more specifically, to a method for precise dispensing of additives for PVC plastic tiles. Background Technology

[0002] PVC plastic roofing sheets are widely used as building roofing materials in industrial plants, warehousing and logistics facilities, and agricultural facilities. Their weather resistance, mechanical strength, and service life are directly related to building safety and economic benefits. In the production process of PVC plastic roofing sheets, the precision of the proportioning and the uniformity of mixing of various additives such as stabilizers, plasticizers, lubricants, fillers, and colorants are the core factors determining product quality.

[0003] Traditional production methods rely on manual experience to judge the amount of material to be added and the mixing speed. When faced with switching between multiple varieties and small batches of orders, this operating mode makes it difficult to quickly adjust the formula parameters, resulting in significant performance fluctuations in different batches of products. This increases quality inspection costs and also causes a rise in the rate of defective products. Summary of the Invention

[0004] In order to overcome the defects of the existing technology, the present invention provides a method for precise dosing of PVC plastic tile additives, which aims to solve the problems in the above-mentioned existing technology.

[0005] The technical solution adopted by this invention to solve its technical problem is: a method for precise dosing of additives for PVC plastic tiles, comprising the following steps: S1: Obtain the target amount of each additive to be added each time after the order is switched and the order of additive addition; S2: According to the order of additive addition, compare the target addition amount with the addition weight collected by the weighing sensor to determine the cumulative error value of each additive in the whole addition process. S3: Based on the cumulative error of each additive, the material density of each additive, and the shear rate of the material in the mixing stage, a support vector machine model is used to predict the flow difference feature vector of the material in the mixing stage. S4: Generate equipment control commands for the mixing equipment based on the flow difference feature vector, adjust the mixing speed of the mixing equipment according to the equipment control commands, and acquire images of the mixed materials; S5: Pixel-level feature extraction is performed on the acquired images using a random forest model, and the defect level of the acquired images with stratification is determined. S6: Generate the corresponding compensation feed weight based on the defect level, and input the target feed amount plus the compensation feed weight as the new target feed amount into the formula adjustment data.

[0006] Preferably, in step S1, the steps of obtaining the target amount of each additive to be added each time after the order switch and the order of additive addition include: Obtain the order switching instruction and extract the target product identifier from the order switching instruction; The formula data corresponding to the target product identifier is obtained by matching the target product identifier in a relational database. The formula data includes the order of additive addition and the target amount of each additive to be added each time.

[0007] Optionally, in step S2, the step of determining the cumulative error value for each additive throughout the feeding process includes: The real-time weight of each additive is collected by the corresponding weighing sensor according to the order in which the additives are added. The single feeding error value for each feeding is obtained by subtracting the target feeding amount from the real-time feeding weight. For the current production batch, multiple single-feeding error values ​​of the same additive are obtained, and these multiple single-feeding error values ​​are input into an accumulator. The accumulator performs an accumulation operation on the multiple single-feeding error values ​​to obtain the cumulative error value corresponding to each additive.

[0008] Specifically, in step S3, the step of using a support vector machine model to predict the flow difference feature vectors of materials in the mixing stage includes: Obtain the material density values ​​and particle size distribution data of each additive, and then weight and fuse the cumulative error value, the material density values, and the particle size distribution data to determine the comprehensive characteristic quantity of the additive. The shear rate of the material in the mixing stage is obtained, and the comprehensive characteristic value of the additive is multiplied by the shear rate to obtain the coefficient parameter corresponding to each additive. A feature matrix is ​​constructed by array concatenation based on the coefficient parameters, the cumulative error value, and the material density value. Obtain the pre-established support vector machine classification hyperplane; The vertical distance from each row of vectors in the feature matrix to the classification hyperplane of the support vector machine is calculated using a support vector machine model. The hyperplane distance corresponding to each row in the feature matrix is ​​then used as the hyperplane distance for each additive. For each additive, the hyperplane distance is obtained by setting a preset numbering order for each additive. The hyperplane distances for each additive are then concatenated into an array according to the preset numbering order to obtain a flow difference feature vector.

[0009] It is worth noting that, in step S3, the step of determining the comprehensive characteristic quantity of the additive includes: The median diameter value is obtained from the particle size distribution data; For the cumulative error value, the material density value, and the median diameter value, obtain the first preset weight corresponding to the cumulative error value, the second preset weight corresponding to the material density value, and the third preset weight corresponding to the median diameter value; The cumulative error value is multiplied by the first preset weight to obtain the first weighted value, the material density value is multiplied by the second preset weight to obtain the second weighted value, the median diameter value is multiplied by the third preset weight to obtain the third weighted value, and the first weighted value, the second weighted value and the third weighted value are summed to obtain the comprehensive characteristic quantity of the additive.

[0010] Preferably, in step S4, the step of generating equipment control commands for the mixing equipment based on the flow difference feature vector includes: The flow difference feature vector is input into a numerical comparator. The numerical comparator compares each element of the flow difference feature vector with a pre-established multi-level threshold interval. If an element of the flow difference feature vector falls into the target threshold interval, the material imbalance level corresponding to that element is determined. The average imbalance level is obtained by calculating the average of all material imbalance levels. Obtain a pre-established hash map table, which contains the correspondence between multiple imbalance average levels and equipment control commands; The corresponding device control command is retrieved from the hash map table based on the average imbalance level.

[0011] Optionally, in step S5, the step of determining the defect level of the acquired image exhibiting layering includes: The acquired image is converted to grayscale to obtain a grayscale pixel matrix; For the grayscale pixel matrix, a sliding window of a preset size is used to traverse the grayscale pixel matrix to obtain the local window matrix corresponding to each pixel. For the local window matrix, the contrast features and energy features of the local window matrix are extracted by the gray-level co-occurrence matrix, and the contrast features and energy features are concatenated to obtain the feature vector corresponding to each pixel. The feature vector is input into a pre-established random forest algorithm. The feature vector is classified by multiple decision trees in the random forest algorithm to obtain the voting results of the multiple decision trees. The stratification probability corresponding to each pixel is determined. Based on the stratification probability, it is determined whether the stratification probability is greater than a preset probability threshold. If the stratification probability is greater than the preset probability threshold, the pixel is determined to be an abnormal pixel. Based on the abnormal pixels, the pixel values ​​of the abnormal pixels in the grayscale pixel matrix are set to 1, and the pixel values ​​of the non-abnormal pixels are set to 0, resulting in an abnormal binary image. For the abnormal binary image, an eight-connected labeling tool is used to label the pixels with a value of 1 in the abnormal binary image as connected components, resulting in at least one connected region. For the connected region, the total number of pixels within the connected region is counted to obtain the connected area corresponding to the connected region; Determine whether there are connected regions with a connected area greater than a preset area threshold. If so, generate a determination result that there is a layering phenomenon in the acquired image, and take all connected regions with a connected area greater than the preset area threshold as valid statistical regions. Calculate the sum of the connected areas of all valid statistical regions as the defect area. Obtain the pre-established mapping relationship between area ranges and defect levels, and determine the defect level corresponding to the area range in which the defect area is located based on the mapping relationship between area ranges and defect levels.

[0012] Specifically, in step S6, the step of generating the corresponding compensation feed weight based on the defect level includes: For each target product identifier, retrieve the mapping relationship between the defect level and the compensation weight corresponding to the current target product identifier from the relational database; The compensation feed weight corresponding to the current defect level is determined based on the mapping relationship between defect level and compensation weight.

[0013] The beneficial effects of this invention are as follows: In the method for precise addition of PVC plastic tile additives, addressing the business scenario problem of cumulative errors in the addition of multiple additives during order switching leading to differences in the fluidity and stratification of the mixed materials, the method obtains the target addition amount and order of each additive after order switching. The target addition amount is then compared sequentially with the real-time addition weight collected by the weighing sensor to determine the cumulative error value of each additive throughout the addition process. Combined with the material density values ​​of each additive and the shear rate of the material during the mixing stage, a support vector machine model is used to predict the flow difference feature vector of the material during the mixing stage. Based on this, equipment control instructions for the mixing equipment are generated to dynamically adjust the mixing speed. Subsequently, images of the mixed material are acquired, and pixel-level features are extracted from the acquired images using a random forest model to determine whether stratification features exist in the high-definition acquired images. A corresponding compensation addition weight is generated based on the defect level. The target addition amount plus the compensation addition weight is used as the new target addition amount and input into the formula adjustment data, thereby achieving closed-loop correction of the addition error and significantly improving the uniformity of the mixed materials and the accuracy of the formula. Attached Figure Description

[0014] Figure 1A flowchart illustrating the method for precisely dispensing additives for PVC plastic roofing tiles.

[0015] Figure 2 A topological diagram of a method for precisely dispensing additives for PVC plastic roofing tiles.

[0016] Figure 3 A topological graph for predicting the flow difference eigenvectors of materials during the mixing stage.

[0017] Figure 4 A topology map to determine the defect level of acquired images exhibiting stratification. Detailed Implementation

[0018] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0019] Combination Figures 1 to 4 The method for precisely dispensing additives for PVC plastic roofing sheets, as shown, includes the following steps: S1: Obtain the target amount of each additive to be added each time after the order is switched and the order of additive addition; S2: According to the order of additive addition, compare the target addition amount with the addition weight collected by the weighing sensor to determine the cumulative error value of each additive in the whole addition process. S3: Based on the cumulative error of each additive, the material density of each additive, and the shear rate of the material in the mixing stage, a support vector machine model is used to predict the flow difference feature vector of the material in the mixing stage. S4: Generate equipment control commands for the mixing equipment based on the flow difference feature vector, adjust the mixing speed of the mixing equipment according to the equipment control commands, and acquire images of the mixed materials; S5: Pixel-level feature extraction is performed on the acquired images using a random forest model, and the defect level of the acquired images with stratification is determined. S6: Generate the corresponding compensation feed weight based on the defect level, and input the target feed amount plus the compensation feed weight as the new target feed amount into the formula adjustment data.

[0020] In the method for precise addition of additives to PVC plastic tiles, addressing the business scenario where accumulated errors in the addition of multiple additives during order switching lead to differences in the fluidity and stratification of the mixed materials, the method obtains the target addition amount and order of each additive after order switching. The target addition amount is then compared sequentially with the real-time addition weight collected by the weighing sensor to determine the cumulative error value of each additive throughout the addition process. Combining the material density values ​​of each additive with the shear rate of the material during the mixing stage, a support vector machine model is used to predict the flow difference feature vector of the material during the mixing stage. Based on this, equipment control instructions are generated for the mixing equipment to dynamically adjust the mixing speed. Subsequently, images of the mixed material are acquired, and pixel-level features are extracted from the acquired images using a random forest model to determine whether stratification features exist in the high-definition acquired images. A corresponding compensation addition weight is generated based on the defect level. The target addition amount plus the compensation addition weight is used as the new target addition amount and input into the formula adjustment data, thereby achieving closed-loop correction of the addition error and significantly improving the uniformity of the mixed materials and the accuracy of the formula.

[0021] It is worth noting that, in step S1, the steps of obtaining the target amount of each additive to be added each time after the order switch and the order of additive addition include: Obtain the order switching instruction and extract the target product identifier from the order switching instruction; The formula data corresponding to the target product identifier is obtained by matching the target product identifier in a relational database. The formula data includes the order of additive addition and the target amount of each additive to be added each time.

[0022] For example, the production control system receives an order switching instruction from the upstream production scheduling system and extracts the target product identifier from the instruction. In one possible implementation, the system performs a precise match in a relational database based on the extracted target product identifier to obtain the corresponding formula data. The relational database uses a two-dimensional table structure to store standard production parameters for various products. By using the target product identifier as the primary key for querying, a unique formula record can be quickly located. The obtained formula data specifies in detail the core control logic of the production process, specifically including the order of additive addition and the target amount of each additive to be added each time.

[0023] The formulation data specifies that stabilizers should be added first, followed by plasticizers. Regarding the addition of plasticizers, the formulation data does not pre-determine a single, all-in-one addition; instead, the target amount is broken down into multiple additions. The first addition targets 40 parts of the total amount, and the second addition targets the remaining 60 parts. Therefore, the resulting order of additive addition is: stabilizer, plasticizer, lubricant, filler, colorant. This strict order of addition and precise control of the target amount for each addition, based on the formulation data, ensures that each additive reacts under optimal physical conditions.

[0024] Preferably, in step S2, the step of determining the cumulative error value for each additive throughout the feeding process includes: The real-time weight of each additive is collected by the corresponding weighing sensor according to the order in which the additives are added. The single feeding error value for each feeding is obtained by subtracting the target feeding amount from the real-time feeding weight. For the current production batch, multiple single-feeding error values ​​of the same additive are obtained, and these multiple single-feeding error values ​​are input into an accumulator. The accumulator performs an accumulation operation on the multiple single-feeding error values ​​to obtain the cumulative error value corresponding to each additive.

[0025] The control unit activates the weighing sensors of the corresponding hoppers sequentially according to the order of material feeding, collecting the real-time feeding weight of the corresponding additives. It should be noted that the process of obtaining the single feeding error value is based on a comparison and analysis between real-time data and the formulation standard. Specifically, after obtaining the real-time feeding weight of a certain additive, the control unit calculates the difference between it and the target feeding amount specified in the formulation data for that feeding. Subtracting the target feeding amount from the real-time feeding weight yields the single feeding error value for each feeding.

[0026] In actual production batches, certain key additives, such as plasticizers, often need to be added in multiple stages. For the current production batch, the control unit continuously tracks and acquires multiple single-dosing error values ​​generated by the same additive in multiple independent dosing actions. In one possible implementation, the system inputs the multiple single-dosing error values ​​into an accumulator, and the accumulator performs an algebraic accumulation operation to obtain the cumulative error value corresponding to each additive. For an additive with only one single-dosing error value, the single-dosing error value is directly used as the cumulative error value corresponding to that additive.

[0027] Optionally, in step S3, the step of using a support vector machine model to predict the flow difference feature vectors of materials in the mixing stage includes: Obtain the material density values ​​and particle size distribution data of each additive, and then weight and fuse the cumulative error value, the material density values, and the particle size distribution data to determine the comprehensive characteristic quantity of the additive. The shear rate of the material in the mixing stage is obtained, and the comprehensive characteristic value of the additive is multiplied by the shear rate to obtain the coefficient parameter corresponding to each additive. A feature matrix is ​​constructed by array concatenation based on the coefficient parameters, the cumulative error value, and the material density value. Obtain the pre-established support vector machine classification hyperplane; The vertical distance from each row of vectors in the feature matrix to the classification hyperplane of the support vector machine is calculated using a support vector machine model. The hyperplane distance corresponding to each row in the feature matrix is ​​then used as the hyperplane distance for each additive. For each additive, the hyperplane distance is obtained by setting a preset numbering order for each additive. The hyperplane distances for each additive are then concatenated into an array according to the preset numbering order to obtain a flow difference feature vector.

[0028] After determining the comprehensive characteristic parameters of the additives, the system obtains the shear rate of the material during the mixing stage, which characterizes the mechanical shear strength applied to the material by the agitator. Multiplying the comprehensive characteristic parameters of the additives by the shear rate yields the coefficient parameters corresponding to each additive. It should be noted that the system constructs a feature matrix based on the coefficient parameters, the cumulative error value, and the material density value, where each row vector represents an additive.

[0029] For the feature matrix, the system invokes a pre-established support vector machine (SVM) classification hyperplane. The SVM model is a pattern recognition algorithm, and its classification hyperplane is the decision boundary that divides the material flow state in a multi-dimensional space. The system calculates the vertical distance from each row vector representing the characteristics of each additive in the feature matrix to the classification hyperplane. This distance maps the degree to which the additive deviates from the baseline flow characteristics, thus obtaining the hyperplane distance corresponding to each additive. For example, the system obtains a preset numbering order for each additive, such as stabilizer 1, plasticizer 2, lubricant 3, filler 4, and colorant 5. The hyperplane distances corresponding to each additive are then concatenated in this order (e.g., from 1 to 5) to obtain a flow difference feature vector.

[0030] Specifically, in step S3, the step of determining the comprehensive characteristic quantity of the additive includes: The median diameter value is obtained from the particle size distribution data; For the cumulative error value, the material density value, and the median diameter value, obtain the first preset weight corresponding to the cumulative error value, the second preset weight corresponding to the material density value, and the third preset weight corresponding to the median diameter value; The cumulative error value is multiplied by the first preset weight to obtain the first weighted value, the material density value is multiplied by the second preset weight to obtain the second weighted value, the median diameter value is multiplied by the third preset weight to obtain the third weighted value, and the first weighted value, the second weighted value and the third weighted value are summed to obtain the comprehensive characteristic quantity of the additive.

[0031] For example, the system acquires the cumulative error value, material density value, and particle size distribution data of each additive, and extracts the median diameter value. Specifically, the median diameter value reflects the average fineness of the powder material. The system performs weighted fusion on the cumulative error value, the material density value, and the median diameter value to obtain the comprehensive characteristic quantity of the additive.

[0032] It is worth noting that, in step S4, the step of generating equipment control commands for the mixing equipment based on the flow difference feature vector includes: The flow difference feature vector is input into a numerical comparator. The numerical comparator compares each element of the flow difference feature vector with a pre-established multi-level threshold interval. If an element of the flow difference feature vector falls into the target threshold interval, the material imbalance level corresponding to that element is determined. The average imbalance level is obtained by calculating the average of all material imbalance levels. Obtain a pre-established hash map table, which contains the correspondence between multiple imbalance average levels and equipment control commands; The corresponding device control command is retrieved from the hash map table based on the average imbalance level.

[0033] For example, the system inputs the aforementioned generated flow difference feature vector into a numerical comparator, which is a logic operation unit for performing discretized interval matching.

[0034] Specifically, the numerical comparator pre-configures multiple threshold ranges for each additive. These ranges are continuous and non-overlapping numerical ranges set based on historical mixing experimental data. The system extracts each element from the flow difference feature vector and compares it with the upper and lower limits of the multi-level threshold range corresponding to the additive for that element.

[0035] Assume that the multi-level threshold intervals are divided into mild imbalance, moderate imbalance, and severe imbalance intervals. The material imbalance level corresponding to the mild imbalance interval is 1, the moderate imbalance interval is 2, and the severe imbalance interval is 3. If an element falls into the moderate imbalance interval, the system will mark the material imbalance level of the additive corresponding to that element as 3. This multi-level threshold judgment mechanism completes the conversion of feature vector values ​​into imbalance state labels. Then, the average imbalance level of all material imbalance levels is calculated to obtain the average imbalance level.

[0036] It should be noted that after determining the average imbalance level, the system invokes a pre-established hash mapping table. A hash mapping table is a data dictionary based on a key-value pair storage structure, which associates the average imbalance level as the key with the corresponding equipment control command as the value. The system uses the currently determined average imbalance level as the search key, calculates its storage address in the mapping table using a hash function, and then directly extracts the corresponding equipment control command. This command contains specific control parameters; for example, for the current average imbalance level, the extracted command might be to increase the spindle stirring speed by a specific number of revolutions and extend the mixing time. Through the key-value pair lookup method of the hash mapping table, the system completes the matching of the imbalance level to the mechanical action command.

[0037] For example, the system sends a control signal to the motor controller of the mixing equipment to adjust the mixing speed according to the parsed equipment control instructions, and triggers the industrial camera to acquire images of the mixed materials after the adjustment is completed.

[0038] Preferably, in step S5, the step of determining the defect level of the acquired image exhibiting layering includes: The acquired image is converted to grayscale to obtain a grayscale pixel matrix; For the grayscale pixel matrix, a sliding window of a preset size is used to traverse the grayscale pixel matrix to obtain the local window matrix corresponding to each pixel. For the local window matrix, the contrast features and energy features of the local window matrix are extracted by the gray-level co-occurrence matrix, and the contrast features and energy features are concatenated to obtain the feature vector corresponding to each pixel. The feature vector is input into a pre-established random forest algorithm. The feature vector is classified by multiple decision trees in the random forest algorithm to obtain the voting results of the multiple decision trees. The stratification probability corresponding to each pixel is determined. Based on the stratification probability, it is determined whether the stratification probability is greater than a preset probability threshold. If the stratification probability is greater than the preset probability threshold, the pixel is determined to be an abnormal pixel. Based on the abnormal pixels, the pixel values ​​of the abnormal pixels in the grayscale pixel matrix are set to 1, and the pixel values ​​of the non-abnormal pixels are set to 0, resulting in an abnormal binary image. For the abnormal binary image, an eight-connected labeling tool is used to label the pixels with a value of 1 in the abnormal binary image as connected components, resulting in at least one connected region. For the connected region, the total number of pixels within the connected region is counted to obtain the connected area corresponding to the connected region; Determine whether there are connected regions with a connected area greater than a preset area threshold. If so, generate a determination result that there is a layering phenomenon in the acquired image, and take all connected regions with a connected area greater than the preset area threshold as valid statistical regions. Calculate the sum of the connected areas of all valid statistical regions as the defect area. Obtain the pre-established mapping relationship between area ranges and defect levels, and determine the defect level corresponding to the area range in which the defect area is located based on the mapping relationship between area ranges and defect levels.

[0039] For example, the system performs grayscale processing on the acquired image to obtain a grayscale pixel matrix, and uses a sliding window traversal to obtain a local window matrix, wherein the grayscale level of the local window matrix is... Grayscale value Specifically, let the original co-occurrence matrix be denoted as... This indicates that within the local window matrix, for the current pixel grayscale... Pairing pixel grayscale values ​​at positions along a set direction and distance d pixel grayscale Total number of occurrences; Gray-level co-occurrence matrix The gray-level co-occurrence matrix (GLCM) is a statistical method used to describe texture by studying the correlation characteristics of gray-level values ​​between two pixels at a specific distance in image space.

[0040] The system extracts contrast and energy features of the local window matrix using the gray-level co-occurrence matrix. Contrast features In images of mixed materials, contrast features reflect the magnitude of local grayscale contrast; if uneven mixing leads to high local color contrast, the contrast value will be high. Energy features This reflects the uniformity of grayscale distribution; when materials are uniformly mixed, the grayscale distribution is concentrated and the energy value is high. The system concatenates the contrast feature and the energy feature to obtain the feature vector corresponding to each pixel.

[0041] It should be noted that the system inputs the feature vectors into the random forest algorithm. The random forest consists of multiple decision trees, each independently classifying the feature vectors based on different feature subsets to determine whether a pixel belongs to a hierarchical region. The system aggregates the classification results and uses the percentage of trees that classify a pixel as belonging to a hierarchical region as the hierarchical probability. If the probability is greater than a threshold, the pixel is marked as an anomalous pixel and an anomalous binary image is generated.

[0042] Then, an eight-connectedness labeling tool is used, which checks the eight adjacent pixels in the eight directions around the current pixel. If they are all abnormal, they are grouped into the same connected region and labeled. Finally, the system counts the total number of pixels in the connected region to obtain the connected area. Regions with areas greater than a threshold are summed to obtain the defect area and determine the defect level.

[0043] Optionally, in step S6, the step of generating the corresponding compensation feed weight based on the defect level includes: For each target product identifier, retrieve the mapping relationship between the defect level and the compensation weight corresponding to the current target product identifier from the relational database; The compensation feed weight corresponding to the current defect level is determined based on the mapping relationship between defect level and compensation weight.

[0044] For example, the system obtains the target product identifier of the current batch of materials and initiates a data query request to a relational database. Specifically, the relational database is a database that uses a relational model to organize data, storing business data in a two-dimensional table of rows and columns. In the material proportioning compensation business, the system pre-builds a product formula data table in the database. This data table uses the target product identifier as a unique primary key and stores multiple sets of mapping relationships between defect levels and compensation weights. This mapping relationship is essentially a conditional matching structure that reflects the additional raw material weight values ​​required for different additives for material mixing defects of different severity. The system uses a structured query language to parse the target product identifier, performs row-level scanning and precise matching in the data table, and extracts the set of compensation strategies for that specific product under different defect states.

[0045] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A method for precisely dispensing additives for PVC plastic roofing sheets, characterized in that, Includes the following steps: S1: Obtain the target amount of each additive to be added each time after the order is switched and the order of additive addition; S2: According to the order of additive addition, compare the target addition amount with the addition weight collected by the weighing sensor to determine the cumulative error value of each additive in the whole addition process. S3: Based on the cumulative error of each additive, the material density of each additive, and the shear rate of the material in the mixing stage, a support vector machine model is used to predict the flow difference feature vector of the material in the mixing stage. S4: Generate equipment control commands for the mixing equipment based on the flow difference feature vector, adjust the mixing speed of the mixing equipment according to the equipment control commands, and acquire images of the mixed materials; S5: Pixel-level feature extraction is performed on the acquired images using a random forest model, and the defect level of the acquired images with stratification is determined. S6: Generate the corresponding compensation feed weight based on the defect level, and input the target feed amount plus the compensation feed weight as the new target feed amount into the formula adjustment data.

2. The method for precise dosing of additives for PVC plastic roofing tiles according to claim 1, characterized in that, In step S1, the steps of obtaining the target amount of each additive to be added each time after the order switch and the order of additive addition include: Obtain the order switching instruction and extract the target product identifier from the order switching instruction; The formula data corresponding to the target product identifier is obtained by matching the target product identifier in a relational database. The formula data includes the order of additive addition and the target amount of each additive to be added each time.

3. The method for precise dosing of PVC plastic tile additives according to claim 1, characterized in that, In step S2, the step of determining the cumulative error value for each additive throughout the feeding process includes: The real-time weight of each additive is collected by the corresponding weighing sensor according to the order in which the additives are added. The single feeding error value for each feeding is obtained by subtracting the target feeding amount from the real-time feeding weight. For the current production batch, multiple single-feeding error values ​​of the same additive are obtained, and these multiple single-feeding error values ​​are input into an accumulator. The accumulator performs an accumulation operation on the multiple single-feeding error values ​​to obtain the cumulative error value corresponding to each additive.

4. The method for precise dosing of additives for PVC plastic roofing tiles according to claim 1, characterized in that, In step S3, the step of predicting the flow difference feature vector of materials in the mixing stage using a support vector machine model includes: Obtain the material density values ​​and particle size distribution data of each additive, and then weight and fuse the cumulative error value, the material density values, and the particle size distribution data to determine the comprehensive characteristic quantity of the additive. The shear rate of the material in the mixing stage is obtained, and the comprehensive characteristic value of the additive is multiplied by the shear rate to obtain the coefficient parameter corresponding to each additive. A feature matrix is ​​constructed by array concatenation based on the coefficient parameters, the cumulative error value, and the material density value. Obtain the pre-established support vector machine classification hyperplane; The vertical distance from each row of vectors in the feature matrix to the classification hyperplane of the support vector machine is calculated using a support vector machine model. The hyperplane distance corresponding to each row in the feature matrix is ​​then used as the hyperplane distance for each additive. For each additive, the hyperplane distance is obtained by setting a preset numbering order for each additive. The hyperplane distances for each additive are then concatenated into an array according to the preset numbering order to obtain a flow difference feature vector.

5. The method for precise dosing of PVC plastic tile additives according to claim 4, characterized in that, In step S3, the step of determining the comprehensive characteristic quantity of the additive includes: The median diameter value is obtained from the particle size distribution data; For the cumulative error value, the material density value, and the median diameter value, obtain the first preset weight corresponding to the cumulative error value, the second preset weight corresponding to the material density value, and the third preset weight corresponding to the median diameter value; The cumulative error value is multiplied by the first preset weight to obtain the first weighted value, the material density value is multiplied by the second preset weight to obtain the second weighted value, the median diameter value is multiplied by the third preset weight to obtain the third weighted value, and the first weighted value, the second weighted value and the third weighted value are summed to obtain the comprehensive characteristic quantity of the additive.

6. The method for precise dosing of additives for PVC plastic roofing tiles according to claim 1, characterized in that, In step S4, the step of generating equipment control commands for the stirring equipment based on the flow difference feature vector includes: The flow difference feature vector is input into a numerical comparator. The numerical comparator compares each element of the flow difference feature vector with a pre-established multi-level threshold interval. If an element of the flow difference feature vector falls into the target threshold interval, the material imbalance level corresponding to that element is determined. The average imbalance level is obtained by calculating the average of all material imbalance levels. Obtain a pre-established hash map table, which contains the correspondence between multiple imbalance average levels and equipment control commands; The corresponding device control command is retrieved from the hash map table based on the average imbalance level.

7. The method for precise dosing of additives for PVC plastic roofing tiles according to claim 1, characterized in that, In step S5, the step of determining the defect level of the acquired image exhibiting layering includes: The acquired image is converted to grayscale to obtain a grayscale pixel matrix; For the grayscale pixel matrix, a sliding window of a preset size is used to traverse the grayscale pixel matrix to obtain the local window matrix corresponding to each pixel. For the local window matrix, the contrast features and energy features of the local window matrix are extracted by the gray-level co-occurrence matrix, and the contrast features and energy features are concatenated to obtain the feature vector corresponding to each pixel. The feature vector is input into a pre-established random forest algorithm. The feature vector is classified by multiple decision trees in the random forest algorithm to obtain the voting results of the multiple decision trees. The stratification probability corresponding to each pixel is determined. Based on the stratification probability, it is determined whether the stratification probability is greater than a preset probability threshold. If the stratification probability is greater than the preset probability threshold, the pixel is determined to be an abnormal pixel. Based on the abnormal pixels, the pixel values ​​of the abnormal pixels in the grayscale pixel matrix are set to 1, and the pixel values ​​of the non-abnormal pixels are set to 0, resulting in an abnormal binary image. For the abnormal binary image, an eight-connected labeling tool is used to label the pixels with a value of 1 in the abnormal binary image as connected components, resulting in at least one connected region. For the connected region, the total number of pixels within the connected region is counted to obtain the connected area corresponding to the connected region; Determine whether there are connected regions with a connected area greater than a preset area threshold. If so, generate a determination result that there is a layering phenomenon in the acquired image, and take all connected regions with a connected area greater than the preset area threshold as valid statistical regions. Calculate the sum of the connected areas of all valid statistical regions as the defect area. Obtain the pre-established mapping relationship between area ranges and defect levels, and determine the defect level corresponding to the area range in which the defect area is located based on the mapping relationship between area ranges and defect levels.

8. The method for precise dosing of additives for PVC plastic roofing tiles according to claim 1, characterized in that, In step S6, the step of generating the corresponding compensation feed weight based on the defect level includes: For each target product identifier, retrieve the mapping relationship between the defect level and the compensation weight corresponding to the current target product identifier from the relational database; The compensation feed weight corresponding to the current defect level is determined based on the mapping relationship between defect level and compensation weight.

Images