Cigarette process quality evaluation method and device based on structural excellent value
By using a structurally optimal model and the expectation-maximization algorithm, the problems of high fitting bias and misjudgment rate of traditional models in cigarette quality assessment are solved, thus achieving accurate assessment and improved adaptability of cigarette quality.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-08
- Publication Date
- 2026-03-10
AI Technical Summary
Traditional statistical models cannot accurately reflect the process quality assessment needs of the cigarette industry under multiple indicators, multiple deduction levels, and cross-brand and cross-time dimensions. They suffer from problems such as extreme value fitting bias, high misjudgment rate of abnormal data, and poor project adaptability.
A structural optimality model is adopted, and the quality inspection data is iteratively solved using the expectation-maximization algorithm. The structural optimality probability and the intensity of natural defects are separated, and the number of quality stability optimality values is calculated to evaluate the quality of cigarettes.
It achieves accurate fitting of extreme value data, reasonably distinguishes between reasonable extreme values and true outliers, improves the explanatory power and adaptability of cigarette quality assessment, and meets the assessment needs of different project scenarios.
Smart Images
Figure CN121639028A_ABST
Abstract
Description
Technical Field
[0001] The embodiments in this specification pertain to the field of process quality evaluation, and specifically relate to a method and apparatus for evaluating the process quality of cigarettes based on structural optimality values. Background Technology
[0002] Cigarette manufacturing accumulates multi-source, heterogeneous, and hierarchically nested inspection data through factory inspection, process monitoring, and self-inspection. Defect indicators are recorded in the form of counts or discrete deductions. However, actual data commonly exhibits extreme values (e.g., defective cigarette count = 0, score = 100) that are far higher than predicted by classic Poisson / normal models, constituting extreme value inflation. In this type of data, extreme values (i.e., values exceeding the normal distribution range) not only occur more frequently than predicted by traditional statistical models (such as normal distribution models and ordinary extreme value models), but their distribution characteristics also differ significantly from conventional data, directly leading to the following technical defects in traditional models:
[0003] Large fitting bias for extreme values: Traditional extreme value models, such as the Generalized Pareto Distribution Model (GPD) or the Generalized Extreme Value Distribution Model (GEV), assume that extreme values follow a single tail distribution, which cannot adapt to the scenario of "dense occurrence of extreme values". This results in an estimation error of more than 30% for the probability density of extreme values, and fails to accurately reflect the true extreme characteristics of the data.
[0004] High misjudgment rate of abnormal data: Traditional models uniformly classify data that "exceeds the normal threshold" as outliers, without distinguishing between "reasonable extreme values" and "true outliers". In actual projects, the misjudgment rate is generally higher than 25%, which seriously affects the accuracy of decision-making.
[0005] Poor project adaptability: Existing models are mostly designed for single scenarios and lack adaptability adjustments for different project data characteristics. When applied across scenarios, a large number of parameters need to be readjusted, which is time-consuming and inefficient, and cannot meet the requirements of "efficiency" and "targeting" of data processing in project phase reports.
[0006] In summary, this phenomenon leads to fitting bias and distortion of quality capability assessment in traditional models, making it difficult to meet the process quality assessment needs of the cigarette industry under multiple indicators, multiple deduction levels, and across brands and time dimensions. Summary of the Invention
[0007] The embodiments of this disclosure provide a method and apparatus for evaluating the quality of cigarette manufacturing processes based on structural optimality values, aiming to solve one or more of the above-mentioned problems and other potential problems.
[0008] According to a first aspect of this disclosure, a method for assessing the quality of cigarette production processes based on structural optimality values is provided. The method includes: acquiring quality inspection data of the target cigarette to be assessed, whereby the quality inspection data characterizes the number of defective cigarettes in the tested samples; constructing a structural optimality model, and iteratively solving the structural optimality model using an expectation-maximization algorithm based on the quality inspection data to obtain the structural optimality probability and natural defect intensity parameters corresponding to the structural optimality model, whereby the structural optimality model characterizes the probability of quality defect counts occurring under the assumption that zero-defect samples originate from structural zero defects and natural zero defects; and calculating a quality stability optimality number based on the structural optimality probability and natural defect intensity parameters, to assess the quality of the target cigarette based on the quality stability optimality number, where the quality stability optimality number represents the overall probability of observing zero defects.
[0009] According to a second aspect of this disclosure, a cigarette process quality assessment device based on structural optimality values is provided. The device includes a data acquisition module configured to acquire quality inspection data of the target cigarette to be assessed, the quality inspection data being used to characterize the number of defective cigarettes in the test sample; a model solving module configured to construct a structural optimality model and, based on the quality inspection data, iteratively solve the structural optimality model using an expectation-maximization algorithm to obtain the structural optimality probability and natural defect intensity parameters corresponding to the structural optimality model, the structural optimality model being used to characterize the probability of quality defect counts occurring under the assumption that zero-defect samples originate from structural zero defects and natural zero defects; and a quality assessment module configured to calculate a quality stability optimality number based on the structural optimality probability and natural defect intensity parameters, to assess the quality of the target cigarette based on the quality stability optimality number, the quality stability optimality number being the overall probability of observing zero defects.
[0010] According to a third aspect of this disclosure, an electronic device is provided, including one or more processors and a memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform a method provided according to a first scheme.
[0011] According to a fourth aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements the method provided according to the first aspect.
[0012] The solution provided in the embodiments of this specification can display the separation of structural optimality probability and natural defect intensity through a structural optimality model, and calculate the number of quality stability optimal values to evaluate the quality of target cigarettes. This overcomes the shortcomings of traditional statistical models, such as low accuracy in handling extreme values, high misjudgment rate of outlier data, and poor project adaptability. It improves the interpretability and fitting degree of optimality set data, achieves accurate fitting of extreme value data, and effectively distinguishes between reasonable extreme values and true outliers, thus adapting to the cigarette quality evaluation needs of different project scenarios. Attached Figure Description
[0013] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein:
[0014] Figure 1 A flowchart illustrating a cigarette process quality assessment method based on structural excellence values, according to some embodiments of this disclosure, is shown.
[0015] Figure 2 The diagram illustrates the data distribution of structural extremum probabilities and natural defect strength parameters across different brands of cigarettes, representing some embodiments of this disclosure.
[0016] Figure 3 This diagram illustrates SOI trend analysis of different brands of cigarettes according to some embodiments of the present disclosure.
[0017] Figure 4 A schematic diagram of the structure of a cigarette process quality assessment device based on structural excellence values, according to some embodiments of the present disclosure, is shown.
[0018] Figure 5 A schematic block diagram of an electronic device according to some embodiments of the present disclosure is shown. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.
[0020] The terms “comprising” and “having”, and any variations thereof, in this specification, claims, and the foregoing drawings are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the steps or units listed, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. Depending on the context, the word “if” as it applies herein may be interpreted as “when”, “in response to determination”, or “in response to detection”.
[0021] Figure 1 A flowchart illustrating a cigarette process quality assessment method 100 based on structural optimality values, representing some embodiments of this disclosure, is shown. Method 100 can be executed, for example, by a terminal, which may include, but is not limited to, mobile phones, tablets, desktop computers, servers, etc. Figure 1 As shown, in method 100, step 102 can obtain the quality inspection data of the target cigarette to be evaluated, which is used to characterize the number of defective cigarettes in the test sample.
[0022] In this embodiment, for the target cigarettes that need to be quality assessed, the corresponding quality inspection data will first be obtained. The quality inspection data can be the test results of the test samples after sampling from the target cigarettes, based on the quality indicators, and can also include the number of defective cigarettes in the test samples (i.e., the number of defective cigarettes as indicated by the test results). Subsequently, the quality inspection status of each defect will be mainly determined by the number of defective cigarettes.
[0023] In method 100, step 104 can construct a structurally optimal model and, based on the quality inspection data, iteratively solve the structurally optimal model using the expectation-maximization algorithm to obtain the structurally optimal probability and natural defect strength parameters corresponding to the structurally optimal model. The structurally optimal model is used to characterize the probability of quality defect counts occurring under the assumption that zero-defect samples originate from structurally zero defects and natural zero defects.
[0024] In this embodiment, the formation process of cigarette physical quality data involves multiple factors such as raw materials, equipment, processes, and environment. The test results often exhibit obvious characteristics of "full score concentration" and "defect sparsity." Traditional statistical methods (such as normal or Poisson models) struggle to simultaneously characterize both "high proportion of full-score samples" and "low-frequency discrete defects," resulting in insufficient explanatory power for process stability. Therefore, this application introduces the concept of the Extremum Inflation Model (EIM) based on existing process quality statistical methods to structurally model cigarette physical quality defect data. The Extremum Inflation Model is a heterogeneous mixed distribution model, derived from Zero-Inflated Poisson (ZIP) and its improved forms. Its core idea is that the "zero-defect" portion in the observed samples not only comes from the random process (the natural zero of the Poisson distribution) but also includes a portion of "structural zero values" caused by the inherent goodness of the production process. By decomposing these two types of zero samples, we can simultaneously characterize the structural stability and random fluctuations of the process, thereby achieving a comprehensive evaluation and trend prediction of the quality of the roll packaging.
[0025] Specifically, in the inspection data of the coiling and packaging process, the defect count exhibits a typical zero-inflation characteristic, with most batches showing no defects (full marks), while a few batches have a small number of defects. Traditional Poisson models can only describe random defect events and cannot explain the excessive concentration of "zero-defect" samples. However, the structurally optimal model in this application assumes that zero-defect samples originate from two mechanisms: structurally zero defects (Structural Full Score), i.e., the inherent high-quality stable state of the production process, whose corresponding structurally optimal probability is denoted as... Natural zero defect intensity refers to the defect occurrence process under random fluctuations, and its corresponding natural defect intensity parameter is denoted as... Therefore, according to the structural optimality model, the probability distribution of the quality defect count Y can be expressed as:
[0026]
[0027] in, It represents the structural extreme value probability, describing the proportion of samples that naturally reach an extreme state due to process stability under current production conditions; The parameter represents the intensity of natural defects and describes the average frequency of defect occurrence under random fluctuations; together, they characterize the structural stability and fluctuation characteristics of process quality.
[0028] Since the likelihood function of the structurally optimal model is nonlinear, it cannot be solved directly analytically. To obtain robust parameter estimation results, an Expectation-Maximization (EM) algorithm is used for iterative solution to obtain the structural extremum probability. and natural defect strength parameters .
[0029] In method 100, step 106 can calculate the number of quality stability optimal values based on the structural extreme value probability and the natural defect intensity parameter, so as to evaluate the quality of the target cigarette according to the number of quality stability optimal values, where the number of quality stability optimal values is the overall probability of observing zero defects.
[0030] In this embodiment, the Stability of Optimality Index (SOI), i.e., the overall probability of observing zero defects, can be calculated based on the structural extremum probability and the natural defect strength parameter. The calculation formula is as follows:
[0031]
[0032] The quality stability optimality (SOI) value can be used to evaluate the quality of a target cigarette, measuring its overall performance in terms of structural goodness and fluctuation stability. The higher the SOI, the more stable the quality and the closer the process is to a zero-defect state.
[0033] In one possible implementation, after obtaining quality inspection data of cigarettes during the manufacturing process, the method further includes:
[0034] The quality inspection data is preprocessed, including field standardization, outlier removal, and validity verification.
[0035] In this embodiment, the quality inspection data first needs to be preprocessed. Field unification is used to standardize heterogeneous data from different production lines, batches, periods, or systems into a consistent format and definition, making it integrateable and comparable. Specifically, this may include unifying the same indicator with different expressions (e.g., unifying circumference, perimeter, and perimeter as circumference) and converting all physical quantities to standard units. Outlier removal involves identifying and removing outlier data points that do not reflect the true process status due to measurement errors, equipment malfunctions, non-standard operations, etc. Specifically, this can be done by using hard boundaries of data ranges set based on industry knowledge and technical standards, or by using methods such as box plots, Z-scores, and median absolute deviation to identify and remove outliers. Legality verification may include verifying whether the data source equipment is certified and legitimate, and verifying according to business rules (e.g., the inspection time cannot be earlier than the start time of the production batch). Subsequent processing will be based on the preprocessed quality inspection data.
[0036] In one possible implementation, the quality defect indicators of the test sample include hardness, length, other physical indicators, and appearance indicators.
[0037] In this embodiment, the present application may cover 22 types of quality defects as shown in the table below:
[0038]
[0039] These include six physical indicators (single cigarette weight, length, circumference, hardness, draw resistance, and total ventilation rate) and 16 appearance indicators (such as uneven cigarette overlap, color difference in the stamp, staining of the cigarette paper, misalignment of the small box seal, and wrinkles in the transparent paper of the box). All defects are recorded in the form of "number of defective cigarettes". However, because the physical indicators and appearance indicators are stored in different data tables and the detection frequency differs, the number of records for the two types of data is not completely consistent, and there is no corresponding relationship. In addition, the length and hardness of the physical indicators rely on independent testing equipment, and the detection frequency is significantly lower than that of the other items, resulting in differences in the number of test samples. On the other hand, appearance inspection is usually completed manually on-site or in the laboratory, and the frequency is significantly different from that of physical indicator inspection. Therefore, the physical indicator data are finally distinguished according to length, hardness, and other physical indicators, while the appearance indicators are grouped separately, forming four categories of quality defect indicators.
[0040] In one possible implementation, a structurally optimal model is constructed, and based on quality inspection data, the structurally optimal model is iteratively solved using an expectation-maximization algorithm to obtain the structural extremum probability and natural defect intensity parameters corresponding to the structurally optimal model, including:
[0041] Construct a structurally optimal model, which is a joint model composed of sub-models corresponding to each quality defect index;
[0042] Based on the quality inspection data, the expectation-maximization algorithm is used to iteratively solve each sub-model to obtain the first probability and first parameter for each sub-model. The first probability is the probability that the sub-model has zero structural defects, and the first parameter is the strength parameter when the sub-model has zero natural defects; and
[0043] The first probability and first parameter of each sub-model are weighted hierarchically to obtain the structural extreme value probability and natural defect intensity parameter corresponding to the structurally optimal model.
[0044] In this embodiment, four types of quality defect indicators are each constructed into a sub-model for separate processing, and the joint model composed of the four sub-models serves as the overall structural optimal model. Thus, each sub-model calculates its corresponding structural extreme value probability and natural defect intensity parameter, i.e., the first probability and the first parameter, based on the quality inspection data under the corresponding quality defect indicator. To improve the representativeness of the indicators and their ability to characterize global trends, a hierarchical weighting method is used for each first probability. and the first parameter Perform weighted fusion, that is:
[0045]
[0046] in, For structural extrema probability; Here, is the natural defect strength parameter, and i represents the different pre-defined deduction levels. When there are multiple deduction levels, the calculation will be performed separately for each deduction level. This represents the percentage of valid test samples in the k-th group of test items; k is used to represent different quality defect index groups.
[0047] In one possible implementation, the natural defect strength parameter includes each second parameter based on the deduction level division based on the number of defect branches, and each second parameter is obtained by hierarchical weighting based on each first parameter under the corresponding deduction level.
[0048] In this embodiment, different deduction levels (e.g., 0.2 points, 0.5 points, 1.0 points) are set to further divide the number of defective items according to the different degrees of defects. The total number of defective items is divided into the number of deductions for the corresponding deduction level according to the deduction level, and the second parameter under each deduction level is calculated in this way.
[0049] As an example, taking cigarettes of brand A and brand B from February 2025 to May 2025 as examples, the corresponding experimental data are shown in the table below:
[0050]
[0051] The following table shows the fitting results of the structural excellence probabilities corresponding to length, hardness, other physical indicators, and appearance indicators with the natural defect strengths for three deduction levels: 0.2, 0.5, and 1.0:
[0052]
[0053] Based on the model estimation results for various indicators of grade A, length and other physical indicators exhibit strong structural optimality characteristics. Specifically, the structural optimality probability of the length indicator in each month... And the corresponding Poisson strength parameter The value is at a moderate level of 0.98–2.82, indicating that the best samples for this quality indicator represent stable outputs dominated by structural mechanisms. Other physical indicators... The values fluctuated between 0.64 and 0.92, exhibiting certain structurally optimal characteristics, and , and The parameters were all less than 0.5 between February and May 2025, further indicating that the defect control level was dominated by structural factors. In contrast, the structural superiority of the hardness parameter was relatively limited. The values remained only within the range of 0.42–0.66, indicating that only a portion of the excellent samples likely originated from structural control, while the other portion was still a product of the naturally low defect rate under the Poisson process. (Appearance index item) The value being close to 0 for a long period indicates that this type of defect control does not exhibit significant structural characteristics, and the best samples mainly come from the fluctuations in the natural defect rate.
[0054] Compared to grade A, grade B exhibits greater fluctuations in its structurally superior characteristics across various indicators. For example, in terms of hardness, grade B... The value fluctuates between 0.41 and 0.87, corresponding to The value is between 0.8 and 3.05, much higher than grade A, indicating significant fluctuations in hardness. Regarding length and other physical properties... The value remained at a high level, but , and They exhibited significant fluctuations, indicating some structural support, but also accompanied by considerable volatility. Regarding appearance indicators, The value is between 0.5 and 1.0. Compared with the near-zero level of Grade A, some of its excellent samples may have a certain structural origin, and the appearance quality is better than that of Grade A.
[0055] After weighted fusion using a hierarchical weighting method, the structurally optimal parameters are obtained. and natural defect strength parameters The distribution is as follows Figure 2 As shown. From the structural optimal parameters Looking at the changing trends, grade B consistently outperformed grade A throughout the entire period, indicating a more stable "structural superb support," which is more conducive to achieving defect-free results. (Regarding the natural defect strength parameters...) Grade B showed significant fluctuations in deduction scores between 0.5 and 1.0, indicating that while its structure is excellent, it suffers from periodic quality instability in terms of medium-to-high strength defects. In contrast, Grade A... The fact that it remains at a high level indicates that its frequency of minor defects is significantly higher than that of grade B, reflecting insufficient control over defects with low deduction scores.
[0056] The trend chart of the SOI (Supreme Excellence Indices) for each grade is as follows: Figure 3 As shown. SOI integrates structurally optimal parameters. and natural defect strength parameters The results showed that grade B had a higher SOI in each month than grade A, indicating that its products had a higher overall probability of achieving a "perfect score / no defects" status in actual testing.
[0057] In one possible implementation, the structurally optimal model is solved iteratively using an expectation-maximization algorithm, including:
[0058] We set up latent variables to characterize whether any sample originates from a structured full distribution or a Poisson distribution, and then alternately perform expectation step calculation and maximization step calculation in the iterative solution based on the latent variables until the preset convergence condition is met.
[0059] In this embodiment, the specific calculation process of the expectation-maximization algorithm is as follows: introducing latent variables. , indicating whether the i-th sample originates from a structured full-score distribution ( or Poisson distribution The expected step (E-step) and the maximum step (M-step) are executed alternately in each iteration until convergence.
[0060] The expectation step is used to calculate the posterior probability of a perfect score in the structure, and the formula is:
[0061]
[0062] The maximization step is used to update parameters, and its calculation formula is:
[0063]
[0064] Where n is the total number of samples, and is the actual count value of the i-th sample.
[0065] Convergence conditions can be set, for example, when the parameter increment... and Or the change in the log-likelihood function When the algorithm converges, it is considered to have converged. , It can be preset. In addition, to ensure the comparability of estimation results between different machines and different testing batches, the model can also use a weighted method based on sample size (N) to correct parameters, so as to reduce the bias caused by the difference in the number of tests.
[0066] In one possible implementation, the method further includes:
[0067] The Kolmogorov-Smirnov test was used to verify the consistency of the structurally optimal model.
[0068] In this embodiment, after determining the structural extreme value probability and natural defect strength parameters of the structurally optimal model, the model can also simulate deductions based on the probability of quality defect counts. For example, it can determine the probability corresponding to each deduction level and then weight the deduction values for each level according to the probability. To verify the consistency between the simulated and actual deduction values in terms of distribution, the Kolmogorov-Smirnov test can be used to test the consistency of the cumulative distribution function of the simulated and actual data. The consistency between the simulated and actual deduction values in terms of mean, standard deviation, and distribution can also be determined. Taking products of grades A and B from February 2025 to May 2025 as examples, the comparative analysis of the two is shown in the table below.
[0069]
[0070] The results show that the differences between the two in terms of mean and standard deviation are small. In particular, the actual deduction mean and the simulated mean remain within ±0.02 in each month, indicating that the model has a good fit to the overall deviation trend. From the KS test statistic and p-value results, the p-values for most samples are greater than 0.77, and some reach 1.00, indicating that the deduction distribution generated by the model is not significantly different from the actual data distribution, verifying that the proposed optimal inflation model has a good distribution fitting ability in terms of overall quality performance.
[0071] Figure 4 The diagram illustrates the structure of a cigarette process quality assessment device 400 based on structural optimality values according to some embodiments of this disclosure. The various embodiments in this specification are described in a progressive manner; similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments. Figure 4As shown, the device 400 includes a data acquisition module 401, configured to acquire quality inspection data of the target cigarette to be evaluated, the quality inspection data being used to characterize the number of defective cigarettes in the test sample; a model solving module 402, configured to construct a structural optimal model, and based on the quality inspection data, iteratively solve the structural optimal model using an expectation-maximization algorithm to obtain the structural extreme value probability and natural defect intensity parameters corresponding to the structural optimal model, the structural optimal model being used to characterize the probability of quality defect count occurring under the assumption that zero-defect samples originate from structural zero defects and natural zero defects; and a quality assessment module 403, configured to calculate the quality stability optimal value number based on the structural extreme value probability and natural defect intensity parameters, to assess the quality of the target cigarette based on the quality stability optimal value number, the quality stability optimal value number being the overall probability of observing zero defects.
[0072] In one possible implementation, the data acquisition module 401 is further configured to preprocess the quality inspection data, including field unification, outlier removal, and validity verification.
[0073] In one possible implementation, the quality defect indicators of the test sample include hardness, length, other physical indicators, and appearance indicators.
[0074] In one implementation, the model solving module 402 is further configured to construct a structurally optimal model, which is a joint model composed of sub-models corresponding to each quality defect index; based on the quality inspection data, the expectation-maximization algorithm is used to iteratively solve each sub-model to obtain the first probability and first parameter corresponding to each sub-model, where the first probability is the probability of the sub-model having zero structural defects, and the first parameter is the intensity parameter when the sub-model has zero natural defects; and the first probability and first parameter of each sub-model are respectively weighted hierarchically to obtain the structural extreme value probability and natural defect intensity parameter corresponding to the structurally optimal model.
[0075] In one possible implementation, the natural defect strength parameter includes each second parameter based on the deduction level division based on the number of defect branches, and each second parameter is obtained by hierarchical weighting based on each first parameter under the corresponding deduction level.
[0076] In one possible implementation, the model solving module 402 is further configured to characterize whether any sample originates from a latent variable of a structured full-score distribution or a Poisson distribution, so as to alternately perform expectation step calculation and maximization step calculation in the iterative solution according to the latent variable until a preset convergence condition is met.
[0077] In one possible implementation, the apparatus further includes a consistency verification module configured to perform consistency verification on the structurally optimal model based on the Kolmogorov-Smirnov test.
[0078] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this specification is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).
[0079] Figure 5 A block diagram of an electronic device 500 that can implement various embodiments of the present disclosure is shown. For example... Figure 5 As shown, the electronic device 500 includes a processor 510, a disk drive 520, an input / output interface 530, a network interface 540, and a memory 550. The processor 510, disk drive 520, input / output interface 530, network interface 540, and memory 550 can communicate with each other via a communication bus 560.
[0080] The processor 510 can be implemented using a general-purpose CPU, microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs and implement the technical solution provided in this application.
[0081] The memory 550 can be implemented in the form of ROM (Read Only Memory), RAM (Read Access Memory), static memory, dynamic storage devices, etc. The memory 550 can store the operating system 551 used to control the operation of the electronic device 500, and the basic input / output system (BIOS) 552 used to control the low-level operations of the electronic device 500. Additionally, it can store a web browser 553, a data storage management system 554, etc. In summary, when implementing the technical solution provided in this application through software or firmware, the relevant program code is stored in the memory 550 and is called and executed by the processor 510.
[0082] Input / output interface 530 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.
[0083] Network interface 540 is used to connect a communication module (not shown in the figure) to enable communication and interaction between the device and other devices. The communication module can communicate via wired means (e.g., USB, Ethernet cable) or wireless means (e.g., mobile network, Wi-Fi, Bluetooth).
[0084] Bus 560 includes a pathway for transmitting information between various components of the device, such as processor 510, disk drive 520, input / output interface 530, network interface 540, and memory 550.
[0085] It should be noted that although the above-described device only shows the processor 510, disk drive 520, input / output interface 530, network interface 540, memory 550, bus 560, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the method of this application, and does not necessarily include all the components shown in the figures.
[0086] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.
[0087] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing. Furthermore, although operations are depicted in a specific order, this should be understood as requiring that such operations be performed in the specific order shown or in sequential order, or requiring that all illustrated operations be performed to achieve the desired result. In certain environments, multitasking and parallel processing may be advantageous. Similarly, while several specific implementation details are included in the foregoing discussion, these should not be construed as limiting the scope of this disclosure. Certain features described in the context of individual embodiments may also be implemented in combination in a single implementation. Conversely, various features described in the context of a single implementation may also be implemented individually or in any suitable sub-combination in multiple implementations.
[0088] Although the subject matter has been described using language specific to structural features and / or methodological logic, it should be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or actions described above. Rather, the specific features and actions described above are merely illustrative examples of implementing the claims.
Claims
1. A method for evaluating the quality of a cigarette process based on a structural extreme value, characterized by, The method comprises: obtaining quality inspection data of a target cigarette to be evaluated, the quality inspection data being used to represent a defect count in a detection sample; constructing a structural extreme value model, and iteratively solving the structural extreme value model by an expectation maximization algorithm according to the quality inspection data to obtain a structural extreme value probability and a natural defect intensity parameter corresponding to the structural extreme value model, the structural extreme value model being used to represent a probability of a quality defect count under an assumption that a zero defect sample is derived from structural zero defects and natural zero defects; and based on the structural extreme value probability and the natural defect intensity parameter, calculating a quality stability extreme value number to evaluate a quality of the target cigarette according to the quality stability extreme value number, the quality stability extreme value number being an overall probability of an observed zero defect.
2. The method according to claim 1, wherein, After the quality inspection data of the cigarette in the manufacturing process is obtained, the method further comprises: preprocessing the quality inspection data, the preprocessing comprising field unification, abnormal value elimination, and legality verification.
3. The method according to claim 1, wherein, The quality defect indicators of the detection sample comprise hardness, length, other physical indicators, and appearance indicators.
4. The method according to claim 3, wherein, The construction of the structural extreme value model and the iterative solving of the structural extreme value model by the expectation maximization algorithm to obtain the structural extreme value probability and the natural defect intensity parameter corresponding to the structural extreme value model comprise: constructing a structural extreme value model, the structural extreme value model being a joint model composed of sub-models corresponding to each quality defect indicator; iteratively solving each sub-model by an expectation maximization algorithm according to the quality inspection data to obtain a first probability and a first parameter corresponding to each sub-model, the first probability being a probability of a structural zero defect of a sub-model, and the first parameter being an intensity parameter of a natural zero defect of a sub-model; and layered weighting the first probability and the first parameter of each sub-model respectively to obtain the structural extreme value probability and the natural defect intensity parameter corresponding to the structural extreme value model.
5. The method according to claim 4, wherein, The natural defect intensity parameter comprises second parameters based on defect count grade division, each second parameter being obtained by layered weighting of each first parameter under a corresponding defect count grade.
6. The method according to claim 1, wherein, The iterative solving of the structural extreme value model by the expectation maximization algorithm comprises: setting a hidden variable used to represent whether a sample is derived from a structural full score distribution or a Poisson distribution to alternately perform an expectation step calculation and a maximization step calculation in the iterative solving according to the hidden variable until a preset convergence condition is met.
7. The method according to claim 1, wherein, The method further comprises: performing consistency testing on the structural extreme value model based on Kolmogorov-Smirnov testing.
8. A cigarette process quality evaluation device based on structural extreme value, characterized in that, The device comprises: a data acquisition module configured to obtain quality inspection data of a target cigarette to be evaluated, the quality inspection data being used to represent a defect count in a detection sample; a model solving module configured to construct a structural MLE model, and to perform iterative solving on the structural MLE model by an expectation-maximization algorithm according to the quality inspection data, to obtain a structural MLE probability and a natural defect intensity parameter corresponding to the structural MLE model, the structural MLE model being used to represent a probability of occurrence of a quality defect count under a condition that a zero defect sample is assumed to be derived from a structural zero defect and a natural zero defect; and a quality evaluation module configured to calculate a quality stability MLE value number based on the structural MLE probability and the natural defect intensity parameter, to evaluate a quality of the target cigarette according to the quality stability MLE value number, the quality stability MLE value number being an overall probability of observation of a zero defect.
9. An electronic device, comprising: comprising: one or more processors, and a memory associated with the one or more processors, the memory for storing program instructions, which, when read and executed by the one or more processors, perform the steps of the method for evaluating a quality of a cigarette process based on a structural MLE value according to any one of claims 1-7.
10. Computer program product, characterized in that, a computer program, which, when executed by a processor, implements the method for evaluating a quality of a cigarette process based on a structural MLE value according to any one of claims 1-7.