Glass bottle production process parameter optimization method based on black spot detection and related equipment

Abstract

The invention belongs to the technical field of glass bottle production quality detection, and discloses a glass bottle production process parameter optimization method based on black spot detection and related equipment, and the method comprises the steps: obtaining first black spot detection data of a glass bottle before a cleaning link; acquiring second black spot detection data of the glass bottle after the cleaning link; for the same glass bottle, calculating a dynamic distance threshold according to the size of the black spot in the first black spot detection data; determining corresponding black spot pairs in the first black spot detection data and the second black spot detection data based on a dynamic distance threshold; according to a black spot size threshold value, performing type classification on black spots in the corresponding black spot pairs; based on the black spot type classification result, performing weighted statistical difference analysis to obtain a difference analysis result; based on a difference analysis result, optimizing glass bottle production process parameters; therefore, scientificity and effectiveness of technological parameter optimization can be improved, and it is guaranteed that the technological parameter optimization can really improve the glass bottle forming quality.

Description

Technical Field

[0001] The present application relates to the technical field of quality inspection in glass bottle production. Specifically, it relates to a method for optimizing glass bottle production process parameters based on black spot detection and related equipment. Background Art

[0002] During the high-speed production of food-grade glass bottles, the detection and control of surface black spot defects are directly related to product quality and food safety. For food-grade glass bottles, after molding, they usually need to be subjected to high-pressure cleaning. During the molding process, black spot defects and unexposed molding defects (such as internal bubbles) may be caused by unreasonable settings of molding process parameters (such as furnace temperature, cooling rate, etc.); during the high-pressure cleaning process, water flow impact may cause new black spot defects (such as microcracks or impurity residues) on the bottle surface, or enlarge the size of the original black spots. The new black spot defects or enlarged black spot sizes caused by water flow impact may be due to improper cleaning parameters, or may be caused by the exposure of originally unexposed molding defects due to water flow impact.

[0003] The prior art usually classifies all the black spot defects finally present in the glass bottle as molding defects, and thus adjusts the molding process parameters according to the final black spot defect data, which may lead to unreasonable adjustment of the molding process parameters. For example, if a large number of small-sized black spots are introduced only due to unreasonable setting of cleaning parameters in the cleaning process, the system may misjudge it as a raw material impurity problem in the molding process and wrongly increase the furnace temperature, which will instead exacerbate energy consumption and bottle body stress. This defect of confusing problems leads to the deviation of process optimization from actual requirements, not only unable to effectively improve the molding quality, but also may introduce new production fluctuations.

[0004] The prior art lacks a dynamic discrimination mechanism for the source of defects, making it difficult to ensure the scientificity and effectiveness of optimizing production process parameters.

[0005] In view of the above problems, the prior art urgently needs to be improved. Summary of the Invention

[0006] The purpose of the present application is to provide a method for optimizing glass bottle production process parameters based on black spot detection and related equipment, which can improve the scientificity and effectiveness of process parameter optimization and ensure that the process parameter optimization can truly improve the glass bottle molding quality.

[0007] In a first aspect, the present application provides a method for optimizing glass bottle production process parameters based on black spot detection, and the technical solution is as follows: Obtain the first black spot detection data of the glass bottle before the cleaning process, where the first black spot detection data includes the position and size of the black spots; Obtain the second black spot detection data of the glass bottle after the cleaning process, where the second black spot detection data includes the position and size of the black spots; For the same glass bottle, calculate a dynamic distance threshold based on the sizes of the black dots in the first black dot detection data; Based on the dynamic distance threshold, determine the corresponding black dot pairs in the first black dot detection data and the second black dot detection data; According to the black dot size threshold, classify the black dots in the corresponding black dot pairs; Based on the black dot type classification results, perform weighted statistical difference analysis to obtain the difference analysis results; Based on the difference analysis results, optimize the production process parameters of the glass bottle.

[0008] This method, through the dynamic matching, classification, and difference analysis of the black dot data before and after cleaning, can avoid the confusion of defect causes leading to the deviation of production process parameter optimization from actual requirements, improve the scientificity and effectiveness of process parameter optimization, and ensure that the process parameter optimization can truly improve the forming quality of glass bottles.

[0009] Furthermore, this application also proposes that the steps of calculating a dynamic distance threshold based on the sizes of the black dots in the first black dot detection data for the same glass bottle include: Real-time monitor the production line status parameters and product characteristic parameters; Determine a parameter adjustment model, which is used to determine the mapping relationship between the production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula; Based on the parameter adjustment model and real-time parameters, adjust the parameters in the dynamic distance threshold calculation formula; Use the adjusted parameters to calculate the dynamic distance threshold.

[0010] Since the parameters in the dynamic distance threshold calculation formula are dynamically adjusted according to the production line status and product characteristics, the calculated dynamic distance threshold is also more accurate and adaptable, can more effectively cope with various changes in the production process, and improve the accuracy of determining the black dot correspondence.

[0011] Furthermore, this application also proposes that the steps of determining a parameter adjustment model, which is used to determine the mapping relationship between the production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula, include: Real-time collect the production line status parameters of the glass bottle production line and the product characteristic parameters of the glass bottle. The production line status parameters include at least one of the furnace temperature, cooling rate, and production line speed, and the product characteristic parameters include at least one of the glass bottle wall thickness, bottle body size consistency, and historical black dot detection data statistical information; Based on the collected production line status parameters and product characteristic parameters, a machine learning algorithm or a statistical regression method is used to train or update a parameter adjustment model, and the parameter adjustment model is a mapping relationship model between the parameters in the dynamic distance threshold calculation formula and the production line status parameters and product characteristic parameters; Verify the performance of the parameter adjustment model, monitor the prediction accuracy of the parameters in the dynamic distance threshold calculation formula, and when the prediction accuracy is lower than the preset accuracy threshold, trigger the retraining or update of the parameter adjustment model.

[0012] This method makes the calculation of the dynamic distance threshold no longer depend on fixed empirical values, but can be adaptively adjusted according to the actual situation of the production line. Through the verification of the model performance and the iterative update mechanism, the accuracy of the model prediction and the reliability of the dynamic distance threshold calculation are ensured, and finally the effectiveness of black point detection and subsequent process parameter optimization is improved.

[0013] Furthermore, the present application also proposes that the steps of training or updating a parameter adjustment model based on the collected production line status parameters and product characteristic parameters by using a machine learning algorithm or a statistical regression method, where the parameter adjustment model is a mapping relationship model between the parameters in the dynamic distance threshold calculation formula and the production line status parameters and product characteristic parameters include: Analyze the correlation between the collected production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula; Based on the parameter correlation analysis results, select the production line status parameters and product characteristic parameters with a correlation higher than the preset threshold to form a subset of characteristic parameters; Use a machine learning algorithm or a statistical regression method to train or update the parameter adjustment model based on the subset of characteristic parameters.

[0014] Furthermore, the present application also proposes that the steps of analyzing the correlation between the collected production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula include: Monitor the execution period of the parameter correlation analysis in real time; If the current moment meets the preset periodic analysis trigger condition, then execute the parameter correlation analysis step; Alternatively, monitor the fluctuation amplitude of the production line status parameters and product characteristic parameters in real time; If it is detected that the fluctuation amplitude of at least one production line status parameter or product characteristic parameter exceeds the corresponding preset fluctuation threshold, then execute the parameter correlation analysis step.

[0015] Furthermore, the present application also proposes that the steps of determining the corresponding black point pairs in the first black point detection data and the second black point detection data based on the dynamic distance threshold include: Perform pose alignment processing on the second black dot detection data based on the glass bottle pose features extracted from the first black dot detection data and the second black dot detection data; After the pose alignment processing, determine the corresponding black dot pairs in the first black dot detection data and the second black dot detection data after the pose alignment processing based on the dynamic distance threshold.

[0016] Further, the present application also proposes that the steps of performing weighted statistical difference analysis based on the black dot type classification result to obtain the difference analysis result include: For each black dot type, respectively count the number, position coordinates, and size of the black dots before and after cleaning; For each black dot type, calculate at least one difference index; Assign weight coefficients to the difference indexes of each black dot type; Based on the weight coefficients, synthesize the difference indexes of different black dot types to generate a comprehensive difference analysis result.

[0017] Further, the present application also proposes that the steps of optimizing the glass bottle production process parameters based on the difference analysis result and the first black dot detection data include: Based on the difference analysis result, select a parameter adjustment strategy from the preset set of parameter adjustment strategies; Execute the selected parameter adjustment strategy to adjust the glass bottle production process parameters.

[0018] Further, the present application also proposes an electronic device, including a processor and a memory, where the memory stores a computer program executable by the processor, and when the processor executes the computer program, it runs the steps in the method for optimizing the glass bottle production process parameters based on black dot detection as described above.

[0019] Further, the present application also proposes a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it runs the steps in the method for optimizing the glass bottle production process parameters based on black dot detection as described above.

[0020] Beneficial effects: The method for optimizing the glass bottle production process parameters based on black dot detection and related devices provided by the present application can avoid the deviation of the production process parameter optimization from the actual requirements caused by the confusion of defect causes through the dynamic matching, classification, and difference analysis of the black dot data before and after cleaning, improve the scientificity and effectiveness of the process parameter optimization, and ensure that the process parameter optimization can truly improve the forming quality of the glass bottle. Description of the Drawings

[0021] Figure 1 It is a flowchart of the method for optimizing the glass bottle production process parameters based on black dot detection provided by the embodiment of the present application.

[0022] Figure 2 It is a schematic structural diagram of the electronic device provided by the embodiment of the present application.

[0023] Label description: 301, processor; 302, memory; 303, communication bus. Specific implementation manners

[0024] Next, the technical solutions in the present application will be clearly and completely described in conjunction with the accompanying drawings in the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Usually, the components of the present application described and shown in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application to be protected, but only represents the selected embodiments of the present application. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative efforts belong to the scope of protection of the present application.

[0025] It should be noted that: similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of the present application, the terms "first", "second", etc. are only used for differential description and cannot be understood as indicating or implying relative importance.

[0026] The present application proposes an optimization method for the production process parameters of glass bottles based on black dot detection. The method includes the following steps: A1. Obtain the first black dot detection data of the glass bottle before the cleaning link. The first black dot detection data includes the position and size of the black dots. A2. Obtain the second black dot detection data of the glass bottle after the cleaning link. The second black dot detection data includes the position and size of the black dots. A3. For the same glass bottle, calculate the dynamic distance threshold according to the size of the black dots in the first black dot detection data. A4. Based on the dynamic distance threshold, determine the corresponding black dot pairs in the first black dot detection data and the second black dot detection data. A5. According to the black dot size threshold, classify the black dots in the corresponding black dot pairs by type. A6. Based on the black dot type classification result, perform weighted statistical difference analysis to obtain the difference analysis result. A7. Based on the difference analysis result, optimize the production process parameters of the glass bottle.

[0027] Specifically, the present technical solution aims to solve the problem that during the production process of glass bottles, new black spots introduced in the cleaning process or the expansion of existing black spots interfere with the optimization of process parameters based on the black spot detection data before cleaning. By comparing the black spot detection data before and after cleaning, the impact of the cleaning process on black spot defects is distinguished and quantified, so as to more accurately optimize the process parameters based on the actual defect situation in the glass bottle forming process. First, obtain the black spot detection data before and after cleaning, including the position and size information of the black spots, as the basis for comparative analysis. Then, for the same glass bottle, calculate the dynamic distance threshold according to the size of the black spots before cleaning, making the corresponding black spot matching more flexible and accurate. Next, based on the dynamic distance threshold, determine the corresponding black spot pairs in the detection data before and after cleaning, providing an object basis for subsequent difference analysis. After that, classify the black spots in the corresponding black spot pairs according to the black spot size threshold, distinguishing black spots of different causes or natures. Further, based on the black spot type classification results, perform weighted statistical difference analysis to comprehensively evaluate the impact degree of the cleaning process on different types of black spots. Finally, based on the difference analysis results, optimize the glass bottle production process parameters, taking into account the impact of the cleaning process, and adjusting the parameters as much as possible based on the data reflecting the actual defect situation in the forming process, avoiding the deviation of the production process parameter optimization from the actual requirements caused by the confusion of defect causes, improving the accuracy and effectiveness of the process parameter optimization, and enhancing the product quality of the glass bottles.

[0028] Among them, the black spot detection data can be obtained by using existing black spot detection methods.

[0029] In some embodiments, in step A3, for the same glass bottle, the steps of calculating the dynamic distance threshold according to the size of the black spots in the first black spot detection data include: A301. Monitor the production line status parameters and product characteristic parameters in real time; A302. Determine the parameter adjustment model, which is used to determine the mapping relationship between the production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula; A303. Adjust the parameters in the dynamic distance threshold calculation formula based on the parameter adjustment model and the real-time parameters; A304. Calculate the dynamic distance threshold using the adjusted parameters.

[0030] Among them, in step A301, the status parameters of the production line and the product characteristic parameters are monitored in real time, specifically referring to collecting various parameters during the operation of the production line and the characteristic parameters of the product itself through sensors in real time. The status parameters of the production line may include, for example, parameters directly reflecting the operation status of the production line such as furnace temperature, cooling rate, and production line speed. The product characteristic parameters may include, for example, parameters characterizing the physical properties or historical defect conditions of the product itself such as the wall thickness of the glass bottle, the consistency of the bottle body size, and the statistical information of historical black spot detection data. The real-time monitoring of these parameters provides a data basis for the subsequent adjustment of the dynamic distance threshold.

[0031] In step A302, a parameter adjustment model is determined, specifically referring to establishing a mathematical model that can describe how the status parameters of the production line and the product characteristic parameters affect the parameters in the dynamic distance threshold calculation formula. The establishment of the parameter adjustment model can be achieved in various ways. As one way, based on expert experience or experimental data, the parameter adjustment model can be preset. For example, when the furnace temperature rises, a certain parameter value in the dynamic distance threshold calculation formula is correspondingly increased. As another way, machine learning or statistical regression methods can be used to train the parameter adjustment model using historical production data. The training process can be offline or updated online to ensure the accuracy and adaptability of the model. The core role of the parameter adjustment model is to establish the mapping relationship between the status parameters of the production line, the product characteristic parameters, and the parameters in the dynamic distance threshold calculation formula, providing model support for the adaptive adjustment of the dynamic distance threshold.

[0032] In step A303, based on the parameter adjustment model and the real-time parameters, the parameters in the dynamic distance threshold calculation formula are adjusted, specifically referring to inputting the real-time monitored status parameters of the production line and the product characteristic parameters into the parameter adjustment model, and the model calculates the parameter values that need to be adjusted in the dynamic distance threshold calculation formula according to the input parameters. Thus, the parameters in the dynamic distance threshold calculation formula are no longer fixed, but can be dynamically adjusted according to the actual production situation. This dynamic adjustment ensures that the dynamic distance threshold can better adapt to the fluctuations of the production line and product differences.

[0033] In step A304, the dynamic distance threshold is calculated using the adjusted parameters, specifically referring to substituting the parameters adjusted by the parameter adjustment model into the preset dynamic distance threshold calculation formula to calculate the final dynamic distance threshold. The dynamic distance threshold will be used to determine the corresponding relationship between the black spot data before and after cleaning. Since the parameters in the dynamic distance threshold calculation formula are dynamically adjusted according to the production line status and product characteristics, the calculated dynamic distance threshold is also more accurate and adaptable, and can more effectively cope with various changes in the production process, improving the accuracy of determining the black spot correspondence.

[0034] Specifically, on a high-speed production line for food-grade glass bottles, to solve the problem that the introduction of new or enlarged black spot defects in the cleaning process leads to the deviation of the process parameter optimization direction from the actual requirements, a method for dynamically adjusting the dynamic distance threshold is proposed. First, various sensors deployed on the production line are used to collect parameters such as the furnace temperature, cooling section temperature, production line speed, and glass bottle wall thickness in real time. Then, a parameter adjustment model is pre-trained. For example, a regression model that takes the furnace temperature, cooling section temperature, and glass bottle wall thickness as inputs and outputs the parameters that control the threshold size in the dynamic distance threshold calculation formula. The model training data can be sourced from the data collected on the historical production line, including the black spot detection data and the corresponding production line parameters under different production states. During the actual production process, the furnace temperature, cooling section temperature, and glass bottle wall thickness are obtained in real time and input into the pre-trained regression model, and the model outputs the adjusted parameter values. The adjusted parameter values are substituted into the dynamic distance threshold calculation formula to calculate the dynamic distance threshold under the current production state. The resulting dynamic distance threshold is used to determine the corresponding relationship between black spots before and after subsequent cleaning, thus ensuring that the process parameter optimization based on black spot difference analysis can truly improve the forming quality of glass bottles.

[0035] In some specific embodiments, the dynamic distance threshold calculation formula can be preset as: D = A * d + B, where D is the dynamic distance threshold, d is the black spot size, and A and B are adjustable parameters. The parameter adjustment model can be a linear regression model used to adjust the parameters A and B in the formula according to the real-time monitored production line state parameters and product characteristic parameters. For example, when the furnace temperature increases, the parameter adjustment model predicts that the value of parameter A increases and the value of parameter B decreases. The adjusted parameters A and B are substituted into the dynamic distance threshold calculation formula to obtain the dynamic distance threshold suitable for the current furnace temperature. This method enables the dynamic distance threshold to be adaptively adjusted with the change of the production line state, improves the accuracy of determining the corresponding relationship of black spots, and thus enhances the effect of process parameter optimization.

[0036] In a possible embodiment, in step A302, the steps of determining the parameter adjustment model, where the parameter adjustment model is used to determine the mapping relationship between the production line state parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula, include: Collect the state parameters of the glass bottle production line and the product characteristic parameters of the glass bottle in real time. The production line state parameters include at least one of the furnace temperature, cooling rate, and production line speed, and the product characteristic parameters include at least one of the glass bottle wall thickness, bottle body size consistency, and historical black spot detection data statistical information; Based on the collected production line status parameters and product characteristic parameters, a machine learning algorithm or a statistical regression method is used to train or update a parameter adjustment model, where the parameter adjustment model is a mapping relationship model between the parameters in the dynamic distance threshold calculation formula and the production line status parameters and product characteristic parameters; Verify the performance of the parameter adjustment model, monitor the prediction accuracy of the parameters in the dynamic distance threshold calculation formula, and trigger the retraining or update of the parameter adjustment model when the prediction accuracy is lower than the preset accuracy threshold.

[0037] Among them, for the determination of the parameter adjustment model, the specific implementation method can be: First, various sensors deployed on the glass bottle production line are used to collect status parameters such as furnace temperature, cooling rate, and production line speed in real time. Product characteristic parameters such as the wall thickness of the glass bottle and the consistency of the bottle body size can be obtained by on-line detection equipment or sampling inspection. The statistical information of historical black spot detection data can be extracted from the quality management system.

[0038] Furthermore, the collected data is used to train the parameter adjustment model. As a preferred implementation method, machine learning algorithms such as linear regression, support vector machine, or neural network can be adopted. The input of the model can be the production line status parameters and product characteristic parameters, and the output is the parameters in the dynamic distance threshold calculation formula. The training process aims to learn the mapping relationship between the input parameters and the output parameters.

[0039] Thus, a model performance verification link is introduced. The prediction accuracy can be evaluated by calculating the error between the model prediction value and the actual value. The preset accuracy threshold can be set according to actual production requirements and quality standards. When the prediction accuracy is lower than the preset accuracy threshold, the model retraining or update process is triggered to ensure the effectiveness and adaptability of the model.

[0040] Specifically, by collecting the status parameters during the operation of the production line and the product's own characteristic parameters in real time, the real-time status of the production line and the quality characteristics of the product are comprehensively reflected. Using these data, the parameter adjustment model trained by machine learning or statistical regression methods can establish a quantitative relationship between the production line status, product characteristics, and the parameters in the dynamic distance threshold calculation formula. This method makes the calculation of the dynamic distance threshold no longer rely on fixed empirical values, but can be adaptively adjusted according to the actual situation of the production line. Through the verification and iterative update mechanism of the model performance, the accuracy of the model prediction and the reliability of the dynamic distance threshold calculation are ensured, and finally the effectiveness of black spot detection and subsequent process parameter optimization is improved.

[0041] In some specific embodiments, for example, on a glass bottle production line, a furnace temperature sensor, a speedometer, and a wall thickness measuring instrument are installed to collect real-time data on furnace temperature, production line speed, and glass bottle wall thickness. Historical black spot detection data is extracted from a database. The collected data is input into a pre-set linear regression model for training, and the model learns the linear relationship between furnace temperature, production line speed, wall thickness, historical black spot data, and the parameters in the dynamic distance threshold calculation formula. After the model training is completed, the prediction accuracy is evaluated by calculating the root mean square error. A preset accuracy threshold is set to 0.05. When the root mean square error exceeds 0.05, the model is retrained using newly collected data. Thus, the parameters in the dynamic distance threshold calculation formula can be dynamically adjusted according to the actual operating state of the production line, ensuring the accuracy of the dynamic distance threshold calculation.

[0042] In some preferred embodiments, based on the collected production line state parameters and product characteristic parameters, a machine learning algorithm or a statistical regression method is used to train or update a parameter adjustment model. The parameter adjustment model is a mapping relationship model between the parameters in the dynamic distance threshold calculation formula and the production line state parameters and product characteristic parameters, and specifically includes: Analyze the correlation between the collected production line state parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula; Based on the parameter correlation analysis results, select the production line state parameters and product characteristic parameters with a correlation higher than the preset threshold to form a subset of characteristic parameters; Use a machine learning algorithm or a statistical regression method to train or update the parameter adjustment model based on the subset of characteristic parameters.

[0043] Among them, the parameter correlation analysis step is performed. Specifically, statistical methods such as Pearson correlation coefficient, Spearman rank correlation coefficient, or mutual information can be used to quantify the degree of association between the production line state parameters, product characteristic parameters, and the parameters in the dynamic distance threshold calculation formula. For example, production line state parameters such as furnace temperature and cooling rate, and product characteristic parameters such as glass bottle wall thickness and bottle body size consistency may have different degrees of correlation with the parameters in the dynamic distance threshold calculation formula. By calculating these correlation coefficients, the influence of each production line state parameter and product characteristic parameter on the parameters in the dynamic distance threshold calculation formula can be evaluated. The parameter correlation analysis results are used to guide the construction of the subset of characteristic parameters.

[0044] Specifically, the preset threshold can be set according to actual needs, for example, it is 0.5. If the correlation coefficient between the production line status parameters and the product characteristic parameters is higher than this threshold, it is considered a high-correlation parameter and is selected to enter the characteristic parameter subset. Thus, the characteristic parameter subset is formed, which only contains the parameters that have a significant impact on the parameters in the dynamic distance threshold calculation formula, while excluding the low-correlation or irrelevant parameters. The model training or update process is executed.

[0045] As a preferred implementation manner, machine learning algorithms such as linear regression, support vector machine, neural network, etc., or statistical regression methods such as multiple regression analysis can be adopted. These methods are used to establish the mapping relationship between the characteristic parameter subset and the parameters in the dynamic distance threshold calculation formula. The input of the model training is the characteristic parameter subset, and the output is the parameters in the dynamic distance threshold calculation formula. By minimizing the error between the predicted value and the actual value, the model parameters are optimized, so as to obtain the parameter adjustment model.

[0046] For example, on a glass bottle production line, the furnace temperature, cooling rate, production line speed, glass bottle wall thickness, bottle body size consistency, and historical black spot detection data statistics information are collected in real time. Parameter correlation analysis is executed, for example, calculating the Pearson correlation coefficient between each parameter and the parameters in the dynamic distance threshold calculation formula. Suppose the analysis results show that the correlation coefficients between the furnace temperature and the glass bottle wall thickness and the parameters in the dynamic distance threshold calculation formula are 0.7 and 0.6 respectively, which are higher than the preset threshold of 0.5, while the correlation coefficients of the cooling rate and the production line speed are lower than 0.5. Thus, the furnace temperature and the glass bottle wall thickness are selected into the characteristic parameter subset. Then, the linear regression algorithm is used to train the parameter adjustment model based on these two characteristic parameters of the furnace temperature and the glass bottle wall thickness. After the training is completed, the parameter adjustment model can predict the parameters in the dynamic distance threshold calculation formula according to the real-time monitored furnace temperature and glass bottle wall thickness, and further realize the adaptive adjustment of the dynamic distance threshold.

[0047] In some specific embodiments, the parameter correlation analysis can be performed periodically. For example, it can be carried out once every hour (referred to as the analysis period). Alternatively, when the fluctuation range of the production line status parameters or product characteristic parameters exceeds the preset range, such as the furnace temperature fluctuation exceeding 5 degrees Celsius, the parameter correlation analysis is immediately triggered. The parameter adjustment model can be updated in an incremental learning manner. When new production data is collected, the model can be fine-tuned on the original basis instead of being completely retrained, thereby improving the model update efficiency. The subset of characteristic parameters is not fixed but can be dynamically adjusted according to the latest parameter correlation analysis results. If the correlation of a certain parameter drops below the threshold, it can be removed from the subset of characteristic parameters; if the correlation of a parameter with originally low correlation significantly increases, it can be added to the subset of characteristic parameters. Through this dynamic adjustment mechanism, the subset of characteristic parameters can always maintain effectiveness and representativeness, thus ensuring the accuracy and adaptability of the parameter adjustment model.

[0048] In some preferred embodiments, the step of analyzing the correlation between the collected production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula includes: Real-time monitoring of the execution period of the parameter correlation analysis; If the current moment meets the preset periodic analysis trigger condition (that is, the time interval between the current moment and the moment when the parameter correlation analysis was last executed reaches the analysis period), then execute the parameter correlation analysis step; Alternatively, real-time monitoring of the fluctuation ranges of the production line status parameters and product characteristic parameters; If it is monitored that the fluctuation range of at least one production line status parameter or product characteristic parameter exceeds the corresponding preset fluctuation threshold (each parameter is set with a corresponding preset fluctuation threshold), then execute the parameter correlation analysis step.

[0049] Correspondingly, the step of selecting the production line status parameters and product characteristic parameters with a correlation higher than the preset threshold based on the parameter correlation analysis results to form a subset of characteristic parameters includes: After each execution of the parameter correlation analysis, based on the latest parameter correlation analysis results, select the production line status parameters and product characteristic parameters with a correlation higher than the preset threshold to form or update the subset of characteristic parameters.

[0050] Among them, the periodic analysis trigger condition is preset to define the regular execution timing of the parameter correlation analysis step. As a preferred embodiment, the periodic analysis trigger condition can be set as a fixed time interval (i.e., the analysis period), such as performing the parameter correlation analysis once every hour, once every day, or once every week. Thus, the system can regularly evaluate the parameter correlation, track the slow changes in the production line status, and ensure the timeliness of the subset of characteristic parameters.

[0051] Furthermore, a preset fluctuation threshold is pre-set to define significant fluctuations in production line status parameters or product characteristic parameters. For example, the preset fluctuation threshold can be set to a furnace temperature change exceeding a preset Celsius value, or a glass bottle wall thickness change exceeding a preset millimeter value. When the fluctuation amplitude of at least one production line status parameter or product characteristic parameter is monitored to exceed the corresponding preset fluctuation threshold, the parameter correlation analysis step will be immediately initiated. As a result, the system can quickly respond to sudden changes in production line status, update the characteristic parameter subset in a timely manner, and avoid the failure of the characteristic parameter subset due to parameter mutations.

[0052] Furthermore, updating the feature parameter subset refers to adjusting the original feature parameter subset according to the latest parameter correlation analysis results. Specifically, if the latest parameter correlation analysis results show that the correlation between some parameters originally selected into the feature parameter subset and the parameters in the dynamic distance threshold calculation formula has decreased to below the preset threshold, then these parameters will be removed from the feature parameter subset. Conversely, if the analysis results show that the correlation between some parameters that were not originally selected into the feature parameter subset and the parameters in the dynamic distance threshold calculation formula has increased to above the preset threshold, then these parameters will be added to the feature parameter subset. As a result, the feature parameter subset can dynamically adapt to the actual operating status of the production line and always include production line status parameters and product characteristic parameters that have a high correlation with the parameters in the dynamic distance threshold calculation formula.

[0053] In some specific embodiments, the execution cycle of the parameter correlation analysis is set to be performed every 8 hours. The preset periodic analysis trigger condition is set to "more than 8 hours from the last analysis time". The fluctuation amplitude threshold is set to 3 degrees Celsius for the furnace temperature, 0.3 mm for the glass bottle wall thickness, and 0.1 m / s for the production line speed. The initial feature parameter subset is set to the furnace temperature and the cooling rate. After the system starts running, every 8 hours, regardless of whether the production line state parameters and product characteristic parameters fluctuate, the parameter correlation analysis will be automatically performed, and the feature parameter subset will be updated according to the analysis results. At the same time, the system monitors the fluctuation amplitude of the furnace temperature, the glass bottle wall thickness and the production line speed in real time. If at a certain moment, the furnace temperature fluctuates by more than 3 degrees Celsius in a short period of time, or the glass bottle wall thickness fluctuates by more than 0.3 mm, or the production line speed fluctuates by more than 0.1 m / s, the system will immediately interrupt the current periodic analysis wait, give priority to performing a parameter correlation analysis, and immediately update the feature parameter subset based on the analysis results. As a result, the system can take into account both periodic tracking and sudden response, and ensure the effectiveness of the feature parameter subset in a more comprehensive and timely manner, thereby improving the accuracy and robustness of the optimization of glass bottle production process parameters based on black spot detection.

[0054] Further, in step A4, the step of determining corresponding black dot pairs in the first black dot detection data and the second black dot detection data based on a dynamic distance threshold includes: A401. Perform pose alignment processing on the second black dot detection data based on the glass bottle pose features extracted from the first black dot detection data and the second black dot detection data; A402. After the pose alignment processing, determine the corresponding black dot pairs in the first black dot detection data and the second black dot detection data after the pose alignment processing based on the dynamic distance threshold.

[0055] Among them, the extraction of glass bottle pose features can be achieved by using a variety of methods. For example, the bottle body contour can be extracted as a pose feature. Edge detection algorithms such as Canny or Sobel can be used to obtain the bottle body contour from the detection data image. Shape matching techniques can be used to compare the changes in the bottle body contour before and after the cleaning process. The position distribution of specific black dots can also be used as a pose feature. If there are easily recognizable black dots on the glass bottle, the relative positions of these black dots can be extracted and used for pose alignment. The implementation of pose alignment processing can calculate pose transformation parameters based on the extracted pose features. For example, if the bottle body contour is used, the iterative closest point algorithm can be used to calculate the rotation angle and translation vector between the contours. If the positions of specific black dots are used, Procrustes analysis can be used to determine the optimal pose transformation. The calculated pose transformation parameters are then used to perform a spatial transformation on the second black dot detection data to achieve pose alignment with the first black dot detection data.

[0056] The dynamic distance threshold is applied after the pose alignment processing to determine the corresponding black dot pairs in the first black dot detection data and the second black dot detection data after the pose alignment processing. For each black dot in the first black dot detection data, search for black dots in the second black dot detection data after the pose alignment processing whose distance is less than the dynamic distance threshold. These black dots are considered to be corresponding black dots.

[0057] For example, on a glass bottle production line, the glass bottles pass through a first detection module and a second detection module before and after the cleaning process respectively, to obtain the first black dot detection data and the second black dot detection data. When it is necessary to determine the correspondence of black dots before and after cleaning, first, the pose features are extracted from the first black dot detection data and the second black dot detection data. As an implementation, the bottle body contour is extracted from the detection image through an edge detection algorithm. Then, the pose alignment process is performed based on the extracted bottle body contour. The iterative closest point algorithm is used to calculate the rotation and translation parameters of the second detection data relative to the first detection data. The second black dot detection data is subjected to rotation and translation transformation based on the calculated parameters to achieve pose alignment. After the pose alignment process is completed, the corresponding black dot pairs are determined using a dynamic distance threshold. For each black dot in the first black dot detection data, in the second black dot detection data after the pose alignment process, search for black dots with an Euclidean distance less than the dynamic distance threshold, and determine the searched black dots as the corresponding black dots. Through the pose alignment process, the pose difference of the glass bottle between the two detections is effectively eliminated. Thus, the accuracy of black dot matching based on the dynamic distance threshold is improved.

[0058] Further, after the pose alignment process, the steps of determining the corresponding black dot pairs in the first black dot detection data and the second black dot detection data after the pose alignment process based on the dynamic distance threshold include: For each black dot in the first black dot detection data, in the second black dot detection data after the pose alignment process, screen out a set of candidate corresponding black dots with a distance less than the dynamic distance threshold; If there is only one candidate corresponding black dot in the set of candidate corresponding black dots, then confirm the candidate corresponding black dot as the corresponding black dot; If there are multiple candidate corresponding black dots in the set of candidate corresponding black dots, then calculate the relative position consistency parameters of the neighborhood black dots between the black dots in the first black dot detection data and each candidate corresponding black dot respectively; Based on the relative position consistency parameters of the neighborhood black dots, determine one corresponding black dot from the multiple candidate corresponding black dots; If the set of candidate corresponding black dots is an empty set, then it is determined that there is no black dot in the second black dot detection data after the pose alignment process corresponding to the current black dot in the first black dot detection data.

[0059] Among them, the purpose of performing the pose alignment process is to correct the possible pose changes of the glass bottle before and after cleaning, such as rotation or translation. The pose features may include geometric features such as the contour of the glass bottle, the position of the bottle mouth, and the shape of the bottle bottom. The pose alignment process can be implemented through an image registration algorithm. For example, the iterative closest point algorithm completes the pose adjustment of the second black dot detection data by minimizing the difference in pose features between the first black dot detection data and the second black dot detection data.

[0060] After attitude alignment processing, for each black dot in the first black dot detection data, search for black dots within a predetermined distance range in the second black dot detection data after attitude alignment processing. This predetermined distance range is determined by a dynamic distance threshold. The dynamic distance threshold is calculated based on the size of the black dots in the first black dot detection data. Larger-sized black dots usually correspond to larger dynamic distance thresholds.

[0061] When, in the second black dot detection data after attitude alignment processing, for a black dot in the first black dot detection data, only one black dot with a distance less than the dynamic distance threshold is found, this black dot is directly confirmed as the corresponding black dot.

[0062] When multiple candidate corresponding black dots with distances less than the dynamic distance threshold are found, it is necessary to further calculate the relative position consistency parameter of the neighboring black dots. Neighboring black dots refer to other black dots within a certain range from the central black dot. The relative position consistency parameter is used to evaluate whether the relative position relationship of the black dots within the neighborhood of the black dot in the first black dot detection data is consistent with the relative position relationship of the black dots within the neighborhood of the candidate corresponding black dot in the second black dot detection data. For example, the relative geometric relationships such as the distances and angles between the black dots within the neighborhood can be calculated and the consistency degree of these relative geometric relationships can be compared.

[0063] Based on the relative position consistency parameter of the neighboring black dots, select the candidate corresponding black dot with the highest degree of consistency as the final corresponding black dot. The evaluation of the degree of consistency can be achieved by setting a threshold or comparing the consistency parameter values of different candidate corresponding black dots.

[0064] If no candidate corresponding black dots with distances less than the dynamic distance threshold are found, it is determined that there is no black dot in the second black dot detection data corresponding to the current black dot in the first black dot detection data.

[0065] Further, if there are multiple candidate corresponding black dots in the candidate corresponding black dot set, the neighborhood black dot relative position consistency parameters between the black dots in the first black dot detection data and each candidate corresponding black dot are calculated respectively. The specific method is, for example: taking the black dot in the first black dot detection data as the center, a neighborhood range is delimited, such as a circular area with a radius of 50 pixels, and the position coordinates of other black dots in this neighborhood are counted. Similarly, for each candidate corresponding black dot in the candidate corresponding black dot set, a neighborhood range of the same size is also delimited, and the position coordinates of other black dots in the neighborhood are counted. Then, the relative position vectors of the black dots in the neighborhood of the black dot in the first black dot detection data (at this time, this black dot is the central black dot) relative to the central black dot are calculated, and the relative position vectors of the black dots in the neighborhood of each candidate corresponding black dot in the candidate corresponding black dot set (at this time, this candidate corresponding black dot is the central black dot) relative to the central black dot are calculated. The similarity of the two groups of relative position vectors is compared, for example, the cosine value of the vector angle or the Euclidean distance is used to measure the similarity. The candidate corresponding black dot with the highest average similarity is selected as the final corresponding black dot.

[0066] By introducing the neighborhood black dot relative position consistency parameter, the accuracy of determining the corresponding relationship of black dots in complex situations can be effectively improved, providing a guarantee for the improvement of the production quality of glass bottles.

[0067] Among them, in step A5, at least one black dot size threshold can be set. By comparing the size of the black dots in the black dot pair with the black dot size threshold, the black dots in the black dot pair are classified by type. For example, a first threshold is set. In the black dot pair, the black dots with a size greater than the first threshold are classified as the first type of black dots, and the black dots with a size less than or equal to the first threshold are classified as the second type of black dots.

[0068] Further, in step A6, the steps of performing weighted statistical difference analysis based on the black dot type classification result to obtain the difference analysis result include: A601. For each type of black dot, count the number, position coordinates, and size of the black dots before and after cleaning respectively; A602. For each type of black dot, calculate at least one difference index; the difference index includes the change rate of the number of black dots before and after cleaning, the average change amount of the size, and the degree of position offset (that is, calculate at least one of these three difference indexes); A603. Assign weight coefficients to the difference indexes of each type of black dot; A604. Based on the weight coefficients, synthesize the difference indexes of different types of black dots to generate a comprehensive difference analysis result.

[0069] Among them, in step A601, for each type of black dot, the number, position coordinates, and size of the black dots before and after cleaning are respectively counted. The purpose is to quantify the impact of the cleaning process on different types of black dots. In specific implementation, for each type of black dot, the number before and after cleaning is obtained by counting. The position coordinates can use image processing technology to determine the pixel coordinates or physical coordinates of the black dots on the surface of the glass bottle. The size can measure the pixel area of the black dots in the image or convert it into a physical size according to the calibration information.

[0070] Among them, in step A602, for each type of black dot, the step of calculating at least one difference index is used to characterize the influence degree of the cleaning process on various black dots. The difference index can include the change rate of the number of black dots before and after cleaning, the average change amount of the size, the degree of position offset, etc. For example, the number change rate can be obtained by dividing the difference between the number of black dots after cleaning and the number of black dots before cleaning by the number of black dots before cleaning.

[0071] Among them, in step A603, a weight coefficient is assigned to the difference index of each type of black dot. The purpose is to distinguish the importance of different types of black dots in quality evaluation. The weight coefficient can be determined according to the influence degree of the black dot type on product quality. For example, for the first type of black dot with a larger size, since it has a greater impact on product quality, a higher weight coefficient can be assigned; for the second type of black dot with a smaller size, a lower weight coefficient can be assigned. The assignment of the weight coefficient can be set based on empirical knowledge, experimental data, or simulation results.

[0072] Among them, in step A604, based on the weight coefficient, the difference indexes of different black dot types are synthesized to generate a comprehensive difference analysis result. The purpose is to obtain an index that can comprehensively reflect the overall impact of the cleaning process on the black dot defects of the glass bottle. The comprehensive difference analysis result can be obtained by weighted summing the difference indexes of different types of black dots. For example, the comprehensive difference analysis result can be equal to the number change rate of the first type of black dot multiplied by its weight coefficient, plus the number change rate of the second type of black dot multiplied by its weight coefficient. Among them, for each difference index, the corresponding comprehensive difference analysis result is calculated.

[0073] This comprehensive difference score is used as the final difference analysis result to evaluate the effectiveness of the cleaning process. The higher the value of the difference analysis result, the more obvious the improvement effect of the cleaning process on the black dot defects.

[0074] For example, assume that the black dots are divided into three types: A, B, and C. Among them, type A black dots are those with a size greater than 0.5 mm, type B black dots are those with a size between 0.2 mm and 0.5 mm, and type C black dots are those with a size less than 0.2 mm. In statistical difference analysis, first, for the black dots of types A, B, and C, count the number of black dots before and after cleaning respectively. For example, before cleaning, there are 10 black dots of type A, 20 black dots of type B, and 50 black dots of type C; after cleaning, there are 5 black dots of type A, 15 black dots of type B, and 40 black dots of type C. Then, calculate the change rate of the number of black dots of each type. The change rate of the number of type A black dots is (5 - 10) / 10 = -50%, the change rate of the number of type B black dots is (15 - 20) / 20 = -25%, and the change rate of the number of type C black dots is (40 - 50) / 50 = -20%. After that, assign weight coefficients to the change rates of the number of black dots of types A, B, and C. For example, the weight coefficient of type A is 0.6, the weight coefficient of type B is 0.3, and the weight coefficient of type C is 0.1. Finally, based on the weight coefficients, synthesize the change rates of the number of black dots of the three types to obtain the comprehensive difference analysis result: (-50%)*0.6 + (-25%)*0.3 + (-20%)*0.1 = -39.5%. This value of -39.5% can be used as a quantitative index for evaluating the cleaning effect.

[0075] In some embodiments, in step A7, the step of optimizing the glass bottle production process parameters based on the difference analysis result includes: A701. Based on the difference analysis result, select a parameter adjustment strategy from a preset set of parameter adjustment strategies; A702. Execute the selected parameter adjustment strategy to adjust the glass bottle production process parameters.

[0076] Among them, the preset set of parameter adjustment strategies refers to a set established in advance that contains various process parameter adjustment schemes for different difference analysis results. This set can be constructed in the form of a rule base, decision tree, or lookup table, etc. Each strategy is associated with a specific range or type of difference analysis result, and clearly defines one or more production process parameters to be adjusted and their adjustment amplitudes and directions.

[0077] For example, when the difference analysis result shows that the number of black dots after cleaning has increased significantly, and mainly small-sized black dots, the set of parameter adjustment strategies may include a strategy to reduce the water pressure in the cleaning process; for another example, when the difference analysis result shows that the size of the black dots after cleaning has increased significantly, and mainly black dots at specific positions, the set of parameter adjustment strategies may include a strategy to adjust the local cooling rate of the glass bottle mold.

[0078] The selection of the parameter adjustment strategy is based on the results of the difference analysis. The results of the difference analysis can be quantified into specific indicator parameters, such as the change rate of the number of black dots before and after cleaning, the average increase in the size of black dots, the change in the proportion of specific types of black dots, etc. The system will compare these indicator parameters with the preset strategy selection thresholds or input them into a pre-trained strategy selection model, so as to automatically select the parameter adjustment strategy that best matches the current difference analysis results from the parameter adjustment strategy set.

[0079] After the parameter adjustment strategy is selected, it will be executed to adjust the glass bottle production process parameters. The execution process can be to directly call the preset parameter adjustment instructions, such as sending instructions to the production line control system to adjust process parameters such as furnace temperature, cooling rate, and production line speed. The parameter adjustment can be a single adjustment or multiple iterative adjustments. In the iterative adjustment method, after each parameter adjustment, the system will re-collect the black dot detection data, conduct a difference analysis, and select and execute the parameter adjustment strategy again according to the new difference analysis results until the difference analysis results meet the preset optimization goals or reach a stable state.

[0080] Specifically, for the optimization problem of glass bottle black dot defects, first, a parameter adjustment strategy set needs to be constructed. Each strategy in this set should clearly target one or more types of difference analysis results and specify a specific process parameter adjustment plan. For example, the strategy set can include the following strategies: when a large number of small-sized black dots are newly added after cleaning, reduce the cleaning water pressure; when the size of black dots generally increases after cleaning, reduce the furnace temperature; when the number of black dots in a specific area increases after cleaning, adjust the angle of the die cooling spray head. Then, during the actual production process, the difference analysis module outputs the difference analysis results. For example, the difference analysis results show that the number of small-sized black dots has increased by 20% after cleaning, and the number of large-sized black dots remains basically unchanged. After receiving this difference analysis result, the system will match it with the strategies in the parameter adjustment strategy set. According to the preset matching rules, for example, "the number of small-sized black dots increases by more than 15% after cleaning" can be used as a trigger condition to match the strategy of "reducing the cleaning water pressure". After the strategy matching is completed, the system will automatically execute the strategy of "reducing the cleaning water pressure". The execution process can be to send a control instruction to the cleaning equipment to reduce the water pressure to within the preset safe threshold range. After the parameter adjustment is executed, the production line continues to run, and the system will continuously monitor the black dot detection data and the difference analysis results, evaluate the parameter adjustment effect, and make further strategy selection and parameter adjustment if necessary, forming a continuous optimization closed-loop control process.

[0081] In some specific embodiments, the construction of the parameter adjustment strategy set can be based on expert experience, experimental data, or machine learning methods. For example, by analyzing historical production data and defect data, the influence law of different process parameter adjustments on black dot defects can be summarized to form parameter adjustment strategies. Another example is to establish a machine learning model, input the differential analysis results and the desired optimization goals, and output the optimal parameter adjustment strategies. As a preferred embodiment, the parameter adjustment strategy set can be designed to be extensible and configurable. Users can customize parameter adjustment strategies according to actual production situations and optimization requirements, or modify and optimize existing strategies to adapt to different production lines and product requirements.

[0082] Please refer to Figure 2 , Figure 2 FIG. [FIG. NUMBER] is a schematic structural diagram of an electronic device provided by an embodiment of the present application. The present application provides an electronic device, including: a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not shown). The memory 302 stores a computer program executable by the processor 301. When the electronic device runs, the processor 301 executes the computer program to perform the method for optimizing the production process parameters of glass bottles based on black dot detection in any optional implementation manner of the above embodiments, so as to implement the following functions: obtaining first black dot detection data of the glass bottle before the cleaning link, where the first black dot detection data includes the position and size of the black dots; obtaining second black dot detection data of the glass bottle after the cleaning link, where the second black dot detection data includes the position and size of the black dots; calculating a dynamic distance threshold according to the size of the black dots in the first black dot detection data for the same glass bottle; determining corresponding black dot pairs in the first black dot detection data and the second black dot detection data based on the dynamic distance threshold; classifying the black dots in the corresponding black dot pairs according to a black dot size threshold; performing weighted statistical differential analysis based on the black dot type classification result to obtain a differential analysis result; and optimizing the production process parameters of the glass bottle based on the differential analysis result.

[0083] Note: In the translation of , the placeholder Figure 2 should be replaced with the actual figure number in the original text. Since it's not provided here, it's left as Figure 2 . Also, the "FIG. [FIG. NUMBER]" in the translation is a placeholder for the actual figure reference in the original Chinese text.An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it executes the method for optimizing the production process parameters of glass bottles based on black dot detection in any optional implementation manner of the above embodiments to implement the following functions: obtaining first black dot detection data of the glass bottle before the cleaning link, where the first black dot detection data includes the position and size of the black dots; obtaining second black dot detection data of the glass bottle after the cleaning link, where the second black dot detection data includes the position and size of the black dots; for the same glass bottle, calculating a dynamic distance threshold according to the size of the black dots in the first black dot detection data; based on the dynamic distance threshold, determining corresponding black dot pairs in the first black dot detection data and the second black dot detection data; classifying the black dots in the corresponding black dot pairs according to a black dot size threshold; based on the black dot type classification result, performing weighted statistical difference analysis to obtain a difference analysis result; and based on the difference analysis result, optimizing the production process parameters of the glass bottles.

[0084] Among them, the computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0085] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some communication interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0086] In addition, the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0087] Furthermore, in each embodiment of this application, the various functional modules may be integrated together to form an independent part, or each module may exist alone, or two or more modules may be integrated to form an independent part.

[0088] In this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0089] The above description is only for the embodiments of this application and does not limit the protection scope of this application. For those skilled in the art, this application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

1. A method for optimizing process parameters of glass bottle production based on black spot detection, characterized in that: The method comprises the following steps: Acquire first black spot detection data of the glass bottle before the cleaning step, wherein the first black spot detection data includes the position and size of the black spot; Acquire second black spot detection data of the glass bottle after the cleaning step, wherein the second black spot detection data includes the position and size of the black spot; For the same glass bottle, a dynamic distance threshold is calculated according to the size of the black spot in the first black spot detection data; Based on the dynamic distance threshold, determining corresponding black dot pairs in the first black dot detection data and the second black dot detection data; Classifying the black dots in the corresponding black dot pairs according to the black dot size threshold; Based on the black spot type classification results, weighted statistical difference analysis is performed to obtain the difference analysis results; Based on the difference analysis results, the glass bottle production process parameters are optimized.

2. The method for optimizing process parameters of glass bottle production based on black spot detection according to claim 1, characterized in that: The step of calculating the dynamic distance threshold according to the size of the black spot in the first black spot detection data for the same glass bottle comprises: Real-time monitoring of production line status parameters and product characteristic parameters; Determine a parameter adjustment model, wherein the parameter adjustment model is used to determine a mapping relationship between production line status parameters and product characteristic parameters and parameters in a dynamic distance threshold calculation formula; Based on the parameter adjustment model and the real-time parameters, adjusting the parameters in the dynamic distance threshold calculation formula; Calculate the dynamic distance threshold using the adjusted parameters.

3. The method for optimizing process parameters of glass bottle production based on black spot detection according to claim 2, characterized in that: The step of determining a parameter adjustment model, wherein the parameter adjustment model is used to determine a mapping relationship between production line status parameters and product characteristic parameters and parameters in a dynamic distance threshold calculation formula comprises: Real-time collection of state parameters of a glass bottle production line and characteristic parameters of glass bottle products, wherein the state parameters of the production line include at least one of furnace temperature, cooling rate, and production line speed, and the product characteristic parameters include at least one of glass bottle wall thickness, bottle body size consistency, and historical black spot detection data statistical information; Based on the collected production line status parameters and product characteristic parameters, a machine learning algorithm or a statistical regression method is used to train or update a parameter adjustment model, wherein the parameter adjustment model is a mapping relationship model between the parameters in the dynamic distance threshold calculation formula and the production line status parameters and the product characteristic parameters; Verify the performance of the parameter adjustment model, monitor the prediction accuracy of the parameters in the dynamic distance threshold calculation formula, and trigger retraining or updating of the parameter adjustment model when the prediction accuracy is lower than the preset accuracy threshold.

4. The method for optimizing process parameters of glass bottle production based on black spot detection according to claim 3, characterized in that: The steps of training or updating the parameter adjustment model based on the collected production line status parameters and product characteristic parameters by using a machine learning algorithm or a statistical regression method, wherein the parameter adjustment model is a mapping relationship model between the parameters in the dynamic distance threshold calculation formula and the production line status parameters and the product characteristic parameters include: Analyze the correlation between the collected production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula; Based on the parameter correlation analysis results, the production line status parameters and product characteristic parameters with correlations higher than a preset threshold are selected to form a characteristic parameter subset; Use machine learning algorithms or statistical regression methods to train or update parameter adjustment models based on a subset of feature parameters.

5. The method for optimizing process parameters of glass bottle production based on black spot detection according to claim 4, characterized in that: The step of analyzing the correlation between the collected production line status parameters and product characteristic parameters and the parameters in the dynamic distance threshold calculation formula includes: Real-time monitoring of the execution cycle of parameter correlation analysis; If the preset periodic analysis trigger condition is met at the current moment, the parameter correlation analysis step is executed; Alternatively, real-time monitoring of the fluctuation range of production line status parameters and product characteristic parameters; If the fluctuation range of at least one production line status parameter or product characteristic parameter is monitored to exceed the corresponding preset fluctuation threshold, the parameter correlation analysis step is performed.

6. The method for optimizing process parameters of glass bottle production based on black spot detection according to claim 1, characterized in that: The step of determining corresponding black dot pairs in the first black dot detection data and the second black dot detection data based on the dynamic distance threshold comprises: Based on the posture features of the glass bottle extracted from the first black spot detection data and the second black spot detection data, performing posture alignment processing on the second black spot detection data; After the posture alignment process, corresponding black dot pairs in the first black dot detection data and the second black dot detection data after the posture alignment process are determined based on the dynamic distance threshold.

7. The method for optimizing process parameters of glass bottle production based on black spot detection according to claim 1, characterized in that: The step of performing weighted statistical difference analysis based on the black spot type classification result to obtain the difference analysis result comprises: For each type of black spot, the number, location coordinates and size of the black spots before and after cleaning are counted; For each black spot type, at least one difference index is calculated; Assign weight coefficients to the difference index of each black spot type; Based on the weight coefficient, the difference indicators of different black spot types are integrated to generate comprehensive difference analysis results.

8. The method for optimizing process parameters of glass bottle production based on black spot detection according to claim 1, characterized in that: The step of optimizing the glass bottle production process parameters based on the difference analysis results comprises: Based on the difference analysis result, selecting a parameter adjustment strategy from a preset parameter adjustment strategy set; Execute the selected parameter adjustment strategy to adjust the glass bottle production process parameters.

9. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it runs the steps in the method for optimizing process parameters of glass bottle production based on black spot detection as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the method for optimizing process parameters for glass bottle production based on black spot detection as described in any one of claims 1 to 8 are executed.

Images