Cigarette packaging quality management method and system, and storage medium

Abstract

The present application relates to the field of cigarette industry data application, and in particular to a cigarette packaging quality management method and system based on cigarette unique identification, and a storage medium, comprising: obtaining historical cigarette package quality association data and preprocessing to obtain a cigarette package dataset; identifying key failure modes according to the cigarette package dataset and constructing a failure mode rule library; comparing and analyzing the parameter distribution of qualified products and rejected products in the cigarette package dataset to obtain an optimal process range; based on the failure mode rule library, real-time monitoring and early warning of cigarette package quality core indicators; and process parameter optimization according to the optimal process range. Through data tracking and intelligent analysis, the present application embodiment realizes accurate attribution of cigarette package quality, predictive maintenance of equipment and intelligent optimization of process parameters, significantly improves quality and reduces loss, and promotes the production from experience-driven to data-driven intelligent closed-loop management.

Description

Technical Field

[0001] This invention relates to the field of data application in the cigarette industry, specifically to a method, system, and storage medium for quality control of cigarette packaging based on unique cigarette identifiers. Background Technology

[0002] In the entire cigarette production chain, the rolling and packaging process, as a crucial step in the formation of the finished product, directly determines the final appearance of the product, and appearance quality is one of the core indicators for measuring cigarette quality. For a long time, quality inspection in the rolling and packaging process has relied heavily on an experience-based system passed down through apprenticeships. Experienced employees, relying on years of accumulated feel, visual memory, and operational intuition, judge the possible causes of defects—for example, if they see wrinkles in the packaging, they prioritize adjusting the label paper conveyor. While this method can solve some common problems, it has significant limitations. On the one hand, experience-based judgment is highly subjective; different personnel have different standards for recognizing defects and different inspection logics, leading to multiple conclusions for the same defect, making it difficult to form a unified solution. On the other hand, experience requires long-term accumulation; new employees often need months or even years to independently handle complex problems, resulting in a long and costly talent development cycle.

[0003] The lack of data support is the core weakness of the current inspection model. In most cigarette production workshops, although the packaging equipment is equipped with basic operational parameter monitoring functions, data collection is mostly limited to macro-level aspects such as equipment start-up and shutdown, and production statistics. Micro-level data collection related to appearance quality is severely insufficient. When appearance defects occur, due to the lack of a complete data flow as a basis, it is impossible to trace back to the time point of the defect and the changes in related parameters. Instead, one can only rely on experience to infer, making the inspection process like the blind men and the elephant. Not only is it difficult to find the root cause, but it may also lead to ineffective adjustments due to misjudgment, increasing production costs. This experience-driven inspection model has become a bottleneck restricting quality improvement in the packaging process. As the cigarette industry continues to increase its requirements for product quality stability and consistency, relying solely on experience to deal with appearance quality problems cannot achieve early warning of defects, nor can it form a standardized inspection process.

[0004] Therefore, establishing a quality control system based on equipment operation data, through online reading and precise tracking of cigarette unique identifiers, and utilizing data analysis models to correlate parameter changes with appearance defects, is essential to shift quality inspection from experience-driven to data-driven, accurately pinpoint the root cause of problems, and promote the upgrading of quality control in the cigarette packaging process towards a more efficient and precise direction. Summary of the Invention

[0005] The purpose of this invention is to provide a method, system, and storage medium for quality control of cigarette packaging based on unique cigarette identifiers. This establishes a quality control system based on equipment operation data. By online reading and precise tracking of unique cigarette identifiers, and utilizing data analysis models to correlate parameter changes with appearance defects, the inspection of cigarette product appearance quality shifts from experience-driven to data-driven, accurately locating the root cause of problems. This promotes a more efficient and precise upgrade of quality control in the cigarette packaging process, thereby solving problems in existing technologies.

[0006] To achieve the above objectives, embodiments of the present invention provide a method for quality control of cigarette packaging, comprising: Historical cigarette pack quality correlation data is obtained and preprocessed to obtain a cigarette pack dataset. Based on the cigarette pack dataset, key failure modes were identified and a failure mode rule base was constructed. By comparing and analyzing the parameter distributions of qualified and rejected products in the aforementioned cigarette pack dataset, the optimal process range can be obtained. Based on the aforementioned fault mode rule base, real-time monitoring and early warning of core indicators of cigarette pack quality are performed. Optimize process parameters based on the optimal process range.

[0007] Optionally, identifying key failure modes and constructing a failure mode rule base based on the cigarette pack dataset includes: The support of frequent itemsets is obtained according to formula (1): (1) in, This is a set of conditional states that may lead to quality problems. For the first The overall condition of the rake pusher within a certain time period. This represents the overall conditional state of the rake pushing process across all time periods. For frequent itemsets Support This represents the number of time periods in historical production where the corresponding state pattern occurred. Candidate association rules are generated based on frequent itemsets, and the confidence level is obtained according to formula (2): (2) in, For confidence level, For candidate association rules Support level; Based on the preset threshold conditions of the support and confidence of the frequent itemsets, strong association rules and their corresponding confidence are selected from the candidate association rules and stored in the failure mode rule base.

[0008] Optionally, comparing and analyzing the parameter distributions of qualified and rejected products in the cigarette pack dataset to obtain the optimal process range includes: Perform single-parameter analysis to obtain the single-parameter safety range; Perform multi-parameter joint analysis to obtain the process safety window.

[0009] Optionally, a single-parameter analysis is performed to obtain the single-parameter safety interval, including: For each key parameter, calculate the sample mean and standard deviation of the qualified set and the rejected set respectively; Independent two-sample t-tests were performed on each key parameter to obtain the corresponding p-value; Determine whether the p-value is less than the preset significance level; If the p-value is less than the preset significance level, calculate the single-parameter safety interval based on the qualified set.

[0010] Optionally, multi-parameter joint analysis can be performed to obtain the process safety window, including: The Mahalanobis distance is obtained according to formula (3): (3) in, The Mahalanobis distance, For process parameter combinations, This is the mean vector of process parameters for qualified products; The process safety window is obtained according to formula (4): (4) in, For process safety window, The soldering iron temperature experienced by the small box. This refers to the heating time experienced by the small box. For process points Mahalanobis distance function, This is the distance threshold.

[0011] Optionally, based on the aforementioned fault mode rule base, real-time monitoring and early warning of core indicators of cigarette pack quality include: The standardized deviation and health index are obtained according to formulas (5) to (6): (5) (6) in, For standardization bias, The rejection rate is the latest unit of time. The average rejection rate. The standard deviation of the rejection rate. For health index, The attenuation coefficient; The failure probability and future health index are obtained according to formulas (7) to (8): (7) (8) in, This represents the probability of failure. For confidence level, For future health index, This is the trend coefficient. The health index at the previous moment. This is the random error term.

[0012] Optionally, optimizing process parameters based on the optimal process range includes: Construct a quality prediction model; An optimization algorithm is used to optimize the quality prediction model to obtain the optimal process parameters; The process standard is updated using the optimal process parameters and then verified.

[0013] Optionally, constructing a quality prediction model includes: Construct a multi-objective constrained optimization function based on formulas (9) to (12). (9) (10) (11) (12) in, For a multi-objective constrained optimization function, For defect prediction models, For energy consumption models, For production efficiency models, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, For the Sigmoid function, For the first The predicted values ​​of each decision tree. For the input feature vector, The heating energy consumption coefficient of the soldering iron. The soldering iron temperature set for the device. For ambient temperature, The energy consumption coefficient for equipment operation. The speed set for the equipment. Based on fixed energy consumption, For the ideal yield coefficient, This is the fault attenuation coefficient. The optimal speed is [the speed of the vehicle].

[0014] On the other hand, the present invention also provides a cigarette packaging quality control system, the system comprising: The online identification reading module is used to read the unique identifier of each small box of semi-finished products online and establish a reading queue according to the workstation. The product movement tracking module is used to accurately track the time and path of each small box of semi-finished products as it passes through each station in the designated process area. The equipment data acquisition module is used to collect key equipment process parameters in real time during the production process. Remove the product collection module to obtain a complete list of unqualified semi-finished products generated in a specified process range; The associated data analysis module is used to perform multi-dimensional feature extraction and analysis on the set of small box semi-finished products; A processor is configured to connect to the online identification reading module, the product shift tracking module, the equipment data acquisition module, the rejected product collection module, and the associated data analysis module, and the processor is configured to perform any of the methods described above.

[0015] In another aspect, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement any of the methods described above.

[0016] The beneficial effects of this invention are: This invention constructs a comprehensive data tracking system that collects key process parameters (such as machine speed and soldering iron temperature) at high frequency, starting from the unique identifier of each cigarette pack. Combined with a kinematic model, it precisely matches the processing status of each cigarette pack at each processing moment, establishing a complete integrated digital archive of "product-equipment-process." This enables precise tracing and objective attribution of quality defects. This fundamentally solves the problem of traditional inspections relying on subjective experience and lacking data support, transforming quality analysis from experience-based speculation to data verification. It effectively unifies inspection standards and provides new employees with clear and repeatable data-driven operational guidelines.

[0017] This invention employs the Apriori algorithm to mine association rules between historical operational data and quality results, obtaining strong association rules and transforming them into a standardized intelligent checklist. Simultaneously, it uses online Z-Score dynamic statistical control for real-time monitoring and early warning of key indicators. This solution transforms the implicit experience of experienced technicians into explicit knowledge of the system, and promotes the upgrade of quality management from reactive post-event processing to proactive predictive maintenance. When the system issues an early warning, maintenance personnel can perform targeted inspections based on precise guidance, greatly improving troubleshooting efficiency, shortening the talent training cycle, and effectively reducing batch quality risks.

[0018] This invention utilizes a combination of single-parameter and multi-parameter analysis with machine learning (GBDT model). Based on massive amounts of historical qualified product data, it obtains a process safety window and uses this as a hard constraint. A genetic algorithm is then employed to intelligently optimize multi-objective process parameters. This solution not only overcomes the blindness and inefficiency of traditional manual trial-and-error parameter tuning, finding optimal production setpoints that surpass experience, but also establishes a continuous improvement mechanism for the production process through a complete data loop of monitoring, analysis, optimization, execution, and verification. This ensures that the production system can dynamically adapt to changes, promoting the synergistic improvement of product quality and production efficiency, and accurately responding to the industry's urgent need for high-quality and stable production.

[0019] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 A flowchart of a cigarette packaging quality control method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a cigarette pack barcode reading station according to one embodiment of the present invention; Figure 3 A flowchart illustrating a method for identifying key failure modes and constructing a failure mode rule base based on a cigarette pack dataset according to an embodiment of the present invention; Figure 4 A flowchart of a single-parameter analysis method according to an embodiment of the present invention; Figure 5 A flowchart illustrating a method for real-time monitoring and early warning of core quality indicators of cigarette packs based on a fault mode rule base, according to an embodiment of the present invention. Figure 6This is a flowchart of a method for optimizing process parameters based on an optimal process range according to an embodiment of the present invention. Detailed Implementation

[0021] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0022] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0023] like Figure 1 The diagram shows a flowchart of a cigarette packaging quality control method according to an embodiment of the present invention. Figure 1 The method for quality control of cigarette packaging may include the following steps: In step S10, historical cigarette pack quality correlation data is obtained and preprocessed to obtain a cigarette pack dataset; In step S11, key failure modes are identified based on the cigarette pack dataset, and a failure mode rule base is constructed. In step S12, the parameter distributions of qualified and rejected products in the cigarette pack dataset are compared and analyzed to obtain the optimal process range; In step S13, based on the fault mode rule base, the core quality indicators of the cigarette pack are monitored and warned in real time. In step S14, process parameters are optimized based on the optimal process range.

[0024] In such Figure 1 In the cigarette packaging quality control method shown, step S10 is used to obtain historical cigarette pack quality correlation data and perform preprocessing to obtain a cigarette pack dataset. In this embodiment, a schematic diagram of the cigarette pack barcode reading station is shown below. Figure 2As shown, fixed barcode readers are installed at the workstations of the cigarette factory packaging machine to read the QR codes of small boxes and cartons online for association. The codes and reading times of all workstations are transmitted to the side server through the end-side industrial control computer and stored synchronously at both ends. The semi-finished small boxes are transported from workstation #1 to the transparent film packaging machine for small boxes via a toothed conveyor belt. After wrapping, folding and heat sealing in the transparent film packaging machine for small boxes (hereinafter referred to as CH), the packaging quality is inspected. Unqualified products are rejected through the CH rejection port, and qualified products enter the next production stage through workstation #2. The quality of transparent film packaging for small boxes is related to at least the following factors: (1) the orientation of the cigarette pack when the semi-finished small box enters CH; (2) the temperature and heating time of the heat sealing iron. Incorrect orientation will cause the seal and pull line to tilt. If the iron temperature is too high or too low, or the heating time is too short or too long, the folded parts will loosen or deform. Therefore, the cigarette pack quality association data includes the cigarette pack motion parameters and the cigarette pack process parameters. Among them, the movement parameters of the cigarette pack are triggered and recorded by the cigarette pack reading code. Specifically, in this example, during the actual production process, the #1 barcode reader reads the QR code of the small box semi-finished product online and establishes the reading code queue S1, and the #2 barcode reader reads the QR code of the small box semi-finished product online and establishes the reading code queue S2. S1 and S2 are stored in real time or at regular intervals. Since there is only one CH rejection port between the #1 and #2 workstations, S1-S2 can obtain the set Sr of rejected small box semi-finished products. By reading the cigarette pack data through the #1 barcode reader, the current pusher number and pusher time are triggered and recorded. After the pusher performs the push action, it is retracted and switched to the next pusher number. The above process is repeated, and finally a pusher number sequence arranged in chronological order is formed. The specific steps are as follows: (1) System initialization: Set the initial value of the pusher number: i=1 (corresponding to the P1 pusher). Ensure that the pusher is in the retracted state and wait for the start signal. (2) Cigarette pack QR code reading trigger: Reader #1 monitors the cigarette packs in real time. When it successfully reads the QR code data of a pack of cigarettes, it obtains the cigarette pack data D and the reading time. (3) Record the pusher number and pusher time: record the current pusher number. (e.g., P1) and the time for pushing the rake Record to the time series database. Usually taken (Assuming the rake is pushed immediately after the code is read), or the precise push-out time can be obtained via a rake sensor. Example recording format: (4) The rake pushes out: control the rake to push out. Push the tobacco pack to the next workstation (such as the small box of semi-finished products). After the pusher is pushed out to the correct position, pause briefly to ensure the tobacco pack enters stably. (5) Pusher retraction and switching: The pusher is retracted to the initial position. Switch the pusher number: If the current i=1 (P1), switch to i=2 (P2); if the current i=2 (P2), switch to i=1 (P1). Use the formula: i=3-i (applicable to two pushers alternating). By recording the movement parameters of the tobacco pack, a one-to-one correspondence between product, equipment, and time can be achieved.

[0025] The cigarette packaging process parameters, including CH speed, end soldering iron, and side soldering iron, are collected at high frequency by a cigarette packaging data acquisition program, and at least the real-time parameter values ​​and acquisition time are stored. During the operation of the packaging machine, the small boxes of semi-finished cigarette packs move from station #1 to station #2. S1 stores the time for each small box of semi-finished product to pass through station #1. By using the real-time speed V and the relative relationship between stations, at least the soldering iron temperature T and the duration D of each small box of semi-finished product being heated by the soldering iron during heat sealing can be obtained from Sr. A set S(T, D) is generated based on Sr.

[0026] After obtaining the motion and process parameters of the tobacco packs, backtracking calculations are performed on the motion and process parameters. The process parameters (T, D) are precisely matched to each specific tobacco pack to form a tobacco pack dataset for subsequent analysis. Specifically, in this example, for each small box i recorded in queue S1 (regardless of whether it is ultimately rejected), its time passing through station #1 is used as the dataset. Based on the real-time vehicle speed V(t) and the fixed distance L from workstation #1 to the target soldering iron workstation, the precise time for the vehicle to reach the soldering iron workstation is calculated using the integral equation of the motion model. ,satisfy .exist At any given moment, the soldering iron temperature experienced by the small box is obtained by interpolating the high-frequency acquired temperature data. And based on real-time vehicle speed With the effective length of the soldering iron Calculate heating time ( Generate process parameter records for small box i: ( After CH processing and quality inspection, qualified cigarette packs are scanned by barcode reader S2 and enter queue S2. The system uses the set operation Sr=S1-S2 to identify the set Sr of rejected cigarette packs. The final state of each cigarette pack (regardless of whether it's S2 or Sr) is marked, forming a complete set of data containing its full history. , This provides labeled training and validation data for subsequent statistical analysis.

[0027] Step S11 is used to identify key failure modes and construct a failure mode rule base based on the cigarette pack dataset. In this example, an improved Apriori algorithm is applied to mine strong association rules. Specifically, the specific methods for constructing the failure mode rule base may include, for example... Figure 3 The steps shown are described in this. Figure 3 In this context, step S11 may include: In step S20, the support of frequent itemsets is obtained; In step S21, candidate association rules are generated based on frequent itemsets and confidence levels are obtained; In step S22, strong association rules and their corresponding confidence scores are selected from the candidate association rules and stored in the failure mode rule base according to the preset threshold conditions of the frequent itemset support and confidence scores.

[0028] In such Figure 3 In the method shown, step S20 is used to discretize the continuous parameters at equal frequencies to form a transaction itemset T, and to calculate the support of the frequent itemset. Specifically, it can be calculated, for example, using formula (1): (1) in, This is a set of conditional states that may lead to quality problems. For the first The overall condition of the rake pusher within a certain time period. This represents the overall conditional state of the rake pushing process across all time periods. For frequent itemsets Support This represents the number of time periods in historical production where the corresponding state pattern occurred.

[0029] Step S21 is used to generate the association rule X→Y and calculate the confidence score. Specifically, the confidence score can be obtained, for example, using formula (2): (2) in, For confidence level, For candidate association rules Support level.

[0030] Step S22 is used to filter out those that meet the requirements. and The rules are stored in the fault mode rule base F.

[0031] Step S12 is used to compare and analyze the parameter distributions of qualified and rejected products in the cigarette pack dataset to obtain the optimal process range, and is used to quantitatively identify the separation boundary between qualified and rejected samples in terms of process parameters. In this embodiment, the specific method for performing the comparative analysis in step S12 to obtain the optimal process range can be of various forms known to those skilled in the art. In one example of the present invention, step S12 may include: In step S30, a single-parameter analysis is performed to obtain the single-parameter safety range; In step S31, a multi-parameter joint analysis is performed to obtain the process safety window.

[0032] Step S30 is used for preliminary screening using single-parameter analysis to quickly identify which individual parameters may be significantly related to quality issues. In this embodiment, the specific method for performing single-parameter analysis in step S30 can be of various forms known to those skilled in the art. In one example of the present invention, step S30 may include, for example... Figure 4 The steps shown are described in this. Figure 4 In this context, step S30 may include: In step S40, for each key parameter, the sample mean and standard deviation of the qualified set and the rejected set are calculated respectively; In step S41, an independent two-sample t-test is performed on each key parameter to obtain the corresponding p-value; In step S42, it is determined whether the p-value is less than the preset significance level; In step S43, if the p-value is determined to be less than the preset significance level, the single-parameter safety interval is calculated based on the qualified set.

[0033] In such Figure 4 In the method shown, step S40 is used to calculate the sample mean and standard deviation of the qualified set S2 and the rejected set Sr for key parameters (such as soldering iron temperature T). An overlap histogram or density curve is plotted to visually compare the two distributions; a box plot is plotted to compare the median, dispersion, and outliers of the two groups, providing a preliminary observation of the distribution differences between qualified and rejected products on this parameter. Step S41 is used to perform an independent two-sample t-test, first constructing hypotheses, for example: null hypothesis H0: (No significant difference in mean temperature); Alternative hypothesis H1: (There is a significant difference in the mean temperature). Next, the t-statistic and degrees of freedom are calculated, and then the p-value is obtained based on the t-statistic and degrees of freedom. Step S42 determines whether the p-value is less than the preset significance level. In this example, if the p-value < 0.05, the null hypothesis is rejected, indicating a significant difference in the mean temperature between qualified and rejected products, meaning the soldering iron temperature T is significantly correlated with rejection. Step S43, when the p-value is determined to be less than the preset significance level, calculates the single-parameter safety interval based on the qualified set. Based on the distribution of the qualified product samples, a high-probability interval covering qualified products is defined as the suggested safety interval for this parameter. For example, the mean ± K times the standard deviation is used. K is set according to quality requirements (e.g., K=2, covering approximately 95.4% of normally distributed qualified products). Simultaneously, the proportion of rejected products falling into this interval is calculated; if the proportion is very low, the validity of this interval is verified.

[0034] Since quality may be affected by both T and D (or V), a two-dimensional joint safety region needs to be constructed. Multi-parameter joint analysis is performed in step S31 to obtain the process safety window. In this embodiment, the method for obtaining the process safety window can be elliptic model acquisition, rule extraction, or various other methods known to those skilled in the art. In one example of this invention, an elliptic model is used to obtain the process safety window. It is assumed that the qualified product parameters (T, D) follow a bivariate normal distribution, and their mean vector μ and covariance matrix Σ are calculated. The Mahalanobis distance d is defined as follows: (3) in, The Mahalanobis distance, For process parameter combinations, This is the mean vector of process parameters for qualified products.

[0035] Select a distance threshold This ensures that the vast majority (e.g., 95%) of qualified products meet the requirements. Then combine security windows Defined as: (4) in, For process safety window, The soldering iron temperature experienced by the small box. This refers to the heating time experienced by the small box. For process points Mahalanobis distance function, This is the distance threshold. The region appears as an ellipse on a two-dimensional plane.

[0036] Step S13 is used to perform real-time monitoring and early warning of core quality indicators of cigarette packs based on a fault mode rule base. In this embodiment, the specific method for real-time monitoring and early warning of core quality indicators of cigarette packs in step S13 can be of various forms known to those skilled in the art. In one example of the present invention, step S13 may include, for example... Figure 5 The steps shown are described in this. Figure 5 In this context, step S13 may include: In step S50, the mean and standard deviation of the rejection rate within the sliding window are calculated in real time; In step S51, the standardized deviation and health index are calculated; In step S52, fault modes are dynamically matched and trend prediction is performed.

[0037] In such Figure 5 In the method shown, step S50 is used to adjust the currently running pusher. Real-time calculation of the mean and standard deviation of the rejection rate within the sliding window: (13) (14) in, The average rejection rate. Total number of time units Number the current time unit. For the first Removal rate per time unit The standard deviation of the rejection rate.

[0038] Step S51 is used to calculate the standardized deviation and health index. In this example, it can be calculated, for example, using formulas (5) and (6): (5) (6) in, For standardization bias, The rejection rate is the latest unit of time. The average rejection rate. The standard deviation of the rejection rate. For health index, This is the attenuation coefficient.

[0039] Step S52 is used to dynamically match fault modes and predict trends. If the current parameter combination matches a rule premise in the rule base F, the corresponding monitoring mode is activated, and the fault probability is calculated. (7) in, This represents the probability of failure. , where is the confidence level.

[0040] Predicting the health index for the next N periods based on the ARIMA model: (8) in, For future health index, This is the trend coefficient. The health index at the previous moment. This is the random error term.

[0041] Step S14 is used to optimize process parameters according to the optimal process range. In this embodiment, the specific method for optimizing process parameters in step S14 can be of various forms known to those skilled in the art. In one example of the present invention, step S14 may include, for example... Figure 6 The steps shown are described in this. Figure 6 In this context, step S14 may include: In step S60, a quality prediction model is constructed; In step S61, an optimization algorithm is used to optimize the quality prediction model in order to obtain the optimal process parameters; In step S62, the process standard is updated using the optimal process parameters and then verified.

[0042] In such Figure 6 In the method shown, step S60 is used to construct a quality prediction model, and the safety window obtained in step S12 is used... As a hard constraint, the process optimization problem is redefined. Specifically, in this example, a multi-objective constrained optimization function is constructed: (9) in, , , , For a defect prediction model trained on historical data, For energy consumption models, For production efficiency models, , , These are weighting coefficients. The key constraint is the process point. It must fall within the joint security window Inside, and Depend on Decide( D). In this example, the defect prediction model It can be used The GBDT model trained on the data. The calculation formulas for the defect prediction model, energy consumption model, and production efficiency model are as follows: (10) (11) (12) in, For the Sigmoid function, For the first The predicted values ​​of each decision tree. For the input feature vector, The heating energy consumption coefficient of the soldering iron. The soldering iron temperature set for the device. For ambient temperature, The energy consumption coefficient for equipment operation. The speed set for the equipment. Based on fixed energy consumption, For the ideal yield coefficient, This is the fault attenuation coefficient. The optimal speed is [the speed of the vehicle].

[0043] Step S61 is used to optimize the quality prediction model using an optimization algorithm to obtain the optimal process parameters. In this embodiment, a genetic algorithm is used for optimization, encoding the machine speed V and soldering iron temperature T as chromosomes, and using the quality risk score y predicted by the GBDT model as the fitness evaluation standard. Under multiple constraints that satisfy equipment safety and process safety windows, the globally optimal or near-optimal combination of process parameters (V, T) is iteratively searched through selection, crossover, and mutation operations that simulate the biological evolution process. Specifically, the specific method of using a genetic algorithm for optimization includes the following steps: In step S70, the heat sealing process parameters of the equipment are encoded using chromosomes and the population is initialized; In step S71, a fitness function is constructed to select a parent individual based on the fitness of each individual in the population; In step S72, crossover and mutation operations are performed on the parent individuals to obtain the offspring individuals; In step S73, the process returns to the step of constructing the fitness function and selecting parent individuals based on the fitness of each individual in the population until the preset conditions are met, so as to obtain the optimal combination of process control parameters.

[0044] Step S70 is used to encode the continuous variables to be optimized, vehicle speed V and soldering iron temperature T, into a single chromosome, within the allowable range of vehicle speed and temperature. and Within the initial population, N individuals (e.g., N=50) are randomly generated to form the initial population P0. For each randomly generated individual, it is checked whether its corresponding process point falls within the safety window. If the solution is not within the window, then the boundary absorption method or the projection method towards the center of the window is used for repair to ensure that all individuals in the initial population are feasible solutions.

[0045] Step S71 is used to construct a fitness function to evaluate the quality of individuals in order to obtain parent individuals. Specifically, in this example, the fitness of each individual is obtained according to formula (15): (15) in, For each individual's fitness, The quality risk score predicted by the defect prediction model. The penalty coefficient is... This is a penalty item.

[0046] For each individual, a selection operation is performed based on the fitness, and the probability of an individual being selected is obtained according to formula (16): (16) in, For the first The probability of an individual being selected. For the first The fitness of an individual For the first The fitness of an individual This represents the required number of parent individuals.

[0047] Step S72 is used to perform crossover and mutation operations on the parent individuals to obtain the offspring individuals. Specifically, in this example, the first generation offspring individuals are obtained according to formulas (17) to (18): (17) (18) in, For the first generation offspring individual 1, These are the cross-weighting coefficients. As the first parent individual, As the second parent individual, Individual 2 is the first generation offspring. The corresponding equipment heat sealing process parameters obtained from the first generation of offspring individuals are mutated, and the mutated equipment heat sealing process parameters are obtained according to formulas (19) to (20): (19) (20) in, The speed set for the modified equipment. The vehicle speed set for the equipment before the mutation. For the number of iterations, This is the maximum speed of the equipment. The soldering iron temperature set for the modified device. The soldering iron temperature set for the device before the modification. This is the maximum soldering iron temperature of the device.

[0048] Next, the new individuals generated through selection, crossover, and mutation are combined with the best individuals of the present generation to form a new generation of population.

[0049] Step S73 returns to the step of constructing the fitness function, selecting parent individuals based on the fitness of each individual in the population, until a preset condition is met to obtain the optimal combination of process control parameters. In this example, the preset condition may be that the optimal fitness of the population no longer significantly increases (the change is less than a threshold) for several consecutive generations (e.g., 20 generations). When the algorithm terminates, the individual with the highest fitness in the current population is decoded to obtain the optimal combination of process parameters, and its corresponding predicted quality risk score and other key information are output.

[0050] Step S62 is used to update the process standard with optimal process parameters and verify it. A comparative analysis report is generated, showing a comparison chart of parameter distributions for acceptable and rejected products, a suggested safety window range, and the new process setpoint (V) obtained within this window. , T ) and its expected benefits.

[0051] Then, closed-loop verification is performed. After the new standard is implemented, newly generated S2 and Sr data are collected, and the new data points are calculated relative to the original safety window. The location of the safety window is determined by the presence of numerous qualified samples deviating from the original window or a large number of rejected samples appearing within the window. This triggers a recalculation and update of the safety window, enabling the model to evolve itself.

[0052] On the other hand, the present invention also provides a cigarette packaging quality control system, the system comprising: an online identification reading module, a product movement tracking module, an equipment data acquisition module, a rejected product collection module, a correlation data analysis module, and a processor. The online identification reading module is used to read the unique identifier of each small box of semi-finished products online and establish a code reading queue according to workstation. The product movement tracking module is used to accurately track the time and path of each small box of semi-finished products passing through each workstation in a designated process area. The equipment data acquisition module is used to collect key equipment process parameters in real time during the production process. The rejected product collection module is used to obtain a complete list of unqualified semi-finished products generated in a designated process interval. The correlation data analysis module is used to perform multi-dimensional feature extraction and analysis on the set of small box semi-finished products. The processor is connected to the online identification reading module, the product movement tracking module, the equipment data acquisition module, the rejected product collection module, and the correlation data analysis module, and the processor is configured to execute any of the methods described in the cigarette packaging quality control method.

[0053] In another aspect, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement any of the methods described in the cigarette packaging quality control method.

[0054] The beneficial effects of this invention are: This invention constructs a comprehensive data tracking system that collects key process parameters (such as machine speed and soldering iron temperature) at high frequency, starting from the unique identifier of each cigarette pack. Combined with a kinematic model, it precisely matches the processing status of each cigarette pack at each processing moment, establishing a complete integrated digital archive of "product-equipment-process." This enables precise tracing and objective attribution of quality defects. This fundamentally solves the problem of traditional inspections relying on subjective experience and lacking data support, transforming quality analysis from experience-based speculation to data verification. It effectively unifies inspection standards and provides new employees with clear and repeatable data-driven operational guidelines.

[0055] This invention employs the Apriori algorithm to mine association rules between historical operational data and quality results, obtaining strong association rules and transforming them into a standardized intelligent checklist. Simultaneously, it uses online Z-Score dynamic statistical control for real-time monitoring and early warning of key indicators. This solution transforms the implicit experience of experienced technicians into explicit knowledge of the system, and promotes the upgrade of quality management from reactive post-event processing to proactive predictive maintenance. When the system issues an early warning, maintenance personnel can perform targeted inspections based on precise guidance, greatly improving troubleshooting efficiency, shortening the talent training cycle, and effectively reducing batch quality risks.

[0056] This invention utilizes a combination of single-parameter and multi-parameter analysis with machine learning (GBDT model). Based on massive amounts of historical qualified product data, it obtains a process safety window and uses this as a hard constraint. A genetic algorithm is then employed to intelligently optimize multi-objective process parameters. This solution not only overcomes the blindness and inefficiency of traditional manual trial-and-error parameter tuning, finding optimal production setpoints that surpass experience, but also establishes a continuous improvement mechanism for the production process through a complete data loop of monitoring, analysis, optimization, execution, and verification. This ensures that the production system can dynamically adapt to changes, promoting the synergistic improvement of product quality and production efficiency, and accurately responding to the industry's urgent need for high-quality and stable production.

[0057] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0059] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0060] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0061] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0062] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0063] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0064] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0065] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for quality control of cigarette packaging, characterized in that, The control methods include: Historical cigarette pack quality correlation data is obtained and preprocessed to obtain a cigarette pack dataset. Based on the cigarette pack dataset, key failure modes were identified and a failure mode rule base was constructed. By comparing and analyzing the parameter distributions of qualified and rejected products in the aforementioned cigarette pack dataset, the optimal process range can be obtained. Based on the aforementioned fault mode rule base, real-time monitoring and early warning of core indicators of cigarette pack quality are performed. Optimize process parameters based on the optimal process range.

2. The control method according to claim 1, characterized in that, Based on the cigarette pack dataset, key failure modes were identified and a failure mode rule base was constructed, including: The support of frequent itemsets is obtained according to formula (1): ，（1） in, This is a set of conditional states that may lead to quality problems. For the first The overall condition of the rake pusher within a certain time period. This represents the overall conditional state of the rake pushing process across all time periods. For frequent itemsets Support This represents the number of time periods in historical production where the corresponding state pattern occurred. Candidate association rules are generated based on frequent itemsets, and the confidence level is obtained according to formula (2): ，（2） in, For confidence level, For candidate association rules Support level; Based on the preset threshold conditions of the support and confidence of the frequent itemsets, strong association rules and their corresponding confidence are selected from the candidate association rules and stored in the failure mode rule base.

3. The control method according to claim 1, characterized in that, Comparative analysis of the parameter distributions of qualified and rejected products in the aforementioned cigarette pack dataset is used to obtain the optimal process range, including: Perform single-parameter analysis to obtain the single-parameter safety range; Perform multi-parameter joint analysis to obtain the process safety window.

4. The control method according to claim 3, characterized in that, Perform single-parameter analysis to obtain the single-parameter safety interval, including: For each key parameter, calculate the sample mean and standard deviation of the qualified set and the rejected set respectively; Independent two-sample t-tests were performed on each key parameter to obtain the corresponding p-value; Determine whether the p-value is less than the preset significance level; If the p-value is less than the preset significance level, calculate the single-parameter safety interval based on the qualified set.

5. The control method according to claim 3, characterized in that, Performing multi-parameter joint analysis to obtain the process safety window includes: The Mahalanobis distance is obtained according to formula (3): ，（3） in, The Mahalanobis distance, For process parameter combinations, This is the mean vector of process parameters for qualified products; Obtain the process safety window according to formula (4): ，（4） in, For process safety window, The soldering iron temperature experienced by the small box. This refers to the heating time experienced by the small box. For process points Mahalanobis distance function, This is the distance threshold.

6. The control method according to claim 1, characterized in that, Based on the aforementioned fault mode rule base, real-time monitoring and early warning of core indicators of cigarette pack quality include: The standardized deviation and health index are obtained according to formulas (5) to (6): ，（5） ，（6） in, For standardization bias, The rejection rate is the latest unit of time. The average rejection rate. The standard deviation of the rejection rate. For health index, The attenuation coefficient; The failure probability and future health index are obtained according to formulas (7) to (8): ，（7） ，（8） in, This represents the probability of failure. For confidence level, For future health index, This is the trend coefficient. The health index at the previous moment. This is the random error term.

7. The control method according to claim 1, characterized in that, Optimizing process parameters based on the optimal process range includes: Build a quality prediction model; An optimization algorithm is used to optimize the quality prediction model to obtain the optimal process parameters; The process standard is updated using the optimal process parameters and then verified.

8. The control method according to claim 7, characterized in that, Building a quality prediction model includes: Construct a multi-objective constrained optimization function based on formulas (9) to (12). ，（9） ，（10） ，（11） ，（12） in, For multi-objective constrained optimization functions, For defect prediction models, For energy consumption models, For production efficiency models, for The weighting coefficients, for The weighting coefficients, for The weighting coefficients, For the Sigmoid function, For the first The predicted values ​​of each decision tree. For the input feature vector, The heating energy consumption coefficient of the soldering iron. The soldering iron temperature set for the device. For ambient temperature, The energy consumption coefficient for equipment operation. The speed set for the equipment. Based on fixed energy consumption, For the ideal yield coefficient, This is the fault attenuation coefficient. The optimal speed is [the speed of the vehicle].

9. A cigarette packaging quality control system, characterized in that, The system includes: The online identification reading module is used to read the unique identifier of each small box of semi-finished products online and establish a reading queue according to the workstation. The product movement tracking module is used to accurately track the time and path of each small box of semi-finished products as it passes through each station in the designated process area. The equipment data acquisition module is used to collect key equipment process parameters in real time during the production process. Remove the product collection module to obtain a complete list of unqualified semi-finished products generated in a specified process range; The associated data analysis module is used to perform multi-dimensional feature extraction and analysis on the set of small box semi-finished products; A processor is configured to connect to the online identification reading module, the product shift tracking module, the equipment data acquisition module, the rejected product collection module, and the associated data analysis module, and the processor is configured to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed by a processor, implement the method as described in any one of claims 1 to 8.

Images