Intelligent cigarette packet paperboard production data processing method and system
By calculating the process composite volatility and long-term deterioration index, an adaptive reliability coefficient is constructed, which solves the problems of accuracy and robustness of the quality prediction model in cigarette pack paper production. It enables accurate early warning and adaptive control in complex environments, reducing scrap rate and production costs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUANGDONG MEIKE NEW MATERIALS CO LTD
- Filing Date
- 2026-02-04
- Publication Date
- 2026-05-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the production process of cigarette pack paperboard, the existing quality prediction model has low prediction accuracy and poor robustness when the equipment status changes, making it difficult to achieve accurate prediction and adaptive control, resulting in high scrap rate and increased production cost.
By calculating the process composite volatility and long-term degradation index, an adaptive reliability coefficient is constructed to dynamically evaluate the credibility of the quality prediction model. Combined with the LSTM model, sub-health status and equipment wear are identified, risk assessment weights are adjusted, and early warning accuracy is improved.
It enables accurate quality prediction and early warning under both stable and unstable operating conditions, reduces scrap rate, minimizes false alarms and missed alarms, and optimizes production process control.
Smart Images
Figure CN122046017A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of data processing technology, and in particular to a method and system for processing data in the production of cigarette pack cardboard based on intelligence. Background Technology
[0002] The production process of cigarette pack cardboard is complex, involving multiple continuous steps such as papermaking, coating, printing, hot stamping, and die-cutting, and involving various process parameters. Even slight fluctuations in process parameters can lead to quality problems in the final product, such as color difference, misregistration, surface defects, and poor flatness, resulting in scrap and waste of resources.
[0003] Traditional quality management methods for cigarette pack cardboard production mainly rely on the following approaches: Statistical Process Control (SPC): This involves monitoring the mean and fluctuation range of key process parameters through control charts. When a parameter exceeds the control limit, an alarm is triggered, requiring manual intervention. However, this method primarily focuses on whether the absolute value of the parameter exceeds the limit. It has a weak ability to identify sub-healthy states where parameters are within the control range but exhibit significant fluctuations, easily leading to delayed alarms or false alarms. Manual Judgment: Operators make judgments and adjustments based on observations of equipment operating status and the initial shape of the product. This method heavily relies on individual experience, lacks standardization and quantitative basis, and is difficult to cope with increasingly complex and high-speed production environments. Model-Based Quality Prediction: This utilizes historical production data to train predictive models, such as Support Vector Machines, Decision Trees, Neural Networks, and especially Long Short-Term Memory (LSTM) networks. These models can predict future product quality based on current process parameters, thus achieving a certain degree of early warning. However, while predictive models perform well under stable operating conditions, their accuracy and reliability decrease when the production environment is complex, equipment conditions change (such as wear and aging), or raw material batches fluctuate. Furthermore, existing prediction models struggle to effectively distinguish between instantaneous process fluctuations caused by random noise and transient disturbances, and systemic quality change trends resulting from equipment wear and control system detuning. This inadequate distinction prevents the model from dynamically adjusting its level of confidence in its own predictions.
[0004] In summary, existing technologies for quality management in cigarette pack paper production suffer from low prediction accuracy and poor robustness when the status of production line equipment changes. This makes it difficult to achieve accurate prediction and adaptive control of the production process, further leading to a high scrap rate and increased production operating costs. Summary of the Invention
[0005] To address the issue of low accuracy in prediction results from existing predictive models during cigarette pack paper production, this invention provides an intelligent data processing method and system for cigarette pack paper production.
[0006] In a first aspect, the present invention provides an intelligent data processing method for cigarette pack cardboard production, employing the following technical solution: A data processing method for intelligent cigarette pack cardboard production includes the following steps: The process parameters of the cigarette pack cardboard production process are obtained and input into a preset quality prediction model to obtain predicted quality indicators. The process composite volatility of the process parameters is calculated, representing the degree of change of the process parameters within a time window. The long-term deterioration index of the cigarette pack cardboard is calculated by: calculating the difficulty coefficient of processing the cigarette pack cardboard, which is the ratio of the registration error to the registration length; accumulating the ratio of the process composite volatility to the correction factor determined based on the difficulty coefficient within a batch cycle to obtain the long-term deterioration index; calculating the adaptive reliability coefficient of the quality prediction model, which is the product of the reliability function of the process composite volatility and the health function of the long-term deterioration index; calculating the short-term quality risk index, which is positively correlated with the product of the adaptive reliability coefficient and the quality indicators; when the short-term quality risk index exceeds a preset risk threshold, it indicates that a quality defect is about to occur, and a shutdown signal is issued.
[0007] An adaptive reliability coefficient was constructed by calculating the composite volatility of the process and the long-term deterioration index, enabling dynamic evaluation of the credibility of the quality prediction model. This effectively identifies sub-healthy conditions where production lines are in a state of drastic fluctuation despite parameters not exceeding limits, as well as long-term wear and tear on equipment. By adaptively adjusting the risk assessment weights, the problem of decreased prediction accuracy in traditional models under unstable operating conditions or equipment aging is solved, improving the accuracy and robustness of quality defect early warning in cigarette pack paper production.
[0008] Preferably, the method for calculating the composite process volatility includes: normalizing the process parameters within the time window; calculating the standard deviation of the rate of change of the normalized process parameters; and weighted summing of all standard deviations to obtain the composite process volatility.
[0009] By normalizing the process parameters within the time window, calculating the standard deviation of the rate of change, and performing weighted summation, the dynamic drastic changes of process parameters with different dimensions can be quantified. This allows for the accurate capture of instantaneous disturbances and instabilities in the production process, providing a quantitative basis for subsequent evaluation of the reliability of the prediction model under current operating conditions.
[0010] Preferably, the reliability function is:
[0011] In the formula: Indicates at time The reliability of the prediction results; This indicates the preset sensitivity parameter; Indicates at time The process of time-related composite volatility; The base value represents the composite volatility of the process, and exp() represents an exponential function with base e.
[0012] By utilizing the properties of the Sigmoid function, the composite process volatility is mapped to a value between 0 and 1. When process volatility exceeds the baseline value, reliability decreases, reflecting the negative impact of current operating conditions on the reliability of model predictions. This ensures that the system can automatically reduce its reliance on model predictions when production fluctuations are large.
[0013] Preferably, the health function is:
[0014] In the formula: Indicates at time The health status of the production line system, with a value range of [0,1]. This represents the preset degradation impact factor; Indicates at time Long-term degradation index; This indicates the preset reference degradation threshold.
[0015] By introducing degradation impact factors, the physical health of the production line system was quantified. This allows for the identification of prediction biases caused by equipment performance degradation, preventing blind reliance on model predictions when equipment is in suboptimal condition, and further improving quality risk management throughout the entire lifecycle.
[0016] Preferably, the reference degradation threshold is the historical long-term degradation index. The 90th percentile.
[0017] A data-driven adaptive threshold setting mechanism has been introduced to avoid the subjectivity and inaccuracy of manually setting thresholds, and can automatically define the boundary between health and deterioration based on the historical performance of the production line.
[0018] Preferably, the production data processing method further includes: issuing a maintenance signal in response to the continuous increase of the long-term degradation index in multiple consecutive production batches.
[0019] By identifying trend changes in degradation indices, potential equipment hazards can be detected before quality incidents occur, reducing unplanned downtime.
[0020] The preferred formula for calculating the short-term quality risk index is:
[0021] In the formula, Indicates at time Short-term quality risk index at that time; Indicates at time The adaptive reliability coefficient of the prediction results; Indicates at time The quality indicators output by the time-of-use quality prediction model; This represents the average absolute error of the quality prediction model over the past batch. Indicates at time The time-dependent composite volatility.
[0022] When the system is healthy and stable, it primarily relies on the model's accurate predictions; when the system experiences large fluctuations or severe degradation, it automatically switches to heuristic assessments based on volatility and historical errors. This ensures that the most reasonable risk warnings are provided under any operating condition, significantly reducing false alarm and false negative rates.
[0023] Preferably, the quality prediction model is the LSTM (Long Short-Term Memory) network model.
[0024] LSTM can better handle the time series dependencies in tobacco packaging production data, capture the temporal characteristics of changes in preceding and subsequent processes and parameters, and thus provide high-precision prediction of basic quality indicators when the operating conditions are stable.
[0025] Preferably, the process parameters include at least one of the following: machine speed, ink viscosity, color difference sequence, registration error, and drying temperature in the printing process; die-cutting pressure, hot stamping temperature, and hot stamping speed in the die-cutting and hot stamping process; and defect area obtained by the online visual inspection system.
[0026] The process parameters cover the core variables of key processes in cigarette pack cardboard production, such as printing, die-cutting, and hot stamping, ensuring the comprehensiveness of the data sources and enabling the calculated volatility and risk index to truly reflect the complex multi-process production situation.
[0027] Secondly, the present invention provides an intelligent data processing system for cigarette pack cardboard production, employing the following technical solution: The intelligent cigarette pack cardboard production data processing system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the intelligent cigarette pack cardboard production data processing method described above.
[0028] The above-mentioned intelligent cigarette pack card production data processing method generates a computer program, which is stored in a memory so that it can be loaded and executed by a processor. Thus, a system can be made based on the memory and processor for easy use.
[0029] The present invention has the following technical effects: 1. By introducing process composite volatility and long-term degradation index, and constructing adaptive reliability coefficient, it can determine in real time whether the prediction results of the prediction model should be trusted. The final risk index is a dynamic weighted result of the model prediction value and the volatility risk value. It realizes accurate prediction using the model under stable operating conditions and early warning using volatility index under unstable operating conditions, thereby improving the detection robustness in complex production environments.
[0030] 2. This invention quantifies instantaneous stability by introducing a composite process fluctuation metric, quantifies equipment health status by introducing a long-term degradation index combined with a difficulty coefficient, and constructs a coupled adaptive reliability coefficient, which can dynamically adjust the composition weights of the final quality risk index according to the real-time health status of the production line.
[0031] It solves the problem of poor robustness of traditional quality prediction models under non-steady-state conditions, avoids false alarms under steady-state conditions and missed alarms in the early stage of equipment deterioration, realizes adaptive quality monitoring and early warning of cigarette pack cardboard production process, reduces scrap rate and production and operating costs, and improves the accuracy of monitoring results. Attached Figure Description
[0032] Figure 1 This is a flowchart of the intelligent cigarette pack cardboard production data processing method of the present invention. Detailed Implementation
[0033] 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, not all, of the embodiments of the present invention. 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.
[0034] This invention discloses an intelligent data processing method for cigarette pack cardboard production, referring to... Figure 1 This includes the following steps: S1: Obtain the process parameters during the production of cigarette pack cardboard.
[0035] Acquire production data during the cigarette pack cardboard production process. This data includes machine speed, ink viscosity for each color group, real-time color difference sequence, longitudinal / transverse registration error, and drying unit temperature in the printing process; and die-cutting pressure, hot stamping temperature, and hot stamping speed in the die-cutting and hot stamping process. Utilize an online visual inspection system to obtain the defect area on the cigarette pack cardboard surface, such as the area of color spots and scratches.
[0036] The production data is aligned on the time axis using time window aggregation or interpolation methods. Finally, a process data matrix is constructed, where each row represents a point in time and each column represents a type of production data, namely a process parameter.
[0037] S2: Calculation process composite volatility.
[0038] When a cigarette pack cardboard production line is healthy, its parameters are stable. However, a production line system on the verge of failure may still have parameters within a reasonable range. To maintain stable and continuous operation, its control system needs to frequently and significantly adjust process parameters, resulting in drastic parameter fluctuations. To quantify the volatility of process parameters, it is necessary to calculate the composite process volatility. The calculation method is as follows: Using time t as the origin, construct a time window along the historical direction. The length of the time window is manually set according to the actual situation, for example, the length of the time window is the most recent 1 minute. Obtain the process parameters within the time window, and use a linear normalization algorithm to normalize each process parameter within the time window. The expression for the composite process volatility is:
[0039] In the formula, Indicates at time The process of time-related composite volatility; Indicates the total number of process parameter types; Index indicating process parameters; Indicates the first The preset weights of each process parameter are calculated using the entropy weight method. Indicates the first time within the time window Normalized time series of process parameters; It is a sampling point in a time series; This indicates the rate of change of the process parameters at sampling point k; This function calculates the standard deviation within a time window.
[0040] when A high value indicates that one or more process parameters are experiencing drastic fluctuations and frequent adjustments. Even if the process parameters are within the normal range, this initially suggests that the production line is in a high-risk state, possibly due to changes in raw material batches, early wear and tear of equipment, or early equipment failure. A lower value indicates A value close to 0 indicates that the rate of change of process parameters is relatively stable, suggesting a relatively smooth production process.
[0041] S3: Calculate the long-term degradation index.
[0042] If the process is composite volatility A sustained high level of degradation, even when parameters are within the normal range, indicates a slow deterioration of equipment or processes on the production line, such as bearing wear or pipeline blockage. The long-term degradation index is calculated using the process composite volatility formula:
[0043] In the formula, This represents the long-term degradation index at time t; T represents the period of a batch before time t, for example, the period of a batch is 2 hours. Indicates at time The process composite volatility over time; t represents the index of time. Indicates in The difficulty level of producing cigarette pack cardboard at any given time is a value that, in the historical production process, is comparable to... The ratio of registration errors for the same type of cigarette pack paper produced at any given time is specifically calculated as follows: the ratio of vertical registration error to vertical registration length, and the ratio of horizontal registration error to horizontal registration length are calculated, and the maximum of the two ratios is selected as the difficulty coefficient. For example, For the Type A cigarette pack cardboard produced at any given time, in the previous batch of Type A cigarette pack cardboard, the average horizontal registration error and the average vertical registration error in the overprinting process were 1.5% and 2.0%, respectively. The value is 2.0%.
[0044] The difficulty coefficient is introduced to isolate the influence of process difficulty when calculating the long-term degradation index, allowing the index to solely reflect the deterioration of equipment condition. For example, in the production of high-difficulty cigarette packaging paper products, such as complex hot stamping and high-precision overprinting, the tolerance for process composite fluctuations is greater due to the higher process difficulty. In other words, even if the process composite fluctuations are high, it initially indicates that the condition of each piece of equipment on the production line may be normal. Conversely, in simple processes, even small fluctuations in process composite fluctuations initially indicate that the condition of each piece of equipment on the production line may be abnormal.
[0045] In conclusion, a sustained increase in the long-term degradation index is an early sign that equipment is about to fail.
[0046] S4: Calculate the adaptive reliability coefficient of the coupling.
[0047] In the production of cigarette pack cardboard, process parameters are input into a quality prediction model to obtain quality indicators. Quality indicators This represents the defect rate over a future period, and the quality prediction model is an LSTM (Long Short-Term Memory) network model. A higher value indicates that the production line conditions have deviated from the stable state during the training of the prediction model, and the reliability of the prediction results is low. This step calculates a coefficient to quantify this change in reliability.
[0048] S41: Construct a reliability function for the prediction results based on the process composite volatility.
[0049] The expression is:
[0050] In the formula: Indicates at time The reliability of the prediction results is measured in the range of (0,1); This represents the preset sensitivity parameter, used to control... right The sensitivity to change, with a value of 1; Indicates at time The process of time-related composite volatility; The benchmark value representing the composite volatility of a process represents the level of normal volatility; for example, the composite volatility of the process in the previous batch. The 90th percentile, exp() represents the exponential function with base e.
[0051] when Much larger hour, When the value approaches 0, the prediction model is not trusted, meaning the reliability of the prediction results is low; when... much smaller hour, When the value approaches 1, the reliability of the prediction model's results is relatively high. (Reliability) Reflects the composite volatility based on the process Under certain conditions, the degree of trust in the prediction results of the prediction model.
[0052] S42: Construct a health function for the production line system based on the long-term degradation index.
[0053] The expression is:
[0054] In the formula: Indicates at time The health status of the production line system, with a value range of [0,1]. This represents the preset degradation impact factor, which controls... The degree of impact on the health of the production line system; Indicates at time Long-term degradation index; This represents a preset reference degradation threshold, such as the historical long-term degradation index. The 90th percentile.
[0055] when At lower levels, the production line system is more stable. The value approaches 1; when At higher levels, The value approaches 0. System health. Reflecting long-term deterioration The degree of confidence in the model's predictions.
[0056] S43: Calculate the adaptive reliability coefficient after coupling.
[0057] Multiplying the reliability of the prediction results by the health of the production line system yields the adaptive reliability coefficient, expressed as:
[0058] In the formula: Indicates at time The adaptive reliability coefficient of the prediction results; Indicates at time The reliability of the prediction results; Indicates at time The health status of the production line system.
[0059] Adaptive reliability coefficient Its value approaches 1 only when the production line process parameters are stable and the production line system is healthy in the long term.
[0060] For example, the equipment in one process has become severely worn, resulting in... Even if the process parameters of the equipment have a small composite fluctuation at time t, that is Lower Overall reliability The value is 0.
[0061] S5: Calculate the short-term quality risk index.
[0062] Based on the overall reliability coefficient and the process composite volatility, the short-term quality risk index is calculated using the following formula:
[0063] In the formula, Indicates at time Short-term quality risk index at that time; Indicates at time The adaptive reliability coefficient of the prediction results; Indicates at time The quality indicators output by the time-of-use quality prediction model, for example, predicting a 0.5% blemish defect rate for the next minute; It represents the average absolute error of the quality prediction model over the past batch, which is the absolute value of the difference between the predicted quality index and the actual quality index. Indicates at time The time-dependent composite volatility.
[0064] When the process parameters change little and the production line system is healthy At this point, the quality risk index is dominated by the quality prediction model to avoid false alarms under steady-state conditions; however, when the production process parameters fluctuate drastically or deteriorate over a long period, At this point, the confidence level of the prediction results of the quality prediction model is low, and the quality risk index is dominated by the historical error and process composite volatility of the quality prediction model.
[0065] S6: Determine whether the production line system has malfunctioned based on the short-term quality risk index and the long-term degradation index.
[0066] When the short-term quality risk index exceeds the preset risk threshold, it indicates that a quality defect is about to occur and signals a shutdown. For example, the risk threshold is 0.3.
[0067] When the long-term degradation index A continuous growth trend, for example, when two consecutive batches show continuous growth, indicates that the health of the production line system is declining and signals that maintenance is needed.
[0068] This invention also discloses an intelligent tobacco pack paper production data processing system, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, the intelligent tobacco pack paper production data processing method according to this invention is implemented.
[0069] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.
[0070] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.
Claims
1. A method for processing data from cigarette pack cardboard production based on intelligent technology, characterized in that: Including the following steps: The process parameters of the cigarette pack cardboard production process are obtained and input into the preset quality prediction model to obtain the predicted quality indicators. Calculate the composite process variability of process parameters, which characterizes the degree of change of process parameters within a time window; The calculation method for the long-term deterioration index of cigarette pack cardboard includes: calculating the difficulty coefficient of processing the cigarette pack cardboard, which is the ratio of the registration error to the registration length; accumulating the ratio of the process composite volatility to the correction factor determined based on the difficulty coefficient within a batch cycle to obtain the long-term deterioration index; calculating the adaptive reliability coefficient of the quality prediction model, which is the product of the reliability function with respect to the process composite volatility and the health function with respect to the long-term deterioration index; calculating the short-term quality risk index, which is positively correlated with the product of the adaptive reliability coefficient and the quality indicator; when the short-term quality risk index exceeds the preset risk threshold, it indicates that a quality defect is about to occur, and a shutdown signal is issued.
2. The intelligent cigarette pack cardboard production data processing method according to claim 1, characterized in that, The calculation method for process composite volatility includes: normalizing the process parameters within the time window; calculating the standard deviation of the rate of change of the normalized process parameters; and weighting and summing all the standard deviations to obtain the process composite volatility.
3. The intelligent cigarette pack cardboard production data processing method according to claim 1, characterized in that, The reliability function is: In the formula: Indicates at time The reliability of the prediction results; This indicates the preset sensitivity parameter; Indicates at time The process of time-related composite volatility; The base value represents the composite volatility of the process, and exp() represents an exponential function with base e.
4. The intelligent cigarette pack cardboard production data processing method according to claim 1, characterized in that, The health function is: In the formula: Indicates at time The health status of the production line system, with a value range of [0,1]. This represents the preset degradation impact factor; Indicates at time Long-term degradation index; This indicates the preset reference degradation threshold.
5. The intelligent cigarette pack cardboard production data processing method according to claim 4, characterized in that, The reference degradation threshold is the historical long-term degradation index. The 90th percentile.
6. The intelligent cigarette pack cardboard production data processing method according to claim 1, characterized in that, Production data processing methods also include issuing maintenance signals in response to the continuous increase of the long-term degradation index across multiple production batches.
7. The intelligent cigarette pack cardboard production data processing method according to claim 1, characterized in that, The formula for calculating the short-term quality risk index is: In the formula, Indicates at time Short-term quality risk index at that time; Indicates at time The adaptive reliability coefficient of the prediction results; Indicates at time The quality indicators output by the time-of-use quality prediction model; This represents the average absolute error of the quality prediction model over the past batch. Indicates at time The time-dependent composite volatility.
8. The intelligent cigarette pack cardboard production data processing method according to claim 3, characterized in that, The quality prediction model is an LSTM (Long Short-Term Memory) network model.
9. The intelligent cigarette pack cardboard production data processing method according to claim 1, characterized in that, The process parameters include at least one of the following: printing speed, ink viscosity, color difference sequence, registration error, and drying temperature in the printing process; die-cutting pressure, hot stamping temperature, and hot stamping speed in the die-cutting and hot stamping process; and defect area obtained by the online visual inspection system.
10. A data processing system for intelligent cigarette pack cardboard production, characterized in that: include: The processor and memory, wherein the memory stores computer program instructions, which, when executed by the processor, implement the intelligent cigarette pack cardboard production data processing method according to any one of claims 1 to 9.