A mathematical and statistical analysis method based on quality management decision-making system

By using mathematical statistical analysis methods in the quality management decision system, screening and optimizing air impurity data, the problem of inaccurate control limit setting caused by model hypothesis deviation in the prior art is solved, and the accuracy and reliability of the quality management system are improved. , significantly improves the scientific nature of early warning capabilities and decision-making support.

CN118691161BActive Publication Date: 2024-12-06STATE GRID JIANGSU ECONOMIC RES INST
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411162848.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-12-06
Estimated Expiration
2044-08-23

AI Technical Summary

Technical Problem

In the actual production environment, the existing impurity particle monitoring and early warning methods have deviations from the actual distribution of data and the model assumptions, resulting in inaccurate setting of control limits, affecting the early warning effect, and even leading to invalidation of early warning.

Method used

Using mathematical statistical analysis method based on quality management decision-making system, through dynamic parameter feedback, obtain air impurity data, screen and classify, build training sets, calculate impurity probability data, optimize Poisson distribution model, and generate optimized impurity mean and early warning value , draw quality decision diagrams and make real-time quality judgments.

Benefits of technology

Improve data processing efficiency and model accuracy, reduce noise impact, dynamic adjustment to adapt to different environments, significantly improve the accuracy of data analysis and quality control in production environments, improve early warning capabilities and scientific decision-making support, and ensure the accuracy of parameter selection sex.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118691161B_ABST
    Figure CN118691161B_ABST
Patent Text Reader

Abstract

The invention discloses a mathematical statistics analysis method based on a quality management decision system, relates to the technical field of quality management decision, and comprises the following steps: obtaining air impurity data, constructing an air impurity data training set, and calculating impurity probability data of various impurities; setting a sliding time window, and dividing the air impurity data training set based on the sliding time window to obtain several groups of second air impurity data training sets; respectively calculating corresponding distribution parameters, and respectively calculating the chi-square statistics of Poisson distribution models corresponding to several distribution parameters, and drawing a chi-square statistic change graph; selecting the best distribution parameter according to the chi-square statistic change graph, and combining the impurity probability data to generate the actual impurity number mean and impurity warning value of the optimized Poisson distribution model; the invention solves the problem that in a production environment, the actual distribution of data deviates from the model assumption, which leads to inaccurate control limit setting.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of quality management decision-making systems, and more specifically, to a mathematical statistics analysis method based on quality management decision-making systems. Background Art

[0002] In some production rooms with high requirements for the manufacturing environment, such as semiconductor production workshops, pharmaceutical factories, and high-precision optical equipment manufacturers, the detection of impurities in the air is the key to ensuring product quality and the stability of the production process. Because these production environments have extremely strict restrictions on particulate matter and pollutants in the air, any tiny impurities may cause product defects or production accidents. Therefore, accurate and efficient detection and control of impurities in the air has become an important task in these high-standard production environments.

[0003] For example, the invention patent with publication number CN114781835A discloses a method for wafer processing using statistical process control, including a statistical process control unit, a database and a feedback unit. The statistical process control unit has a data processing module and a judgment module. The statistical process control unit reads the data in the database, firstly processes the processing data and the standard wafer design parameters; then divides and matches and formulates adjustment rules; then the detection data is judged and compared with the wafer design parameters through the judgment module, and the comparison results are divided into close to the standard, in compliance with the standard and poor according to the degree of compliance. Feedback is given based on the comparison results of close to the standard, in compliance with the standard and poor. If adjustment is required, the wafer design parameters corresponding to the processing data are adjusted according to the adjustment rules to form new processing data; the new processing data is fed back to the processing unit through the feedback unit and executed. Wafer processing quality is ensured through statistical process control.

[0004] For example, the invention patent with publication number: CN117252480A publishes a quality warning method for welding workshops, including: setting upper and lower limits of specifications; calculating process standard deviation; calculating CP value; calculating CPK value; calculating CS qualified rate; real-time monitoring and feedback; data visualization and warning. The design of the present invention can detect quality problems in production by collecting and analyzing data in real time; accurately evaluate production quality using statistical process control methods; automatically collect data, calculate quality indicators and issue warnings to reduce manual burden; display quality data in charts to make data easier to understand and analyze; be able to process data at multiple points to adapt to different production environments and needs; be able to automatically issue warnings when CP value and CPK value are lower than the preset threshold or when continuous measurement results are out of tolerance, helping enterprises to promptly detect and deal with quality problems; be able to help enterprises effectively improve production efficiency and reduce costs, and be data-driven to help enterprises make more correct business decisions.

[0005] The above-disclosed technical solution has at least the following technical problems: the existing monitoring and early warning of impurity particles usually uses probability statistics and distribution model training to judge the controllability of air impurities in the production room. The distribution of impurities is usually modeled based on set distribution parameters, but in the actual production environment, the distribution of impurities may not completely conform to this assumption. When the actual distribution of the data deviates from the model assumptions, it will lead to inaccurate control limit settings, affecting the early warning effect, or even invalidating the early warning.

[0006] In view of the above problems, the present invention proposes a solution. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a mathematical statistical analysis method based on a quality management decision system, which uses dynamic parameter feedback to solve the problem that in the actual production environment, the actual distribution of data deviates from the model assumptions, resulting in inaccurate control limit setting.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A mathematical statistics analysis method based on a quality management decision system comprises the following steps: obtaining air impurity data, screening and classifying the air impurity data based on a preset radius, and constructing an air impurity data training set; calculating impurity probability data of various impurities according to the air impurity data training set; obtaining the actual impurity number mean and impurity warning value in a Poisson distribution model of an initial preset distribution parameter based on the impurity probability data and a cyclic algorithm; setting a sliding time window, and dividing the air impurity data training set based on the sliding time window to obtain several groups of second air impurity data training sets; respectively calculating the distribution parameters corresponding to the several groups of second air impurity data training sets, and respectively calculating the chi-square statistics of the Poisson distribution model corresponding to the several distribution parameters, and drawing a chi-square statistic change graph; selecting the best distribution parameter according to the chi-square statistic change graph, and combining the impurity probability data, based on a cyclic algorithm, obtaining an optimized Poisson distribution model, and generating the actual impurity number mean and impurity warning value of the optimized Poisson distribution model; drawing a quality decision graph according to the actual impurity number mean and the impurity warning value, and making a quality judgment.

[0010] In a preferred embodiment, the air impurity data includes the radius of impurities in the workshop air; the air impurity data is screened and classified based on a preset radius, specifically: the preset radius includes a monitoring radius and a radius window; the radius of the impurities in the air impurity data is compared with the monitoring radius, and impurities larger than the monitoring radius are screened out; the radius value is cut into several equidistant radius ranges according to the radius window, and the screened impurities are classified according to the radius of the impurities falling into the corresponding radius range; several equidistant radius ranges are numbered, the number of impurities within the corresponding radius range is counted, and an air impurity data training set is constructed, wherein the air impurity data training set includes several vectors consisting of the labels of the impurity radius ranges and the corresponding impurity numbers.

[0011] In a preferred embodiment, the impurity probability data includes an impurity probability and an impurity probability average.

[0012] In a preferred embodiment, the cyclic algorithm is based on obtaining the actual mean number of impurities and the impurity warning value in the Poisson distribution model with the initial preset distribution parameters, specifically: setting the initial value of the impurity warning value to zero; calculating and obtaining the initial mean parameter based on the initial value of the impurity warning value; calculating the corresponding model impurity probabilities of different types based on the initial mean parameter and the initial value of the impurity warning value and the impurity probability data; comparing the actual impurity probability and the model impurity probability in the impurity probability data to obtain a decision value; setting the impurity warning value of the next cycle to be the initial value of the impurity warning value plus the preset cycle step, repeating the initial decision value calculation process to obtain the decision value of the first cycle; repeating the cycle process until the current decision value is less than the previous decision value, stopping the cycle, and obtaining the actual mean number of impurities and the impurity warning value in the Poisson distribution model with the initial preset distribution parameters.

[0013] In a preferred embodiment, the generation of several groups of more refined second air impurity data training sets is specifically: a preset time window length, a sliding window on the original data set according to a set time interval, and gradually intercepting a data subset within each time period.

[0014] In a preferred embodiment, the specific method for obtaining the distribution parameters is as follows: calculating the distribution parameters by calculating the mean of the second air impurity data training set; respectively calculating the chi-square statistics of the Poisson distribution model corresponding to several distribution parameters, specifically: calculating the expected impurity frequency of each group according to the corresponding Poisson distribution model; calculating the chi-square statistic by comparing the actual impurity probability and the expected impurity frequency in the second air impurity data training set; drawing the chi-square statistic change graph, specifically: sorting the distribution parameters from large to small, and drawing a curve graph in combination with the corresponding chi-square statistics.

[0015] In a preferred embodiment, the method of selecting the best distribution parameter based on the chi-square statistic change graph is specifically as follows: taking the distribution parameter at the point where the curve changes most steadily as the best distribution parameter; taking the best distribution parameter as the initial best distribution parameter when the Poisson distribution model is looped, and performing a loop iteration operation to obtain an optimized Poisson distribution model and the actual impurity number mean and impurity warning value of the optimized Poisson distribution model.

[0016] In a preferred embodiment, the quality decision diagram is drawn according to the actual impurity number mean and the impurity warning value, specifically: the quality decision diagram is drawn with time as the horizontal axis and the impurity number as the vertical axis; the warning line on the quality decision diagram is calculated using the actual impurity number mean and the impurity warning value; the air impurity data collected in real time is input into the quality decision diagram, and when the data is above the warning line, an alarm is issued to indicate that the workshop is unqualified.

[0017] The technical effects and advantages of the mathematical statistics analysis method based on the quality management decision system of the present invention are as follows:

[0018] 1. The present invention screens and classifies air impurity data by setting monitoring radius and radius window, effectively reducing the impact of irrelevant small impurities and constructing a structured data training set. This method not only improves the efficiency of data processing and the accuracy of the model, but also allows dynamic adjustment to adapt to different environmental conditions. By reducing the impact of noise and accurately calculating the probability of impurities and their mean, the model's predictive ability and decision support level are optimized, thereby improving the accuracy and reliability of the quality management system.

[0019] 2. The present invention significantly improves the accuracy of data analysis and quality control in the production environment by dynamically segmenting air impurity data, optimizing model parameters and accurately setting warning values. Specifically, by segmenting the data through a sliding time window, subtle changes and short-term fluctuations in the data can be captured, allowing the model to better adapt to the actual data distribution. Calculating and drawing a chi-square statistic change graph helps to identify and correct the deviation between the model and the actual data, ensuring the accuracy of parameter selection. By optimizing the Poisson distribution model through a cyclic algorithm, a more reliable impurity number mean and warning value can be generated, which improves the warning capability and the scientific nature of decision support, thereby effectively solving the problem of inaccurate control limit setting due to model assumption deviation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 The present invention is a schematic diagram of the structure of a mathematical statistics analysis method based on a quality management decision system. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Embodiment 1, Figure 1 The present invention provides a mathematical statistical analysis method based on a quality management decision system, comprising the following steps:

[0023] S1, obtaining air impurity data, screening and classifying the air impurity data based on a preset radius, and constructing an air impurity data training set.

[0024] In this embodiment, the air impurity data includes the radius of the impurities in the workroom air;

[0025] In this embodiment, the air impurity data is screened and classified based on a preset radius, specifically:

[0026] The preset radius includes a monitoring radius and a radius window;

[0027] Furthermore, the radius of the impurities in the air impurity data is compared with the monitoring radius, and impurities larger than the monitoring radius are screened out;

[0028] Furthermore, the radius value is cut into a number of equidistant radius ranges according to the radius window, and the screened impurities are classified according to the radius of the impurities falling into the corresponding radius range;

[0029] A number of equidistant radius ranges are numbered, the number of impurities within the corresponding radius ranges is counted, and an air impurity data training set is constructed. The air impurity data training set includes a number of vectors including the numbers of the impurity radius ranges and the corresponding impurity numbers.

[0030] It should be noted that constructing an air impurity data training set has the following advantages:

[0031] Preset monitoring radius screening: By setting the monitoring radius, impurities larger than a certain size can be quickly screened out, effectively reducing the amount of data; this allows you to focus on the analysis of larger impurities, saving computing resources while avoiding interference from irrelevant small impurities on the model; Radius window segmentation and equidistant range classification: By using a radius window to cut the impurity radius into several equidistant radius ranges, this classification method can more accurately group impurities of different sizes. This makes the classified data more representative, which helps the accuracy of subsequent statistical analysis;

[0032] Constructing a vectorized training set: Construct the filtered and classified data into a vector containing the impurity radius range label and the corresponding impurity quantity. This structured data format facilitates subsequent data analysis, modeling, and algorithm application; especially when using statistical models or machine learning algorithms, vectorized data can be more easily processed and analyzed;

[0033] Dynamic adjustment of monitoring radius and radius window: This method allows the monitoring radius and radius window to be adjusted dynamically according to actual needs, so as to adapt to the needs of different scenarios or air quality conditions; for example, the screening and classification criteria can be adjusted according to the requirements of a specific workroom to improve the applicability and flexibility of the method;

[0034] Reduce the impact of noise: By screening out impurities smaller than the monitoring radius, the impact of noise data on the model is effectively reduced, thereby improving the robustness and prediction accuracy of the model; this helps to generate a more reliable air quality assessment and early warning model.

[0035] Use of radius labels: The use of radius labels gives each classification a clear physical meaning; by counting the number of impurities within different radius ranges, it can directly reflect the distribution of impurities in the air, providing a strong basis for subsequent quality decisions.

[0036] It should be noted that the selection of monitoring radius and radius window should be based on the data analysis results of specific application scenarios or industry standards to ensure the effectiveness and scientificity of screening and classification; for impurity data smaller than the monitoring radius that are excluded during the screening process, choose to discard it directly or retain it for subsequent noise analysis or model optimization; when the radius value is divided equally, if the boundary problem is that the impurity radius is less than or equal to a certain range, it will fall into that range.

[0037] S2, calculates the impurity probability data of each type of impurity based on the air impurity data training set.

[0038] In this embodiment, the impurity probability data of various impurities are calculated based on the air impurity data training set, specifically:

[0039] The impurity probability data includes impurity probability and impurity probability average value;

[0040] It should be noted that obtaining impurity probability data has the following functions:

[0041] Providing data foundation: Impurity probability and impurity probability average are the basic data of the entire patent process. These data provide the necessary basis for model construction, optimization and quality judgment in subsequent steps;

[0042] Estimation of model parameters: In the subsequent process, preliminary parameter estimation of the Poisson distribution model is performed based on the impurity probability data, especially the impurity probability mean value; accurate impurity probability and mean value help to fit the actual distribution of impurities more accurately, thereby improving the reliability of the model;

[0043] Anomaly detection and early warning: Impurity probability reflects the distribution of different types of impurities in the air, while the impurity probability average provides an overall baseline level; these data can help detect whether there are anomalies (such as the probability of a certain type of impurity is significantly higher than the average), thereby providing timely early warning of possible quality problems;

[0044] Optimization of distribution model: In the subsequent process, the impurity probability data is combined with the selected optimal distribution parameters to optimize the Poisson distribution model; this step relies on accurate probability values ​​and average values ​​to ensure that the optimized model is more in line with the actual situation and improve the model's predictive ability;

[0045] Support for quality decision-making: Impurity probability and impurity probability average are important inputs for drawing quality decision diagrams; these data help to indicate the normal and abnormal ranges in the diagram, supporting the accuracy and reliability of quality judgments;

[0046] Effectiveness of data classification: The calculation of impurity probability helps to effectively screen and classify air impurity data in S1; accurate probability estimation can ensure the rationality of classification, thereby improving the quality of data training sets;

[0047] Decision support and optimization: By calculating the probability and average value of impurities, we can better understand the distribution characteristics of impurities, thereby providing a scientific basis for adjusting monitoring parameters and optimizing production processes, and improving the responsiveness and efficiency of the quality management decision-making system.

[0048] S3, according to the impurity probability data, based on a cyclic algorithm, obtaining the actual impurity number mean and impurity warning value in the Poisson distribution model of the initial preset distribution parameters.

[0049] The initial preset distribution parameter refers to the assumption that the number of impurities obeys Poisson distribution, and the mean parameter Z of the Poisson distribution needs to be set, that is, the preset distribution parameter, which represents the average value of the number of impurities in a certain unit time or unit space.

[0050] It should be noted that the mean parameter Z plays a key role in the Poisson distribution model of impurities, which directly affects the expected value of the number of impurities; a larger Z value means that the average number of impurities is larger within a given time or space, and the model will predict more impurity events; conversely, a smaller Z value indicates a smaller number of impurities; the accurate setting of Z is crucial to the effectiveness of the model because it determines the shape of the Poisson distribution, thereby affecting the calculation of the impurity warning value and the degree of fit to the actual impurity mean data; if Z deviates from the actual situation, the prediction and warning performance of the model may be significantly affected.

[0051] It should be noted that the parameters of the initial probability distribution are usually set based on historical data or the experience of domain experts; these initial parameters provide a basis for the model so that the model can begin to fit and optimize the actual data.

[0052] In this embodiment, based on the cyclic algorithm, the actual impurity number mean and impurity warning value in the Poisson distribution model of the initial preset distribution parameters are obtained, specifically:

[0053] S31, setting the initial value of the impurity warning value to zero;

[0054] S32, calculating and obtaining an initial mean parameter according to the initial value of the impurity warning value;

[0055] The specific calculation formula of the initial mean parameter is as follows:

[0056]

[0057] In the formula, is the initial mean parameter, The degrees of freedom are The lower point of the initial impurity probability of the Poisson distribution, is the initial model impurity probability, is the initial value of the impurity warning value;

[0058] S33, calculating corresponding different types of model impurity probabilities according to the initial mean parameter, the initial value of the impurity warning value, and the impurity probability data;

[0059] The specific calculation formulas for the impurity probabilities of different types of models are as follows:

[0060]

[0061] In the formula, is the model impurity probability of impurity type p;

[0062] S34, comparing the actual impurity probability in the impurity probability data with the model impurity probability to obtain a decision value;

[0063] The specific calculation formula of the decision value is as follows:

[0064]

[0065] In the formula, is the initial decision value, L is the total number of impurity types, is the average impurity probability value, is the actual impurity probability.

[0066] S35, setting the impurity warning value of the next cycle to the initial value of the impurity warning value plus the preset cycle step length, repeating the initial decision value calculation process to obtain the decision value of the first cycle;

[0067] S36, repeat the above cycle process until the current decision value is less than the previous decision value, stop the cycle, and obtain the actual impurity number mean and impurity warning value in the Poisson distribution model of the initial preset distribution parameters.

[0068] It should be noted that the purpose of the Poisson distribution model of the initial preset distribution parameters is:

[0069] Initialize model parameters: By setting the initial mean parameter Z and warning value, steps S31 and S32 provide a starting point for the Poisson distribution model; these initial parameters are based on historical data or domain expert experience, laying the foundation for model iteration and optimization;

[0070] Preliminary model fitting: Step S33 calculates different types of model impurity probabilities and performs preliminary fitting on the actual impurity data based on the initial mean parameters and warning values; this process helps the model establish preliminary prediction capabilities and provides a basis for subsequent optimization;

[0071] Evaluate model accuracy: Step S34 compares the actual impurity probability with the impurity probability predicted by the model and calculates the decision value; the purpose of this step is to evaluate the prediction accuracy of the initial model and identify parameters that need to be adjusted;

[0072] Optimizing model parameters: Steps S35 and S36 continuously adjust the warning value through a loop algorithm to optimize the model's predictive ability; each loop recalculates the decision value based on the new warning value until the optimal parameter setting is found; this process ensures that the model can more accurately reflect the actual impurity data and improve the accuracy of warning and decision-making;

[0073] Improve model accuracy and reliability: optimize mean parameters and warning values ​​through a cyclic algorithm, and finally obtain the actual impurity number mean and the optimized warning value in step S36; these optimized parameters enable the Poisson distribution model to better adapt to actual data, and improve the effectiveness and reliability of the model in impurity monitoring;

[0074] Support decision making: The actual impurity count mean and warning value output by the optimized model provide accurate data support for the quality management decision system; with these data, the system can make more accurate quality judgments and warnings, thereby improving production process control and quality management.

[0075] This embodiment screens and classifies air impurity data by setting the monitoring radius and radius window, effectively reducing the impact of irrelevant small impurities and building a structured data training set. This method not only improves the efficiency of data processing and the accuracy of the model, but also allows dynamic adjustment to adapt to different environmental conditions. By reducing the impact of noise and accurately calculating the impurity probability and its mean, the model's predictive ability and decision support level are optimized, thereby improving the accuracy and reliability of the quality management system.

[0076] Embodiment 2, S4, setting a sliding time window, and dividing the air impurity data training set based on the sliding time window to obtain several groups of second air impurity data training sets.

[0077] In this implementation, the main purpose of setting the sliding time window is to dynamically segment the air impurity data, thereby generating several groups of more refined second air impurity data training sets.

[0078] The generating of several groups of more detailed second air impurity data training sets is specifically as follows:

[0079] The preset time window length slides the window on the original data set according to the set time interval and gradually intercepts the data subset within each time period.

[0080] In this way, each subset represents the air impurity status within a specific time range, which can more accurately capture the changing trends in time series data. This method allows for more detailed analysis of the data and identification of short-term change patterns or abnormal events. This segmentation not only improves the timeliness and representativeness of the data set, but also helps to more accurately calculate the probability distribution of impurities in each time period, providing reliable support for subsequent model optimization and quality judgment.

[0081] S5, respectively calculating the distribution parameters corresponding to several groups of second air impurity data training sets, respectively calculating the chi-square statistics of the Poisson distribution model corresponding to several distribution parameters, and drawing a chi-square statistic change graph.

[0082] In this implementation, the specific method for obtaining the distribution parameters is as follows:

[0083] The distribution parameter is calculated by calculating the mean of the second air impurity data training set. The specific calculation formula is as follows:

[0084]

[0085] In the formula, is the distribution parameter corresponding to the kth group of second air impurity data training set, is the total number of data in the second air impurity data training set, is the impurity probability in the second air impurity data training set; b is the label of the data in the second air impurity data training set.

[0086] The chi-square statistics of the Poisson distribution model corresponding to several distribution parameters are calculated respectively, specifically:

[0087] Calculate the expected impurity frequency of each group according to the corresponding Poisson distribution model;

[0088] The chi-square statistic is calculated by comparing the actual impurity probability and the expected impurity frequency in the second air impurity data training set;

[0089] The specific calculation method of the chi-square statistic is as follows:

[0090]

[0091] In the formula, is the chi-square statistic of the second air impurity data training set of the rth group, is the expected impurity frequency of the bth item in the rth group.

[0092] The drawing of the chi-square statistic change graph is specifically as follows:

[0093] Sort the distribution parameters from large to small and draw a curve graph based on the corresponding chi-square statistic.

[0094] The calculation of the chi-square statistic has the following effects on solving the problem of inaccurate control limit setting caused by the deviation between the actual distribution of data and the model assumptions in the actual production environment:

[0095] Identify model deviations: By calculating the distribution parameters of each set of second air impurity data training set, the distribution of actual data can be carefully analyzed; the chi-square statistic helps identify the deviation between the data and the assumed model by comparing the actual impurity probability with the expected frequency predicted by the model. If the chi-square statistic is large, it means that the model assumptions are poorly matched with the actual data, thus exposing the inaccuracy of the control limit setting;

[0096] Optimize model assumptions: Draw a graph of chi-square statistics changes. By comparing the chi-square statistics under different distribution parameters, you can intuitively find out which parameter settings are closer to the actual data distribution. This method allows you to adjust and optimize model assumptions in an actual production environment, thereby reducing control limit setting errors caused by assumption deviations and improving the accuracy and reliability of the model.

[0097] Improve the accuracy of control limit setting: By selecting the distribution parameters that minimize the chi-square statistic, a model that better fits the actual data distribution can be obtained; this makes the control limit setting more accurate, helping to reduce false alarms or missed alarms caused by deviations from model assumptions, thereby improving quality control in the production process;

[0098] Dynamic adjustment and adaptation: This step provides a means to dynamically adjust model assumptions so that the model can adapt to the characteristics of data under different time periods or conditions; this adaptability ensures that in a production environment, the model and control limits can effectively respond to changes in data distribution and maintain the stability and reliability of quality management;

[0099] Through the above steps, the match between model assumptions and actual data distribution can be significantly improved, thereby improving the accuracy of quality control and early warning in the production environment.

[0100] S6, selecting the best distribution parameters according to the chi-square statistic change graph, and combining with the impurity probability data, based on the cyclic algorithm, obtaining the optimized Poisson distribution model, and generating the actual impurity number mean and impurity warning value of the optimized Poisson distribution model.

[0101] The method of selecting the best distribution parameter according to the chi-square statistic change graph is as follows: taking the distribution parameter at the point where the curve changes most steadily as the best distribution parameter;

[0102] The optimal distribution parameter is used as the initial optimal distribution parameter when the Poisson distribution model is circulated, and a cyclic iteration operation is performed to obtain an optimized Poisson distribution model and an actual impurity number mean value and an impurity warning value of the optimized Poisson distribution model;

[0103] It should be noted that the optimal distribution parameters have the following effects on solving the problem of inaccurate control limit setting in actual production environments where the actual distribution of data deviates from the model assumptions:

[0104] Accurate parameter selection: By selecting the best distribution parameters based on the chi-square statistic change graph, we can ensure that the parameters that most accurately describe the distribution of the actual data are selected; selecting the distribution parameters at the most stable point of the curve as the best parameters can reduce the deviation between the model assumptions and the actual data and improve the prediction accuracy of the model;

[0105] Optimize model performance: Combined with impurity probability data, the Poisson distribution model can be further refined and adjusted through an optimization process based on a loop algorithm; the optimized model can better reflect the characteristics of actual data and improve the accuracy and reliability of the actual impurity number mean and warning value;

[0106] Improve prediction accuracy: The loop algorithm can refine the model parameters in continuous iterations, so that the final optimized model has higher prediction accuracy; this method can effectively reduce the gap between the model prediction value and the actual value, thereby enhancing the practicality of the model;

[0107] Robustness and adaptability: By selecting the distribution parameters where the chi-square statistic changes most steadily, the model's adaptability and robustness to data fluctuations can be enhanced; the optimized model can better cope with the natural fluctuations of data in the production environment and improve the stability of the quality management system;

[0108] Enhanced early warning capability: The optimized Poisson distribution model can generate more accurate impurity warning values, which helps improve the early warning capability during the production process; this is crucial for timely detection and handling of potential quality issues, and helps avoid production quality problems caused by inaccurate early warnings;

[0109] Improve decision support: The optimized model provides accurate actual impurity count mean and warning value, which can provide strong support for quality management decisions. With the optimized data, managers can make more scientific decisions and improve the control and management level of the production process.

[0110] S7, draw a quality decision diagram based on the actual impurity number mean and the impurity warning value, and make a quality judgment.

[0111] The quality decision diagram is drawn according to the actual impurity number mean value and the impurity warning value, specifically:

[0112] Draw a quality decision diagram with time as the horizontal axis and the amount of impurities as the vertical axis;

[0113] The warning line on the quality decision diagram is calculated by using the mean value of the actual impurity number and the impurity warning value;

[0114] The air impurity data collected in real time is input into the quality decision diagram. When the data is above the warning line, an alarm is sounded to indicate that the workshop is unqualified.

[0115] This embodiment significantly improves the accuracy of data analysis and quality control in the production environment by dynamically segmenting air impurity data, optimizing model parameters, and accurately setting warning values. Specifically, by segmenting the data through a sliding time window, subtle changes and short-term fluctuations in the data can be captured, allowing the model to better adapt to the actual data distribution. Calculating and drawing a chi-square statistic change graph helps to identify and correct the deviation between the model and the actual data, ensuring the accuracy of parameter selection. By optimizing the Poisson distribution model through a cyclic algorithm, a more reliable impurity number mean and warning value can be generated, which improves the warning capability and the scientific nature of decision support, thereby effectively solving the problem of inaccurate control limit setting due to model assumption deviations.

[0116] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0117] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0118] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0119] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0120] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0121] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A mathematical and statistical analysis method based on the quality management decision system, It is characterized in that The steps include: Acquire air impurity data, filter and classify the air impurity data based on a preset radius, and construct an air impurity data training set; Calculate the impurity probability data of various impurities based on the air impurity data training set; According to the impurity probability data, based on the cyclic algorithm, the actual impurity number mean and impurity warning value in the Poisson distribution model of the initial preset distribution parameters are obtained; The actual impurity number mean and impurity warning value in the Poisson distribution model of the initial preset distribution parameters are obtained based on the cyclic algorithm, specifically: Set the initial value of the impurity warning value to zero; According to the initial value of the impurity warning value, the initial mean parameter is calculated and obtained; Calculate the corresponding impurity probabilities of different types of models according to the initial mean parameter, the initial value of the impurity warning value and the impurity probability data; Compare the actual impurity probability in the impurity probability data with the model impurity probability to obtain a decision value; The impurity warning value of the next cycle is set to the initial value of the impurity warning value plus the preset cycle step length, and the calculation process of the initial decision value is repeated to obtain the decision value of the first cycle; Repeat the cycle until the current decision value is less than the previous decision value, stop the cycle, and obtain the actual impurity number mean and impurity warning value in the Poisson distribution model with the initial preset distribution parameters; Setting a sliding time window, and dividing the air impurity data training set based on the sliding time window to obtain several groups of second air impurity data training sets; Calculate the distribution parameters corresponding to several groups of second air impurity data training sets respectively, calculate the chi-square statistics of the Poisson distribution model corresponding to several distribution parameters respectively, and draw a chi-square statistic change graph; Select the best distribution parameters according to the chi-square statistic change chart, and combine with the impurity probability data, based on the cyclic algorithm, obtain the optimized Poisson distribution model, and generate the actual impurity number mean and impurity warning value of the optimized Poisson distribution model; Draw a quality decision diagram based on the actual impurity number average and impurity warning value, and make quality judgments; The specific calculation formula of the initial mean parameter is as follows: In the formula, is the initial mean parameter, The degrees of freedom are The lower point of the initial impurity probability of the Poisson distribution, is the initial model impurity probability, is the initial value of the impurity warning value; The specific calculation formulas for the impurity probabilities of different types of models are as follows: In the formula, is the model impurity probability of impurity type p, and e is a natural number; The specific calculation formula of the decision value is as follows: In the formula, is the initial decision value, L is the total number of impurity types, is the average impurity probability value, is the actual impurity probability.

2. The mathematical and statistical analysis method based on the quality management decision system according to claim 1, It is characterized in that The air impurity data includes the radius of impurities in the air of the workshop, and the preset radius includes a monitoring radius and a radius window; The air impurity data is screened and classified based on a preset radius, specifically: Compare the radius of impurities in the air impurity data with the monitoring radius, and filter out impurities larger than the monitoring radius; The radius value is cut into several equidistant radius ranges according to the radius window, and the screened impurities are classified according to the radius of the impurities falling into the corresponding radius range; A number of equidistant radius ranges are numbered, the number of impurities within the corresponding radius ranges is counted, and an air impurity data training set is constructed. The air impurity data training set includes a number of vectors including the numbers of the impurity radius ranges and the corresponding impurity numbers.

3. The mathematical and statistical analysis method based on the quality management decision system according to claim 2, It is characterized in that The impurity probability data includes an impurity probability and an impurity probability average value.

4. The mathematical and statistical analysis method based on the quality management decision system according to claim 3, It is characterized in that The obtaining of several sets of second air impurity data training sets is specifically: According to the preset time window length, the window is slid on the original data set according to the set time interval, and the data subset within each time period is gradually intercepted.

5. The mathematical and statistical analysis method based on the quality management decision system according to claim 4, It is characterized in that The chi-square statistics of the Poisson distribution model corresponding to several distribution parameters are calculated respectively, and a chi-square statistic change graph is drawn, specifically: Calculating the distribution parameter by calculating the mean of the second air impurity data training set; Calculate the expected impurity frequency of each group according to the corresponding Poisson distribution model; The chi-square statistic is calculated by comparing the actual impurity probability and the expected impurity frequency in the second air impurity data training set; Sort the distribution parameters from large to small and draw a curve graph based on the corresponding chi-square statistics.

6. The mathematical statistical analysis method based on the quality management decision system according to claim 5, It is characterized in that The method of selecting the best distribution parameter according to the chi-square statistic change graph is as follows: The distribution parameter at the point where the curve changes most steadily is taken as the optimal distribution parameter; The optimal distribution parameters are used as the initial optimal distribution parameters when the Poisson distribution model is cycled, and a cycle iteration operation is performed to obtain an optimized Poisson distribution model and an actual impurity number mean and an impurity warning value of the optimized Poisson distribution model.

7. The mathematical and statistical analysis method based on the quality management decision system according to claim 6, It is characterized in that The quality decision diagram is drawn according to the actual impurity number mean value and the impurity warning value, specifically: Draw a quality decision diagram with time as the horizontal axis and the amount of impurities as the vertical axis; The warning line on the quality decision diagram is calculated by using the mean value of the actual impurity number and the impurity warning value; The air impurity data collected in real time is input into the quality decision diagram. When the data is above the warning line, an alarm is sounded to indicate that the workshop is unqualified.

8. The mathematical and statistical analysis method based on the quality management decision system according to claim 7, It is characterized in that The distribution parameter is calculated by calculating the mean of the second air impurity data training set. The specific calculation formula is as follows: In the formula, is the distribution parameter corresponding to the kth group of second air impurity data training set, is the total number of data in the second air impurity data training set, is the impurity probability in the second air impurity data training set; b is the label of the data in the second air impurity data training set; The specific calculation method of the chi-square statistic is as follows: In the formula, is the chi-square statistic of the second air impurity data training set of the rth group, is the expected impurity frequency of the bth item in the rth group.