CN-121998194-A - Data collaborative optimization method and system for cross-cigarette factory

CN121998194ACN 121998194 ACN121998194 ACN 121998194ACN-121998194-A

Abstract

The application discloses a data collaborative optimization method and a system crossing cigarette factories, and relates to the technical field of tobacco, wherein the method comprises the steps of collecting data of multiple factories MES (manufacturing execution system) and performing ETL (extract-transform-load) cleaning standardization to form a unified quality data pool; grouping according to brand, model and standard dimensions, calculating the sequence similarity of different factory quality indexes with the same dimension by using a dynamic time warping algorithm, integrating the sequence similarity into a multi-dimensional standard matching similarity matrix, inputting the matrix into a characteristic importance evaluation model based on a gradient lifting decision tree, extracting a key technological parameter characteristic set influencing quality fluctuation, constructing a collaborative optimization objective function by the key characteristic set, searching a pareto optimal solution set by using a multi-objective global optimization algorithm to generate a balanced multi-objective optimal technological parameter adjustment scheme, and issuing the pareto optimal solution set to MES control ends of each factory as a scheme to realize closed loop of collaborative scheme generation and implementation and improve data utilization rate and accuracy.

Inventors

LI QIAN
WANG SHIHONG
LI YUN
LI SHUMING
HU TAO
ZHU CHEN
ZHAO YIFAN

Assignees

红云红河烟草(集团)有限责任公司

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. A data collaborative optimization method across cigarette factories, the data collaborative optimization method being applied to a plurality of cigarette factories having MES systems, the data collaborative optimization method comprising: step S1, collecting factory quality inspection data, process special inspection data and equipment information data through an MES system of each cigarette factory, and performing cleaning and standardization processing through an ETL flow to obtain a factory-crossing quality data pool with a uniform format; Step S2, grouping the cross-factory quality data pools according to brand, model and standard dimensions based on preset data dimensions, and integrating to obtain a grouping result set; Step S3, calculating the quality index sequence similarity among different factories of the same brand, different factories of the same model and different factories of the same standard in the grouping result set through a dynamic time warping algorithm, and integrating to obtain a multidimensional standard matching similarity matrix; s4, inputting the multidimensional scaling similarity matrix into a feature importance evaluation model based on a gradient lifting decision tree to extract key process parameter features influencing quality fluctuation, and integrating to obtain a key process parameter feature set; S5, constructing a cross-factory collaborative optimization objective function based on the key process parameter feature set, and searching a pareto optimal solution set of the cross-factory collaborative optimization objective function through a multi-objective global optimization algorithm; And S6, issuing the pareto optimal solution set to a control end of an MES system of each cigarette factory as an optimal process parameter adjustment scheme.
2. The data collaborative optimization method according to claim 1, wherein step S1, collecting factory quality inspection data, process specific inspection data, and equipment information data through an MES system of each cigarette factory, and performing cleaning and standardization processing through an ETL flow, to obtain a factory-crossing quality data pool with a uniform format, comprises: Step S11, extracting factory quality inspection data, process special inspection data and equipment information data in real time through an automatic data interface of an MES system of each cigarette factory, and obtaining an original data set based on one cigarette factory; step S12, preprocessing an original data set of each cigarette factory and filling missing values through outlier rejection and median filling, and obtaining a primary cleaning data set based on the original data set; Step S13, mapping heterogeneous fields of the current preliminary cleaning data set into unified names and types according to the cross-factory unified metadata specification, and carrying out dimension normalization processing on the numerical data through Min-Max normalization to obtain an intermediate data set with unified format; Step S14, classifying the current intermediate data set according to the factory identification to obtain a classified data set; and S15, integrating the classified data sets of all cigarette factories to obtain a factory-crossing quality data pool with a uniform format.
3. The method of collaborative optimization of data according to claim 1, wherein step S2, grouping the cross-factory quality data pools by brand, model, and standard dimensions based on a preset data dimension, and integrating to obtain a grouping result set, includes: Step S21, extracting field data of three dimensions of brands, machine types and standards from the cross-factory quality data pool to form a grouping reference data set comprising identification records and dimension fields; S22, merging the identification records with the same brand value in the grouping reference data set into a group by using the brand field as a grouping key through a grouping aggregation algorithm to obtain a brand dimension grouping result set; S23, merging the identification records with the same model value in the grouping reference data set into a group by using the model field as a grouping key through a grouping aggregation algorithm to obtain a model dimension grouping result set; s24, merging the identification records with the same standard value in the group reference data set into a group by using the standard field as a grouping key through a grouping aggregation algorithm to obtain a standard dimension grouping result set; step S25, integrating the brand dimension grouping result set, the model dimension grouping result set and the standard dimension grouping result set into grouping result sets which are grouped according to brand, model and standard dimensions.
4. The method of claim 1, wherein step S3 calculates quality index sequence similarities among different factories of the same brand, different factories of the same model, and different factories of the same standard in the grouping result set by a dynamic time warping algorithm, and integrates the quality index sequence similarities to obtain a multidimensional scaling similarity matrix, and the method comprises the following steps: Step S31, extracting the grouping result set based on quality index sequences of different factories in a brand dimension, quality index sequences of different factories in a model dimension and quality index sequences of different factories in a standard dimension, and sequentially obtaining a brand dimension factory quality sequence set, a model dimension factory quality sequence set and a standard dimension factory quality sequence set; step S32, calculating the similarity of the brand dimension factory quality sequence set based on quality index sequences among different factories of the same brand through a dynamic time warping algorithm to obtain a brand dimension factory similarity matrix; Step S33, calculating the similarity of the model dimension factory quality sequence set based on the quality index sequences among different factories of the same model through a dynamic time warping algorithm to obtain a model dimension factory similarity matrix; Step S34, calculating the similarity of the quality sequence set of the standard dimension factory based on the quality index sequences among different factories of the same standard by a dynamic time warping algorithm to obtain a standard dimension factory similarity matrix; Step S35, integrating the brand dimension factory similarity matrix, the model dimension factory similarity matrix and the standard dimension factory similarity matrix according to dimensions to obtain a multi-dimensional standard comparison similarity matrix.
5. The data collaborative optimization method according to claim 1, wherein step S4, inputting the multi-dimensional benchmarking similarity matrix into a feature importance evaluation model based on a gradient lifting decision tree to extract key process parameter features affecting quality fluctuation, and integrating to obtain a key process parameter feature set, includes: step S41, extracting process parameter characteristic data and corresponding quality fluctuation index data from the cross-factory quality data pool, and integrating to obtain a process parameter quality data set; Step S42, performing dimension association on the multi-dimensional benchmarking similarity matrix and the technological parameter quality data set according to factory identifications, brands, models and standards to obtain a similarity and characteristic fusion target data set; Step S43, inputting the similarity and feature fusion target data set into a feature importance assessment model based on a gradient lifting decision tree, taking quality fluctuation index data as a target variable and process parameter feature data as input features, and training the feature importance assessment model through a gradient lifting decision tree algorithm to obtain a pre-trained feature importance assessment model; Step S44, calculating importance scores of the process parameter feature data through a pre-trained gradient lifting decision tree model; And step S45, sorting the importance scores, and extracting process parameter feature data with the median higher than the feature importance score to obtain a key process parameter feature set.
6. The data collaborative optimization method according to claim 1, wherein step S5, constructing a cross-factory collaborative optimization objective function based on the key process parameter feature set, and searching for a pareto optimal solution set of the cross-factory collaborative optimization objective function through a multi-objective global optimization algorithm, comprises: Step S51, extracting historical adjustment quantity data and quality index data corresponding to the key process parameter feature set from the cross-factory quality data pool; Step S52, taking the historical adjustment quantity data as a decision variable, and constructing a cross-factory collaborative optimization objective function by taking the minimum variance of quality index data of each cigarette factory and the minimum mean difference of quality index data under the same brand, same model or same standard dimension as a target; step S53, the upper limit and the lower limit of the historical adjustment quantity data are used as boundary constraints of the decision variables, and a constrained decision variable definition domain is formed; Step S54, inputting the cross-factory collaborative optimization objective function and the decision variable definition domain into a non-dominant order genetic algorithm, and performing global optimization on the cross-factory collaborative optimization objective function through the non-dominant order genetic algorithm to obtain a non-dominant solution set; and step S55, repeated solution rejection and feasibility verification are carried out on the non-dominant solution set, and the pareto optimal solution set is obtained.
7. The method according to claim 1, wherein step S6, wherein issuing the pareto optimal solution set as an optimal process parameter adjustment scheme to the control end of the MES system of each cigarette factory includes: Step S61, using a factory mark as a grouping key, merging the technological parameter adjustment schemes corresponding to each cigarette factory in the pareto optimal solution set into independent subsets through a grouping aggregation algorithm to form an adjustment scheme set grouped according to the cigarette factories; step S62, converting each cigarette factory subset in the adjustment scheme set into a structural transmission format supported by a MES system control end of each cigarette factory respectively through a data serialization algorithm to obtain a factory adaptation adjustment scheme set; step S63, pushing the factory adaptation adjustment scheme set to the control end of the MES system of the corresponding cigarette factory through the open control end automation data interface of the MES system of each cigarette factory.
8. A data collaborative optimization system across a cigarette factory, the data collaborative optimization system being applied to the data collaborative optimization method of any one of claims 1-7, the data collaborative optimization system comprising: the cross-factory quality data pool construction module is used for acquiring factory quality inspection data, process special inspection data and equipment information data through an MES system of each cigarette factory, and performing cleaning and standardization processing through an ETL flow to obtain a cross-factory quality data pool with a uniform format; The cross-factory quality data grouping module is used for grouping the cross-factory quality data pools according to brand, model and standard dimensions based on preset data dimensions, and integrating to obtain a grouping result set; The quality data similarity calculation module is used for calculating the quality index sequence similarity among different factories of the same brand, different factories of the same model and different factories of the same standard in the grouping result set through a dynamic time warping algorithm, and integrating the quality index sequence similarity to obtain a multidimensional standard matching similarity matrix; The key process parameter feature acquisition module is used for inputting the multidimensional benchmarking similarity matrix into a feature importance evaluation model based on a gradient lifting decision tree so as to extract key process parameter features influencing quality fluctuation and integrate the key process parameter features to obtain a key process parameter feature set; The cross-factory collaborative optimization module is used for constructing a cross-factory collaborative optimization objective function based on the key process parameter feature set, and searching a pareto optimal solution set of the cross-factory collaborative optimization objective function through a multi-objective global optimization algorithm; The optimal solution set issuing module is used for issuing the pareto optimal solution set serving as an optimal process parameter adjustment scheme to the control end of the MES system of each cigarette factory.
9. An electronic device comprising a processor and a memory coupled to the processor, the memory storing program instructions executable by the processor, the processor implementing the data co-optimization method of any one of claims 1-7 when executing the program instructions stored by the memory.
10. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein program instructions, which when executed by a processor are capable of implementing the data co-optimization method according to any one of claims 1 to 7.

Description

Data collaborative optimization method and system for cross-cigarette factory Technical Field The application relates to the technical field of tobacco, in particular to a data collaborative optimization method and system for a cross-cigarette factory. Background At present, the upper department mainly adopts a sampling inspection mode to check the wrapping workshop of the cigarette factory, and statistics of full rate, top quality rate, appearance defect rate, physical index defect rate and the like is carried out, and as the sample is sampled once every month, the sample quantity in one checking period is less, and the sampled sample has larger uncertainty, a certain factory checking index can have larger fluctuation, so that the data value mining effect of the upper data notification on the cigarette factory is limited, and the quality fluctuation condition of each factory is not favorable to be mastered in time. The existing upper-level data notification adopts file issuing, even if the data is displayed through sail softwares, the physical indexes only check whether the data accords with the design range or not, and the appearance data is not analyzed in place, so that the cigarette factory needs to analyze the notified data again for a second time, the data utilization efficiency is low, the quality trend and key problem points cannot be intuitively reflected, meanwhile, the existing notification does not establish multidimensional association, and cross-factory, cross-model and cross-quality comparison cannot be realized, so that the data utilization efficiency is difficult to improve. Disclosure of Invention The application mainly aims to provide a data collaborative optimization method and a system for a cross-cigarette factory, which are used for solving the problem of low data utilization efficiency of cross-factory, cross-model and cross-quality comparison in the prior art. In order to achieve the above object, the present application provides the following technical solutions: A data collaborative optimization method across cigarette factories, the data collaborative optimization method being applied to a plurality of cigarette factories having MES systems, the data collaborative optimization method comprising: step S1, collecting factory quality inspection data, process special inspection data and equipment information data through an MES system of each cigarette factory, and performing cleaning and standardization processing through an ETL flow to obtain a factory-crossing quality data pool with a uniform format; Step S2, grouping the cross-factory quality data pools according to brand, model and standard dimensions based on preset data dimensions, and integrating to obtain a grouping result set; Step S3, calculating the quality index sequence similarity among different factories of the same brand, different factories of the same model and different factories of the same standard in the grouping result set through a dynamic time warping algorithm, and integrating to obtain a multidimensional standard matching similarity matrix; s4, inputting the multidimensional scaling similarity matrix into a feature importance evaluation model based on a gradient lifting decision tree to extract key process parameter features influencing quality fluctuation, and integrating to obtain a key process parameter feature set; S5, constructing a cross-factory collaborative optimization objective function based on the key process parameter feature set, and searching a pareto optimal solution set of the cross-factory collaborative optimization objective function through a multi-objective global optimization algorithm; And S6, issuing the pareto optimal solution set to a control end of an MES system of each cigarette factory as an optimal process parameter adjustment scheme. Step S1 to step S6 has the beneficial effects that: The method comprises the steps of collecting multi-factory MES system data through an ETL cleaning standardization, forming a unified quality data pool, laying a cross-factory data collaborative basis, realizing data structuring organization according to brand, model and product rule dimensions in a grouping mode, enabling data structuring organization to support dimension alignment, calculating similarity of different factory quality index sequences in the same dimension through a dynamic time warping algorithm in a step S3, integrating the similarity of the multi-dimensional alignment similarity matrix, quantifying factory quality performance relevance, inputting the matrix into a feature importance evaluation model based on a gradient lifting decision tree in a step S4, extracting key process parameter feature sets affecting quality fluctuation, clearly optimizing directions, constructing a collaborative optimization objective function according to the key feature sets, searching a pareto optimal solution set through a multi-objective global optimizing algorithm, generating an optimal proces