Strip tobacco processing quality dynamic evaluation and abnormal tracing method and system and storage medium

Abstract

The present application relates to the technical field of strip tobacco quality control, in particular to a method and system for dynamic evaluation and abnormal tracing of strip tobacco processing quality consistency and a storage medium, comprising: obtaining multi-dimensional quality index monitoring data and constructing an original index detection vector; constructing a quality consistency baseline based on the original index detection vector; obtaining a dynamic quality consistency evaluation based on the original index detection vector and the quality consistency baseline; performing quality consistency difference analysis on the original index detection vector to obtain abnormal shift and date positioning; and calculating the contribution rate of each quality index based on the dynamic quality consistency evaluation to obtain abnormal sources. Through the above technical solution, the present application achieves dynamic and multi-dimensional quality consistency evaluation and accurately traces abnormal time periods and shifts, thereby improving the quality control level of the strip tobacco processing process.

Description

Technical Field

[0001] This invention relates to the field of tobacco leaf quality control technology, specifically to a dynamic evaluation and anomaly tracing method, system, and storage medium for tobacco leaf processing quality consistency. Background Technology

[0002] After initial curing, tobacco leaves need to undergo threshing and re-drying to remove stems and balance moisture content, resulting in finished tobacco sheets suitable for use in cigarette production. In the threshing and re-drying process, the final quality of the tobacco sheet module is the core factor in evaluating the overall quality of the re-drying process and plays a crucial role in maintaining stable cigarette quality. Currently, the evaluation of the tobacco sheet module quality is mostly post-processing, meaning that after all module processing is completed, statistical measures such as the mean and standard deviation of all tested samples in that batch are calculated to macroscopically evaluate its overall quality level.

[0003] Currently, in evaluating the consistency of tobacco leaf quality, the industry generally uses the nicotine coefficient of variation (CV) to measure the quality consistency of a particular batch of processed tobacco leaves. However, the balance and coordination of carbohydrates (total sugars, reducing sugars) and nitrogenous compounds (total alkaloids, total nitrogen) in the conventional chemical components of tobacco are important material bases for the differences in tobacco leaf aroma and sensory quality. Existing methods for evaluating tobacco leaf quality typically only calculate the nicotine coefficient of variation, resulting in a single dimension for quality consistency evaluation. This fails to comprehensively utilize multi-dimensional quality indicators to construct a comprehensive process consistency evaluation model, leading to insufficient comprehensiveness in the evaluation results. Furthermore, evaluating the stability of all processed tobacco leaves lacks dynamic process evaluation, failing to reflect the fluctuation trend of quality indicators over continuous processing time, and making it difficult to identify specific potential quality problems at specific times or shifts. Moreover, the location and tracking of abnormal quality are not precise enough. When overall quality is found to be poor, it is impossible to quickly and accurately pinpoint which production shift, day, or even production interval caused the quality abnormality, hindering rapid tracing of process problems and timely adjustment of process parameters.

[0004] Therefore, there is an urgent need for a dynamic evaluation and anomaly tracing method and system for the consistency of tobacco processing quality, to achieve dynamic, visualized, and multi-dimensional quality consistency evaluation, and to accurately trace abnormal periods and shifts, so as to improve the quality control level of tobacco processing. Summary of the Invention

[0005] The purpose of this invention is to provide a dynamic evaluation and anomaly tracing method and system for the consistency of tobacco processing quality, in order to solve the problem that the existing technology, which only uses the nicotine variation coefficient for single-dimensional static evaluation, results in unclear quality fluctuation trends and makes it difficult to accurately locate and trace anomalies.

[0006] To achieve the above objectives, embodiments of the present invention provide a dynamic evaluation and anomaly tracing method for the consistency of tobacco processing quality, including: Acquire multidimensional quality indicator monitoring data and construct original indicator detection vectors; Based on the original indicator detection vector, a quality consistency baseline is constructed; Based on the original indicator detection vector and the quality consistency baseline, a dynamic quality consistency evaluation is obtained; A quality consistency difference analysis is performed on the original indicator detection vector to obtain the location of abnormal shifts and dates; Based on the dynamic quality consistency evaluation, the contribution rate of each quality indicator is calculated to identify the source of anomalies.

[0007] Optionally, constructing a quality consistency baseline based on the original indicator detection vector includes: Calculate the overall mean and overall standard deviation of each quality indicator in the original indicator detection vector to construct a quality evaluation benchmark; The original indicator detection vector is standardized, and the mean of each quality indicator is calculated to construct a comprehensive quality center.

[0008] Optionally, obtaining a dynamic quality consistency evaluation based on the original indicator detection vector and the quality consistency baseline includes: Draw a single-value-moving range control chart for each quality indicator in the original indicator detection vector; An abnormal situation is determined when a single value or moving range exceeds the preset upper or lower limit.

[0009] Optionally, obtaining a dynamic quality consistency evaluation based on the original indicator detection vector and the quality consistency baseline includes: Calculate the process capability index of each quality indicator in the original indicator detection vector within the sliding window, and plot the trend graph of the process capability index of each quality indicator. When the process capability index is lower than a preset threshold or If the decline continues over a period of time, it is considered an abnormal situation.

[0010] Optionally, obtaining a dynamic quality consistency evaluation based on the original indicator detection vector and the quality consistency baseline includes: Calculate the squared Mahalanobis distance of each sample relative to the quality consistency baseline, and plot the trend of the squared Mahalanobis distance; The early warning threshold and the comprehensive anomaly threshold are obtained based on the chi-square distribution quantiles; Based on the warning threshold and the comprehensive anomaly threshold, anomaly determination rules are constructed, and warning and comprehensive anomaly samples are obtained.

[0011] Optionally, based on the warning threshold and the comprehensive anomaly threshold, anomaly determination rules are constructed, and warning and comprehensive anomaly samples are obtained, including: Anomaly detection rules are constructed using formula (1). (1) in, The squared Mahalanobis distance, To determine the comprehensive anomaly threshold, This is the warning threshold.

[0012] Optionally, performing a quality consistency difference analysis on the original indicator detection vector to obtain the location of abnormal shifts and dates includes: Using the processing shift as the grouping factor, a one-way ANOVA was performed on each quality indicator in the original indicator detection vector to test whether there are significant differences in the mean values ​​of each quality indicator among different shifts. For quality indicators that show significant differences, post-hoc testing is used to compare the mean values ​​of each shift pairwise to identify the shifts that deviate significantly. Using time as the grouping factor, one-way ANOVA and post-hoc tests were repeatedly performed on each quality indicator in the original indicator detection vector to obtain the dates of specific significant deviations.

[0013] Optionally, based on the dynamic quality consistency evaluation, the contribution rate of each quality indicator is calculated, and the sources of anomalies are identified, including: The contribution rate of each quality indicator is calculated using formula (2). (2) in, Let be the contribution rate of the j-th quality index in the i-th box sample to the overall deviation. Let be the standardized deviation of the i-th box sample on the j-th quality index.

[0014] On the other hand, the present invention also provides a dynamic evaluation and anomaly traceability system for the consistency of tobacco processing quality, the system comprising: The quality indicator collection and data construction module is used to acquire multi-dimensional quality indicator monitoring data and construct the original indicator detection vector; A quality consistency baseline construction module is used to construct a quality consistency baseline based on the original indicator detection vector. The dynamic quality consistency evaluation module is used to obtain dynamic quality consistency evaluation based on the original indicator detection vector and the quality consistency baseline. The abnormal shift and date location module is used to perform quality consistency difference analysis on the original indicator detection vector to obtain the location of abnormal shifts and dates. The anomaly source tracing module is used to calculate the contribution rate of each quality indicator based on the dynamic quality consistency evaluation in order to obtain the source of the anomaly.

[0015] In another aspect, the present invention provides a machine-readable storage medium storing instructions that cause a machine to perform any of the methods described above in this application.

[0016] The beneficial effects of this invention are: This invention obtains dynamic quality consistency evaluation based on original indicator detection vectors and quality consistency baselines, and perceives quality fluctuation trends in real time during processing. This provides a basis for decision-making regarding material feeding and processing management and timely adjustment of process parameters, effectively preventing batch quality accidents. By analyzing the quality consistency differences of the original indicator detection vectors, the abnormal shifts and dates are located, allowing for a scientific and quantitative determination of whether quality fluctuations are caused by specific shifts or time periods, which is beneficial for implementing quality responsibility and rapid process troubleshooting. Furthermore, by calculating the contribution rate of each quality indicator, the source of anomalies is identified, enabling a comprehensive quantitative evaluation of single-box quality and automatic marking of abnormal boxes.

[0017] 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

[0018] 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 This is a flowchart of a method for dynamic evaluation and anomaly tracing of the consistency of tobacco processing quality according to an embodiment of the present invention; Figure 2 This is a structural block diagram of a dynamic evaluation and anomaly traceability system for the consistency of tobacco processing quality according to an embodiment of the present invention; Figure 3 This is a single-value-moving range dynamic control chart of total sugar, nicotine, reducing sugar and total nitrogen content of tobacco module according to an embodiment of the present invention; Figure 4 This is a trend chart of the sliding window process capability index of various indicators of the tobacco module according to an embodiment of the present invention; Figure 5 This is a box plot of four quality indicators for a tobacco sheet module under different processing shifts according to an embodiment of the present invention; Figure 6 This is a box plot of four quality indicators for a tobacco sheet module under different processing dates according to an embodiment of the present invention; Figure 7 This is a trend chart of square Mahalanobis distance for all box samples of the tobacco module according to an embodiment of the present invention, based on multidimensional quality indicators. Figure 8This is a radar chart of multi-dimensional quality indicators for the comprehensive abnormal box No. 5440 according to an embodiment of the present invention. Figure 9 This is a box plot of the overall quality deviation of the tobacco leaf module for different processing shifts according to an embodiment of the present invention; Figure 10 This is a graph showing the trend of the overall quality consistency score of the tobacco module as a function of the processing sequence, according to an embodiment of the present invention. Detailed Implementation

[0019] 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.

[0020] 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.

[0021] like Figure 1 The diagram shows a flowchart of a method for dynamic evaluation and anomaly tracing of tobacco processing quality consistency according to an embodiment of the present invention. Figure 1 In this context, the prediction method may include the following steps: In step S10, multidimensional quality indicator monitoring data are acquired, and the original indicator detection vector is constructed. In step S11, a quality consistency baseline is constructed based on the original indicator detection vector; In step S12, dynamic quality consistency evaluation is obtained based on the original index detection vector and quality consistency baseline; In step S13, a quality consistency difference analysis is performed on the original indicator detection vector to obtain the location of abnormal shifts and dates; In step S14, based on the dynamic quality consistency evaluation, the contribution rate of each quality indicator is calculated to obtain the source of the anomaly.

[0022] In such Figure 1In the dynamic evaluation and anomaly tracing method for the consistency of tobacco processing quality shown, step S10 is used to acquire multi-dimensional quality indicator monitoring data and construct the original indicator detection vector. In this embodiment, step S10 acquires multi-dimensional quality indicator detection data of all box samples collected at fixed box intervals during the processing of the target tobacco module, including: detection time, production shift, box number, and the actual quality indicator detection values ​​of each quality indicator. In this embodiment, each quality indicator can be one of four indicators: total sugar, nicotine, reducing sugar, and total nitrogen.

[0023] Step S11 is used to establish a quality consistency baseline. In this embodiment, step S11 first calculates the overall mean and overall standard deviation of all qualified box samples in the module for each quality indicator, as a benchmark for quality evaluation. Subsequently, in order to eliminate the influence of different indicator units, the original indicator detection vector needs to be preprocessed. In this embodiment, Z-score standardization is used, and the standardization formula is as follows: ; in, Let be the standardized deviation of the i-th box sample on the j-th index. This represents the original detection value of the i-th box sample on the j-th indicator. Let j be the mean of all valid box samples on the j-th index. Let be the standard deviation of all valid box samples on the j-th index.

[0024] Furthermore, the standardized multidimensional index mean vector is used as the comprehensive quality center point. , where k is the number of indicators.

[0025] Step S12 is used to obtain a dynamic quality consistency evaluation based on the original indicator detection vector and the quality consistency baseline. In this embodiment, step S12 includes: drawing a single-value-moving range control chart for each quality indicator in the original indicator detection vector in the order of processing time; when a single value or moving range exceeds a preset upper limit or a preset lower limit, an abnormal situation is determined to have occurred. In this embodiment, the preset upper limit can be defined as: The preset lower limit can be defined as: .in, The overall mean for each quality indicator for all qualified box samples. The overall standard deviation for each quality indicator for all qualified box samples.

[0026] Furthermore, the process capability index of each quality indicator in the original indicator detection vector within the sliding window is calculated according to the following formula. and draw Trend chart over time: .

[0027] A Level 1 alarm is triggered when the process capability index falls below a preset threshold or shows a continuous downward trend. In this embodiment, the preset threshold can be set to 0.9.

[0028] Furthermore, the weighted Euclidean distance of each box sample relative to the overall mass center point is calculated according to the following formula, and a trend chart of the squared Mahalanobis distance is plotted: ; in, Let be the squared Mahalanobis distance statistic for the i-th box sample. Let be the standardized quality feature vector of the i-th box sample. As a standardized integrated quality center point, The covariance matrix of the standardized quality indicators, is the inverse of the covariance matrix.

[0029] Since the square Mahalanobis distance approximately follows a chi-square distribution with k degrees of freedom, where k is the number of quality indicators involved in the evaluation, the warning threshold and the comprehensive anomaly threshold can be determined based on the quantiles of the chi-square distribution. The warning threshold can be defined as: The comprehensive anomaly threshold can be defined as: .in, As the warning threshold, To determine the comprehensive anomaly threshold, For a chi-square distribution with k degrees of freedom at a confidence level The quantiles below, and These are the confidence levels used for early warning determination and comprehensive anomaly determination, respectively. Higher than .

[0030] Anomaly detection rules are constructed based on formula (1), and early warning and comprehensive anomaly samples are obtained. When the box sample... When the value exceeds the comprehensive anomaly threshold, it is judged as a multi-indicator comprehensive anomaly box; when When the value is between the warning threshold and the comprehensive anomaly threshold, it is identified as a warning concern box; when When the temperature is below the warning threshold, the container is considered normal. (1) Furthermore, a comprehensive quality consistency score is constructed based on the squared Mahalanobis distance to intuitively characterize the overall quality stability of the box samples. The calculation formula is as follows: ; in, Let denot be the overall quality consistency score of the i-th bin sample. The smaller the squared Mahalanobis distance, the higher the overall quality consistency score, indicating that the bin is closer to the overall quality center and has better quality consistency. The larger the squared Mahalanobis distance, the lower the overall quality consistency score, indicating that the bin has a greater degree of overall deviation in the multi-index space.

[0031] Step S13 is used to perform a quality consistency difference analysis on the original indicator detection vector to obtain the location of abnormal shifts and dates. In this embodiment, the specific method for performing the quality consistency difference analysis on the original indicator detection vector in step S13 can be various forms known to those skilled in the art. In this embodiment, step S13 includes: using the processing shift as a grouping factor, performing a one-way ANOVA on each quality indicator in the original indicator detection vector to test whether there is a significant difference in the mean of each quality indicator between different shifts. Specifically, the sample data is divided into several groups according to processing time and shift, and a one-way ANOVA is performed on each key quality indicator to calculate the p-value. The p-value measures the probability of observing such a difference under the assumption that there is no difference in quality performance between different shifts. When the p-value is less than the significance level, the possibility of sampling error causing such a large difference is extremely low, and therefore it can be considered that there is indeed a significant difference in this indicator between different shifts. In this embodiment, the significance level can be 0.05.

[0032] Subsequently, for quality indicators showing significant differences, post-hoc tests were used to compare the means of each shift pairwise to identify the shifts with significant deviations. Specifically, Tukey HSD post-hoc tests were used to compare the means of each shift pairwise to determine which shift significantly deviated from the others. Simultaneously, box plots were generated for each shift's indicators to visually display the median, interquartile range, and outliers.

[0033] Furthermore, using time as a grouping factor, one-way ANOVA and post-hoc tests were repeatedly performed on each quality indicator in the original indicator detection vector to obtain the dates of specific significant deviations and pinpoint the specific processing dates or periods that caused the quality fluctuations.

[0034] To further explain the main sources of deviation in the abnormal boxes, it is necessary to calculate the contribution rate of each quality indicator in step S14. In this embodiment, the absolute values ​​of the standardized deviations of each indicator of the box sample are normalized to obtain the contribution ratio of each indicator to the overall deviation of the sample. Specifically, the contribution rate of each quality indicator after standardization of the multidimensional feature vector is calculated using formula (2): (2) in, Let be the contribution rate of the j-th quality index in the i-th box sample to the overall deviation. Let be the standardized deviation of the i-th box sample on the j-th quality index.

[0035] By ranking the contribution rates, the main quality indicators that cause the overall abnormality or warning of the box can be identified. Combined with the direction of the indicators being too high or too low, a basis can be provided for subsequent tracing of material feeding status, processing time, and fluctuations in process parameters.

[0036] Finally, an automatic quality consistency evaluation report for tobacco processing is generated. The report includes: basic information such as the tobacco module name, testing cycle, sample size, testing indicators, sampling interval, production shift, and processing date; an overall quality overview of each quality indicator, including the mean, standard deviation, upper control limit, lower control limit, process capability index, and overall quality consistency evaluation results; dynamic trend charts for each key quality indicator, including single-value moving range control charts, sliding window process capability index trend charts, and comprehensive quality deviation trend charts; shift-based one-way ANOVA results, Tukey HSD post-test results, box plots for each shift's indicators, date-based one-way ANOVA results, and date-based Tukey HSD post-test results. Post-hoc inspection results of HSD and comparative analysis of shifts and dates in box plots of key date indicators; a list of abnormal boxes including all box samples marked as warning concern or comprehensive anomalies, along with their box number, processing date, production shift, square Mahalanobis distance, comprehensive quality consistency score, anomaly level, and main deviation indicators; standardized deviations and contribution rates of each quality indicator for comprehensive anomaly boxes and critical warning boxes, and explanations of the anomaly contribution of the corresponding multi-dimensional indicator radar charts; conclusions and recommendations on the main quality consistency issues and corresponding process traceability suggestions obtained based on dynamic control charts, shift and date difference analysis, multi-indicator comprehensive anomaly identification, and indicator contribution analysis results.

[0037] On the other hand, embodiments of the present invention also provide a dynamic evaluation and anomaly traceability system for the consistency of tobacco processing quality. The structural block diagram of this system is shown below. Figure 2 As shown. In this Figure 2 The system comprises a quality indicator collection and data construction module, a quality consistency baseline construction module, a dynamic quality consistency evaluation module, an abnormal shift and date location module, and an anomaly source tracing module. Specifically, the quality indicator collection and data construction module acquires multi-dimensional quality indicator monitoring data and constructs original indicator detection vectors. The quality consistency baseline construction module constructs a quality consistency baseline based on the original indicator detection vectors. The dynamic quality consistency evaluation module obtains a dynamic quality consistency evaluation based on the original indicator detection vectors and the quality consistency baseline. The abnormal shift and date location module performs quality consistency difference analysis on the original indicator detection vectors to locate abnormal shifts and dates. The anomaly source tracing module calculates the contribution rate of each quality indicator based on the dynamic quality consistency evaluation to identify the source of the anomaly.

[0038] In another aspect, the present invention provides a machine-readable storage medium storing instructions for causing a machine to execute any of the methods described in a dynamic evaluation and anomaly tracing method for the consistency of tobacco processing quality.

[0039] Example: Taking the SC1125CFW tobacco sheet module processed by a certain re-drying plant as an example, the method of the present invention will be described in detail.

[0040] (1) Data acquisition: This module retrieved data from 141 cases sampled at intervals of 40 cases each during the production period from November 1st to November 8th, 2024. The data included: testing date, production shift (Shift A, B, C, and D), case number, and four indicators: total sugar, nicotine, reducing sugar, and nitrogen. Some examples of the raw data are shown in Table 1. Table 1. Test results of quality indicators for re-dried tobacco sheets (example)

[0041] (2) Data preprocessing: The four quality indicators—total sugar, nicotine, reducing sugar, and total nitrogen—were standardized using Z-scores. The standardized data are shown in Table 2. Table 2. Standardized processing results of quality index test data for re-dried tobacco sheets (example)

[0042] (3) Establishment of a baseline for overall processing quality consistency: The population mean vector and standard deviation vector for all 141 samples were calculated, as shown in Table 3: Table 3. Mean and standard deviation of each indicator for the sample

[0043] Calculate the overall quality center point of this module: Use the standardized multidimensional index mean vector as the center point in the multidimensional space. The specific statistical data is shown in Table 4. Table 4. Central points of various indicators in standardized data

[0044] Process capability index baseline construction: .in, , The results for each indicator, UCL and LCL, are shown in Table 5: Table 5 Statistical results of UCL and LCL for each indicator

[0045] (4) Dynamic quality consistency evaluation: Plot a single-value moving range control chart for the indicators "total sugar, nicotine, reducing sugar, and total nitrogen". The control chart is shown below. Figure 3 As shown, the total sugar content (31.49%) of 5640 boxes processed by shift C on November 8th exceeded the lower limit control (LCL=31.53), triggering a single-point loss-of-control alarm. Looking at the moving range control chart, an anomaly occurred near box number 3440, with a range value reaching approximately 2.3, far exceeding the upper limit control. This indicates a significant difference in the total sugar content between this batch and the previous batch, suggesting a sudden and large fluctuation in the process. The focus should be on checking whether there were any sudden changes in the raw materials, equipment parameters, and operating procedures for this batch.

[0046] Calculate the sliding window process capability index for four core indicators, including total sugar and nicotine. The trend chart of the sliding window process capability index is shown below. Figure 4 The data shows a slow upward trend in the early stages, rising from around 2.0 to a peak of approximately 2.5; subsequently, it slowly declined, eventually stabilizing between 1.5 and 2.0. Although process capability fluctuated slightly, it remained at a high level overall, without significant decline or anomalies. This corresponds to the increased volatility in the later stages of the moving range control chart. The increased volatility in the moving range control chart led to a slight decrease in the process capability index, but it remained within a controllable range. The process capability indices for nicotine and total nitrogen remained consistently stable between 5.0 and 7.5, representing the two strongest performers among the four indicators. The process capability index for reducing sugar remained stable between 3.5 and 5.0, indicating excellent process capability with manageable fluctuations.

[0047] (5) Analysis of differences between shifts and times: To analyze the impact of different processing shifts on the consistency of tobacco flake quality, 141 carton samples were divided into four groups (A, B, C, and D) according to processing shifts. The mean and standard deviation of four quality indicators—total sugar, nicotine, reducing sugar, and total nitrogen—were calculated for each group. The comparison results of the mean chemical indicators for each shift are shown in Table 6. Table 6 Comparison of the average chemical indicators for each shift

[0048] Table 6 shows that there are certain differences in the mean values ​​of quality indicators among different processing shifts. Regarding total sugar, shift B has the highest mean value at 33.48, while shifts A and C have relatively lower values. Regarding nicotine, shifts A and C have slightly higher mean values, while shift D has relatively lower values. Regarding reducing sugar, shifts B and D have higher mean values, while shift C has relatively lower values. Regarding total nitrogen, shift C has the highest mean value at 2.00, while shifts A and B have relatively lower values. These results indicate that there are certain differences in quality distribution among different shifts, and further analysis of variance and post-hoc testing are needed to determine the specific sources of these differences.

[0049] Subsequently, using processing shift as the grouping factor, one-way ANOVA was conducted on the four indicators of total sugar, nicotine, reducing sugar, and total nitrogen to test whether there were significant differences in the means of the indicators among different shifts. The results of the ANOVA are shown in Table 7. Table 7 Results of One-Way ANOVA for Quality Indicators of Each Shift

[0050] Table 7 shows that there were statistically significant differences in the four indicators—total sugar, nicotine, reducing sugar, and total nitrogen—across different shifts. The differences in total sugar and total nitrogen were particularly pronounced, with p-values ​​of 0.0081 for total sugar and 0.0003 for total nitrogen, both reaching highly significant levels. Nicotine and reducing sugar also showed significant differences. These results indicate that shift factors have a certain impact on the quality indicators of tobacco processing, and that different shifts may have differences in raw material conditions, equipment operation, process control, or operating conditions during the corresponding processing periods.

[0051] Furthermore, for indicators showing significant differences according to the analysis of variance, the Tukey HSD post-hoc test was used to compare the means of each shift pairwise to identify the shifts with specific significant deviations. The results of the Tukey HSD post-hoc test for significant differences are shown in Table 8: Table 8. Results of Post-hoc Significance Tests for Various Quality Indicators in Tukey HSD

[0052] Table 8 shows that Class B had relatively high levels of total sugar and reducing sugar, while Class C had relatively high levels of nicotine and total nitrogen. Specifically, Class B's total sugar was significantly higher than that of Class C and Class A, and its reducing sugar was significantly higher than that of Class C, indicating that Class B mainly exhibited higher levels of sugar indicators. Class C's nicotine was significantly higher than that of Class D, and its total nitrogen was significantly higher than that of Class B and Class A, indicating that Class C mainly exhibited higher levels of nitrogen-containing indicators. Therefore, Class B can be identified as a key class to be traced for its significantly higher sugar indicators, and Class C can be identified as a key class to be traced for its significantly higher nitrogen-containing indicators.

[0053] Furthermore, to visually demonstrate the differences in the distribution of quality indicators across different processing shifts, box plots were created for four indicators: total sugar, nicotine, reducing sugar, and total nitrogen. The box plots are shown below. Figure 5As shown in the figure, the box plot, with the processing shift as the horizontal axis and the quality index test value as the vertical axis, displays the median, interquartile range, dispersion, and outliers for each shift's samples. It shows the distribution of four shifts (A, B, C, and D) in terms of total sugar, nicotine, reducing sugar, and total nitrogen. For total sugar and reducing sugar, shift B's overall box position is relatively high, while shift C's is relatively low, consistent with the Tukey HSD test result showing shift B significantly higher than shift C. For nicotine, shift C's overall level is higher than shift D's. For total nitrogen, shift C's box position is significantly higher, and significantly higher than both shifts A and B. This figure visually demonstrates the differences in the distribution of quality indicators across different shifts, corroborating the results of one-way ANOVA and the Tukey HSD post-hoc test, providing a visual basis for locating outlier shifts.

[0054] The results of combined shift averages, one-way ANOVA, Tukey HSD post-hoc tests, and box plots show that shift factors have a clear impact on the quality indicators of tobacco sheets. Specifically, shift B mainly exhibits higher total sugar and reducing sugar levels, while shift C mainly exhibits higher nicotine and total nitrogen levels. Therefore, in subsequent anomaly tracing, the focus should be on the raw material ratio, leaf conditioning and feeding, and moisture balance during the processing period corresponding to shift B, as well as the source of raw materials and processing parameter control for nitrogen-containing indicators during the processing period corresponding to shift C.

[0055] Furthermore, based on the shift-based stratified analysis, the samples were further grouped by date using "day" as the time unit. One-way ANOVA and Tukey HSD post-hoc tests were repeatedly performed to pinpoint the specific processing dates or periods causing quality fluctuations. In this embodiment, 141 box samples were divided into eight processing date groups (1101, 1102, 1103, 1104, 1105, 1106, 1107, and 1108) based on the testing date. Date-specific variance analyses were performed on four indicators: total sugar, nicotine, reducing sugar, and total nitrogen. The ANOVA results for each indicator under different processing dates are shown in Table 9. Table 9. Results of One-Way ANOVA for Quality Indicators at Different Processing Dates

[0056] Table 9 shows that there were significant differences in total sugar, reducing sugar, and total nitrogen among different processing dates, while nicotine levels did not differ significantly. The differences in total sugar and total nitrogen were more pronounced over time, indicating periodic quality fluctuations during processing. The nicotine index did not differ significantly across dates, suggesting relatively small diurnal fluctuations, and that the date factor had a weaker impact on nicotine than on total sugar, reducing sugar, and total nitrogen.

[0057] For total sugar, reducing sugar, and total nitrogen indicators that showed significant date differences, Tukey HSD post-hoc tests were used to perform pairwise comparisons between different processing dates. The results of significant differences are shown in Table 10. Table 10. Results of Post-hoc Significance Tests for Tukey HSD Processing Dates

[0058] Table 10 shows that the date differences are mainly concentrated in total sugar and total nitrogen. Specifically, the total sugar on November 8th was significantly lower than on November 2nd, 3rd, 4th, 5th, and 6th, indicating a significant downward shift in total sugar during the corresponding processing stage on November 8th, making it the main abnormal date for total sugar fluctuations in this batch. The total sugar on November 7th was significantly lower than on November 3rd, and the total nitrogen was also significantly lower than on November 2nd, indicating that November 7th also had a certain risk of quality deviation and can be considered a secondary monitoring date. Regarding reducing sugars, November 6th was significantly higher than November 4th, suggesting fluctuations in sugar indicators on certain dates, but its anomaly indication was weaker than the downward shift in total sugar on November 8th.

[0059] Furthermore, to visually demonstrate the differences in the distribution of key quality indicators under different processing dates, box plots of total sugar and total nitrogen indicators under different processing dates were drawn. The box plots are shown below. Figure 6 As shown. Compared to simply grouping by shift, date box plots can further pinpoint the specific processing dates where quality fluctuations occurred. According to... Figure 6 The distribution of total sugar, nicotine, reducing sugar, and total nitrogen across different processing dates reveals that the overall position of the sample box for November 8th is significantly lower than that of most previous dates. Furthermore, the Tukey HSD test shows that the total sugar level on November 8th is significantly lower than that on November 2nd, 3rd, 4th, 5th, and 6th, indicating a significant downward shift in total sugar levels during the processing stage corresponding to November 8th. This result pinpoints the fluctuation in total sugar quality to a specific processing date, providing a basis for subsequent tracing of the raw material batch, processing parameters, and equipment operating status corresponding to November 8th. Meanwhile, the overall total nitrogen levels on November 7th and 8th are relatively low, with the Tukey HSD test showing that the total nitrogen level on November 7th is significantly lower than that on November 2nd, suggesting a certain deviation in nitrogen-containing indicators during the processing stage corresponding to November 7th. Given the significant decrease in total sugar on November 8th, it can be concluded that there is a risk of periodic quality fluctuations in the later stages of processing. Special attention should be paid to the feeding status, leaf moistening and feeding, temperature and humidity control during the re-drying period corresponding to November 7th and 8th, as well as online moisture fluctuations.

[0060] Based on the analysis results from both shift and date dimensions, this embodiment identifies Shift B, Shift C, and dates 1108 and 1107 as key traceability targets. Shift B primarily exhibits significantly higher levels of sugar indicators such as total sugar and reducing sugar, while Shift C primarily exhibits significantly higher levels of nitrogen-containing indicators such as nicotine and total nitrogen. From a date perspective, date 1108 shows significantly lower total sugar levels than most previous processing dates, making it the primary date for the abnormal shift in total sugar in this batch. Date 1107 shows relatively lower total sugar and total nitrogen levels, making it a secondary date of focus. These results demonstrate that this method can pinpoint quality fluctuations from an overall statistical level to specific production organization units and specific processing dates, providing a clear direction for subsequent anomaly tracing based on single-box Mahalanobis distance anomaly identification and indicator contribution analysis.

[0061] It should be noted that the shift and date difference analysis is mainly used to determine whether there are differences in the distribution of quality indicators between different groups and to locate shifts or dates with significant deviations; whether a specific individual container constitutes a comprehensive anomaly of multiple indicators needs to be further determined by combining square Mahalanobis distance analysis.

[0062] (6) Abnormal box identification based on multi-dimensional indicator fusion: This embodiment selects four key quality indicators—total sugar, nicotine, reducing sugar, and total nitrogen—to conduct a comprehensive quality consistency evaluation of 141 valid chamber samples. First, the sample mean, standard deviation, and control limits for the four indicators are calculated, and the results are shown in Table 11. Table 11 Basic Statistical Results of Four Quality Indicators

[0063] Subsequently, a correlation analysis was conducted on the four indicators, as shown in Table 12: Table 12 Correlation Coefficient Matrix of Four Quality Indicators

[0064] The results showed that total sugar and reducing sugar were significantly positively correlated (correlation coefficient 0.64); reducing sugar and total nitrogen were negatively correlated (correlation coefficient -0.49); and nicotine and total nitrogen were positively correlated (correlation coefficient 0.30). These results indicate that the four indicators are not completely independent. Using Mahalanobis distance can simultaneously consider the variance of the indicators and the covariance relationship between them when evaluating the overall deviation of the sample, thereby improving the rationality of multi-indicator anomaly identification.

[0065] Furthermore, using the mean vector of 141 valid bin samples as the overall quality center, the four quality indicators were standardized. Based on the standardized covariance matrix, the squared Mahalanobis distance of each bin sample was calculated, and a trend chart of the squared Mahalanobis distance was plotted. This trend chart is shown below. Figure 7As shown in the figure. In this embodiment, the number of indicators is 4, therefore the statistical threshold for the squared Mahalanobis distance adopts the chi-square distribution quantile with 4 degrees of freedom. Among them, the 95% warning threshold is 9.49, and the 99% comprehensive anomaly threshold is 13.28; the corresponding Mahalanobis distance thresholds are 3.08 and 3.64, respectively. The threshold settings and overall identification results are shown in Table 13: Table 13 Mahalanobis distance threshold settings and overall recognition results

[0066] Furthermore, the early warning and comprehensive anomaly boxes identified based on Mahalanobis distance are shown in Table 14: Table 14 Early Warning and Comprehensive Anomaly Boxes Identified Based on Mahalanobis Distance

[0067] As shown in Table 14, the squared Mahalanobis distance for sample 5440 is 14.21, which is higher than the 99% comprehensive anomaly threshold of 13.28, thus classifying it as a multi-indicator comprehensive anomaly box. Further analysis of the direction of deviation reveals that this box mainly exhibits simultaneous low levels of total sugar, reducing sugar, and total nitrogen, with the most significant deviation in total sugar. This indicates that the anomaly is not caused by fluctuations in a single quality indicator, but rather by a multi-indicator joint anomaly formed by multiple quality indicators deviating from the comprehensive quality center. This result demonstrates that the Mahalanobis distance method can identify comprehensive quality deviations that are not easily reflected in single-indicator control charts.

[0068] For sample 5440 of the comprehensive anomaly test chamber, the standardized deviation and contribution rate of each quality indicator for this sample were calculated, and a corresponding multidimensional indicator radar chart was generated. The radar chart is shown below. Figure 8 As shown in the multi-dimensional indicator radar chart, the changes in each quality indicator can be seen intuitively.

[0069] The squared Mahalanobis distance of sample 4320 is 13.05. Although it does not exceed the 99% comprehensive anomaly threshold, it is close to the anomaly judgment boundary and belongs to the critical warning sample. This sample mainly shows low total sugar, high nicotine, and low total nitrogen, suggesting that there may be short-term fluctuations in the quality composition or processing status of tobacco flakes during the corresponding processing period of this box. Samples 5600 and 5640 are both from Class C and both show low total sugar and low reducing sugar, indicating that there may be a certain continuous deviation trend in the later stage of processing.

[0070] To further analyze the differences in quality consistency from the perspective of shift, the squared Mahalanobis distance of different processing shifts was summarized and statistically analyzed. The results are shown in Table 15: Table 15 Summary Results of Overall Quality Deviation for Different Train Routes

[0071] Table 15 shows that the overall quality deviation varies across different shifts. A box plot of the squared Mahalanobis distance for the multidimensional indicators of each production shift is shown below. Figure 9 As shown in the figure, the average squared Mahalanobis distance of the samples from shift A was 4.78, higher than that of shifts D, C, and B. Shift A also exhibited the only comprehensive anomaly in this batch, with a maximum squared Mahalanobis distance of 14.21. This indicates that the overall quality deviation during the processing period corresponding to shift A was relatively high, requiring priority verification of the feeding status, tobacco source composition, processing parameter settings, and online quality control records for that period. Although shifts B and C did not exhibit any comprehensive anomaly, each had two samples requiring early warning, indicating localized short-term fluctuations in these shifts. Further investigation should be conducted based on the box number sequence and processing time. Shift D did not exhibit any early warning or anomaly boxes, and the maximum squared Mahalanobis distance was below the 95% early warning threshold, indicating relatively good overall quality consistency during the processing period corresponding to shift D. Looking at the box number sequence, early warning and anomaly samples were relatively concentrated in the later stages of processing, especially around boxes 5440-5640, where multiple indicators showed consecutive deviations. This indicates that the method can not only identify individual anomalies but also reveal the trend of declining quality consistency in localized periods during processing. By associating the abnormal box number with processing time, production shift, and process records, possible sources of abnormality can be traced, including changes in feeding structure, fluctuations in leaf feeding status, and deviations in re-drying temperature or moisture control.

[0072] The overall quality consistency score for each box sample was calculated based on the squared Mahalanobis distance, and a trend chart of the overall quality consistency score of the tobacco sheet module as a function of processing sequence was plotted to visually demonstrate the trend of the overall quality consistency score of each box as a function of processing sequence. This trend chart is shown below. Figure 10 As shown.

[0073] (7) Conclusion generation and report output: The system automatically generates an evaluation report, with the following key conclusions: Main issues: The overall processing quality of the SC1125CFW tobacco flake module was relatively consistent, with 135 out of 141 valid chamber samples falling within the normal fluctuation range, accounting for 95.74%; 5 samples reached the warning concern level, and 1 sample reached the comprehensive abnormal level. The abnormal and warning samples mainly showed a combined deviation of multiple indicators such as total sugar, reducing sugar, and total nitrogen.

[0074] Dynamic monitoring conclusions: The single-value moving range control chart shows that the total sugar index of 5640 boxes processed by shift C on November 8th was below the lower control limit, triggering a single-point loss-of-control alarm; the moving range near box number 3440 increased significantly, indicating a sudden fluctuation in the total sugar index between adjacent boxes. The sliding window process capability index results show that the overall process capability of the four indicators is at a good level, but the fluctuation of some indicators in the later stages of processing has increased, suggesting that further tracing is needed by combining shift, date, and single-box comprehensive deviation results.

[0075] Shift-specific analysis: From a shift perspective, Shift B primarily exhibited significantly higher levels of carbohydrate indicators such as total sugar and reducing sugar, while Shift C primarily exhibited significantly higher levels of nitrogen-containing indicators such as nicotine and total nitrogen. Tukey HSD post-hoc testing results showed that Shift B had significantly higher total sugar and reducing sugar levels than Shift C; Shift C had significantly higher nicotine and total nitrogen levels than Shift D. Therefore, Shifts B and C can be considered key shifts for quality difference tracing, with a focus on the reasons for higher carbohydrate indicators in Shift B and the reasons for higher nitrogen-containing indicators in Shift C.

[0076] Date Identification: From a date perspective, November 8th is the most noteworthy anomaly date for this batch. The total sugar content on November 8th was significantly lower than on November 2nd, 3rd, 4th, 5th, and 6th, indicating a significant downward shift in total sugar levels during the corresponding processing stage. Furthermore, the total sugar content on November 7th was significantly lower than on November 3rd, and the total nitrogen content was significantly lower than on November 2nd, suggesting a certain risk of quality deviation on November 7th as well. Therefore, November 8th can be identified as the primary anomaly date with significantly lower total sugar levels, while November 7th can be identified as a secondary date of concern for lower total sugar and total nitrogen levels.

[0077] Comprehensive Anomaly Detection: Based on the comprehensive evaluation results of multiple indicators in a single container, container number 5440 is the only container with comprehensive anomalies in this batch, with a squared Mahalanobis distance of 14.21, which is higher than the 99% comprehensive anomaly threshold of 13.28. This container mainly exhibits simultaneously low levels of total sugar, reducing sugar, and total nitrogen, indicating that its anomaly is not due to a single indicator exceeding the limit, but rather a combined anomaly caused by multiple quality indicators deviating from the comprehensive quality center. Container number 4320 is close to the 99% anomaly threshold and belongs to the critical warning sample category. Container numbers 5600 and 5640 both come from Class C and show low levels of total sugar and reducing sugar, suggesting a certain continuous deviation trend in the later stages of processing.

[0078] Comprehensive Traceability Conclusion: Based on the combined results of dynamic control charts, shift / date difference analysis, and Mahalanobis distance evaluation, a relatively clear direction for anomaly tracing can be established. At the shift level, shifts B and C showed significant deviations in sugar and nitrogen content indicators, respectively. At the date level, total sugar was significantly lower on November 8th, and total nitrogen was relatively lower on November 7th. At the individual box level, box 5440 was the only box with comprehensive multi-indicator anomalies, and warning and anomaly samples were relatively concentrated around boxes 5440-5640. These results indicate that the quality fluctuations in this batch are mainly concentrated in the later stages of processing, and particular attention needs to be paid to the processing procedures corresponding to November 7th-8th and their adjacent boxes.

[0079] Improvement suggestions: It is recommended to conduct traceability verification in three directions. First, verify the raw material ratio, leaf-moistening feeding, and moisture balance status for the corresponding period of Shift B to determine whether the high total sugar and reducing sugar levels are related to the feed composition or processing status. Second, verify the sources of raw materials and processing parameters related to nicotine and total nitrogen for the corresponding period of Shift C to determine whether the high levels of nitrogen-containing indicators are continuous. Third, focus on verifying the batch feeding, re-drying temperature and humidity control, online moisture fluctuations, and adjacent container detection records for the processing stages around day 1108 and boxes 5440-5640 to determine whether the simultaneous low levels of total sugar, reducing sugar, and total nitrogen are caused by changes in the state of raw materials in the later stages of processing or fluctuations in process control.

[0080] This embodiment demonstrates that the present invention can form a complete technical chain: dynamic monitoring of single indicators, significant analysis of shifts / dates, comprehensive anomaly identification of single boxes, explanation of the contribution of anomaly indicators, and report-based output. Compared with traditional ex-post evaluation methods based solely on the coefficient of variation of nicotine, the present invention can not only evaluate overall quality consistency but also pinpoint quality fluctuations to specific shifts, dates, and boxes, and further explain the sources of anomalies, providing quantitative support for the dynamic evaluation and anomaly traceability of tobacco processing quality consistency.

[0081] The beneficial effects of this invention are: The embodiments of the present invention achieve real-time and dynamic monitoring of multi-dimensional quality indicators of tobacco flakes during continuous processing by drawing dynamic control charts and process capability index trend charts within sliding windows. This intuitively displays the quality fluctuation trend, provides a basis for decision-making for the management of the feeding and processing process and the timely adjustment of process parameters, effectively prevents batch quality accidents, and solves the problem of lack of process dynamism in traditional post-event evaluation.

[0082] The embodiments of this invention perform stratified statistical analysis and significance testing on the box inspection data by shift and processing date, quantifies the impact of different factors on quality consistency, and quickly identifies the key shifts or time periods that cause overall quality fluctuations. This solves the problem that traditional methods cannot quickly and accurately locate shifts and time periods, and is conducive to implementing quality responsibility and rapid process troubleshooting.

[0083] The embodiments of this invention utilize the square Mahalanobis distance to fuse multidimensional indicators, avoiding the biased evaluation results caused by relying on a single indicator to measure the quality consistency of processed batches of tobacco leaves. This improves the comprehensiveness and accuracy of single-box quality evaluation, enabling more sensitive and accurate identification of abnormal quality boxes that deviate from the center point in multidimensional space. This provides support for the automation and intelligence of quality evaluation, and promotes the improvement of tobacco leaf processing quality control.

[0084] 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.

[0085] 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.

[0086] 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.

[0087] 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.

[0088] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0089] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, like read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0090] 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.

[0091] 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.

[0092] 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 dynamic evaluation and anomaly tracing of consistent quality in tobacco processing, characterized in that, The method includes: Acquire multidimensional quality indicator monitoring data and construct original indicator detection vectors; Based on the original indicator detection vector, a quality consistency baseline is constructed; Based on the original indicator detection vector and the quality consistency baseline, a dynamic quality consistency evaluation is obtained; A quality consistency difference analysis is performed on the original indicator detection vector to obtain the location of abnormal shifts and dates; Based on the dynamic quality consistency evaluation, the contribution rate of each quality indicator is calculated to identify the source of anomalies.

2. The method according to claim 1, characterized in that, Based on the original indicator detection vector, constructing a quality consistency baseline includes: Calculate the overall mean and overall standard deviation of each quality indicator in the original indicator detection vector to construct a quality evaluation benchmark; The original indicator detection vector is standardized, and the mean of each quality indicator is calculated to construct a comprehensive quality center.

3. The method according to claim 1, characterized in that, Based on the original indicator detection vector and the quality consistency baseline, obtaining the dynamic quality consistency evaluation includes: Draw a single-value-moving range control chart for each quality indicator in the original indicator detection vector; An abnormal situation is determined when a single value or moving range exceeds the preset upper or lower limit.

4. The method according to claim 1, characterized in that, Based on the original indicator detection vector and the quality consistency baseline, obtaining the dynamic quality consistency evaluation includes: Calculate the process capability index of each quality indicator in the original indicator detection vector within the sliding window, and plot the trend graph of the process capability index of each quality indicator. When the process capability index is lower than a preset threshold or If the decline continues over a period of time, it is considered an abnormal situation.

5. The method according to claim 1, characterized in that, Based on the original indicator detection vector and the quality consistency baseline, obtaining the dynamic quality consistency evaluation includes: Calculate the squared Mahalanobis distance of each sample relative to the quality consistency baseline, and plot the trend of the squared Mahalanobis distance; The early warning threshold and the comprehensive anomaly threshold are obtained based on the chi-square distribution quantiles; Based on the warning threshold and the comprehensive anomaly threshold, anomaly determination rules are constructed, and warning and comprehensive anomaly samples are obtained.

6. The method according to claim 5, characterized in that, Based on the warning threshold and the comprehensive anomaly threshold, anomaly determination rules are constructed, and warning and comprehensive anomaly samples are obtained, including: Anomaly detection rules are constructed using formula (1). ，（1） in, The squared Mahalanobis distance, To determine the comprehensive abnormal threshold, This is the warning threshold.

7. The method according to claim 1, characterized in that, Performing a quality consistency difference analysis on the original indicator detection vector to obtain the location of abnormal shifts and dates includes: Using the processing shift as the grouping factor, a one-way ANOVA was performed on each quality indicator in the original indicator detection vector to test whether there are significant differences in the mean values ​​of each quality indicator among different shifts. For quality indicators that show significant differences, post-hoc testing is used to compare the mean values ​​of each shift pairwise to identify the shifts that deviate significantly. Using time as the grouping factor, one-way ANOVA and post-hoc tests were repeatedly performed on each quality indicator in the original indicator detection vector to obtain the dates of specific significant deviations.

8. The method according to claim 1, characterized in that, Based on the dynamic quality consistency evaluation, the contribution rate of each quality indicator is calculated, and the sources of anomalies are identified, including: The contribution rate of each quality indicator is calculated using formula (2). ，（2） in, Let be the contribution rate of the j-th quality index in the i-th box sample to the overall deviation. Let be the standardized deviation of the i-th box sample on the j-th quality index.

9. A dynamic evaluation and anomaly traceability system for the consistency of tobacco processing quality, characterized in that, The system includes: The quality indicator collection and data construction module is used to acquire multi-dimensional quality indicator monitoring data and construct the original indicator detection vector; A quality consistency baseline construction module is used to construct a quality consistency baseline based on the original indicator detection vector. The dynamic quality consistency evaluation module is used to obtain dynamic quality consistency evaluation based on the original indicator detection vector and the quality consistency baseline. The abnormal shift and date location module is used to perform quality consistency difference analysis on the original indicator detection vector to obtain the location of abnormal shifts and dates. The anomaly source tracing module is used to calculate the contribution rate of each quality indicator based on the dynamic quality consistency evaluation in order to obtain the source of the anomaly.

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