One-key production change control system and method for cosmetic glass bottle production line

Abstract

The present application relates to the technical field of production change control, and specifically discloses a one-key production change control system and method for a cosmetic glass bottle production line, which comprises a parameter acquisition module for recording production process parameter combinations and production quality data; a production quality quantification module for calculating production quality influence coefficients based on the production quality data; a quality analysis module for determining whether the production process parameter combinations are qualified based on the production quality influence coefficients; a production change mechanism construction module for establishing a production change mechanism in combination with the qualified determination results, a Gaussian process regression model and SHAP value analysis; and a production change control module for executing the production change mechanism. The present application selects / optimizes the production process parameter combinations in combination with the qualified determination, the Gaussian process regression model and the SHAP value analysis, so that the established production change mechanism is directly applicable to the production and manufacturing of target types of cosmetic glass bottles without frequent adjustments.

Description

Technical Field

[0001] This invention relates to the field of production changeover control technology, specifically to a one-click production changeover control system and method for cosmetic glass bottle production lines. Background Technology

[0002] The one-click production changeover control system for cosmetic glass bottle production lines aims to achieve rapid switching between different specifications and models of glass bottles through highly automated and intelligent technologies, in order to meet the market's demand for product diversification and customization.

[0003] In existing technologies, traditional one-click production changeover control systems for cosmetic glass bottle production lines typically establish a database containing various glass bottle production process parameters, including raw material ratio data and molding process data. When changing production, the corresponding parameter set is retrieved from the database according to the type of cosmetic glass bottle, thus enabling rapid production changeover.

[0004] Traditional cosmetic glass bottle production lines typically use a one-click changeover control system to change products according to predetermined production parameters. This requires frequent adjustments to molding process parameters to gradually reach the optimal production state, which is often lengthy and affects the pass rate and output targets. Furthermore, the high number of changeovers per day significantly impacts production efficiency. Summary of the Invention

[0005] The purpose of this invention is to provide a one-click production changeover control system and method for cosmetic glass bottle production lines, solving the following technical problems:

[0006] How to optimize the one-click production changeover of a glass bottle production line.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] The one-click production changeover control system for cosmetic glass bottle production lines includes:

[0009] The parameter acquisition module is used to determine the type of cosmetic glass bottle to be replaced as the target type, and to use the historical production rounds used to produce the target type of cosmetic glass bottle as the target production rounds. The module collects and records the combination of production process parameters and production quality data under each target production round.

[0010] The production quality quantification module is used to analyze production quality data to calculate the production quality impact coefficient under various combinations of production process parameters. It is used to reflect the quality of the target type of cosmetic glass bottles produced under the corresponding combination of production process parameters.

[0011] The quality analysis module is used to compare each production quality influence coefficient with the reference value of the quality influence coefficient one by one in order to determine whether each combination of production process parameters is qualified.

[0012] The production switch mechanism construction module establishes a production switch mechanism for the target type of cosmetic glass bottles based on the results of the qualification judgment and in combination with the combination of production process parameters, Gaussian process regression model and SHAP value analysis.

[0013] The production changeover control module is used to execute the production changeover mechanism to perform one-click production changeover control for target type cosmetic glass bottles.

[0014] Furthermore, the production quality data includes the finish, transparency, and wall thickness of the finished cosmetic glass bottles of the target type produced in the target production round;

[0015] The combination of production process parameters consists of various production process parameters executed in the target production cycle, including molding temperature, molding pressure, droplet weight, molding speed, annealing temperature, and annealing time.

[0016] Furthermore, the production quality influence coefficient is inversely proportional to the finished smoothness of the target type of cosmetic glass bottle, inversely proportional to the transparency of the target type of cosmetic glass bottle, and directly proportional to the wall thickness deviation of the target type of cosmetic glass bottle. For any target type of cosmetic glass bottle produced in this target production cycle, its wall thickness deviation is the absolute difference between its wall thickness and the design value of the wall thickness of the target type of cosmetic glass bottle.

[0017] Furthermore, for any combination of production process parameters, if the corresponding production quality influence coefficient is not greater than the reference value of the quality influence coefficient, then the combination of production process parameters is considered qualified; otherwise, the combination of production process parameters is considered unqualified.

[0018] Furthermore, the production switching mechanism includes: if at least one combination of production process parameters is deemed qualified, then from all qualified combinations of production process parameters, the combination of production process parameters with the smallest production quality impact coefficient is selected and output as the production model.

[0019] Furthermore, the production switching mechanism also includes: if all production process parameter combinations are deemed unqualified, the production process parameter combination with the smallest production quality impact coefficient is selected as the candidate production process parameter combination. By combining SHAP value calculation and Gaussian process regression model, the candidate production process parameter combination is fine-tuned and compared, and the optimal production process parameter combination is output as the production model.

[0020] Furthermore, the Gaussian process regression model is trained using the combination of production process parameters under each target production round as input and the corresponding production quality impact coefficient as output, so that when the Gaussian process regression model receives the combination of production process parameters, it outputs the predicted value and standard deviation of the corresponding production quality impact coefficient.

[0021] Furthermore, the logic for outputting the optimal combination of production process parameters is as follows:

[0022] 1) In the Gaussian process regression model, calculate the SHAP value of each production process parameter in the candidate production process parameter combination. Based on the SHAP value, determine whether to fine-tune the production process parameters to generate a replacement scheme. If no replacement scheme is generated, the candidate production process parameter combination is taken as the optimal production process parameter combination. Otherwise, based on the replacement scheme, replace the production process parameters in the candidate production process parameter combination to generate the preferred production process parameter combination.

[0023] 2) Analyze each optimal combination of production process parameters one by one based on the Gaussian process regression model to output the predicted value and standard deviation of the corresponding production quality influence coefficient. For any optimal combination of production process parameters, add the predicted value and standard deviation of the corresponding production quality influence coefficient to obtain the most unfavorable value of the production quality influence coefficient corresponding to the optimal combination of production process parameters. Divide the predicted value and standard deviation of the corresponding production quality influence coefficient to obtain the confidence value of the production quality influence coefficient corresponding to the optimal combination of production process parameters.

[0024] 3) Determine if there is an optimal combination of production process parameters with a confidence value lower than the confidence threshold. If not, take the candidate combination of production process parameters as the optimal combination of production process parameters; otherwise, proceed to step 4).

[0025] 4) From all the preferred production process parameter combinations with a confidence value lower than the confidence threshold, select the preferred production process parameter combination with the smallest worst value, and determine whether its worst value is less than the production quality influence coefficient under the candidate production process parameter combination. If it is, update it as a new candidate production process parameter combination and return to step 1), until no new candidate production process parameter combinations are generated. The last generated candidate production process parameter combination is taken as the optimal production process parameter combination. Otherwise, the candidate production process parameter combination is taken as the optimal production process parameter combination.

[0026] Furthermore, the logic for generating the optimal combination of production process parameters is as follows:

[0027] 1.1) For any production process parameter in the combination of candidate production process parameters, determine whether its SHAP value is not greater than 0. If it is, then it is not considered as a production process parameter to be fine-tuned and is not fine-tuned. Otherwise, it is considered as a production process parameter to be fine-tuned, and based on the first derivative of its SHAP value, it is determined whether to fine-tune it to generate a replacement value for the production process parameter.

[0028] Specifically, for any production process parameter to be fine-tuned, if its first derivative of SHAP value is greater than 0, it is reduced downward by a preset number of percentage points to generate a corresponding production process parameter replacement value; if its first derivative of SHAP value is less than 0, it is increased upward by a preset number of percentage points to generate a corresponding production process parameter replacement value; if its first derivative of SHAP value is equal to 0, it is determined whether its SHAP value is a maximum or a minimum value in the SHAP dependency graph. If it is a minimum value, it is not adjusted; if it is a maximum value, it is reduced downward by a preset number of percentage points to generate a corresponding production process parameter replacement value.

[0029] 1.2) Determine if the total number of replacement values ​​for production process parameters is 0. If it is 0, no replacement scheme is generated, and the candidate production process parameter combination is taken as the optimal production process parameter combination; otherwise, proceed to step 1.3).

[0030] 1.3) First, with the constraint that the replacement scheme includes one production process parameter replacement value, arrange and combine the replacement values ​​of each production process parameter to form several replacement schemes. Then, with the constraint that the replacement scheme includes two production process parameter replacement values, arrange and combine the replacement values ​​of each production process parameter to form several replacement schemes. Continue in this manner until the number of production process parameter replacement values ​​included in the replacement scheme is consistent with the total number of production process parameter replacement values.

[0031] 1.4) For each replacement scheme, replace the same type of production process parameters in the candidate production process parameter combination with the production process parameter replacement value within it, so as to generate the preferred production process parameter combination after the replacement scheme has been processed.

[0032] A one-click changeover control method for a cosmetic glass bottle production line, used to execute the aforementioned one-click changeover control system for the cosmetic glass bottle production line, includes the following steps:

[0033] S1: Determine the type of cosmetic glass bottle to be replaced as the target type, and use the historical production rounds used to produce the target type of cosmetic glass bottle as the target production rounds. Collect and record the combination of production process parameters and production quality data under each target production round.

[0034] S2: Analyze production quality data to calculate the production quality impact coefficient under each combination of production process parameters. This coefficient is used to reflect the quality of the target type of cosmetic glass bottles produced under the corresponding combination of production process parameters.

[0035] S3: Compare each production quality influence coefficient with the reference value of the quality influence coefficient one by one to determine whether each combination of production process parameters is qualified;

[0036] S4: Based on the results of the qualification judgment, and in combination with the combination of production process parameters, establish a production change mechanism for the target type of cosmetic glass bottles;

[0037] S5: Implement the production changeover mechanism to control the production changeover of target type cosmetic glass bottles with one click.

[0038] The beneficial effects of this invention are as follows:

[0039] This invention first uses a production quality influence coefficient, which reflects the quality of target type cosmetic glass bottles, to determine the passability of production process parameter combinations for each target production round. Then, it introduces a Gaussian process regression model and SHAP value analysis to select / optimize production process parameter combinations. This allows the established production changeover mechanism to be directly applicable to the production of target type cosmetic glass bottles, without the need for frequent adjustments to molding process parameters during production to gradually reach the optimal production state. This greatly improves product pass rate and production efficiency, ensuring the smooth completion of output targets. Attached Figure Description

[0040] The invention will now be further described with reference to the accompanying drawings.

[0041] Figure 1 This is a schematic block diagram of the one-click production change control system for the cosmetic glass bottle production line in this invention;

[0042] Figure 2 This is a flowchart of the one-click production change control method for cosmetic glass bottle production lines in this invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Please see Figure 1 As shown, in one embodiment, this application provides a one-click production changeover control system for a cosmetic glass bottle production line, including:

[0045] The parameter acquisition module is used to determine the type of cosmetic glass bottle to be replaced as the target type, and to use the historical production rounds used to produce the target type of cosmetic glass bottle as the target production rounds. The module collects and records the combination of production process parameters and production quality data under each target production round.

[0046] As one implementation method, production quality data includes the finish, transparency, and wall thickness of the finished cosmetic glass bottles of the target type produced in the target production round;

[0047] Specifically, the finish, transparency, and wall thickness of cosmetic glass bottles are important indicators for measuring their quality. Since the appearance of cosmetic glass bottles directly affects brand value, their finish requires higher standards. Furthermore, cosmetics need to clearly display the texture of the product inside, so higher transparency is also required. Finally, because cosmetics need to maintain a balance between airtightness and lightweight design, their wall thickness also has certain requirements. Based on this, the finish, transparency, and wall thickness data are the core data for judging the quality of cosmetic glass bottles. Based on this data, the quality of cosmetics can be analyzed, thus providing reliable data support for establishing a replacement production mechanism.

[0048] It should be noted that the finished gloss of cosmetic glass bottles is obtained based on a gloss meter. Specifically, light is emitted onto the surface of the target type of cosmetic glass bottle at a specific angle (e.g., 60 degrees) using a gloss meter, and the intensity of the reflected light is measured. Under the same test conditions, light is emitted onto the surface of a target type of cosmetic glass bottle with a known acceptable finished gloss, and the intensity of the reflected light is measured as a baseline value. The ratio of the reflected light intensity at the target type of cosmetic glass bottle to the baseline value is calculated as the finished gloss of the target type of cosmetic glass bottle. The higher the value, the better the finished gloss index. Transparency can be characterized based on the transmittance measured by a spectrophotometer. The higher the value, the higher the transparency. Wall thickness can be obtained using an ultrasonic thickness gauge. The above data acquisition methods are all existing technologies and will not be elaborated on here.

[0049] As one implementation method, the combination of production process parameters consists of various production process parameters executed in the target production cycle, including molding temperature, molding pressure, droplet weight, molding speed, annealing temperature, and annealing time.

[0050] Specifically, molding temperature affects the fluidity of glass and molding quality; excessively high or low temperatures can lead to glass bottle deformation or uneven wall thickness. Molding pressure determines the shape and dimensional accuracy of the glass bottle; excessive or insufficient pressure will adversely affect the product dimensions. Droplet weight affects the weight and wall thickness of the glass bottle; excessively high or low droplet weight will adversely affect the product weight and wall thickness. Molding speed affects production efficiency and product quality; excessively high speed may result in incomplete glass bottle molding. Annealing temperature eliminates internal stress in the glass bottle; improper temperature control can lead to glass bottle breakage or deformation. Annealing time affects the stress elimination effect; insufficient or excessive time will affect product quality. Therefore, finding the optimal combination of production process parameters is particularly important for the production quality of cosmetic glass bottles.

[0051] The production quality quantification module is used to analyze production quality data to calculate the production quality impact coefficient under various combinations of production process parameters. It is used to reflect the quality of the target type of cosmetic glass bottles produced under the corresponding combination of production process parameters.

[0052] Among them, the production quality influence coefficient is inversely proportional to the finish smoothness of the target type of cosmetic glass bottle and inversely proportional to the transparency of the target type of cosmetic glass bottle, and directly proportional to the wall thickness deviation of the target type of cosmetic glass bottle. For any target type of cosmetic glass bottle produced in the target production cycle, its wall thickness deviation is the absolute difference between its wall thickness and the design value of the wall thickness of the target type of cosmetic glass bottle. This ensures that the better the quality of the target type of cosmetic glass bottle, the smaller the corresponding production quality influence coefficient.

[0053] As one implementation method, the mathematical expression for the production quality impact coefficient is as follows:

[0054]

[0055] In the formula, For the first The production quality impact coefficient corresponding to the combination of production process parameters under each target production round; the larger the value, the better the impact of the production quality. The better the combination of production process parameters for each target production round, the better. Index the target production round. For the first Under a target production cycle, the finished surface finish of the s-th target type cosmetic glass bottle sample, where s is the index of the target type cosmetic glass bottle sample. Specifically, from the target type cosmetic glass bottles produced under the target production cycle, a number of target type cosmetic glass bottles are randomly selected as target type cosmetic glass bottle samples. The total number of cosmetic glass bottle samples of the target type selected from the same target production round can generally be taken as 5% of the total number of cosmetic glass bottle samples of the target type produced in the target production round. This ensures representativeness of the sample selection while minimizing the number of samples to reduce workload. The standard value for the finished surface finish of the target type of cosmetic glass bottles can be determined by first selecting multiple target type cosmetic glass bottles that are rated as superior through expert evaluation, and then taking the average of their finished surface finish values ​​as the standard value for the finished surface finish of the target type of cosmetic glass bottles. For the first Under a target production round, the transparency of the s-th target type cosmetic glass bottle sample... To determine the standard transparency value for the target type of cosmetic glass bottles, an expert evaluation process can be used to select several bottles that are rated as superior. The average transparency value of these selected bottles can then be used as the standard transparency value for the target type of cosmetic glass bottles. The wall thickness, This refers to the designed wall thickness value for the target type of cosmetic glass bottle, which should be set according to actual needs. , and The weighting coefficient is set based on the core needs of the product and the allowable error in the empirical data. Specifically, quality engineers, production managers, marketing personnel, etc. can be invited to score the importance of the finish, transparency and wall thickness of the finished product (e.g., 1-10 points). The weighted average is then normalized to obtain the result. The higher the score, the more important it is.

[0056] Through the above technical solution, this example provides a mathematical expression for the production quality impact coefficient. Clearly, the lower the gloss and transparency of the target type of cosmetic glass bottle, and the greater the difference between the wall thickness and the designed wall thickness value, the greater the corresponding production quality impact coefficient. Specifically, since the appearance of cosmetic glass bottles directly affects brand value, the finished product's gloss needs to meet higher standards. Therefore, the lower the gloss of the finished cosmetic glass bottle, the lower its quality. Furthermore, cosmetics need to clearly display the texture of the internal product, so transparency also requires higher standards. Based on this, when cosmetics... The lower the transparency of the glass bottle, the lower the quality of the cosmetic glass bottle. Finally, since cosmetics need to balance airtightness and lightweight, their wall thickness also has certain requirements. The greater the difference between the wall thickness of the cosmetic glass bottle and the preset wall thickness, the lower its quality will be. Based on this, since the finished product's smoothness, transparency, and wall thickness are the core data for judging the quality of cosmetic glass bottles, the production quality impact coefficient corresponding to each combination of production process parameters can be analyzed based on the above data. This provides reliable data support for the subsequent establishment of a production change mechanism, ensuring the rationality of the establishment of the production change mechanism.

[0057] It should be noted that, and Part of this can be understood as first quantifying the differences in the finish and transparency of each sample in the form of a difference measurement, that is, calculating the difference between the transparency of each sample and the standard transparency value, and the difference between the finish and the standard finish value. One part involves calculating the difference between the wall thickness of each sample and the designed wall thickness, which can reflect quality problems of cosmetic glass bottles. Based on this, by setting different weighting coefficients for each part, it is possible to reflect... , as well as The three values ​​respectively affect the production quality impact coefficient, thereby improving the rationality of the data calculation results.

[0058] The quality analysis module is used to compare each production quality influence coefficient with the reference value of the quality influence coefficient one by one in order to determine whether each combination of production process parameters is qualified.

[0059] For any combination of production process parameters, if the corresponding production quality influence coefficient is not greater than the reference value of the quality influence coefficient, it indicates that the target type of cosmetic glass bottles produced based on that combination of production process parameters are of superior quality, and the combination of production process parameters is deemed qualified. Conversely, if the coefficient is greater than the reference value, it indicates that the target type of cosmetic glass bottles produced based on that combination of production process parameters are not of superior quality, and the combination of production process parameters is deemed unqualified.

[0060] It should be noted that when the finished product's smoothness meets the standard value, its transparency meets the standard value, and its wall thickness matches the design value, the target type of cosmetic glass bottle is considered a superior product. Therefore, the quality influence coefficient reference value can be set to... Of course, the reference value for the quality influence coefficient can also be adjusted according to actual needs. It can be slightly adjusted upwards or downwards based on the existing basis, without any restrictions;

[0061] The production switch mechanism construction module establishes a production switch mechanism for the target type of cosmetic glass bottles based on the results of the qualification judgment and in combination with the combination of production process parameters, Gaussian process regression model and SHAP value analysis.

[0062] The production switching mechanism includes: if at least one combination of production process parameters is deemed qualified, then from all qualified combinations of production process parameters, the combination of production process parameters with the smallest production quality impact coefficient is selected and output as the production model. This ensures that qualified and the most excellent combination of production process parameters is used to produce the target type of cosmetic glass bottles, thereby providing reliable data support for the subsequent establishment of the production switching mechanism and ensuring the rationality of the establishment result of the production switching mechanism.

[0063] The production switch mechanism also includes: if all production process parameter combinations are deemed unqualified, it indicates that the quality of the target type of cosmetic glass bottles produced in the past was generally low. Based on this, it means that there was no qualified production process parameter combination in the historical production process. It is necessary to produce a suitable production model by adjusting the production process parameters in advance. Therefore, this comparison method can provide reliable data support for the subsequent establishment of the production switch mechanism to ensure the rationality of the establishment result of the production switch mechanism. The production process parameter combination with the smallest production quality impact coefficient is selected as the candidate production process parameter combination. That is, the most excellent production process parameter combination in the past production, although it failed to meet the requirements, is used as the optimization basis. Combined with SHAP value calculation and Gaussian process regression model, the candidate production process parameter combination is fine-tuned and compared, and the optimal production process parameter combination is output as the production model.

[0064] The training process of the Gaussian process regression model is as follows: The Gaussian process regression model is trained using combinations of production process parameters for each target production round as input and the corresponding production quality impact coefficient as output. This allows the Gaussian process regression model to output the predicted value and standard deviation of the corresponding production quality impact coefficient when receiving combinations of production process parameters. The specific implementation plan is as follows:

[0065] The production quality impact coefficient and each production process parameter in the combination of production process parameters under each target production round are preprocessed using Z-score standardization or maximum-minimum normalization to eliminate the adverse effects of different dimensions on the training process of the Gaussian process regression model. The robust Matern kernel is selected as the kernel function of the Gaussian process regression model, and the smoothness parameter of the kernel function is set to a conventional 1.5 or 2.5, depending on the actual needs. The preprocessed combination of production process parameters and production quality impact coefficient under each target production round are input into the Gaussian process regression model. The optimal hyperparameters (such as length scale, signal variance, and noise variance) of the kernel function are found based on the "maximum likelihood estimation" algorithm. The specific construction and training process of the Gaussian process regression model is common knowledge in the art and will not be elaborated here. After the training of the Gaussian process regression model is completed, each time a new combination of production process parameters is received, the predicted value and standard deviation of the corresponding production quality impact coefficient will be output. The larger the standard deviation, the lower the confidence of the output predicted value in the Gaussian process regression model.

[0066] The logic for outputting the optimal combination of production process parameters is as follows:

[0067] 1) In the Gaussian process regression model, the SHAP value of each production process parameter in the candidate production process parameter combination is calculated. Based on the SHAP value, it is determined whether to fine-tune the production process parameters to generate a replacement scheme. If no replacement scheme is generated, it means that there is currently no clear direction for optimizing the candidate production process parameter combination. In order to avoid blind optimization leading to low quality or even unqualified production of the target type of cosmetic glass bottles, the candidate production process parameter combination is taken as the optimal production process parameter combination. That is, the best production process parameter combination in the target production batch is selected as the optimal production process parameter combination. Although the target type of cosmetic glass bottles produced by it are not superior products, they are still relatively good qualified products, thus ensuring the normal operation of production. However, a large number of small-scale production tests are needed to find qualified production process parameter combinations or candidate production process parameter combinations with clear optimization directions, laying the foundation for the subsequent production of superior target type cosmetic glass bottles. Otherwise, based on the replacement scheme, the production process parameters in the candidate production process parameter combination are replaced to generate the optimal production process parameter combination.

[0068] The logic for generating the optimal combination of production process parameters is as follows:

[0069] 1.1) For any production process parameter in the combination of candidate production process parameters, determine whether its SHAP value is not greater than 0. If it is, then it is not considered as a production process parameter to be fine-tuned and is not fine-tuned. Otherwise, it is considered as a production process parameter to be fine-tuned, and based on the first derivative of its SHAP value, it is determined whether to fine-tune it to generate a replacement value for the production process parameter.

[0070] It should be noted that the SHAP value is specifically calculated using the KernelSHAP algorithm, which is existing technology and will not be elaborated here. For any production process parameter in the candidate production process parameter combination, if its SHAP value is positive, it means that the production process parameter has improved the production quality influence coefficient in the candidate production process parameter combination; otherwise, it means that the production process parameter has not improved the production quality influence coefficient in the candidate production process parameter combination. As can be seen from the above description, the lower the production quality influence coefficient, the better the quality of the target type of cosmetic glass bottle. Therefore, the production process parameters with SHAP values ​​greater than 0 in the candidate production process parameter combination are selected as production process parameters to be fine-tuned. This is to select production process parameters that have the necessity of optimization, and then the first derivative of their SHAP values ​​is used to guide the fine-tuning.

[0071] As an implementation method, for any production process parameter to be fine-tuned, if its first derivative of the SHAP value is greater than 0, it means that in the combination of production process parameters to be fine-tuned, increasing the production process parameter to be fine-tuned will cause its SHAP value to continue to increase. Therefore, it is necessary to reduce it based on the combination of production process parameters to improve the superiority of the production quality of the target type of cosmetic glass bottles. Then, it is reduced by a preset number of percentage points to generate the corresponding production process parameter replacement value.

[0072] If the first derivative of its SHAP value is less than 0, it means that in the combination of production process parameters to be fine-tuned, increasing the production process parameter to be fine-tuned will lead to a decrease in its SHAP value. Therefore, it is necessary to increase it on the basis of the combination of production process parameters to improve the quality of the target type of cosmetic glass bottle production. Then, increase it upward by a preset number of percentage points to generate the corresponding production process parameter replacement value.

[0073] If the first derivative of its SHAP value is equal to 0, then determine whether its SHAP value is a maximum or minimum value in the SHAP dependency graph. If it is a minimum value, it means that the production process parameter to be fine-tuned is at its optimal value in the combination of candidate production process parameters, so it will not be adjusted, that is, no corresponding production process parameter replacement value will be generated. If it is a maximum value, it means that the production process parameter to be fine-tuned is at its worst value in the combination of candidate production process parameters, so adjusting it up or down is to optimize it. In order to save production costs, it will be reduced down by a preset number of percentage points to generate the corresponding production process parameter replacement value. In addition, there is another possibility that the SHAP dependency graph is a horizontal straight line. This situation is extremely rare. Once it occurs, it means that the production process parameter to be fine-tuned is not important at present, and it will not be adjusted either.

[0074] It should be noted that the preset quantity is generally set between 1 and 5. That is, each time a fine-tuning is performed, the production process parameter to be fine-tuned is adjusted up or down by 1 to 5 percentage points to avoid the problem of over-adjustment. This is a routine operation for those skilled in the art.

[0075] 1.2) Determine if the total number of replacement values ​​for production process parameters is 0. If it is 0, it means that there are no replacement values ​​for production process parameters generated after fine-tuning. Therefore, no replacement scheme is generated, and the selected combination of production process parameters cannot be optimized to generate the preferred combination of production process parameters. The selected combination of production process parameters is then used as the optimal combination of production process parameters. Otherwise, proceed to step 1.3).

[0076] 1.3) First, with the constraint that the replacement scheme includes one production process parameter replacement value, the replacement values ​​of each production process parameter are arranged and combined to form several replacement schemes. Then, with the constraint that the replacement scheme includes two production process parameter replacement values, the replacement values ​​of each production process parameter are arranged and combined to form several replacement schemes. This process is repeated until the number of production process parameter replacement values ​​included in the replacement scheme is consistent with the total number of production process parameter replacement values. This setting traverses all possible combinations of production process parameter replacement values, laying the foundation for a complete analysis of all possibilities in the subsequent process.

[0077] 1.4) For each replacement scheme, the production process parameter replacement value within it is used to replace the same type of production process parameter in the candidate production process parameter combination to generate the preferred production process parameter combination after the replacement scheme is processed. In this way, all optimization directions of the candidate production process parameter combination are traversed, laying the foundation for a complete analysis of all possibilities in the future.

[0078] 2) Analyze each optimal combination of production process parameters one by one based on the Gaussian process regression model to output the predicted value and standard deviation of the corresponding production quality influence coefficient. For any optimal combination of production process parameters, add the predicted value and standard deviation of the corresponding production quality influence coefficient to obtain the most unfavorable value of the production quality influence coefficient corresponding to the optimal combination of production process parameters. Divide the predicted value and standard deviation of the corresponding production quality influence coefficient to obtain the confidence value of the production quality influence coefficient corresponding to the optimal combination of production process parameters.

[0079] It should be noted that when analyzing combinations of production process parameters, the production quality impact coefficient output by the Gaussian process regression model is not a unique value, but rather follows a Gaussian distribution. The model uses the mean of this distribution as the predicted value for the production quality impact coefficient. The sum of the predicted value and the standard deviation represents the 84.1% quantile, meaning that 84.1% of the data in the Gaussian distribution is lower than the sum of the predicted value and the standard deviation. This can be roughly considered as the actual value of the production quality impact coefficient corresponding to the optimal combination of production process parameters being lower than the sum of the corresponding predicted value and the standard deviation. Therefore, this sum is considered the corresponding production quality impact coefficient. The worst-case value of the coefficient represents the less-than-ideal state of the corresponding production quality impact coefficient. If a strict constraint is desired, the sum of the predicted value and three times the standard deviation can be used as the worst-case value. However, this constraint is too strict and not conducive to exploring and discovering better combinations of production process parameters. Therefore, a more lenient sum of the predicted value and the standard deviation is used here as the worst-case value. In addition, the confidence value obtained by dividing the predicted value by the standard deviation can be regarded as the coefficient of variation. The larger the value, the worse the accuracy of the Gaussian process regression model in predicting the corresponding candidate combinations of production process parameters, the more dispersed the predicted production quality impact coefficient, and the lower the confidence.

[0080] 3) Determine if there is an optimal combination of production process parameters with a confidence value lower than the confidence threshold. The confidence threshold is generally between 5% and 15%, and can be set to 10%. When the confidence value is lower than 10%, the data is very concentrated and the fluctuation is very small, indicating that the Gaussian process regression model is very confident in the predicted value. If there is no such combination, it means that the prediction output of the Gaussian process regression model for all optimal combinations of production process parameters is unreliable, that is, the predicted value of the production quality impact coefficient is unreliable and may have a large error with the actual value. In order to avoid blindly adjusting the combination of production process parameters, which may lead to low quality or even unqualified production of the target type of cosmetic glass bottles, the candidate combination of production process parameters is taken as the optimal combination of production process parameters. Otherwise, it means that there is an optimal combination of production process parameters with reliable output, and then proceed to step 4).

[0081] 4) From all the preferred production process parameter combinations with a confidence value lower than the confidence threshold, select the preferred production process parameter combination with the smallest worst-case value, and determine whether its worst-case value is less than the production quality influence coefficient under the candidate production process parameter combination. If it is, it means that the candidate production process parameter combination is still better than the candidate production process parameter combination under the undesirable state. Then update it as a new candidate production process parameter combination and return to step 1). Determine whether to continue to optimize it until no new candidate production process parameter combinations are generated. Take the last generated candidate production process parameter combination as the optimal production process parameter combination. Otherwise, take the candidate production process parameter combination as the optimal production process parameter combination.

[0082] The production changeover control module is used to execute the production changeover mechanism to perform one-click production changeover control for the target type of cosmetic glass bottles. Specifically, it adjusts various production process parameters based on the production model output by the production changeover mechanism to ensure the production quality of the target type of cosmetic glass bottles.

[0083] Please see Figure 2 As shown, in one embodiment, this application provides a one-click production changeover control method for a cosmetic glass bottle production line, used to execute the aforementioned one-click production changeover control system for the cosmetic glass bottle production line, including the following steps:

[0084] S1: Determine the type of cosmetic glass bottle to be replaced as the target type, and use the historical production rounds used to produce the target type of cosmetic glass bottle as the target production rounds. Collect and record the combination of production process parameters and production quality data under each target production round.

[0085] S2: Analyze production quality data to calculate the production quality impact coefficient under each combination of production process parameters. This coefficient is used to reflect the quality of the target type of cosmetic glass bottles produced under the corresponding combination of production process parameters.

[0086] S3: Compare each production quality influence coefficient with the reference value of the quality influence coefficient one by one to determine whether each combination of production process parameters is qualified;

[0087] S4: Based on the results of the qualification judgment, and in combination with the combination of production process parameters, establish a production change mechanism for the target type of cosmetic glass bottles;

[0088] S5: Implement the production changeover mechanism to control the production changeover of target type cosmetic glass bottles with one click.

[0089] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A one-button production changeover control system for a cosmetic glass bottle production line, characterized in that, include: The parameter acquisition module is used to determine the type of cosmetic glass bottle to be replaced as the target type, and to use the historical production rounds used to produce the target type of cosmetic glass bottle as the target production rounds. The module collects and records the combination of production process parameters and production quality data under each target production round. The production quality quantification module is used to analyze production quality data to calculate the production quality impact coefficient under various combinations of production process parameters. It is used to reflect the quality of the target type of cosmetic glass bottles produced under the corresponding combination of production process parameters. The quality analysis module is used to compare each production quality influence coefficient with the reference value of the quality influence coefficient one by one in order to determine whether each combination of production process parameters is qualified. The production switch mechanism construction module establishes a production switch mechanism for the target type of cosmetic glass bottles based on the results of the qualification judgment and in combination with the combination of production process parameters, Gaussian process regression model and SHAP value analysis. The production changeover control module is used to execute the production changeover mechanism to perform one-click production changeover control for the target type of cosmetic glass bottles; The production switching mechanism also includes: if all production process parameter combinations are deemed unqualified, the production process parameter combination with the smallest production quality impact coefficient is selected as the candidate production process parameter combination. By combining SHAP value calculation and Gaussian process regression model, the candidate production process parameter combination is fine-tuned and compared, and the optimal production process parameter combination is output as the production model.

2. The one-button production change control system for a cosmetic glass bottle production line according to claim 1, characterized in that: The production quality data includes the finish, transparency, and wall thickness of the finished cosmetic glass bottles produced in the target production round for the target type. The combination of production process parameters consists of various production process parameters executed in the target production cycle, including molding temperature, molding pressure, droplet weight, molding speed, annealing speed, and annealing time.

3. The one-click production change control system for the cosmetic glass bottle production line according to claim 2, characterized in that, The production quality influence coefficient is inversely proportional to the finish smoothness of the target type of cosmetic glass bottle, inversely proportional to the transparency of the target type of cosmetic glass bottle, and directly proportional to the wall thickness deviation of the target type of cosmetic glass bottle. For any target type of cosmetic glass bottle produced in the target production cycle, its wall thickness deviation is the absolute difference between its wall thickness and the design value of the wall thickness of the target type of cosmetic glass bottle.

4. The one-click production change control system for a cosmetic glass bottle production line according to claim 1, characterized in that, For any combination of production process parameters, if the corresponding production quality influence coefficient is not greater than the reference value of the quality influence coefficient, then the combination of production process parameters is considered qualified; otherwise, the combination of production process parameters is considered unqualified.

5. The one-click production change control system for the cosmetic glass bottle production line according to claim 3, characterized in that, The production switching mechanism includes: if at least one combination of production process parameters is deemed qualified, then from all qualified combinations of production process parameters, the combination of production process parameters with the smallest production quality impact coefficient is selected and output as the production model.

6. The one-click production change control system for the cosmetic glass bottle production line according to claim 5, characterized in that: The Gaussian process regression model is trained using the combination of production process parameters for each target production round as input and the corresponding production quality impact coefficient as output, so that when the Gaussian process regression model receives the combination of production process parameters, it outputs the predicted value and standard deviation of the corresponding production quality impact coefficient.

7. The one-button production change control system for a cosmetic glass bottle production line according to claim 6, characterized in that, The logic for outputting the optimal combination of production process parameters is as follows: 1) In the Gaussian process regression model, calculate the SHAP value of each production process parameter in the candidate production process parameter combination. Based on the SHAP value, determine whether to fine-tune the production process parameters to generate a replacement scheme. If no replacement scheme is generated, the candidate production process parameter combination is taken as the optimal production process parameter combination. Otherwise, based on the replacement scheme, replace the production process parameters in the candidate production process parameter combination to generate the preferred production process parameter combination. 2) Analyze each optimal combination of production process parameters one by one based on the Gaussian process regression model to output the predicted value and standard deviation of the corresponding production quality influence coefficient. For any optimal combination of production process parameters, add the predicted value and standard deviation of the corresponding production quality influence coefficient to obtain the most unfavorable value of the production quality influence coefficient corresponding to the optimal combination of production process parameters. Divide the predicted value and standard deviation of the corresponding production quality influence coefficient to obtain the confidence value of the production quality influence coefficient corresponding to the optimal combination of production process parameters. 3) Determine if there is an optimal combination of production process parameters with a confidence value lower than the confidence threshold. If not, take the candidate combination of production process parameters as the optimal combination of production process parameters; otherwise, proceed to step 4). 4) From all the preferred production process parameter combinations with a confidence value lower than the confidence threshold, select the preferred production process parameter combination with the smallest worst value, and determine whether its worst value is less than the production quality influence coefficient under the candidate production process parameter combination. If it is, update it as a new candidate production process parameter combination and return to step 1), until no new candidate production process parameter combinations are generated. The last generated candidate production process parameter combination is taken as the optimal production process parameter combination. Otherwise, the candidate production process parameter combination is taken as the optimal production process parameter combination.

8. The one-button production change control system for a cosmetic glass bottle production line according to claim 7, characterized in that, The logic for generating the optimal combination of production process parameters is as follows: 1.1) For any production process parameter in the combination of candidate production process parameters, determine whether its SHAP value is not greater than 0. If it is, then it is not considered as a production process parameter to be fine-tuned and is not fine-tuned. Otherwise, it is considered as a production process parameter to be fine-tuned, and based on the first derivative of its SHAP value, it is determined whether to fine-tune it to generate a replacement value for the production process parameter. Specifically, for any production process parameter to be fine-tuned, if its first derivative of SHAP value is greater than 0, it is reduced downward by a preset number of percentage points to generate a corresponding production process parameter replacement value; if its first derivative of SHAP value is less than 0, it is increased upward by a preset number of percentage points to generate a corresponding production process parameter replacement value; if its first derivative of SHAP value is equal to 0, it is determined whether its SHAP value is a maximum or a minimum value in the SHAP dependency graph. If it is a minimum value, it is not adjusted; if it is a maximum value, it is reduced downward by a preset number of percentage points to generate a corresponding production process parameter replacement value. 1.2) Determine if the total number of replacement values ​​for production process parameters is 0. If it is 0, no replacement scheme is generated, and the candidate production process parameter combination is taken as the optimal production process parameter combination; otherwise, proceed to step 1.3). 1.3) First, with the constraint that the replacement scheme includes one production process parameter replacement value, arrange and combine the replacement values ​​of each production process parameter to form several replacement schemes. Then, with the constraint that the replacement scheme includes two production process parameter replacement values, arrange and combine the replacement values ​​of each production process parameter to form several replacement schemes. Continue in this manner until the number of production process parameter replacement values ​​included in the replacement scheme is consistent with the total number of production process parameter replacement values. 1.4) For each replacement scheme, replace the same type of production process parameters in the candidate production process parameter combination with the production process parameter replacement value within it, so as to generate the preferred production process parameter combination after the replacement scheme has been processed.

9. A one-click production changeover control method for a cosmetic glass bottle production line, used to execute the one-click production changeover control system for the cosmetic glass bottle production line as described in any one of claims 1-8, characterized in that, Includes the following steps: S1: Determine the type of cosmetic glass bottle to be replaced as the target type, and use the historical production rounds used to produce the target type of cosmetic glass bottle as the target production rounds. Collect and record the combination of production process parameters and production quality data under each target production round. S2: Analyze production quality data to calculate the production quality impact coefficient under each combination of production process parameters. This coefficient is used to reflect the quality of the target type of cosmetic glass bottles produced under the corresponding combination of production process parameters. S3: Compare each production quality influence coefficient with the reference value of the quality influence coefficient one by one to determine whether each combination of production process parameters is qualified; S4: Based on the results of the qualification judgment, and combined with the combination of production process parameters, Gaussian process regression model and SHAP value analysis, establish a production change mechanism for the target type of cosmetic glass bottles; S5: Implement the production changeover mechanism to control the production changeover of target type cosmetic glass bottles with one click.

Images