Production quality dynamic analysis method and system based on golden batch optimization
By determining and pre-processing multiple finished product evaluation indicators in biological product production, using multiple empowerment algorithms and preference coefficients to dynamically update the gold batch, the problem of inconsistent product quality in biological product batch production is solved, and more accurate quality assessment and market adaptability.
Patent Information
- Application Number
- CN202411246136.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-06
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-09-06
AI Technical Summary
In the production of biological products batches, factors such as environmental variables and material status lead to inconsistent product quality, making it difficult for enterprises to determine which variables have a more significant impact on product quality, resulting in complex management of gold batch configuration files and difficulty in achieving stability in production quality.
By determining multiple finished product evaluation indicators, obtaining and preprocessing production data of historical batches, using multiple empowerment algorithms to determine the weight of each indicator, initializing the preference coefficient and setting the incremental step size, dynamically adjusting the weight allocation, and accurately identifying gold. batches and dynamic updates of gold batches are performed through target preference coefficients.
Achieve comprehensive coverage and accurate evaluation of product quality, eliminate data noise, improve data accuracy, dynamically adjust weight allocation, accurately identify gold batches, help enterprises capture market changes and user needs, improve product market competitiveness, and promote Continuous improvement of products and production processes.
Smart Images

Figure CN118761685B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of data processing, and in particular to a production quality dynamic analysis method and system based on golden batch optimization. Background Art
[0002] In the batch production of biological products, companies need to add raw materials to unit equipment tanks, mixers and reactors to process them into new products or states, and then discharge them to another unit for further processing or place them in another tank for storage. In each batch operation, the company's goal is to obtain the ideal product. However, many environmental variables, material states, operating parameters, etc. will cause differences in product quality between different batches, such as boundary temperature, additive brand, water content of a certain ingredient, or any other measurable variables that may affect the output product. At the same time, over time, the values of many variables will also change, affecting product quality. For example, in the reactor, as the material ages, the insulation effect of the entire reaction stage begins to deteriorate. In order to obtain consistency in product quality in each batch of production, the company has proposed the concept of "golden batch", which aims to enable the system to consistently produce high-quality batches and identify and avoid the impact of environmental variables on batch quality. To achieve this, we need to record process data and analyze the results to determine the key variables. The analysis results generate a configuration file that maintains relatively stable quality, which is called the "golden batch file". In the actual production process, comparing the key variables of production with the "golden batch" configuration file can achieve monitoring, early warning and alarm of production quality. With the development and application of intelligent technology, the entire production system can automatically adjust the production process parameters by matching the "golden batch" production curve, ensuring that the quality of the products produced is consistent with the "golden batch". The production management model of the "golden batch" is the common development direction for the intelligent control of production quality in industrial enterprises such as biomedicine, food industry, and chemical industry. The determination of the golden batch will be affected by factors such as production conditions, production quality indicators, and process changes. Although enterprises are aware of the impact of environmental variables, material status, operating parameters and other factors on product quality, there may be a certain degree of unclear association in practice, that is, which variables have a more significant impact on product quality, which are secondary and which are primary. These issues require further scientific analysis and verification. There may be certain complexities and challenges in establishing and managing the "golden batch" configuration file. It is necessary to determine which variables should be included in the "golden batch" file and how to set and adjust these variables to ensure quality stability during the production process.
[0003] Therefore, it is necessary to provide a dynamic analysis method and system for production quality based on golden batch optimization, so as to quickly and accurately determine the golden batch, thereby improving the quality stability in the production process. Summary of the invention
[0004] The present invention provides a production quality dynamic analysis method based on golden batch optimization, including: determining multiple finished product evaluation indicators; obtaining production data and finished product inspection data of multiple historical batches; based on the multiple finished product evaluation indicators, preprocessing the production data and finished product inspection data of the multiple historical batches to obtain the preprocessed production data and finished product evaluation indicator data of the multiple historical batches; determining the weight of each finished product evaluation indicator corresponding to each weighting algorithm through multiple weighting algorithms; initializing the preference coefficient and setting the incremental step size of the preference coefficient; based on the preference coefficient, the incremental step size and the weight of each finished product evaluation indicator corresponding to each weighting algorithm, determining the quality score number of each historical batch corresponding to the multiple preference coefficients; based on each The quality score numbers of the historical batches corresponding to the multiple preference coefficients are used to determine the golden batch from the multiple historical batches; based on the golden batch, the target preference coefficient is determined; the production data and finished product inspection data of the current batch are obtained; based on the multiple finished product evaluation indicators, the production data and finished product inspection data of the current batch are preprocessed to obtain the preprocessed production data and finished product evaluation indicator data of the current batch; based on the target preference coefficient and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms, the quality score number of the current batch corresponding to the target preference coefficient is determined; based on the quality score number of the current batch corresponding to the target preference coefficient and the quality score number of the golden batch corresponding to the target preference coefficient, the golden batch is dynamically updated.
[0005] Further, determining a plurality of finished product evaluation indicators includes: determining a plurality of candidate finished product evaluation indicators; calculating the correlation between any two of the candidate finished product evaluation indicators; and determining the plurality of finished product evaluation indicators based on the correlation between any two of the candidate finished product evaluation indicators.
[0006] Furthermore, based on the multiple finished product evaluation indicators, the production data and finished product inspection data of the multiple historical batches are preprocessed, including: determining invalid historical batches among the multiple historical batches, deleting the production data and finished product inspection data of the invalid historical batches; and filling in missing data, normalizing and standardizing the production data and finished product inspection data of valid historical batches.
[0007] Furthermore, the weight of each finished product evaluation index corresponding to each weighting algorithm is determined through a plurality of weighting algorithms, including: determining the weight of each finished product evaluation index corresponding to the hierarchy analysis algorithm through a hierarchy analysis algorithm.
[0008] Furthermore, the weight of each finished product evaluation index corresponding to each weighting algorithm is determined through a plurality of weighting algorithms, including: determining the weight of each finished product evaluation index corresponding to the entropy method through an entropy method.
[0009] Furthermore, based on the preference coefficient, the incremental step and the weight of each finished product evaluation indicator corresponding to each of the weighting algorithms, the quality score number of each of the historical batches corresponding to the multiple preference coefficients is determined, including: S11, based on the preference coefficient and the weight of each finished product evaluation indicator corresponding to each of the weighting algorithms, the comprehensive weight corresponding to each of the finished product evaluation indicators is calculated; S12, based on the comprehensive weight corresponding to each of the finished product evaluation indicators, the quality score number of each of the historical batches corresponding to the preference coefficient is determined; S13, judging whether the preference coefficient meets the preset conditions, if so, ending the process, if not, executing S14; S14, updating the preference coefficient based on the incremental step, executing S11.
[0010] Furthermore, the comprehensive weight corresponding to each of the finished product evaluation indicators is calculated based on the following formula, including: ,in, is the comprehensive weight corresponding to the j-th finished product evaluation index, is the preference coefficient, is the weight of the hierarchical analysis algorithm corresponding to the j-th finished product evaluation index, is the weight of the entropy method corresponding to the j-th finished product evaluation index.
[0011] Furthermore, based on the quality score numbers of the plurality of preference coefficients corresponding to each of the historical batches, a golden batch is determined from the plurality of historical batches, including: for each of the preference coefficients, based on the quality score numbers of the preference coefficients corresponding to each of the historical batches, determining a maximum value of the quality score number corresponding to the preference coefficient; based on the maximum value of the quality score number corresponding to each of the preference coefficients, determining an array of maximum value of the quality score numbers; based on the array of maximum value of the quality score numbers, determining a golden batch from the plurality of historical batches.
[0012] Further, based on the golden batch, a target preference coefficient is determined, including: determining the target preference coefficient based on quality score numbers of the golden batch corresponding to multiple preference coefficients.
[0013] The present invention provides a production quality dynamic analysis system based on golden batch optimization, including: an index determination module, used to determine multiple finished product evaluation indicators; a data acquisition module, used to obtain production data and finished product inspection data of multiple historical batches; a data processing module, used to pre-process the production data and finished product inspection data of the multiple historical batches based on the multiple finished product evaluation indicators, and obtain the pre-processed production data and finished product evaluation index data of the multiple historical batches; an index weighting module, used to determine the weight of each finished product evaluation indicator corresponding to each weighting algorithm through multiple weighting algorithms; a parameter setting module, used to initialize the preference coefficient and set the incremental step size of the preference coefficient; a score calculation module, used to determine the quality score number of each historical batch corresponding to multiple preference coefficients based on the preference coefficient, the incremental step size and the weight of each finished product evaluation indicator corresponding to each weighting algorithm; batch A determination module is used to determine a golden batch from the multiple historical batches based on the quality score number of each of the historical batches corresponding to the multiple preference coefficients, and determine the target preference coefficient based on the golden batch; the data acquisition module is also used to obtain the production data and finished product inspection data of the current batch; the data processing module is also used to preprocess the production data and finished product inspection data of the current batch based on the multiple finished product evaluation indicators to obtain the preprocessed production data and finished product evaluation indicator data of the current batch; the score calculation module is also used to determine the quality score number of the current batch corresponding to the target preference coefficient based on the target preference coefficient and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms; a dynamic optimization module is used to dynamically update the golden batch based on the quality score number of the current batch corresponding to the target preference coefficient and the quality score number of the golden batch corresponding to the target preference coefficient.
[0014] Compared with the prior art, the production quality dynamic analysis method and system based on golden batch optimization provided by the present invention have at least the following beneficial effects:
[0015] By determining multiple finished product evaluation indicators, it is possible to comprehensively cover all aspects of product quality, such as performance, reliability, durability, etc., so as to more accurately reflect the overall quality level of the product. Preprocessing production data and finished product inspection data can eliminate noise and outliers in the data, improve the accuracy and reliability of the data, and provide a solid data foundation for subsequent quality evaluation. Using multiple weighting algorithms to determine the weight of each evaluation indicator can fully consider the advantages of different algorithms and avoid the bias or limitations that may be brought by a single algorithm. By initializing the preference coefficient and setting the incremental step size, combined with the quality score number under different preference coefficients, the weight distribution can be dynamically adjusted to make the evaluation more in line with the actual situation. Based on the quality score number of multiple preference coefficients, the best performing golden batch can be accurately identified, setting a quality benchmark for the enterprise, and providing a clear direction and goal for the subsequent production and quality management of the enterprise. By comparing the quality score number of the current batch with the golden batch under the target preference coefficient, the dynamic update of the golden batch is realized. This helps enterprises to capture market changes and user needs in a timely manner, adjust production strategies, and improve the market competitiveness of products. The dynamic update mechanism of the golden batch encourages enterprises to continuously pursue higher quality standards and promote continuous improvement of products and production processes. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:
[0017] Figure 1 It is a schematic diagram of a process of a dynamic analysis method of production quality based on golden batch optimization according to some embodiments of this specification;
[0018] Figure 2 It is a schematic diagram of a process for determining a quality score number corresponding to a plurality of preference coefficients for each historical batch according to some embodiments of this specification;
[0019] Figure 3 is a schematic diagram of a golden batch reference curve according to some embodiments of this specification;
[0020] Figure 4 It is a module schematic diagram of a production quality dynamic analysis system based on golden batch optimization according to some embodiments of this specification. DETAILED DESCRIPTION
[0021] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.
[0022] Figure 1 is a flow chart of a method for dynamic analysis of production quality based on golden batch optimization according to some embodiments of this specification, such as Figure 1 As shown, the dynamic analysis method for production quality based on golden batch optimization may include the following process.
[0023] Step 111, determining a plurality of finished product evaluation indicators.
[0024] Specifically include:
[0025] Determine multiple candidate finished product evaluation indicators. For example, taking the alcohol precipitation production of Danhong injection as an example, the multiple candidate finished product evaluation indicators may at least include a total solid removal rate evaluation indicator, a hydroxysafflor yellow A removal rate evaluation indicator, a salvianolic acid B removal rate evaluation indicator, a total phenolic acid removal rate evaluation indicator, a total sugar removal rate evaluation indicator, a solid removal rate evaluation indicator, etc.;
[0026] Calculate the correlation between any two candidate product evaluation indicators;
[0027] Based on the correlation between any two candidate finished product evaluation indicators, multiple finished product evaluation indicators are determined.
[0028] Specifically, finished product inspection data of multiple sample batches may be obtained, and finished product evaluation index data of the sample batches may be determined based on the finished product inspection data of the sample batches, wherein the finished product evaluation index data of the sample batches may include the values of the sample batches in multiple candidate finished product evaluation indexes. The correlation between any two candidate finished product evaluation indexes may be calculated based on the finished product evaluation index data of multiple sample batches.
[0029] For example, the correlation between two candidate finished product evaluation indicators can be calculated based on the following formula:
[0030] ,
[0031] in, is the correlation between the evaluation index of the e-th candidate finished product and the evaluation index of the f-th candidate finished product, G is the total number of sample batches, is the value of the evaluation index of the e-th candidate product of the g-th sample batch, is the mean value of the evaluation index of the e-th candidate product for all sample batches, is the value of the evaluation index of the fth candidate product of the gth sample batch, It is the mean value of the evaluation index of the fth candidate finished product for all sample batches.
[0032] Determine multiple finished product evaluation indicators according to the following process:
[0033] S21. For each candidate finished product evaluation indicator, based on the correlation between the candidate finished product evaluation indicator and each other candidate finished product evaluation indicator, calculate the correlation mean and correlation variance corresponding to the candidate finished product evaluation indicator, and based on the correlation mean and correlation variance corresponding to the candidate finished product evaluation indicator, calculate the screening score corresponding to the candidate finished product evaluation indicator;
[0034] S22, determining whether there is at least one candidate finished product evaluation indicator whose corresponding screening score is less than the screening score threshold, if so, executing S23, if not, taking the current remaining candidate finished product evaluation indicators as finished product evaluation indicators;
[0035] S23. For each candidate finished product evaluation indicator, when the screening score corresponding to the candidate finished product evaluation indicator is less than the screening score threshold, the candidate finished product evaluation indicator is determined to be an invalid candidate finished product evaluation indicator, and the invalid candidate finished product evaluation indicator is deleted. After deleting all invalid candidate finished product evaluation indicators, S21 is executed.
[0036] For example, the correlation mean, correlation variance and screening score corresponding to the candidate finished product evaluation indicators are calculated based on the following formula:
[0037] ,
[0038] in, is the correlation mean corresponding to the e-th candidate product evaluation index, F is the total number of candidate product evaluation indexes, is the correlation variance corresponding to the e-th candidate product evaluation index, is the preset correlation mean, is the preset parameter, greater than 0, is the screening score corresponding to the evaluation index of the e-th candidate finished product.
[0039] By calculating the correlation between each candidate finished product evaluation indicator and other indicators, and calculating the screening score based on these correlations, those candidate indicators that are highly correlated with other indicators can be identified. Removing these highly correlated candidate indicators can reduce the redundancy and duplication of the evaluation system, making the final evaluation system more streamlined and effective. Retaining those evaluation indicators with strong independence and large amount of information can make the entire evaluation system more accurately reflect the characteristics of multiple aspects of the finished product, thereby improving the accuracy and reliability of the evaluation. This is of great significance for product quality control, market positioning and subsequent improvements. In the evaluation process, if a large number of redundant or highly correlated indicators are included, it will not only increase the difficulty and complexity of the evaluation, but also reduce the efficiency of the evaluation. Excluding these unnecessary indicators by screening scores can simplify the evaluation process and improve the evaluation efficiency.
[0040] Step 112, obtaining production data and finished product inspection data of multiple historical batches.
[0041] Specifically, the production data and finished product inspection data of historical batches include the production data involved in each process step starting from the material. These data can be collected by management software such as the material management system, production management system (Manufacturing Execution System, MES), laboratory information management system (Laboratory Information Management System, LIMIS), quality management system (Quality Management System, QMS), batch release system, etc., including the material components, production process parameters, intermediate product inspection and testing, finished product inspection and other production data of each production batch. Taking the alcohol precipitation production of Danhong Injection as an example, the finished product inspection data of historical batches may at least include the total solid removal rate, hydroxysafflor yellow A (HSYA) removal rate, tanshinone B removal rate, total phenolic acid removal rate, total sugar removal rate, and solid removal rate.
[0042] Step 113, based on multiple finished product evaluation indicators, preprocessing the production data and finished product inspection data of multiple historical batches to obtain preprocessed production data and finished product evaluation indicator data of multiple historical batches.
[0043] Specifically include:
[0044] Determine invalid historical batches among multiple historical batches, and delete production data and finished product inspection data of invalid historical batches;
[0045] Perform missing data filling, normalization and standardization on the production data and finished product inspection data of valid historical batches.
[0046] Specifically, the data is compared with the production control conditions or required ranges. If there are batches that exceed the condition range and have no product qualification indicators, they will be deleted as invalid historical batches. If the scan finds that a historical batch has multiple (can be set to more than 3) missing data, and it is confirmed that it cannot be compensated, the historical batch will be deemed as an invalid historical batch.
[0047] If there is a single missing data, it is handled in the following two cases:
[0048] If it is discrete data, fill it with the mode and update the data set;
[0049] If it is continuous, use the mean value to fill and update the data set.
[0050] The updated production data set X and the finished product inspection data set Y are normalized to eliminate the dimension difference so that different indicators can affect the results more equally. The normalization process uses the following formula:
[0051] ,
[0052] In the formula, For the A certain production data or finished product verification data of a valid historical batch, It is the minimum value of the production data or finished product verification data in all valid historical batches. It is the maximum value of the production data or finished product verification data in all valid historical batches; The production data or finished product verification data of the valid historical batch after normalization calculation. .
[0053] After calculation according to the above formulas, the data table is updated to obtain the new normalized production data set X and finished product inspection data set Y.
[0054] The updated X and Y are standardized to reduce the impact of different data scales, and the following formula is used for calculation:
[0055] ,
[0056] In the formula, The first Data values of valid historical batches, is the mean of the finished product evaluation index data , is the standard deviation of the finished product evaluation index data , It is the finished product evaluation index data after standardized calculation.
[0057] Step 114, using a plurality of weighting algorithms, determining the weight of each finished product evaluation index corresponding to each weighting algorithm.
[0058] Specifically include:
[0059] Through the hierarchical analysis algorithm, the weight of each finished product evaluation index corresponding to the hierarchical analysis algorithm is determined. Specifically, by establishing a judgment matrix, a pairwise comparison between production data items and quality inspection expert scoring items is achieved, the maximum eigenvalue of the judgment matrix is solved, and the weight value of each finished product evaluation index in each batch is quickly calculated. The consistency test is performed on it. If it passes, the weight of each finished product evaluation index corresponding to each weighting algorithm is determined;
[0060] The entropy method is used to determine the weight of each finished product evaluation index corresponding to the entropy method. Specifically, in the quotient method, the entropy value of each finished product evaluation index is calculated ( ) to evaluate its information content and then determine its weight in the entire evaluation system. The smaller the entropy value, the greater the amount of information of the indicator and the greater its contribution to the evaluation results. Suppose there is an evaluation matrix ,in, represents the score of the i-th historical batch on the j-th finished product evaluation index, , , is the total number of historical batches, is the total number of evaluation indicators for finished products. is the normalized value, indicating the relative importance of the i-th batch on the j-th finished product evaluation index. The calculation formula is: , define the constant ,calculate , calculate the coefficient of variation , represents the difference degree of the jth finished product evaluation index. Finally, the weight of each finished product evaluation index is calculated. , which is the ratio of the difference coefficient of the finished product evaluation index to the sum of the difference coefficients of all indicators: .
[0061] Step 115, initializing the preference coefficient and setting the incremental step size of the preference coefficient.
[0062] Specifically, the preference coefficient can be initialized to 0, the preference coefficient .
[0063] The incremental step size of the preference coefficient can be determined based on the following process:
[0064] S31, initializing the incremental step size, for example, setting the initial value of the incremental step size to 0.1;
[0065] S32, according to the current incremental step size, determine the preference coefficients of multiple consecutive samples, for example, 1.1, 1.2, 1.3, 1.4, etc., calculate the comprehensive weight of the preference coefficient of each sample corresponding to each finished product evaluation index, and then determine the weight sequence corresponding to the preference coefficient of each sample, wherein an element of the weight sequence represents the comprehensive weight of a finished product evaluation index corresponding to the preference coefficient of the sample;
[0066] S33, calculating a sequence fluctuation parameter based on a weight sequence corresponding to a plurality of consecutive sampled preference coefficients;
[0067] S34, if the sequence fluctuation parameter is less than the sequence fluctuation threshold, execute S35, if the sequence fluctuation parameter is greater than or equal to the sequence fluctuation threshold, use the current incremental step size as the incremental step size of the preference coefficient;
[0068] S35. Adjust the incremental step size based on the sequence fluctuation parameter and execute S32.
[0069] For example, the series volatility parameter can be calculated according to the following formula:
[0070] ,
[0071] in, is the sequence fluctuation parameter corresponding to the current incremental step size, is the total number of sampled preference coefficients, is the comprehensive weight corresponding to the jth finished product evaluation index in the weight sequence corresponding to the preference coefficient of the nth sample, It is the mean of the comprehensive weight corresponding to the jth finished product evaluation index in the weight sequence corresponding to the preference coefficient of the n+1th sample.
[0072] The incremental step size can be adjusted based on the sequence fluctuation parameter according to the following formula:
[0073] ,
[0074] in, is the adjusted incremental step size, is the current incremental step size, is the sequence fluctuation threshold.
[0075] When the sequence fluctuation parameter is greater than or equal to the threshold, it means that the current preference coefficient setting may not fully reflect the quality performance of the batch corresponding to multiple preference coefficients. At this time, by adjusting the incremental step size and recalculating the weight sequence fluctuation parameter, a more accurate preference coefficient setting can be gradually approached, thereby enhancing adaptability and flexibility. By determining the incremental step size of the preference coefficient through an automated process, the time cost of manual intervention and decision-making can be significantly reduced. This not only improves decision-making efficiency, but also reduces the risk of decision-making errors due to human factors, and makes the quality score number corresponding to multiple preference coefficients of each subsequent historical batch better able to characterize the quality characteristics of the historical batch, thereby making the final golden batch more accurate.
[0076] Step 116, based on the preference coefficient, the incremental step size and the weight of each finished product evaluation index corresponding to each weighting algorithm, determine the quality score number corresponding to multiple preference coefficients for each historical batch.
[0077] Figure 2 is a flow chart of determining the quality score number corresponding to multiple preference coefficients for each historical batch according to some embodiments of this specification, such as Figure 2 As shown, in some embodiments, based on the preference coefficient, the incremental step size and the weight of each finished product evaluation index corresponding to each weighting algorithm, the quality score number corresponding to multiple preference coefficients of each historical batch is determined, including:
[0078] S11. Calculate the comprehensive weight corresponding to each finished product evaluation indicator based on the preference coefficient and the weight of each weighting algorithm corresponding to each finished product evaluation indicator;
[0079] S12. Determine the quality score number of each historical batch corresponding to the preference coefficient based on the comprehensive weight corresponding to each finished product evaluation indicator;
[0080] S13, determining whether the preference coefficient meets a preset condition (for example, the absolute value of the difference between the preference coefficient and 1 is less than the incremental step size), if so, ending the process, if not, executing S14;
[0081] S14. Update the preference coefficient based on the incremental step size and execute S11.
[0082] In some embodiments, the comprehensive weight corresponding to each finished product evaluation indicator is calculated based on the following formula, including:
[0083]
[0084] ,in, is the comprehensive weight corresponding to the j-th finished product evaluation index, is the preference coefficient, is the weight of the hierarchical analysis algorithm corresponding to the j-th finished product evaluation index, is the weight of the entropy method corresponding to the j-th finished product evaluation index.
[0085] The quality score number of the historical batch corresponding to the preference coefficient can be calculated based on the following formula:
[0086] ,
[0087] in, is the quality score number of the i-th historical batch corresponding to the k-th preference coefficient, is the value of the evaluation index of the i-th historical batch at the j-th finished product.
[0088] The preference coefficient can be updated based on the following formula:
[0089] ,
[0090] in, is the updated preference coefficient, is the preference coefficient before updating.
[0091] Step 117, determining a golden batch from multiple historical batches based on the quality score number of each historical batch corresponding to multiple preference coefficients.
[0092] Specifically include:
[0093] For each preference coefficient, based on the quality score number of each historical batch corresponding to the preference coefficient, determine the maximum quality score number corresponding to the preference coefficient;
[0094] Based on the maximum value of the quality score number corresponding to each preference coefficient, determine a maximum value array of the quality score number, wherein an array element in the maximum value array of the quality score number represents the maximum value of the quality score number corresponding to a preference coefficient;
[0095] Determine the golden batch from multiple historical batches based on an array of maximum quality score numbers.
[0096] For example, determine the historical batch corresponding to each array element in the maximum value array of the quality score number, determine the maximum value times of each historical batch, and take the historical batch with the largest maximum value times as the golden batch. For example, the maximum value array of the quality score number includes 5 array elements, among which the first array element, the second array element, and the fourth array element all correspond to historical batch 3, and the third array element and the fifth array element all correspond to historical batch 2. Then the maximum value times of historical batch 3 is 3, the maximum value times of historical batch 2 is 2, and the maximum values times of other historical batches are all 0. The maximum value times of historical batch 3 is the largest, so historical batch 3 is the golden batch.
[0097] For another example, the array elements included in the array of maximum quality score values are averaged to obtain the mean of the maximum quality score values, the array element in the array of maximum quality score values having the smallest difference with the mean of the maximum quality score values is determined, and the historical batch corresponding to the array element having the smallest difference with the mean of the maximum quality score values is taken as the golden batch.
[0098] Step 118, determining a target preference coefficient based on the golden batch.
[0099] Specifically include:
[0100] Based on the quality scores of the golden batch corresponding to multiple preference coefficients, the target preference coefficient is determined.
[0101] For example, for each preference coefficient, the quality scores of all historical batches form a curve along the time axis, called the batch-quality curve, the batch-quality curve corresponding to the golden batch at the maximum point is used as the target batch-quality curve, and the preference coefficient corresponding to the target batch-quality curve is used as the target preference coefficient. Under the target preference coefficient, the quality scores of all historical batches form a curve along the time axis (Time Batch—Q), called the golden batch reference curve; the golden batch is at the highest point of the golden batch reference curve. The quality trend of production batches over the years can be analyzed based on the golden batch reference curve.
[0102] Step 119, obtaining the production data and finished product inspection data of the current batch.
[0103] Step 120, based on a plurality of finished product evaluation indicators, preprocess the production data and finished product inspection data of the current batch to obtain preprocessed production data and finished product evaluation indicator data of the current batch.
[0104] Specifically, the method of preprocessing the production data and finished product inspection data of the current batch based on multiple finished product evaluation indicators is similar to the method of preprocessing the production data and finished product inspection data of multiple historical batches based on multiple finished product evaluation indicators, which will not be repeated here.
[0105] Step 121, based on the target preference coefficient and the weight of each finished product evaluation index corresponding to each weighting algorithm, determine the quality score number of the current batch corresponding to the target preference coefficient.
[0106] Step 122, dynamically update the golden batch based on the quality score number of the current batch corresponding to the target preference coefficient and the quality score number of the golden batch corresponding to the target preference coefficient.
[0107] Specifically, if the quality score of the current batch corresponding to the target preference coefficient is greater than the quality score of the golden batch corresponding to the target preference coefficient, the golden batch is updated to the current batch. At the same time, the golden batch can be updated on the golden batch reference curve, so that the changes of the optimal golden batch can be intuitively analyzed.
[0108] As an example only, companies usually need to conduct a comprehensive evaluation of the quality of finished products from multiple production batches to identify the golden batch. The finished product inspection data of Danhong Injection after alcohol precipitation obtained from the actual production of a pharmaceutical factory is shown in Table 1. Table 1 shows that the finished product of the drug mainly includes total solid removal rate, hydroxysafflor yellow A removal rate, tannic acid B removal rate, total phenolic acid removal rate, total sugar removal rate, and solid removal rate. The six quality variables are the finished product evaluation indicators, but their different dimensions will affect the comprehensive evaluation results, so data preprocessing will be performed to obtain a normalized data set:
[0109] Table 1
[0110]
[0111] Based on the analytic hierarchy process and entropy method, the weight of each finished product evaluation index corresponding to the analytic hierarchy process, the weight corresponding to the entropy method, and the comprehensive weight are shown in Table 2:
[0112] Table 2
[0113]
[0114] Based on the comprehensive weight corresponding to each finished product evaluation index, the quality score number corresponding to the preference coefficient of each historical batch is determined as follows: Figure 3 shown.
[0115] Figure 4 is a module schematic diagram of a production quality dynamic analysis system based on golden batch optimization according to some embodiments of this specification, such as Figure 4 As shown, the production quality dynamic analysis system based on golden batch optimization can include an indicator determination module, a data acquisition module, a data processing module, an indicator weighting module, a parameter setting module, a score calculation module, a batch determination module and a dynamic optimization module.
[0116] The indicator determination module can be used to determine multiple finished product evaluation indicators;
[0117] The data acquisition module can be used to obtain production data and finished product inspection data of multiple historical batches;
[0118] The data processing module can be used to preprocess the production data and finished product inspection data of multiple historical batches based on multiple finished product evaluation indicators, and obtain the preprocessed production data and finished product evaluation indicator data of multiple historical batches;
[0119] The indicator weighting module can be used to determine the weight of each finished product evaluation indicator corresponding to each weighting algorithm through multiple weighting algorithms;
[0120] The parameter setting module can be used to initialize the preference coefficient and set the incremental step size of the preference coefficient;
[0121] The score calculation module can be used to determine the quality score number corresponding to multiple preference coefficients for each historical batch based on the preference coefficient, incremental step size and the weight of each weighting algorithm corresponding to each finished product evaluation index;
[0122] The batch determination module may be used to determine a golden batch from multiple historical batches based on the quality score number of each historical batch corresponding to multiple preference coefficients, and determine a target preference coefficient based on the golden batch;
[0123] The data acquisition module can also be used to obtain the production data of the current batch and the finished product inspection data;
[0124] The data processing module can also be used to preprocess the production data and finished product inspection data of the current batch based on multiple finished product evaluation indicators to obtain the preprocessed production data and finished product evaluation indicator data of the current batch;
[0125] The score calculation module can also be used to determine the quality score number of the current batch corresponding to the target preference coefficient based on the target preference coefficient and the weight of each finished product evaluation index corresponding to each weighting algorithm;
[0126] The dynamic optimization module can be used to dynamically update the golden batch based on the quality score number of the current batch corresponding to the target preference coefficient and the quality score number of the golden batch corresponding to the target preference coefficient.
[0127] The production quality dynamic analysis system based on golden batch optimization can be used to execute the production quality dynamic analysis method based on golden batch optimization. For more descriptions of the production quality dynamic analysis system based on golden batch optimization, please refer to the relevant descriptions of the production quality dynamic analysis method based on golden batch optimization, which will not be repeated here.
[0128] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.
Claims
1. A dynamic analysis method for production quality based on golden batch optimization, characterized in that: include: Determine multiple finished product evaluation indicators; Obtain production data and finished product inspection data for multiple historical batches; Based on the multiple finished product evaluation indicators, preprocessing the multiple historical batches of production data and finished product inspection data to obtain preprocessed multiple historical batches of production data and finished product evaluation indicator data; Determine the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms through a plurality of weighting algorithms; Initialize the preference coefficient and set the incremental step size of the preference coefficient; Based on the preference coefficient, the incremental step size, and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms, determine the quality score number of each of the historical batches corresponding to the plurality of preference coefficients; Determining a golden batch from the multiple historical batches based on the quality score number of each of the historical batches corresponding to the multiple preference coefficients; Based on the golden batch, determining a target preference coefficient; Obtain production data and finished product inspection data for the current batch; Based on the multiple finished product evaluation indicators, preprocessing the production data and finished product inspection data of the current batch to obtain preprocessed production data and finished product evaluation indicator data of the current batch; Based on the target preference coefficient and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms, determine the quality score number of the current batch corresponding to the target preference coefficient; Dynamically update the golden batch based on the quality score number of the current batch corresponding to the target preference coefficient and the quality score number of the golden batch corresponding to the target preference coefficient; Wherein, the determination of multiple finished product evaluation indicators includes: Determine multiple candidate finished product evaluation indicators; Acquire finished product inspection data of multiple sample batches, and determine finished product evaluation index data of the sample batches based on the finished product inspection data of the sample batches, wherein the finished product evaluation index data of the sample batches includes values of the sample batches in multiple candidate finished product evaluation indexes; Based on the finished product evaluation index data of multiple sample batches, the correlation between any two candidate finished product evaluation indicators is calculated; Based on the correlation between any two candidate finished product evaluation indicators, multiple finished product evaluation indicators are determined, including: S21. For each candidate finished product evaluation indicator, based on the correlation between the candidate finished product evaluation indicator and each other candidate finished product evaluation indicator, calculate the correlation mean and correlation variance corresponding to the candidate finished product evaluation indicator, and based on the correlation mean and correlation variance corresponding to the candidate finished product evaluation indicator, calculate the screening score corresponding to the candidate finished product evaluation indicator; S22, determining whether there is at least one candidate finished product evaluation indicator whose corresponding screening score is less than the screening score threshold, if so, executing S23, if not, taking the current remaining candidate finished product evaluation indicators as finished product evaluation indicators; S23, for each candidate finished product evaluation indicator, when the screening score corresponding to the candidate finished product evaluation indicator is less than the screening score threshold, the candidate finished product evaluation indicator is determined to be an invalid candidate finished product evaluation indicator, and the invalid candidate finished product evaluation indicator is deleted. After deleting all invalid candidate finished product evaluation indicators, S21 is executed; The initializing the preference coefficient and setting the incremental step of the preference coefficient include: S31, initializing the incremental step size; S32, according to the current incremental step size, determine the preference coefficients of multiple consecutive samples, calculate the comprehensive weight of each finished product evaluation index corresponding to each sampled preference coefficient, and then determine the weight sequence corresponding to each sampled preference coefficient, wherein an element of the weight sequence represents the comprehensive weight of a finished product evaluation index corresponding to the sampled preference coefficient; S33, calculating a sequence fluctuation parameter based on a weight sequence corresponding to a plurality of consecutive sampled preference coefficients; S34, if the sequence fluctuation parameter is less than the sequence fluctuation threshold, execute S35, if the sequence fluctuation parameter is greater than or equal to the sequence fluctuation threshold, use the current incremental step size as the incremental step size of the preference coefficient; S35, adjusting the incremental step size based on the sequence fluctuation parameter, and executing S32; The series volatility parameter is calculated according to the following formula: , in, is the sequence fluctuation parameter corresponding to the current incremental step size, is the total number of sampled preference coefficients, is the comprehensive weight corresponding to the jth finished product evaluation index in the weight sequence corresponding to the preference coefficient of the nth sample, is the mean of the comprehensive weight corresponding to the jth finished product evaluation index in the weight sequence corresponding to the preference coefficient of the n+1th sample; The increment step size is adjusted based on the sequence fluctuation parameter according to the following formula: , in, is the adjusted incremental step size, is the current incremental step size, is the sequence fluctuation threshold; Based on the preference coefficient, the incremental step size, and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms, the quality score number corresponding to the plurality of preference coefficients of each of the historical batches is determined, including: S11. Calculate the comprehensive weight corresponding to each finished product evaluation indicator based on the preference coefficient and the weight of each weighting algorithm corresponding to each finished product evaluation indicator; S12. Determine the quality score number of each historical batch corresponding to the preference coefficient based on the comprehensive weight corresponding to each finished product evaluation indicator; S13, determine whether the preference coefficient meets the preset conditions, if yes, end the process, if not, execute S14; S14. Update the preference coefficient based on the incremental step size and execute S11.
2. The method for dynamic analysis of production quality based on golden batch optimization according to claim 1, characterized in that: Based on the multiple finished product evaluation indicators, preprocessing the production data and finished product inspection data of the multiple historical batches includes: Determine invalid historical batches among the multiple historical batches, and delete production data and finished product inspection data of the invalid historical batches; Perform missing data filling, normalization and standardization on the production data and finished product inspection data of valid historical batches.
3. The method for dynamic analysis of production quality based on golden batch optimization according to claim 1, characterized in that: The weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms is determined through a plurality of weighting algorithms, including: The weight of each finished product evaluation index corresponding to the hierarchical analysis algorithm is determined through the hierarchical analysis algorithm.
4. The method for dynamic analysis of production quality based on golden batch optimization according to claim 3, characterized in that: The weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms is determined through a plurality of weighting algorithms, including: By using the entropy method, the weight of each finished product evaluation index corresponding to the entropy method is determined.
5. The method for dynamic analysis of production quality based on golden batch optimization according to claim 4, characterized in that: The comprehensive weight corresponding to each of the finished product evaluation indicators is calculated based on the following formula, including: , in, is the comprehensive weight corresponding to the j-th finished product evaluation index, is the preference coefficient, is the weight of the hierarchical analysis algorithm corresponding to the j-th finished product evaluation index, is the weight of the entropy method corresponding to the j-th finished product evaluation index.
6. The method for dynamic analysis of production quality based on golden batch optimization according to any one of claims 1 to 5, characterized in that: Determining a golden batch from the multiple historical batches based on the quality score number of each of the historical batches corresponding to the multiple preference coefficients includes: For each of the preference coefficients, based on the quality score number of each of the historical batches corresponding to the preference coefficient, determining the maximum value of the quality score number corresponding to the preference coefficient; Determine an array of maximum quality score values based on the maximum quality score values corresponding to each of the preference coefficients; A golden batch is determined from the plurality of historical batches based on the maximum quality score number array.
7. The method for dynamic analysis of production quality based on golden batch optimization according to any one of claims 1 to 5, characterized in that: Based on the golden batch, a target preference coefficient is determined, including: The target preference coefficient is determined based on the quality score numbers of the golden batch corresponding to the plurality of preference coefficients.
8. The production quality dynamic analysis system based on golden batch optimization is characterized by: include: An indicator determination module, used to determine multiple finished product evaluation indicators; Data acquisition module, used to obtain production data of multiple historical batches and finished product inspection data; A data processing module, configured to preprocess the production data and finished product inspection data of the multiple historical batches based on the multiple finished product evaluation indicators, and obtain the preprocessed production data and finished product evaluation indicator data of the multiple historical batches; An indicator weighting module, used to determine the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms through multiple weighting algorithms; The parameter setting module is used to initialize the preference coefficient and set the incremental step size of the preference coefficient; A score calculation module, for determining the quality score number corresponding to the plurality of preference coefficients for each of the historical batches based on the preference coefficient, the incremental step size and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms; A batch determination module, configured to determine a golden batch from the plurality of historical batches based on the quality score number of each of the historical batches corresponding to the plurality of preference coefficients, and determine a target preference coefficient based on the golden batch; The data acquisition module is also used to obtain the production data and finished product inspection data of the current batch; The data processing module is also used to preprocess the production data and finished product inspection data of the current batch based on the multiple finished product evaluation indicators to obtain the preprocessed production data and finished product evaluation indicator data of the current batch; The score calculation module is also used to determine the quality score number of the current batch corresponding to the target preference coefficient based on the target preference coefficient and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms; A dynamic optimization module, used for dynamically updating the golden batch based on the quality score number of the current batch corresponding to the target preference coefficient and the quality score number of the golden batch corresponding to the target preference coefficient; Wherein, the determination of multiple finished product evaluation indicators includes: Determine multiple candidate finished product evaluation indicators; Acquire finished product inspection data of multiple sample batches, and determine finished product evaluation index data of the sample batches based on the finished product inspection data of the sample batches, wherein the finished product evaluation index data of the sample batches includes values of the sample batches in multiple candidate finished product evaluation indexes; Based on the finished product evaluation index data of multiple sample batches, the correlation between any two candidate finished product evaluation indicators is calculated; Based on the correlation between any two candidate finished product evaluation indicators, multiple finished product evaluation indicators are determined, including: S21. For each candidate finished product evaluation indicator, based on the correlation between the candidate finished product evaluation indicator and each other candidate finished product evaluation indicator, calculate the correlation mean and correlation variance corresponding to the candidate finished product evaluation indicator, and based on the correlation mean and correlation variance corresponding to the candidate finished product evaluation indicator, calculate the screening score corresponding to the candidate finished product evaluation indicator; S22, determining whether there is at least one candidate finished product evaluation indicator whose corresponding screening score is less than the screening score threshold, if so, executing S23, if not, taking the current remaining candidate finished product evaluation indicators as finished product evaluation indicators; S23, for each candidate finished product evaluation indicator, when the screening score corresponding to the candidate finished product evaluation indicator is less than the screening score threshold, the candidate finished product evaluation indicator is determined to be an invalid candidate finished product evaluation indicator, and the invalid candidate finished product evaluation indicator is deleted. After deleting all invalid candidate finished product evaluation indicators, S21 is executed; The initializing the preference coefficient and setting the incremental step of the preference coefficient include: S31, initializing the incremental step size; S32, according to the current incremental step size, determine the preference coefficients of multiple consecutive samples, calculate the comprehensive weight of each finished product evaluation index corresponding to each sampled preference coefficient, and then determine the weight sequence corresponding to each sampled preference coefficient, wherein an element of the weight sequence represents the comprehensive weight of a finished product evaluation index corresponding to the sampled preference coefficient; S33, calculating a sequence fluctuation parameter based on a weight sequence corresponding to a plurality of consecutive sampled preference coefficients; S34, if the sequence fluctuation parameter is less than the sequence fluctuation threshold, execute S35, if the sequence fluctuation parameter is greater than or equal to the sequence fluctuation threshold, use the current incremental step size as the incremental step size of the preference coefficient; S35, adjusting the incremental step size based on the sequence fluctuation parameter, and executing S32; The series volatility parameter is calculated according to the following formula: , in, is the sequence fluctuation parameter corresponding to the current incremental step size, is the total number of sampled preference coefficients, is the comprehensive weight corresponding to the jth finished product evaluation index in the weight sequence corresponding to the preference coefficient of the nth sample, is the mean of the comprehensive weight corresponding to the jth finished product evaluation index in the weight sequence corresponding to the preference coefficient of the n+1th sample; The increment step size is adjusted based on the sequence fluctuation parameter according to the following formula: , in, is the adjusted incremental step size, is the current incremental step size, is the sequence fluctuation threshold; Based on the preference coefficient, the incremental step size, and the weight of each of the finished product evaluation indicators corresponding to each of the weighting algorithms, the quality score number corresponding to the plurality of preference coefficients of each of the historical batches is determined, including: S11. Calculate the comprehensive weight corresponding to each finished product evaluation indicator based on the preference coefficient and the weight of each weighting algorithm corresponding to each finished product evaluation indicator; S12. Determine the quality score number of each historical batch corresponding to the preference coefficient based on the comprehensive weight corresponding to each finished product evaluation indicator; S13, determine whether the preference coefficient meets the preset conditions, if yes, end the process, if not, execute S14; S14. Update the preference coefficient based on the incremental step size and execute S11.
