CN113962554A - Slow feature clustering-based dual-granularity cigarette quality online evaluation method - Google Patents
Slow feature clustering-based dual-granularity cigarette quality online evaluation method Download PDFInfo
- Publication number
- CN113962554A CN113962554A CN202111232270.1A CN202111232270A CN113962554A CN 113962554 A CN113962554 A CN 113962554A CN 202111232270 A CN202111232270 A CN 202111232270A CN 113962554 A CN113962554 A CN 113962554A
- Authority
- CN
- China
- Prior art keywords
- quality
- slow
- subclass
- evaluation
- grade
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000011156 evaluation Methods 0.000 title claims abstract description 176
- 235000019504 cigarettes Nutrition 0.000 title claims abstract description 84
- 230000003068 static effect Effects 0.000 claims abstract description 49
- 238000004519 manufacturing process Methods 0.000 claims abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 24
- 230000002159 abnormal effect Effects 0.000 claims description 38
- 239000011159 matrix material Substances 0.000 claims description 21
- 241000208125 Nicotiana Species 0.000 claims description 11
- 235000002637 Nicotiana tabacum Nutrition 0.000 claims description 11
- 238000001514 detection method Methods 0.000 claims description 11
- 238000005056 compaction Methods 0.000 claims description 8
- 238000009423 ventilation Methods 0.000 claims description 8
- 238000012216 screening Methods 0.000 claims description 7
- 239000000126 substance Substances 0.000 claims description 7
- 238000012163 sequencing technique Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 230000007774 longterm Effects 0.000 claims description 4
- 238000003825 pressing Methods 0.000 claims description 4
- 238000012545 processing Methods 0.000 claims description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 3
- 238000013441 quality evaluation Methods 0.000 abstract description 15
- 230000008569 process Effects 0.000 abstract description 12
- 238000004458 analytical method Methods 0.000 description 6
- 238000007689 inspection Methods 0.000 description 5
- 238000013210 evaluation model Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000002087 whitening effect Effects 0.000 description 4
- 230000005856 abnormality Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 3
- 238000001303 quality assessment method Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 235000019505 tobacco product Nutrition 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 238000003324 Six Sigma (6σ) Methods 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000000053 physical method Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06395—Quality analysis or management
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Game Theory and Decision Science (AREA)
- Primary Health Care (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Manufacturing & Machinery (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a dual-granularity cigarette quality online evaluation method based on slow feature clustering. The invention also provides an idea of dividing the quality grade of the cigarettes by taking the double slow characteristics, the horizontal evaluation index and the speed evaluation index as the basis, and thinly adopts the static slow characteristics and the dynamic slow characteristics in the production process of the cigarettes to respectively represent the quality information and the fluctuation information in the production process of the cigarettes, thereby establishing a scientific online cigarette quality evaluation model and solving the problems that the quality of products is difficult to reasonably evaluate depending on manual drawing check and too much variables in the cigarette making process.
Description
Technical Field
The invention belongs to the field of product quality evaluation, and particularly relates to a slow feature clustering-based dual-granularity cigarette quality online evaluation method.
Background
At present, the labor productivity and the intelligence level of a cigarette making enterprise are low, and a lot of operations still depend on manual work. With the improvement of the social informatization level in recent years, more and more factories convert the tasks which need manual operation in the past into automatic assembly line completion, and the used equipment and algorithms are advanced more and more. The domestic cigarette factory also speeds up the steps of technology upgrading and industry upgrading.
The quality evaluation of cigarette products is an important link in the tobacco production process, and the quality of cigarettes can directly influence the development and economic benefit of a cigarette enterprise. However, the production process is complex and has a lot of variables, and it is difficult to establish a scientific evaluation system and algorithm to reflect the quality of the cigarette products in time.
At present, the inspection of cigarette factories in the cigarette packet receiving process depends on manual spot inspection, the intelligentization level is low, a large amount of manpower is consumed, the inspection depends on experience knowledge, and no clear standard exists, so that the inspection report of cigarette products excessively depends on the result of manual inspection. At the present stage, the quality grade of the product in the production process can only be screened and divided manually, and the product quality evaluation accuracy is low due to the fact that large uncertainty exists. Meanwhile, due to the limited manpower, the sampling frequency and the number of samples are restricted by the workload of personnel. The quality of the cigarette product cannot be reflected in time when the sampling frequency is too low, and the current production process cannot be guided; an excessively small number of samples may result in an unrepresentative test sample and may be difficult to reflect on the true quality of the product. The cigarette samples which are manually detected are difficult to recycle, so if the frequency of manual detection and the number of the samples are increased, the cigarette scrap quantity is increased.
As a typical multi-quality-variable product, the physical measurement indexes of the cigarette in the rolling process comprise quality parameters such as suction resistance, tobacco shred moisture, tobacco shred belt position and the like, and correlation of different degrees exists among the quality indexes. The quality of products in different batches and between machines is influenced by environmental conditions, so that the quality fluctuates. At present, the quality evaluation method of cigarette products is still more traditional, such as six sigma method and Delphi expert consultation method. The method is mostly based on expert experience and fails to consider the coupling relationship among variables, each variable is evaluated in an isolated manner, and meanwhile, the quality evaluation obtained by the traditional method is usually delayed and has no real-time property. Therefore, the quality evaluation of cigarette products based on the traditional method is difficult to objectively and comprehensively reflect the product quality and the production condition at the current moment.
In conclusion, the automation and intelligence level of the cigarette making enterprises in the aspect of quality evaluation needs to be improved urgently, and better treatment cannot be performed in the prior art aiming at the problems that detection depends on manpower, sampling frequency, sample quantity is restricted by the workload of personnel and the like in the cigarette production process.
Disclosure of Invention
In view of the above, the invention aims to provide a slow feature clustering-based dual-granularity cigarette quality online evaluation method, so as to realize rapid and accurate evaluation of cigarette grades based on cigarette related measurement data.
A slow feature clustering-based dual-granularity cigarette quality online evaluation method comprises the following steps:
acquiring historical data of a cigarette product production process, and preprocessing and analyzing slow characteristics of the historical data at each moment to obtain static slow characteristics and dynamic slow characteristics;
performing coarse-grained classification on quality grades based on static slow characteristics, comprising: clustering the static slow characteristics to divide quality grades, and determining a grade control limit of each quality grade according to a level evaluation index of an evaluation product corresponding to each quality grade;
performing fine-grained quality grade subclass division based on the dynamic slow characteristic, comprising the following steps: clustering the dynamic slow characteristics to divide quality grade subclasses, and determining a subclass control limit of each quality grade subclass according to a speed evaluation index of an evaluation product corresponding to each quality grade subclass;
respectively calculating a level evaluation index and a speed evaluation index according to static slow characteristics and dynamic slow characteristics corresponding to the online data aiming at the online data of the current moment in the production process of the cigarette products;
and after coarse-grained evaluation is carried out according to the level evaluation index and the grade control limit to determine the quality grade, fine-grained evaluation is carried out according to the speed evaluation index and the subclass control limit to determine the quality grade subclass.
In one embodiment, the pre-processing of the data includes data normalization and first order difference processing of the data at adjacent time instants. The data generated in the production process of the cigarette product comprises the suction resistance, the tobacco shred moisture, the circumference actual value, the circumference detection _ pressing plate actual position, the circumference deviation value, the circumference detection pollution value, the current weight average value, the current weight deviation, the long-term standard deviation, the short-term standard deviation, the position of a compaction end, the current compaction amount, the position of a tobacco shred suction belt, the current value of the total cigarette ventilation degree of the previous cigarette and the current value of the total cigarette ventilation degree of the next cigarette, and each type of data is used as a dimension to form a multi-dimensional matrix.
In one embodiment, a density-based clustering algorithm is employed to cluster the static slow features and the dynamic slow features to determine quality levels and quality level sub-classes. Preferably, the determined quality classes are 4 classes, respectively good, medium, bad, and the corresponding quality class subclasses are 2 classes, respectively normal subclass (normal) and abnormal subclass (abnormal), respectively denoted by n and a.
In one embodiment, the determining the grade control limit of each quality grade according to the level evaluation index of the evaluation product corresponding to each quality grade includes:
aiming at each quality grade, calculating a level evaluation index according to the static slow characteristic of an evaluation product, sequencing serial numbers and screening confidence degrees of all water evaluation indexes, and then taking the level evaluation index corresponding to the evaluation product serial number for determining a control limit as a grade control limit corresponding to the quality grade;
the determining the subclass control limit of each quality grade subclass according to the speed evaluation index of the evaluation product corresponding to each quality grade subclass comprises the following steps:
and aiming at each quality grade subclass, calculating speed evaluation indexes according to the dynamic slow characteristics of the evaluation items, sequencing the serial numbers of all the speed evaluation indexes and screening confidence degrees, and then taking the speed evaluation index corresponding to the evaluation item serial number for determining the control limit as the subclass control limit corresponding to the quality grade subclass.
In one embodiment, the level assessment indicator is calculated from the static slow feature using equation (1) below:
wherein q iscIndicating the overall level evaluation index corresponding to the c-th quality grade,represents the static slow characteristics of all the evaluators belonging to the c quality grades, tr () represents the trace of the matrix, the level evaluation index of the ith evaluator belonging to the c quality grade Representing a static slow feature of an ith assessor belonging to a c-th quality class;
calculating a speed evaluation index according to the dynamic slow characteristic by adopting the following formula (2):
wherein the content of the first and second substances,denotes the cOverall speed evaluation indexes corresponding to the normal subclasses of the quality grades,represents the dynamic slow characteristics of all the evaluators belonging to the c-th quality grade normal subclass, tr () represents the trace of the matrix, the speed evaluation index of the i-th evaluators belonging to the c-th quality grade subclass And (4) representing the dynamic slow characteristic of the ith assessment product belonging to the c-th quality grade normal subclass, wherein N is the number of the assessment products.
In one embodiment, the coarse-grained evaluation to determine the quality level according to the level evaluation indicator and the level control limit comprises:
and dividing the online data into quality grades corresponding to the maximum membership degree according to the membership degree of the online data and each quality grade, and then performing abnormity judgment by using grade control limits corresponding to the quality grades and the level evaluation indexes.
In one embodiment, the dividing the online data into quality grades corresponding to the maximum membership degree according to the membership degree of the online data and each quality grade comprises:
firstly, calculating membership according to Euclidean distance between static slow features corresponding to online data of an evaluation product and a static slow feature mean value corresponding to each quality grade:
wherein the content of the first and second substances,indicating the ith evaluation article for the c-th qualityThe degree of membership of the magnitude classes,expressing the Euclidean distance between the static slow feature of the ith assessment product and the static slow feature mean value corresponding to the c-th quality grade, and k represents the number of the quality grades;
then, according to the membership degree, dividing the online data of the evaluation product into quality grades corresponding to the maximum membership degree so as to determine the quality grades of the online data;
the abnormal judgment by using the grade control limit corresponding to the quality grade and the level evaluation index comprises the following steps:
and when the level evaluation index of the online data exceeds the control limit of the corresponding quality grade, the online data is considered to belong to an abnormal grade, otherwise, the online data is a normal quality grade, and the quality grade subclass evaluation is required.
In one embodiment, the determining the quality level subclass by performing fine-grained evaluation according to the speed evaluation index and the subclass control limit includes:
and dividing the dynamic slow characteristics into normal subclasses and abnormal subclasses according to the membership degree of the dynamic slow characteristics and each quality grade subclass, and then performing abnormal judgment according to the subclass control limit and the speed evaluation index corresponding to the normal subclasses.
In one embodiment, the dividing the dynamic slow features into normal subclasses and abnormal subclasses according to the membership of the dynamic slow features and each quality class subclass includes: firstly, calculating membership according to Euclidean distance between dynamic slow characteristics of an evaluation product and a dynamic slow characteristic mean value corresponding to each quality grade subclass:
where cn denotes the normal sub-class of the c-th quality class, ca denotes the anomaly subclass of the c-th quality level,representing the degree of membership of the ith assessor to the c-th quality class subclass,representing the degree of membership to the c-th quality level anomaly subclass,representing the Euclidean distance between the dynamic slow characteristic of the ith assessment product and the dynamic slow characteristic mean value corresponding to the c quality grade subclass,representing the Euclidean distance between the dynamic slow feature of the ith assessment product and the dynamic slow feature mean value corresponding to the c quality grade abnormal subclass;
then, the dynamic slow characteristics of the evaluation product are classified into normal or abnormal according to the membership degree.
In one embodiment, the performing an abnormality determination according to the subclass control limit and the speed evaluation index corresponding to the normal subclass includes:
and when the speed evaluation index of the online data exceeds the control limit of the normal subclass of the corresponding quality grade, the online data is considered to belong to the abnormal subclass, otherwise, the online data is considered to be the normal subclass.
Compared with the prior art, the invention has the beneficial effects that at least:
by using multivariate data in the cigarette process as variables, extracting static slow characteristics and dynamic slow characteristics in the production process by using slow characteristic analysis, using the static characteristics as the basis of coarse granularity evaluation, using the dynamic characteristics as the basis of fine granularity evaluation, and obtaining different quality grades based on double granularity clustering. A level evaluation index and a speed evaluation index are defined to comprehensively evaluate the product quality. The product quality evaluation task is completed from thick to thin by combining dynamic and static. And finally, calculating the membership degree of the online cigarette data according to the quality grade determined offline, establishing an online cigarette quality evaluation strategy, and evaluating the quality characteristics of cigarette products in real time and in a refined manner.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow diagram providing off-line modeling according to an embodiment;
FIG. 2 is a diagram of an online evaluation trip provided by an embodiment;
FIG. 3 is a diagram of a partial slow feature distribution corresponding to four classes of quality levels provided by an embodiment;
FIG. 4 is a graph illustrating membership of an evaluator to each quality class according to one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the detailed description and specific examples, while indicating the scope of the invention, are intended for purposes of illustration only and are not intended to limit the scope of the invention.
The method aims at the conditions that the product quality is difficult to evaluate due to a plurality of variables in the cigarette production process, and also aims at the problems that the product quality cannot be fed back on line and is inaccurate due to the fact that the detection in the cigarette production process depends on manpower, the sampling frequency and the number of samples are limited by the workload of personnel. The embodiment provides a dual-granularity cigarette quality online evaluation method based on slow feature clustering. The online evaluation method comprises an offline evaluation model establishing stage and an online quality evaluation stage, wherein the offline evaluation model establishing stage is used for dividing quality grades and quality grade subclasses according to multivariate historical data in the cigarette production process, and meanwhile, the grade control limit of the quality grades and the subclass control limit of the quality grade subclasses are also determined. And the online quality evaluation stage is used for performing coarse-grained quality grade division and fine-grained quality grade division on online data by using the quality grade and the quality grade subclass determined in the off-line evaluation model establishing stage and the corresponding grade control limit and subclass control limit so as to realize online evaluation of cigarette quality. Each stage is explained in detail below.
Stage of establishing evaluation model off line
Fig. 1 is a flowchart of an offline modeling provided in an embodiment, and as shown in fig. 1, the offline modeling stage mainly includes the following steps:
step 1-1, acquiring historical data of a cigarette product production process at multiple moments, and after preprocessing and slow characteristic analysis are carried out on the historical data at each moment, obtaining static slow characteristics and dynamic slow characteristics.
In the embodiment, the acquired historical data generated in the production process of the tobacco products has j measured variables, and t historical data are described as a two-dimensional matrix X (t X j). Wherein j represents the number of variables, t represents the number of samples, and j and t are positive integers. The actual measurement variables are state parameters which can be measured in the operation process and comprise suction resistance, tobacco shred moisture, a circumference actual value, a circumference detection _ pressing plate actual position, a circumference deviation value, a circumference detection pollution value, a current weight average value, a current weight deviation, a long-term standard deviation, a short-term standard deviation, a position of a compaction end, a current compaction amount, a tobacco shred suction belt position, a current value of the total ventilation degree of the cigarette in the previous cigarette and a current value of the total ventilation degree of the cigarette in the next cigarette.
After obtaining the historical data, the historical data needs to be standardized and subjected to first-order difference processing of data at adjacent moments. The specific process is as follows:
for input historical dataAre all normalized, i.e. the measured j variables are normalizedObtaining normalized data x (t) ═ x1(t),…,xj(t)]TWhereinRepresenting time-series variablesAverage value over time t.
Calculating the difference of the sample data of the two-dimensional matrix X at each time t: wherein the content of the first and second substances,representing the first order difference of sample data x (t), such that the sample data at each time t is expanded toIn the format, the difference at the time t-1 is calculated from the measurement information at the time t-0.
And (4) performing slow feature analysis on the preprocessed sample data, wherein the slow feature analysis comprises data whitening, slow feature solving and slow feature screening.
Specifically, data whitening refers to singular value decomposition of the covariance matrix of the input data x (t), eliminating the correlation between variables:
<xxT>t=UΛUT (1)
wherein, U is a characteristic vector matrix, Λ is a characteristic value matrix, and Q ═ Λ-1/2UTIs a whitening matrix, symbolIs expressed asEqual to.
And (3) slow feature solving:for a set of non-linear mappings, the input is mapped to a latent variable s by the non-linear mappingj(t)=gj(x (t)), also called slow features, it is desirable to minimize the variation of the slow features:
wherein the symbol: ═ represents is defined as,<·>tsymbol representation of time sequence variableAverage value over time t. Bringing formula (2) into formula (3) to obtain:
wherein, the matrix P is WQ-1The transformation matrix between the slow features s (t) and the whitening data z (t), and the optimization problem of equation (3) can be transformed into:
wherein p isjRepresents pjThe matrix P is the j-th column of the matrix P, and after singular value decomposition is performed on the covariance matrix of the whitened data, the matrix P of the optimal solution is obtained, because s ═ Pz, z (t) ═ Qx, the slow eigen projection matrix can be written as:
W=PQ=PΛ-1/2UT (6)
solving the slow static characteristic from s-WxAnd dynamic slow features derived from first order difference dataWherein s isdAndis extracted slow features, s, which can represent the main trend of the changeeAndis a secondary feature where slow feature changes are relatively fast.
Selecting the number of slow features: the main slow feature s needs to be screened outdAndand (5) further analyzing. Main slow characteristic number h ═ Rm-cnt (F), where cnt (F) denotes the number of elements in the F set, RmFor all slow feature numbers, F is the following set:
F={si|Δ(si)>maxj{Δ(xj)}} (7)
step 1-2, performing coarse-grained quality grading based on static slow characteristics, comprising: and clustering the static slow features to divide quality grades, and determining the grade control limit of each quality grade according to the level evaluation index of the evaluation product corresponding to each quality grade.
Slow characteristics of cigarettes sdThe method characterizes the trend that the quality information changes slowly along with the time in the cigarette production process. Thus, the static slow feature sdThe quality level information of the current product can be well reflected. For static slow characteristics s of cigarettesdCoarse-grained quality assessment using a density-based clustering algorithm (DBSCAN algorithm) to classify multiple quality classes Yc(c ═ 1,2, …, k). The DBSCAN algorithm is started from a certain selected core point and is continuously expanded to a region with reachable density, so that a maximized region comprising the core point and boundary points is obtained, the densities of any two points in the region are connected, and noise data can be well distinguished. In an embodiment, the quality grades can be divided into four grades of good, medium and poorThe grades, i.e. c, take values of 1,2,3 and 4 respectively.
On the basis of obtaining a plurality of quality grades, determining the grade control limit of each quality grade according to the level evaluation index of the evaluation product corresponding to each quality grade, wherein the grade control limit comprises the following steps:
firstly, for each quality grade, calculating a level evaluation index according to the static slow characteristic of an evaluation product, and specifically adopting the following formula (8) to calculate the level evaluation index:
wherein q iscIndicating the overall level evaluation index corresponding to the c-th quality grade,represents the static slow characteristics of all the evaluators belonging to the c quality grades, tr () represents the trace of the matrix, the level evaluation index of the ith evaluator belonging to the c quality grade Representing the static slow signature of the ith assessor belonging to the c-th quality class.
Because the level evaluation index can only reflect the average quality information of different quality grade clusters, and does not have an exact range to process samples with abnormal indexes, the control limit Ct is calculatedcThe metrics are further constrained. The specific process is as follows: sequencing serial numbers and screening confidence degrees of all water evaluation indexes, and then taking a level evaluation index corresponding to the serial number of the evaluation product for determining the control limit as a grade control limit corresponding to the quality grade, wherein the formula is expressed as follows:
qc=sort(qc) (9)
num=confidence*size(qc) (10)
wherein sort (. circle.) represents the sequence number sorting, and size (. circle.) represents qcThe confidence level represents the confidence level, the value range is 95% -99%, in each quality grade, the level evaluation indexes of all products are sorted from small to large by using a formula (9), and then the sample serial number of the control limit is calculated by using a formula (10). And finally, calculating the actual grade control limit of each quality grade according to the formula (11). Level evaluation index qcThe smallest corresponding quality is the best quality level, the second smallest corresponding sub-best quality level, and so on. When the confidence value is preferably 99%, the grade control limit of the obtained four types of quality grades is Ct1=[0.1863,19.9918],Ct2=[13.8758,33.4530],Ct3=[68.1530,111.3914],Ct4=[124.4502,134.1401]。
Step 1-3, performing fine-grained quality grade subclass division based on dynamic slow characteristics, comprising: and clustering the dynamic slow characteristics to divide quality grade subclasses, and determining the subclass control limit of each quality grade subclass according to the speed evaluation index of the evaluation product corresponding to each quality grade subclass.
Dynamic slow characteristics of cigarettesThe fluctuation change of the quality information in the cigarette production process is represented, namely whether the quality of the current product is stable or not and the change is small. Thus dynamic slow featureThe method can well reflect the condition of the fluctuation of the quality of the current product. For the dynamic slow characteristics of the obtained cigarettes of each quality gradeAnd performing fine-grained quality evaluation by using a density-based clustering algorithm (DBSCAN algorithm) to obtain subclasses of corresponding quality grades. Application hairNow, the distribution of the dynamic slow characteristics is obvious, so that the dynamic slow characteristics are clusteredAnd the noise cluster obtained by clustering is regarded as a dynamic characteristic abnormal subclass, and other clusters are dynamic characteristic normal subclasses.
In the examples, the speed evaluation index is used to evaluate the quality fluctuation of the sample. On the basis of obtaining a plurality of quality grade subclasses, determining the subclass control limit of each quality grade subclass according to the speed evaluation index of the evaluation product corresponding to each quality grade subclass, wherein the subclass control limit comprises the following steps:
firstly, for each quality grade subclass, calculating a speed evaluation index according to the dynamic slow characteristic of an evaluation product, and calculating the speed evaluation index by adopting the following formula (12):
wherein the content of the first and second substances,represents the overall speed evaluation index corresponding to the c-th quality grade normal subclass,represents the dynamic slow characteristics of all the evaluators belonging to the c-th quality grade normal subclass, tr () represents the trace of the matrix, the speed evaluation index of the i-th evaluators belonging to the c-th quality grade subclass And (4) representing the dynamic slow characteristic of the ith assessment product belonging to the c-th quality grade normal subclass, wherein N is the number of the assessment products.
Because the level evaluation index can only reflect the average quality information of subclasses of different quality grades and does not have an exact range to process the abnormal sample of the index, the control limit is calculatedThe metrics are further constrained. The specific process is as follows: sequencing the serial numbers and screening the confidence degrees of all the speed evaluation indexes, and then taking the speed evaluation index corresponding to the evaluation product serial number for determining the control limit as a subclass control limit corresponding to a quality grade subclass, wherein the formula is expressed as follows:
wherein, confidence represents the confidence coefficient, and the value range is 95-99%. In each quality class subclass, the speed evaluation indexes of all products are sorted from small to large by using formula (13), and then the sample serial numbers of the control limits are found by using formula (14). And finally, calculating the actual subclass control limit of each quality class subclass according to the formula (15).
On-line quality assessment phase
FIG. 2 is an online evaluation trip graph provided by an embodiment. As shown in fig. 2, the online quality assessment phase includes the following steps:
and 2-1, respectively calculating a level evaluation index and a speed evaluation index according to the static slow characteristic and the dynamic slow characteristic corresponding to the online data aiming at the online data of the current time in the production process of the tobacco products.
In the embodiment, the online data also comprises a suction resistance, tobacco shred moisture, a circumference actual value, a circumference detection _ pressing plate actual position, a circumference deviation value, a circumference detection pollution value, a current weight average value, a current weight deviation, a long-term standard deviation, a short-term standard deviation, a position of a compaction end, a current compaction amount, a tobacco shred suction belt position, a current cigarette total ventilation degree value of a previous cigarette and a current cigarette total ventilation degree value of a next cigarette.
After the online data is obtained, the online data is preprocessed, namely the online data is preprocessedSubtracting the mean of the variables in offline modelingTo obtain xnew(t), simultaneously subtracting the value of the previous moment to obtain difference data to obtain preprocessed dataThe formula is expressed as:
for pre-processed dataIn the embodiment, the slow characteristic analysis process of the step 1-1 is adopted to obtain dataCorresponding static slow feature sdnewAnd dynamic slow featureThen, the processes of the step 1-2 and the step 1-3 are adopted to calculate the horizontal evaluation index q according to the static slow characteristic and the dynamic slow characteristic corresponding to the online datanewAnd speed evaluation index
And 2-2, performing coarse grain evaluation according to the level evaluation index and the grade control limit to determine the quality grade.
In an embodiment, performing coarse-grained evaluation to determine a quality rating in accordance with a level evaluation indicator and a rating control limit includes: and dividing the online data into quality grades corresponding to the maximum membership degree according to the membership degree of the online data and each quality grade, and then performing abnormity judgment by using grade control limits corresponding to the quality grades and the level evaluation indexes.
The method comprises the following steps of dividing online data into quality grades corresponding to the maximum membership degrees according to the membership degrees of the online data and each quality grade, wherein the quality grades comprise: firstly, according to the static slow characteristic s corresponding to the online data of the assessment productdnewEuclidean distance calculation membership degree of static slow feature mean value corresponding to each quality gradeThen, according to the membership degree, dividing the online data of the evaluation product into quality grades corresponding to the maximum membership degree to determine the quality grades of the online data, that is, the quality grades in fig. 2 Wherein, the calculation formula of the membership degree is as follows:
wherein the content of the first and second substances,representing the degree of membership of the ith assessor to the c-th quality level,and k represents the number of the quality grades, wherein the Euclidean distance between the static slow characteristic of the ith assessment product and the static slow characteristic mean value corresponding to the c-th quality grade is represented.
In an embodiment, the determining the abnormality by using the level control limit and the level evaluation index corresponding to the quality level includes: evaluating the index q when the level of the online data isnewExceeding the control limit Ct of the corresponding quality classcWhen (i.e. the) And in time, the online data is considered to belong to an abnormal grade, otherwise, the online data is a normal quality grade, and the quality grade subclass needs to be evaluated.
And 2-3, performing fine-grained evaluation according to the speed evaluation index and the subclass control limit to determine quality grade subclasses.
After coarse grain evaluation, fine grain evaluation is performed on the data of the normal quality grade. In an embodiment, performing fine-grained evaluation to determine a quality level subclass according to a speed evaluation index and a subclass control limit includes: according to dynamic slow characteristicsAnd dividing the dynamic slow characteristics into a normal subclass and an abnormal subclass according to the membership degree of each quality class subclass, and then carrying out abnormal judgment according to the subclass control limit corresponding to the normal subclass and the speed evaluation index.
In one embodiment, the dynamic slow features are assigned according to their membership level to each quality class subclassThe classification into normal subclass and abnormal subclass, i.e. in FIG. 2, whenBelong to the dynamic characteristic exception subclass whenBelong to the dynamic characteristicsAnd characterizing normal subclasses, wherein ca represents abnormal subclasses at the c-th quality level, and cn represents normal subclasses at the c-th quality level.
Wherein, the calculation formula of the membership degree is as follows:
wherein cn represents the normal sub-class of the c-th quality level, ca represents the abnormal sub-class of the c-th quality level,indicating the degree of membership of the ith evaluator to the normal subclass of the c-th quality class,representing the degree of membership to the c-th quality level anomaly subclass,the Euclidean distance between the dynamic slow characteristic of the ith assessment product and the dynamic slow characteristic mean value corresponding to the c quality grade normal subclass is represented,representing the Euclidean distance between the dynamic slow feature of the ith assessment product and the dynamic slow feature mean value corresponding to the c quality grade abnormal subclass;
after the membership value is obtained, the dynamic slow characteristics of the evaluation product are classified into normal class or abnormal class according to the membership value.
In an embodiment, the performing an abnormality determination according to the subclass control limit and the speed evaluation index corresponding to the normal subclass includes: speed evaluation index of on-line dataBeyond the control limit of the normal subclass of the corresponding quality class (i.e. the) And if so, considering that the online data belongs to the abnormal subclass, and otherwise, judging that the online data belongs to the normal subclass.
In one embodiment, if the quality class subclasses are divided into 2, the normal subclasses and abnormal subclasses indicated by the symbols "+" and "-" are respectively assigned, and the online data belonging to the abnormal subclasses are divided into the classes corresponding to the poor quality classes and denoted as "-" classes. And for the online data belonging to the normal subclasses, dividing the online data into better classes under the corresponding quality grades, and marking the better classes as the class of "+", so that the fine-grained classification is completed.
For an online sample, the membership degree is high for a certain quality grade, but the level evaluation index q is highnewThe abnormal samples are classified into abnormal samples and normal samples according to coarse-grained evaluation, namely, the abnormal samples and the normal samples are classified. Evaluating the index q for levelnessnewAnd classifying the normal samples which are not overrun according to the dynamic slow characteristics, and then judging whether the speed evaluation indexes are overrun to finish fine-grained classification. The coarse-grained classification reflects the current quality condition of the product, and the fine-grained classification reflects the fluctuation condition of the product quality under the same quality grade. The quality grades are shown in table 1:
TABLE 1 quality classification rating
In the method for on-line evaluation of the quality of the dual-granularity cigarettes based on slow feature clustering, firstly, multivariate data in the cigarette production process is used as variables, static slow features and dynamic slow features are extracted through slow feature analysis, the static slow features are clustered and divided into different quality grades, and then an on-line cigarette quality evaluation strategy is established according to an off-line model. The invention also provides an idea of dividing the quality grade of the cigarettes by taking the double slow characteristics, the horizontal evaluation index and the speed evaluation index as the basis, and thinly adopts the static slow characteristics and the dynamic slow characteristics in the production process of the cigarettes to respectively represent the quality information and the fluctuation information in the production process of the cigarettes, thereby establishing a scientific online cigarette quality evaluation model and solving the problems that the quality of products is difficult to reasonably evaluate depending on manual drawing check and too much variables in the cigarette making process.
The above-mentioned embodiments are intended to illustrate the technical solutions and advantages of the present invention, and it should be understood that the above-mentioned embodiments are only the most preferred embodiments of the present invention, and are not intended to limit the present invention, and any modifications, additions, equivalents, etc. made within the scope of the principles of the present invention should be included in the scope of the present invention.
Claims (10)
1. A slow feature clustering-based dual-granularity cigarette quality online evaluation method comprises the following steps:
acquiring historical data of a cigarette product production process, and preprocessing and analyzing slow characteristics of the historical data at each moment to obtain static slow characteristics and dynamic slow characteristics;
performing coarse-grained classification on quality grades based on static slow characteristics, comprising: clustering the static slow characteristics to divide quality grades, and determining a grade control limit of each quality grade according to a level evaluation index of an evaluation product corresponding to each quality grade;
performing fine-grained quality grade subclass division based on the dynamic slow characteristic, comprising the following steps: clustering the dynamic slow characteristics to divide quality grade subclasses, and determining a subclass control limit of each quality grade subclass according to a speed evaluation index of an evaluation product corresponding to each quality grade subclass;
respectively calculating a level evaluation index and a speed evaluation index according to static slow characteristics and dynamic slow characteristics corresponding to the online data aiming at the online data of the current moment in the production process of the cigarette products;
and after coarse-grained evaluation is carried out according to the level evaluation index and the grade control limit to determine the quality grade, fine-grained evaluation is carried out according to the speed evaluation index and the subclass control limit to determine the quality grade subclass.
2. The slow feature clustering-based dual-granularity cigarette quality online evaluation method according to claim 1, wherein the preprocessing of the data comprises data standardization and first-order difference processing of data at adjacent moments;
the data generated in the production process of the cigarette product comprises the suction resistance, the tobacco shred moisture, the circumference actual value, the circumference detection _ pressing plate actual position, the circumference deviation value, the circumference detection pollution value, the current weight average value, the current weight deviation, the long-term standard deviation, the short-term standard deviation, the position of a compaction end, the current compaction amount, the position of a tobacco shred suction belt, the current value of the total cigarette ventilation degree of the previous cigarette and the current value of the total cigarette ventilation degree of the next cigarette, and each type of data is used as a dimension to form a multi-dimensional matrix.
3. The dual-granularity cigarette quality online evaluation method based on slow feature clustering of claim 1, characterized in that a density-based clustering algorithm is adopted to cluster static slow features and dynamic slow features to determine quality grades and quality grade subclasses;
preferably, the determined quality classes are 4 classes, which are respectively good, medium and poor, and the corresponding quality class subclasses are 2 classes, which are respectively normal subclass and abnormal subclass.
4. The slow-feature-clustering-based dual-granularity cigarette quality online evaluation method according to claim 1, wherein the step of determining the grade control limit of each quality grade according to the level evaluation index of the evaluation product corresponding to each quality grade comprises the following steps:
aiming at each quality grade, calculating a level evaluation index according to the static slow characteristic of an evaluation product, sequencing serial numbers and screening confidence degrees of all water evaluation indexes, and then taking the level evaluation index corresponding to the evaluation product serial number for determining a control limit as a grade control limit corresponding to the quality grade;
the determining the subclass control limit of each quality grade subclass according to the speed evaluation index of the evaluation product corresponding to each quality grade subclass comprises the following steps:
and aiming at each quality grade subclass, calculating speed evaluation indexes according to the dynamic slow characteristics of the evaluation items, sequencing the serial numbers of all the speed evaluation indexes and screening confidence degrees, and then taking the speed evaluation index corresponding to the evaluation item serial number for determining the control limit as the subclass control limit corresponding to the quality grade subclass.
5. The slow feature clustering-based dual-granularity cigarette quality online evaluation method according to claim 1 or 4, characterized in that the following formula (1) is adopted to calculate a level evaluation index according to static slow features:
wherein q iscIndicating the overall level evaluation index corresponding to the c-th quality grade,represents the static slow characteristics of all the evaluators belonging to the c quality grades, tr () represents the trace of the matrix, the level evaluation index of the ith evaluator belonging to the c quality grade Representing a static slow feature of an ith assessor belonging to a c-th quality class;
calculating a speed evaluation index according to the dynamic slow characteristic by adopting the following formula (2):
wherein the content of the first and second substances,represents the overall speed evaluation index corresponding to the c-th quality grade normal subclass,represents the dynamic slow characteristics of all the evaluators belonging to the c-th quality grade normal subclass, tr () represents the trace of the matrix, the speed evaluation index of the i-th evaluators belonging to the c-th quality grade subclass And (4) representing the dynamic slow characteristic of the ith assessment product belonging to the c-th quality grade normal subclass, wherein N is the number of the assessment products.
6. The slow feature clustering-based dual-granularity cigarette quality online evaluation method according to claim 1, wherein the coarse-granularity evaluation is performed according to a level evaluation index and a grade control limit to determine a quality grade, and the method comprises the following steps:
and dividing the online data into quality grades corresponding to the maximum membership degree according to the membership degree of the online data and each quality grade, and then performing abnormity judgment by using grade control limits corresponding to the quality grades and the level evaluation indexes.
7. The slow-feature-clustering-based dual-granularity cigarette quality online evaluation method according to claim 6, wherein the step of dividing online data into quality grades corresponding to the maximum membership degrees according to the membership degrees of the online data and each quality grade comprises the steps of:
firstly, calculating membership according to Euclidean distance between static slow features corresponding to online data of an evaluation product and a static slow feature mean value corresponding to each quality grade:
wherein the content of the first and second substances,representing the degree of membership of the ith assessor to the c-th quality level,expressing the Euclidean distance between the static slow feature of the ith assessment product and the static slow feature mean value corresponding to the c-th quality grade, and k represents the number of the quality grades;
then, according to the membership degree, dividing the online data of the evaluation product into quality grades corresponding to the maximum membership degree so as to determine the quality grades of the online data;
the abnormal judgment by using the grade control limit corresponding to the quality grade and the level evaluation index comprises the following steps:
and when the level evaluation index of the online data exceeds the control limit of the corresponding quality grade, the online data is considered to belong to an abnormal grade, otherwise, the online data is a normal quality grade, and the quality grade subclass evaluation is required.
8. The slow-feature-clustering-based dual-granularity cigarette quality online evaluation method according to claim 1, wherein performing fine-grained evaluation to determine quality grade subclasses according to a speed evaluation index and a subclass control limit comprises:
and dividing the dynamic slow characteristics into normal subclasses and abnormal subclasses according to the membership degree of the dynamic slow characteristics and each quality grade subclass, and then performing abnormal judgment according to the subclass control limit and the speed evaluation index corresponding to the normal subclasses.
9. The slow feature clustering-based dual-granularity cigarette quality online evaluation method according to claim 8, wherein the step of dividing the dynamic slow features into normal subclasses and abnormal subclasses according to the membership degree of the dynamic slow features and each quality class subclass comprises the steps of:
firstly, calculating membership according to Euclidean distance between dynamic slow characteristics of an evaluation product and a dynamic slow characteristic mean value corresponding to each quality grade subclass:
wherein cn represents the normal sub-class of the c-th quality level, ca represents the abnormal sub-class of the c-th quality level,indicating the degree of membership of the ith evaluator to the normal subclass of the c-th quality class,representing the degree of membership to the c-th quality level anomaly subclass,the Euclidean distance between the dynamic slow characteristic of the ith assessment product and the dynamic slow characteristic mean value corresponding to the c quality grade normal subclass is represented,representing the Euclidean distance between the dynamic slow feature of the ith assessment product and the dynamic slow feature mean value corresponding to the c quality grade abnormal subclass;
then, the dynamic slow characteristics of the evaluation product are classified into normal or abnormal according to the membership degree.
10. The slow-feature-clustering-based dual-granularity cigarette quality online evaluation method according to claim 1, wherein the abnormal judgment is performed according to subclass control limits and speed evaluation indexes corresponding to normal subclasses, and comprises the following steps:
and when the speed evaluation index of the online data exceeds the control limit of the normal subclass of the corresponding quality grade, the online data is considered to belong to the abnormal subclass, otherwise, the online data is considered to be the normal subclass.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111232270.1A CN113962554A (en) | 2021-10-22 | 2021-10-22 | Slow feature clustering-based dual-granularity cigarette quality online evaluation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111232270.1A CN113962554A (en) | 2021-10-22 | 2021-10-22 | Slow feature clustering-based dual-granularity cigarette quality online evaluation method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113962554A true CN113962554A (en) | 2022-01-21 |
Family
ID=79466141
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111232270.1A Pending CN113962554A (en) | 2021-10-22 | 2021-10-22 | Slow feature clustering-based dual-granularity cigarette quality online evaluation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113962554A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1975706A (en) * | 2005-11-28 | 2007-06-06 | 颐中烟草(集团)有限公司 | Cigarette organoleptic quality qualitative index estimating method |
CN104616323A (en) * | 2015-02-28 | 2015-05-13 | 苏州大学 | Space-time significance detecting method based on slow characteristic analysis |
US20200256837A1 (en) * | 2017-10-31 | 2020-08-13 | East China University Of Science And Technology | Electronic nose instrument for sensory quality evaluation of tobacco and tobacco product |
CN111949003A (en) * | 2020-07-17 | 2020-11-17 | 浙江浙能技术研究院有限公司 | Closed-loop control loop performance evaluation method based on SFA and Hellinger distance |
-
2021
- 2021-10-22 CN CN202111232270.1A patent/CN113962554A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1975706A (en) * | 2005-11-28 | 2007-06-06 | 颐中烟草(集团)有限公司 | Cigarette organoleptic quality qualitative index estimating method |
CN104616323A (en) * | 2015-02-28 | 2015-05-13 | 苏州大学 | Space-time significance detecting method based on slow characteristic analysis |
US20200256837A1 (en) * | 2017-10-31 | 2020-08-13 | East China University Of Science And Technology | Electronic nose instrument for sensory quality evaluation of tobacco and tobacco product |
CN111949003A (en) * | 2020-07-17 | 2020-11-17 | 浙江浙能技术研究院有限公司 | Closed-loop control loop performance evaluation method based on SFA and Hellinger distance |
Non-Patent Citations (1)
Title |
---|
LIMING ZHU, ET AL.: "Residual Attention Dual Autoencoder for Anomaly Detection and Localization in Cigarette Packaging", IEEE, 8 November 2020 (2020-11-08), pages 475 - 480 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110223196B (en) | Anti-electricity-stealing analysis method based on typical industry feature library and anti-electricity-stealing sample library | |
CN108805202B (en) | Machine learning method for electrolytic bath fault early warning and application thereof | |
CN110634080B (en) | Abnormal electricity utilization detection method, device, equipment and computer readable storage medium | |
CN115993807B (en) | Production monitoring optimization control method and system for silicon carbide | |
CN111126865B (en) | Technology maturity judging method and system based on technology big data | |
CN110738232A (en) | grid voltage out-of-limit cause diagnosis method based on data mining technology | |
CN113627735A (en) | Early warning method and system for safety risk of engineering construction project | |
CN111967717A (en) | Data quality evaluation method based on information entropy | |
CN111402236A (en) | Hot-rolled strip steel surface defect grading method based on image gray value | |
CN111539657A (en) | Typical electricity consumption industry load characteristic classification and synthesis method combined with user daily electricity consumption curve | |
CN115147208A (en) | Supply chain financial credit risk evaluation method and system based on artificial intelligence | |
CN113642666A (en) | Active enhanced soft measurement method based on sample expansion and screening | |
CN117193222A (en) | Intelligent quality control system based on industrial Internet of things and big data and control method thereof | |
CN111077876B (en) | Power station equipment state intelligent evaluation and early warning method, device and system | |
CN116227786A (en) | Unmanned aerial vehicle comprehensive efficiency evaluation system | |
CN117453764A (en) | Data mining analysis method | |
CN113726558A (en) | Network equipment flow prediction system based on random forest algorithm | |
CN117251814A (en) | Method for analyzing electric quantity loss abnormality of highway charging pile | |
CN113962554A (en) | Slow feature clustering-based dual-granularity cigarette quality online evaluation method | |
CN113298148B (en) | Ecological environment evaluation-oriented unbalanced data resampling method | |
CN112288571B (en) | Personal credit risk assessment method based on rapid construction of neighborhood coverage | |
CN114625781A (en) | Commodity housing value-based batch evaluation method | |
CN115271442A (en) | Modeling method and system for evaluating enterprise growth based on natural language | |
CN114757495A (en) | Membership value quantitative evaluation method based on logistic regression | |
CN112069633A (en) | Power distribution network data preprocessing method based on particle swarm principle and adopting big data clustering |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |