CN115392408B - Method and system for detecting abnormal operation of electronic tablet counting machine - Google Patents

Method and system for detecting abnormal operation of electronic tablet counting machine Download PDF

Info

Publication number
CN115392408B
CN115392408B CN202211341208.0A CN202211341208A CN115392408B CN 115392408 B CN115392408 B CN 115392408B CN 202211341208 A CN202211341208 A CN 202211341208A CN 115392408 B CN115392408 B CN 115392408B
Authority
CN
China
Prior art keywords
data
vector
key index
abnormal
mean value
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.)
Active
Application number
CN202211341208.0A
Other languages
Chinese (zh)
Other versions
CN115392408A (en
Inventor
高芳旺
康等贵
胡小凤
张超
张展
马学龙
刘娟
牛元亮
李娜
牛元玮
朱富兰
刘江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Lukang Pharmaceutical Group Saite Co ltd
Original Assignee
Shandong Lukang Pharmaceutical Group Saite Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Lukang Pharmaceutical Group Saite Co ltd filed Critical Shandong Lukang Pharmaceutical Group Saite Co ltd
Priority to CN202211341208.0A priority Critical patent/CN115392408B/en
Publication of CN115392408A publication Critical patent/CN115392408A/en
Application granted granted Critical
Publication of CN115392408B publication Critical patent/CN115392408B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Complex Calculations (AREA)

Abstract

The invention relates to the technical field of data identification, in particular to a method and a system for detecting abnormal operation of an electronic tablet counting machine. The method comprises the steps of collecting data of key indexes of the electronic tablet counting machine by setting sampling frequency, constructing a similarity matrix according to similarity weights between every two key index data, obtaining a feature matrix corresponding to the key index data through the similarity matrix, and screening abnormal data by comparing a feature matrix row vector corresponding to the key index data with a feature matrix row vector of normal data. According to the method, the similarity weight is calculated through the intersection of the neighbor sets of the key index data and the spatial distance, and the abnormal data is screened out according to the difference between the characteristic matrix corresponding to the key index data and the characteristic matrix corresponding to the normal data, so that the method has high adaptability in different use environments and meanwhile guarantees the accuracy of abnormal data detection.

Description

Method and system for detecting abnormal operation of electronic tablet counting machine
Technical Field
The invention relates to the technical field of data identification, in particular to a method and a system for detecting abnormal operation of an electronic tablet counting machine.
Background
The abnormal operation of the electronic counting machine can affect the packaging quality of the medicines, and further can affect the benefits of the medicine users and the market reputation of the medicine enterprises, so that the electronic counting machine is necessary to detect the abnormal operation of the electronic counting machine. However, the method for manually detecting the operation abnormity of the electronic particle counter cannot realize the real-time detection of the operation condition of the electronic particle counter, so the real-time monitoring of the operation condition of the electronic particle counter can be effectively realized by adopting the existing automatic detection and identification technology.
The method for detecting the abnormal operation of the electronic particle counter by the existing automatic detection and identification technology mainly comprises the following steps: establishing a similarity matrix according to the similarity between the acquired data, obtaining a feature matrix from the similarity matrix through a spectral clustering algorithm, representing the features of the original data through vectors in the feature matrix, and analyzing and screening abnormal data according to the features. However, the traditional similarity matrix is constructed only based on the whole similarity for analysis, the similarity value is used as an element value of the similarity matrix, or when the similarity is greater than a threshold value, a weight value between two elements is set as a fixed value, otherwise, the weight value is set as 0. Therefore, the traditional construction method of the similarity matrix has poor adaptability and cannot accurately reflect the characteristics of the data.
Disclosure of Invention
In order to solve the above technical problems, an object of the present invention is to provide a method and a system for detecting an abnormal operation of an electronic tablet counting machine, wherein the adopted technical scheme is as follows:
the invention provides a method for detecting the abnormal operation of an electronic particle counter, which comprises the following steps:
and acquiring at least two key indexes in the running process of the electronic particle counting machine according to a preset sampling frequency to obtain at least two key index vectors.
Obtaining a neighbor set of each element according to each element value of the target key index vector; obtaining a similarity weight according to the difference between two elements in the target key index vector; if a neighbor set intersection exists between two elements in the target key index vector, obtaining a similarity weight according to the difference between the two elements in the target key index vector and the corresponding elements in the neighbor set intersection;
constructing a feature matrix according to the feature column vectors obtained by spectral clustering according to the similarity weight, and performing clustering analysis on the row vectors of the feature matrix to obtain at least two feature categories and a mean row vector of each feature category; screening abnormal mean value row vectors according to the difference distance between the standard mean value row vectors and the mean value row vectors; obtaining the abnormal degree of the target key index according to the difference distances corresponding to all abnormal mean value row vectors and the number of samples in the feature categories corresponding to the abnormal mean value row vectors;
obtaining the shutdown probability of each key index due to the shutdown of the electronic particle counter caused by the abnormality according to the historical data; obtaining the running state index of the electronic grain counting machine according to the abnormal degree of each key index and the corresponding shutdown probability; and judging the abnormal operation condition of the electronic grain counting machine according to the operation condition index.
Further, obtaining at least two key indicator vectors includes:
screening out noise data according to the difference between adjacent elements of the vectors; taking the mean value of the previous element and the next element of the noise data as the data after the noise data is reconstructed; when the noise data is the initial end of the key index vector, the next element of the noise data is used as the data after the noise data is reconstructed, and when the noise data is the tail end of the key index vector, the previous element of the noise data is used as the data after the noise data is reconstructed;
and correspondingly replacing the noise point data in the key index vector with the data reconstructed from the noise point data to obtain the denoised key index vector.
Further, the screening noise data according to the difference between the adjacent elements of the vector comprises:
screening out noise data in the key index vector through a noise confidence coefficient calculation model, wherein the noise confidence coefficient calculation model comprises the following steps:
Figure 565419DEST_PATH_IMAGE002
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE003
is a noise confidence level, is asserted>
Figure 236571DEST_PATH_IMAGE004
Represents an element in the target key indicator vector, is asserted>
Figure DEST_PATH_IMAGE005
Is the normalized data value corresponding to the target key index at the y moment>
Figure 721648DEST_PATH_IMAGE006
As data->
Figure 138854DEST_PATH_IMAGE005
The one element of the preceding group of elements, device for combining or screening>
Figure DEST_PATH_IMAGE007
Is data->
Figure 441832DEST_PATH_IMAGE005
Is selected, is the next element of (4)>
Figure 475647DEST_PATH_IMAGE008
A preset parameter greater than 0>
Figure DEST_PATH_IMAGE009
Is an integer greater than 0>
Figure 967677DEST_PATH_IMAGE010
Selecting a function for the minimum value;
and taking the element corresponding to the noise confidence higher than the preset noise confidence threshold value as noise data.
Further, the method for obtaining the similarity weight includes:
the similarity weight calculation model comprises the following steps:
Figure 754237DEST_PATH_IMAGE012
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE013
represents an element->
Figure 220990DEST_PATH_IMAGE014
And element +>
Figure DEST_PATH_IMAGE015
The similarity weight between them, is greater than or equal to>
Figure 850423DEST_PATH_IMAGE016
Represents an element->
Figure 772243DEST_PATH_IMAGE014
The cluster set and the element->
Figure 678888DEST_PATH_IMAGE015
The intersection of the located cluster set,. Sup.,. Of the located cluster set>
Figure DEST_PATH_IMAGE017
Represents->
Figure 316543DEST_PATH_IMAGE016
The number of the internal elements is greater or less>
Figure 574218DEST_PATH_IMAGE018
Represents the c-th intersection element within said intersection, is>
Figure DEST_PATH_IMAGE019
Is element->
Figure 955520DEST_PATH_IMAGE014
And intersection element->
Figure 982251DEST_PATH_IMAGE018
In the data on (c), in the Euclidean distance between->
Figure 135015DEST_PATH_IMAGE020
Is element +>
Figure 348827DEST_PATH_IMAGE015
And intersection element->
Figure 409187DEST_PATH_IMAGE018
In the data on (c), in the Euclidean distance between->
Figure DEST_PATH_IMAGE021
Is element->
Figure 211796DEST_PATH_IMAGE014
And element->
Figure 801040DEST_PATH_IMAGE015
The euclidean distance between the corresponding data is,
Figure 33307DEST_PATH_IMAGE022
is element->
Figure 366200DEST_PATH_IMAGE014
And element->
Figure 101943DEST_PATH_IMAGE015
Corresponding to a distance index between the data, the distance index being the sum of the Euclidean distance between the data of the two elements and the Euclidean distance between the positions of the two elements, and->
Figure DEST_PATH_IMAGE023
Is a natural constant.
Further, the screening the abnormal mean value row vector according to the difference distance between the standard mean value row vector and the mean value row vector comprises:
and calculating Euclidean distance between the mean value of the standard row vectors and the mean row vector, and recording the mean row vector of which the corresponding Euclidean distance is higher than a preset first distance threshold as an abnormal mean row vector.
Further, the standard mean line vector includes:
and obtaining a standard key index vector corresponding to normal operation data, obtaining a standard feature matrix through the standard key index vector, and calculating the mean value of standard row vectors of the standard feature matrix to obtain a standard mean value row vector.
Further, the obtaining the abnormal degree of the target key index according to the difference distances corresponding to all the abnormal mean value row vectors and the number of samples in the feature categories corresponding to the abnormal mean value row vectors includes:
and calculating a distance mean value of Euclidean distances between the abnormal mean value row vector and the standard mean value row vector, and taking the product of the distance mean value and the number of samples in the feature category corresponding to the abnormal mean value row vector as the abnormal degree.
Further, the obtaining of the operation condition index of the electronic particle counter according to the abnormal degree of the key index and the probability of shutdown of the electronic particle counter caused by the abnormal key index comprises:
the method comprises the following steps of obtaining an operation condition index of the electronic particle counting machine through an operation abnormity detection model of the electronic particle counting machine, wherein the operation abnormity detection model of the electronic particle counting machine comprises the following steps:
Figure DEST_PATH_IMAGE025
wherein the content of the first and second substances,
Figure 517881DEST_PATH_IMAGE026
for a key indicator which causes an abnormality of the electronic particle counting machine in historical data>
Figure DEST_PATH_IMAGE027
Number of occurrences, <' >>
Figure 909548DEST_PATH_IMAGE028
For the total times of shutdown caused by abnormality of the electronic particle counting machine in historical data, the judgment is carried out>
Figure DEST_PATH_IMAGE029
Is the key indicator->
Figure 436345DEST_PATH_IMAGE027
Is abnormal degree of (4), (v) is abnormal>
Figure 292174DEST_PATH_IMAGE030
Is the operating condition indicator of the electronic pellet counting machine.
The invention also provides a system for detecting the running abnormity of the electronic counting machine, which comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, wherein the processor realizes any step of the method for detecting the running abnormity of the electronic counting machine when executing the computer program.
The invention has the following beneficial effects:
according to the embodiment of the invention, the neighbor set of the elements in the target key index vector is obtained firstly, and the similarity weight is calculated according to the intersection of the neighbor set and the spatial distance between two elements in the target key index vector. And further representing the data of the target key index according to a feature matrix constructed by the similarity weight, extracting features of the data of the target key index by constructing the feature matrix, and detecting abnormal data according to the difference distance between the feature matrix corresponding to the target key index vector and the feature matrix corresponding to the data of the normal key index. The influence of the subjectivity on the abnormal operation detection of the electronic counting machine is effectively avoided, and the self-adaptive detection of different key indexes and different environments can be further realized. In conclusion, the method ensures the accuracy of abnormal data detection while having high adaptability to different use environments.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the embodiments or the description of 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 other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a flowchart of a method for detecting an abnormal operation of an electronic counting machine according to an embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description will be given to a method and a system for detecting abnormal operation of an electronic tablet counting machine according to the present invention, with reference to the accompanying drawings and preferred embodiments, and the detailed implementation, structure, features and effects thereof are described below. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following describes a specific scheme of the method and system for detecting the abnormal operation of the electronic tablet counting machine provided by the present invention in detail with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a method for detecting an abnormal operation of an electronic granule counter according to an embodiment of the present invention is shown, where the method includes:
step S1: and acquiring at least two key indexes in the running process of the electronic particle counting machine according to a preset sampling frequency to obtain at least two key index vectors.
The operation condition of the electronic particle counting machine is influenced by the operation data of multiple key indexes, the embodiment of the invention adopts a sensor or a detector to collect the operation data of the key indexes, the sensor or the detector comprises but is not limited to a power detector, a vibration frequency sensor and a speed sensor, and the key indexes comprise but is not limited to the electromagnetic vibration frequency of the electronic particle counting machine, the setting angle of a vibration plate, the power of the particle counting machine, the counting speed of a counter and the separation speed. It should be noted that the key index is specifically data that causes a failure of the electronic particle counter, and the key index may be determined by historical data that causes a failure of the electronic particle counter, which is not limited herein.
In the embodiment of the invention, a sensor or a detector is adopted to collect corresponding key indexes in real time, and the number of the collected key index types is recorded as m. Because the key index data collected by the sensor or the detector are time sequence data, and the time sequence data are continuous, the analysis of the operation data of the key index is inconvenient. Therefore, the invention carries out discrete processing on the acquired key index data, specifically: for each key index, data acquisition is carried out every T time period, and simultaneously, the data sampling frequency T and the key index acquisition data volume n are set. In the embodiment of the invention, the data acquisition time period T is set to be 10 minutes, the sampling frequency T is 1 time per second, and the number of the key index data acquisition quantity n is 200.
In order to facilitate subsequent data analysis, the collected key index data is normalized, and in the embodiment of the invention, for each key index, 0-mean normalization is adopted to normalize the collected index data to obtain each key index vector. 0 matrix normalization includes:
Figure 223221DEST_PATH_IMAGE032
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE033
is a key finger of the targetMark data->
Figure 554714DEST_PATH_IMAGE034
Is the average value of the target key index data>
Figure DEST_PATH_IMAGE035
Is the standard deviation of the target key index data>
Figure 150780DEST_PATH_IMAGE036
For target key indicator data>
Figure 346270DEST_PATH_IMAGE033
The corresponding normalized data values.
And taking the normalized key index data as elements in the key index vector, wherein the elements in all the key index vectors in the subsequent process are the normalized key index data.
When the key index of the electronic particle counter operates abnormally, the abnormal mode usually lasts for a period of time, so that the abnormal data does not appear in a small number of discrete forms for the data in the key index vector, and the influence of noise data on the data detection in the key index vector needs to be avoided. In order to avoid the influence of noise data, the noise data needs to be reconstructed, and the specific reconstruction method includes:
screening out noise data according to the difference between adjacent elements of the vectors; taking the mean value of the previous element and the next element of the noise data as data after noise data reconstruction; when the noise data is the initial end of the key index vector, the next element of the noise data is used as the data after the noise data is reconstructed, and when the noise data is the tail end of the key index vector, the previous element of the noise data is used as the data after the noise data is reconstructed; and correspondingly replacing the noise point data in the key index vector with the data reconstructed from the noise point data to obtain the denoised key index vector.
The specific method for screening the noise point comprises the following steps: and detecting the noise data by constructing a noise confidence model. The noise confidence model includes:
Figure 697485DEST_PATH_IMAGE002
wherein the content of the first and second substances,
Figure 860482DEST_PATH_IMAGE003
for noise confidence level, <' > based on the number of pixels in the image>
Figure 73289DEST_PATH_IMAGE004
Represents an element in the target key indicator vector, is asserted>
Figure 106973DEST_PATH_IMAGE005
Is the normalized data value corresponding to the target key index at the y moment>
Figure 645402DEST_PATH_IMAGE006
Is data->
Figure 295695DEST_PATH_IMAGE005
The one element of the preceding group of elements, device for combining or screening>
Figure 577771DEST_PATH_IMAGE007
As data->
Figure 997120DEST_PATH_IMAGE005
Is selected, is the next element of (4)>
Figure 424559DEST_PATH_IMAGE009
Is an integer greater than 0>
Figure 312881DEST_PATH_IMAGE008
A preset parameter greater than 0>
Figure 382337DEST_PATH_IMAGE010
Selecting a function for the minimum value, in an embodiment of the invention>
Figure 672504DEST_PATH_IMAGE008
Is set to 1;
the formula aims to screen mutation data in key indexes, and the mutation data can be accurately selected by judging whether the target data is abnormal data or not through the minimum difference value between the target data and adjacent data, so that the selection error is reduced.
And obtaining the noise confidence of the target element data through the minimum value of the difference value of the front element and the rear element of the target element data. It should be noted that, if the target element data is the start end of the key indicator vector, the noise confidence of the target element data is calculated by the difference between the target element data and the next element data of the target element; and if the target element data is the ending end of the key index vector, calculating the noise confidence of the target element data according to the difference value of the target element data and the next element data of the target element data.
And taking the element corresponding to the early point confidence coefficient higher than the preset noise confidence coefficient threshold value as noise data. In the embodiment of the invention, in order to facilitate the selection of the threshold, the noise confidence is normalized, and the threshold of the noise confidence is set to be 0.5.
In the embodiment of the invention, because a plurality of key indexes are obtained, namely a plurality of key index vectors are obtained, the key index vectors are stacked in the data storage process to obtain the key index matrix. In the key index matrix, each row represents discrete data of one key index in sequence time, and each column represents data of different key indexes in the same time.
Step S2: obtaining a neighbor set of each element according to each element value of the target key index vector; obtaining a similarity weight according to the difference between two elements in the target key index vector; and if a neighbor set intersection exists between two elements in the target key index vector, obtaining a similarity weight according to the difference between the two elements in the target key index vector and the corresponding elements in the neighbor set intersection.
Because the conventional similarity matrix is constructed based on the similarity analysis, the distance similarity between the elements is used as the element value of the similarity matrix, or the distance similarity between two elements is set to be a fixed value if the distance similarity is greater than a threshold value, otherwise, the distance similarity is set to be 0. The methods do not combine the relationship between adjacent samples, do not divide the actual situation according to the scene, the obtained similarity matrix can not specifically represent the original data, and the further spectral clustering algorithm obtains the feature matrix according to the similarity matrix to further represent the features of the original data, so that the data in the traditional similarity matrix is not accurately calculated only by the similarity between the data, and therefore, the method for calculating the similarity weight according to the intersection of the adjacent sets and the spatial distance between two elements in the target key index vector combines the relationship between the adjacent data of the target key index, so that the calculated similarity weight of the target key index data is more accurate.
After at least two key index vectors are obtained, a neighbor set of each element is obtained through a K neighbor algorithm according to the element value of the target key index vector. It should be noted that, by the technical means known to those skilled in the art of the K-nearest neighbor algorithm, the value of the specific K may be specifically set according to the length of the key indicator vector and the element value, and the present invention is not further limited and described in detail. Each element of the target key index vector belongs to a corresponding K neighbor set, and for any two elements i and j in the target key index, the corresponding K neighbor sets are respectively obtained
Figure DEST_PATH_IMAGE037
And &>
Figure 208528DEST_PATH_IMAGE038
The intersection of the neighbor set of two elements is computed and recorded as @>
Figure DEST_PATH_IMAGE039
. Note that when K neighbor set @>
Figure 247063DEST_PATH_IMAGE037
And &>
Figure 605364DEST_PATH_IMAGE038
When there is no intersection, corresponding
Figure 999305DEST_PATH_IMAGE039
Is empty, i.e.>
Figure 50437DEST_PATH_IMAGE039
The number of internal elements is 0.
And obtaining a similarity weight according to the difference between two elements in the target key index vector. If a neighbor set intersection exists between two elements in a target key index vector, obtaining a similarity weight according to the difference between the two elements in the target key index vector and the corresponding elements in the neighbor set intersection; if no neighbor set intersection exists between two elements in the target key index, the similarity weight can be directly obtained according to the difference between the two elements in the key index. The specific acquisition method of the similarity weight comprises the following steps:
and obtaining a similarity weight through a similarity weight calculation model according to the element neighbor set intersection of the target key index vector and the element difference of the key index vector. The similarity weight calculation model comprises the following steps:
Figure 162619DEST_PATH_IMAGE040
wherein, the first and the second end of the pipe are connected with each other,
Figure 42719DEST_PATH_IMAGE013
representing an element +>
Figure 307478DEST_PATH_IMAGE014
And element->
Figure 778780DEST_PATH_IMAGE015
A similarity weight therebetween, based on the comparison result>
Figure 597831DEST_PATH_IMAGE016
Represents an element->
Figure 547201DEST_PATH_IMAGE014
The cluster set and the element->
Figure 650155DEST_PATH_IMAGE015
The intersection of the located cluster set,. Sup.,. Of the located cluster set>
Figure 777511DEST_PATH_IMAGE017
Represents->
Figure 598706DEST_PATH_IMAGE016
The number of internal elements, and>
Figure 102499DEST_PATH_IMAGE018
represents the c-th intersection element within said intersection, is>
Figure 325539DEST_PATH_IMAGE019
Is element +>
Figure 623797DEST_PATH_IMAGE014
And intersection element->
Figure 932287DEST_PATH_IMAGE018
In the data on (c), in the Euclidean distance between->
Figure 223460DEST_PATH_IMAGE020
Is element +>
Figure 317318DEST_PATH_IMAGE015
And intersection element +>
Figure 35744DEST_PATH_IMAGE018
Is determined by the Euclidean distance between the data in (c), and (d)>
Figure 582263DEST_PATH_IMAGE021
Is element +>
Figure 411548DEST_PATH_IMAGE014
And element->
Figure 359912DEST_PATH_IMAGE015
Corresponding Euclidean distance between data, <' > v>
Figure 514819DEST_PATH_IMAGE022
Is element->
Figure 283055DEST_PATH_IMAGE014
And element->
Figure 447189DEST_PATH_IMAGE015
Corresponding to the distance index between the data, wherein the distance index is the sum of Euclidean distance between the data of two elements and Euclidean distance between the positions of the two elements, and the length of the Euclidean distance is greater than the length of the Euclidean distance between the data of two elements>
Figure 515639DEST_PATH_IMAGE023
Is a natural constant. It should be noted that the calculation of euclidean distances between vector elements is a means well known to those skilled in the art, and the present invention is not limited thereto.
The first item of the similarity weight calculation model is a calculation model with intersection existing in neighbor sets of two elements, the model represents a similarity weight between the two elements according to Euclidean distance between the two elements and the mean value of the Euclidean distance between the two elements and the element in the intersection of the neighbor sets, and the similarity weight is in inverse proportion to the mean value, so that when the difference between the intersection element and the element is larger, the corresponding similarity weight is smaller; the second term of the similarity weight is a calculation model without intersection according to the neighboring sets of the two elements, the model represents the similarity weight between the two elements according to the Euclidean distance between the data and the position of the two elements, the Euclidean distance is an index taking a natural constant as a base, the exponential function is taken as a denominator in a calculation model expression, and the negative correlation relationship between the Euclidean distance between the data and the position of the two elements and the similarity weight is realized, so that when the Euclidean distance between the data and the position of the two elements is larger, the corresponding similarity weight is smaller.
And step S3: constructing a feature matrix according to the feature column vectors obtained by spectral clustering of the similarity weights, and performing clustering analysis on the row vectors of the feature matrix to obtain at least two feature classes and a mean value row vector of each feature class; screening abnormal mean value row vectors according to the difference distance between the standard mean value vectors and the mean value row vectors; and obtaining the abnormal degree of the key index according to the difference distance corresponding to all abnormal mean value row vectors and the number of samples in the feature category corresponding to the abnormal mean value row vector.
And constructing a feature matrix according to the feature column vectors obtained by spectral clustering according to the obtained similarity weight, wherein the spectral clustering process comprises the following steps:
(1) Constructing according to the obtained similarity weight
Figure DEST_PATH_IMAGE041
Is based on the similarity matrix->
Figure 779130DEST_PATH_IMAGE042
. Note that the->
Figure 283929DEST_PATH_IMAGE041
Is based on the similarity matrix->
Figure 736907DEST_PATH_IMAGE042
And obtaining a similarity weight value between every two n elements in the target key index vector, wherein the similarity weight value is an element value of the similarity matrix. It should be noted that the similarity matrix uses the similarity weight between the key indicator vector elements rather than the similarity.
(2) According to the constructed similarity matrix
Figure 909132DEST_PATH_IMAGE042
And a laplacian matrix is obtained. Obtaining the laplacian matrix from the similarity matrix comprises:
Figure 156573DEST_PATH_IMAGE044
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE045
is a similarity matrix>
Figure 820773DEST_PATH_IMAGE042
Corresponding Laplace matrix,. According to the Laplace matrix,. The corresponding matrix is selected>
Figure 61130DEST_PATH_IMAGE046
Is a similarity matrix>
Figure 838593DEST_PATH_IMAGE042
Corresponding degree matrix is ^ h>
Figure 240625DEST_PATH_IMAGE041
Is based on the diagonal matrix, the degree matrix->
Figure 454437DEST_PATH_IMAGE046
Is a similarity matrix>
Figure 983639DEST_PATH_IMAGE042
The sum of the elements of the corresponding row. It should be noted that, by establishing the laplacian matrix to implement the feature mapping on the target key index, the elements with relatively high similarity in the target key index can be as close as possible after the new space after dimension reduction, and the elements with relatively low similarity can be as far away as possible in the new space after dimension reduction, which is convenient for data detection and identification.
(3) Calculating the eigenvalues of the obtained Laplace matrix, arranging the obtained eigenvalues from small to large, and selecting eigenvectors corresponding to the previous P eigenvalues
Figure DEST_PATH_IMAGE047
Each feature vector being a>
Figure 68138DEST_PATH_IMAGE048
On the basis of the P column vectors, a feature matrix is constructed>
Figure DEST_PATH_IMAGE049
The characteristic matrix is dimensioned>
Figure 578754DEST_PATH_IMAGE050
. Note that>
Figure DEST_PATH_IMAGE051
And each feature value selects only one feature vector, where->
Figure 935655DEST_PATH_IMAGE052
Is an integer greater than 0.
And performing cluster analysis on the row vectors of the feature matrix to obtain at least two feature classes, wherein each feature class corresponds to at least one row vector, and the mean value vector of the row vectors in each feature class is obtained and used as the mean value row vector. In the present example, the mean row vector is noted as
Figure DEST_PATH_IMAGE053
Wherein s is a characteristic class number, based on which>
Figure 658760DEST_PATH_IMAGE054
. It should be noted that the row vector of the feature matrix represents the feature of each element in the original target key indicator vector, and the specific clustering method is not limited herein.
And screening abnormal mean value row vectors according to the mean value of the standard mean value vectors and the difference distance between the mean value row vectors. Wherein the obtaining of the mean of the standard mean vector comprises:
and obtaining a corresponding characteristic matrix according to a key index vector corresponding to normal data in the historical data, recording the characteristic matrix as a standard characteristic matrix, taking each row vector of the standard characteristic matrix as a standard row vector, calculating the mean value of the standard row vectors, and recording the mean value as a standard mean value vector. It should be noted that the obtaining manner of the feature matrix corresponding to the normal data key index vector is consistent with the manner of obtaining the feature matrix through the target key index vector.
After obtaining the mean and the mean row vectors of the standard mean vector, calculating the mean row vectors respectively
Figure 394504DEST_PATH_IMAGE053
The Euclidean distance from the standard mean vector is found->
Figure DEST_PATH_IMAGE055
In order to facilitate subsequent data analysis, the collected Euclidean distance is subjected to normalization processing, and corresponding abnormal mean value row vectors are screened out by setting a distance threshold. In the embodiment of the present invention, the distance threshold is set to be 0.6, the mean row vector corresponding to the normalized euclidean distance higher than the distance threshold is used as the abnormal mean row vector, the feature class corresponding to the abnormal mean row vector is considered as the abnormal class, and the element data in the target key index vector corresponding to each row vector in the abnormal class is used as the abnormal data. It should be noted that the euclidean distance and normalization are well known to those skilled in the art, and are not further limited herein.
After the abnormal data are screened out, judging the abnormal degree of the target key index according to the abnormal data, and taking the product of the distance mean value and the number of samples in the feature category corresponding to the abnormal mean value row vector as the abnormal degree, namely an expression of the abnormal degree is as follows:
Figure DEST_PATH_IMAGE057
wherein the content of the first and second substances,
Figure 731813DEST_PATH_IMAGE058
is the abnormal degree of the key index of the target>
Figure DEST_PATH_IMAGE059
Is the Euclidean distance mean value between the mean value row vector corresponding to the abnormal data and the standard mean value vector, and is used for judging whether the abnormal data is abnormal or not>
Figure 326743DEST_PATH_IMAGE060
The number of abnormal data in the target key indicator vector.
In the embodiment of the invention, the abnormal degree of each key index vector is calculated based on the method, and the abnormal operation condition of the electronic counting machine is further judged according to the abnormal degree.
And step S4: obtaining the shutdown probability of the electronic particle counter shutdown caused by each kind of abnormal key index according to historical data; obtaining the running state index of the electronic grain counting machine according to the abnormal degree and the corresponding shutdown probability of each key index; and judging the abnormal operation condition of the electronic counting machine according to the operation condition index.
The operation condition of the electronic grain counting machine is influenced by the abnormal key indexes in the actual operation process, and the operation condition index of the electronic grain counting machine is further obtained according to the influence of the abnormal key indexes in the historical data on the operation condition of the electronic grain counting machine.
On the basis of obtaining the abnormal degree of each key index, the shutdown probability of the electronic particle counting machine shutdown caused by the abnormal key index in the historical data is obtained, the shutdown probability is the ratio of the number of times of the abnormal key index in the historical data to the total number of times of shutdown caused by the abnormal electronic particle counting machine in the historical data, and the shutdown probability can visually reflect the influence of each key index on the running condition of the electronic particle counting machine. And obtaining the running condition index of the electronic grain counting machine through an electronic grain counting machine running abnormity detection model according to the abnormity degree of each key index and the probability of the electronic grain counting machine caused by the abnormity of the corresponding key index in the historical data. In the embodiment of the invention, the historical data is the running condition data of the electronic counter in one year. The corresponding operation abnormity detection model of the electronic particle counter comprises the following steps:
Figure 204873DEST_PATH_IMAGE062
wherein the content of the first and second substances,
Figure 545855DEST_PATH_IMAGE026
for a key indicator which causes an abnormality of the electronic particle counting machine in historical data>
Figure 460590DEST_PATH_IMAGE027
Number of occurrences, <' >>
Figure 401871DEST_PATH_IMAGE028
For the total times of shutdown caused by abnormality of the electronic particle counting machine in historical data, the judgment is carried out>
Figure 76566DEST_PATH_IMAGE029
Is a key indicator>
Figure 521322DEST_PATH_IMAGE027
W is the number of key indicators, is based on the degree of abnormality of (4), based on the number of key indicators, is based on>
Figure 357691DEST_PATH_IMAGE030
Is an index of the running condition of the electronic particle counting machine. In the embodiment of the invention, the historical data selects the running condition data of the electronic grain counting machine within one year.
In the abnormality detection model, the shutdown probability of the shutdown of the electronic counting machine caused by the abnormality of the key index in the historical data is used as the weight of the operation condition index of the electronic counting machine, so that the influence of the abnormality degree of the key index on the operation condition of the electronic counting machine can be amplified, the error of the abnormality judgment of the operation condition of the electronic counting machine is reduced, and the calculated operation condition index of the electronic counting machine can intuitively express the operation condition of the electronic counting machine.
Normalizing the obtained running state index of the electronic counting machine, ensuring that the value of the running state index of the electronic counting machine is between 0 and 1, and setting a key index threshold value to judge the abnormal running condition of the electronic counting machine. In the embodiment of the present invention, the threshold of the key indicator is set to 0.5. And when the index value of the running state of the electronic counting granulator after normalization is higher than 0.5, the running of the electronic counting granulator is considered to be abnormal. It is to be noted that the normalization is not further limited by the technical means well known to those skilled in the art.
In summary, the invention collects the data of the key indexes of the electronic tablet counting machine by setting the sampling frequency, constructs the similarity matrix according to the similarity weight between every two key index data, obtains the feature matrix corresponding to the key index data through the similarity matrix, and screens out abnormal data by comparing the feature matrix row vector corresponding to the key index data with the feature matrix row vector of the normal data. According to the method, the similarity weight is calculated through the intersection of the neighbor sets of the key index data and the spatial distance, and the abnormal data is screened out according to the difference between the characteristic matrix corresponding to the key index data and the characteristic matrix corresponding to the normal data, so that the method has high adaptability in different use environments and meanwhile guarantees the accuracy of abnormal data detection.
The invention also provides a system for detecting the running abnormity of the electronic particle counter, which comprises a memory, a processor and a computer program which is stored in the memory and can run on the processor, wherein the processor realizes any step of the method for detecting the running abnormity of the electronic particle counter when executing the computer program.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. The processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (7)

1. An abnormal operation detection method for an electronic particle counter, which is characterized by comprising the following steps:
acquiring at least two key indexes in the running process of the electronic particle counter according to a preset sampling frequency to obtain at least two key index vectors; the type of the key index is determined by historical data causing the electronic counting machine to malfunction;
obtaining a neighbor set of each element according to each element value of the target key index vector; obtaining a similarity weight according to the difference between two elements in the target key index vector; if a neighbor set intersection exists between two elements in the target key index vector, obtaining a similarity weight according to the difference between the two elements in the target key index vector and the corresponding elements in the neighbor set intersection;
constructing a feature matrix according to the feature column vectors obtained by spectral clustering according to the similarity weight, and performing clustering analysis on the row vectors of the feature matrix to obtain at least two feature categories and a mean row vector of each feature category; screening abnormal mean value row vectors according to the difference distance between the standard mean value row vectors and the mean value row vectors; obtaining the abnormal degree of the target key index according to the difference distances corresponding to all abnormal mean value row vectors and the number of samples in the feature categories corresponding to the abnormal mean value row vectors;
obtaining the shutdown probability of each key index, which is caused by abnormality, of the electronic counting machine according to historical data; obtaining the running state index of the electronic grain counting machine according to the abnormal degree and the corresponding shutdown probability of each key index; judging the abnormal operation condition of the electronic particle counter according to the operation condition index;
the similarity weight value obtaining method comprises the following steps:
the similarity weight calculation model comprises the following steps:
Figure 472601DEST_PATH_IMAGE001
wherein, the first and the second end of the pipe are connected with each other,
Figure 237557DEST_PATH_IMAGE002
represents an element->
Figure 86565DEST_PATH_IMAGE003
And element->
Figure 398597DEST_PATH_IMAGE004
The similarity weight between them, is greater than or equal to>
Figure 711767DEST_PATH_IMAGE005
Represents an element->
Figure 913203DEST_PATH_IMAGE003
The cluster set and the element->
Figure 46244DEST_PATH_IMAGE004
The intersection of the located cluster set,. Sup.,. Of the located cluster set>
Figure 427547DEST_PATH_IMAGE006
Represents->
Figure 565530DEST_PATH_IMAGE005
The number of internal elements, and>
Figure 108507DEST_PATH_IMAGE007
represents the c-th intersection element within said intersection, is>
Figure 728844DEST_PATH_IMAGE008
Is element->
Figure 648258DEST_PATH_IMAGE003
And intersection element->
Figure 906326DEST_PATH_IMAGE007
In the data on (c), in the Euclidean distance between->
Figure 620205DEST_PATH_IMAGE009
Is element->
Figure 931100DEST_PATH_IMAGE004
And intersection element->
Figure 185364DEST_PATH_IMAGE007
In the data on (c), in the Euclidean distance between->
Figure 297939DEST_PATH_IMAGE010
Is element->
Figure 448297DEST_PATH_IMAGE003
And element->
Figure 980910DEST_PATH_IMAGE004
Corresponds to the Euclidean distance between data, <' > H>
Figure 242127DEST_PATH_IMAGE011
Is element->
Figure 209208DEST_PATH_IMAGE003
And element->
Figure 530468DEST_PATH_IMAGE004
Corresponding to a distance index between the data, the distance index being the sum of the Euclidean distance between the data of the two elements and the Euclidean distance between the positions of the two elements, and->
Figure 143852DEST_PATH_IMAGE012
Is a natural constant;
the obtaining of the running condition index of the electronic grain counting machine according to the abnormal degree of the key index and the probability of shutdown of the electronic grain counting machine caused by abnormal key index comprises:
the method comprises the following steps of obtaining an operation condition index of the electronic particle counting machine through an operation abnormity detection model of the electronic particle counting machine, wherein the operation abnormity detection model of the electronic particle counting machine comprises the following steps:
Figure 943181DEST_PATH_IMAGE013
wherein the content of the first and second substances,
Figure 233610DEST_PATH_IMAGE014
for key indicators in historical data which cause an abnormality in an electronic counting and counting machine>
Figure 725771DEST_PATH_IMAGE015
Number of occurrences, <' >>
Figure 29714DEST_PATH_IMAGE016
For the total times of shutdown caused by abnormality of the electronic counting and counting machine in the historical data, the system is combined>
Figure 632733DEST_PATH_IMAGE017
Is the key indicator->
Figure 156163DEST_PATH_IMAGE015
W is the number of key indicators, is based on the degree of abnormality of (4), based on the number of key indicators, is based on>
Figure 553647DEST_PATH_IMAGE018
Is the operating condition indicator of the electronic particle counter.
2. The method of claim 1, wherein the step of obtaining at least two key indicator vectors comprises:
screening out noise data according to the difference between adjacent elements of the vector; taking the mean value of the previous element and the next element of the noisy point data as data after the noisy point data is reconstructed; when the noise data is the initial end of the key index vector, taking the next element of the noise data as the data after the noise data is reconstructed, and when the noise data is the final end of the key index vector, taking the previous element of the noise data as the data after the noise data is reconstructed;
and correspondingly replacing the noise point data in the key index vector with the data reconstructed from the noise point data to obtain a denoised key index vector.
3. The method of claim 2, wherein the screening noisy data according to the difference between the adjacent elements of the vector comprises:
screening out noise data in the key index vector through a noise confidence coefficient calculation model, wherein the noise confidence coefficient calculation model comprises the following steps:
Figure 79306DEST_PATH_IMAGE019
wherein the content of the first and second substances,
Figure 49798DEST_PATH_IMAGE020
is a noise confidence level, is asserted>
Figure 610093DEST_PATH_IMAGE021
Represents an element in the target key indicator vector, is asserted>
Figure 444056DEST_PATH_IMAGE022
Is the normalized data value corresponding to the target key index at the y moment>
Figure 427318DEST_PATH_IMAGE023
As data->
Figure 372140DEST_PATH_IMAGE022
Is selected, is the previous element of (4)>
Figure 786941DEST_PATH_IMAGE024
Is data->
Figure 791806DEST_PATH_IMAGE022
Is selected, is the next element of (4)>
Figure 590260DEST_PATH_IMAGE025
A preset parameter greater than 0>
Figure 542036DEST_PATH_IMAGE026
Is an integer greater than 0>
Figure 76922DEST_PATH_IMAGE027
Selecting a function for the minimum value;
and taking the element corresponding to the noise confidence higher than the preset noise confidence threshold value as noise data.
4. The method of claim 1, wherein the step of selecting the abnormal average row vector according to the difference distance between the standard average row vector and the average row vector comprises:
and calculating Euclidean distance between the mean value row vector of the standard row vectors and the mean value row vector, and recording the mean value row vector of which the corresponding Euclidean distance is higher than a preset first distance threshold as an abnormal mean value row vector.
5. The method of claim 4, wherein the standard mean row vector comprises:
and obtaining a standard key index vector corresponding to normal operation data, obtaining a standard feature matrix through the standard key index vector, and calculating the mean value of standard row vectors of the standard feature matrix to obtain a standard mean value row vector.
6. The method of claim 1, wherein obtaining the abnormal degree of the target key indicator according to the difference distances corresponding to all abnormal mean value row vectors and the number of samples in the feature categories corresponding to the abnormal mean value row vectors comprises:
and calculating a distance mean value of Euclidean distances between the abnormal mean value row vector and the standard mean value row vector, and taking the product of the distance mean value and the number of samples in the feature category corresponding to the abnormal mean value row vector as the abnormal degree.
7. An electronic counting machine operation anomaly detection system, comprising a memory, a processor and a computer program which is stored in the memory and can be operated on the processor, wherein the steps of the method according to any one of claims 1 to 6 are realized when the processor executes the computer program.
CN202211341208.0A 2022-10-31 2022-10-31 Method and system for detecting abnormal operation of electronic tablet counting machine Active CN115392408B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211341208.0A CN115392408B (en) 2022-10-31 2022-10-31 Method and system for detecting abnormal operation of electronic tablet counting machine

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211341208.0A CN115392408B (en) 2022-10-31 2022-10-31 Method and system for detecting abnormal operation of electronic tablet counting machine

Publications (2)

Publication Number Publication Date
CN115392408A CN115392408A (en) 2022-11-25
CN115392408B true CN115392408B (en) 2023-04-07

Family

ID=84115107

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211341208.0A Active CN115392408B (en) 2022-10-31 2022-10-31 Method and system for detecting abnormal operation of electronic tablet counting machine

Country Status (1)

Country Link
CN (1) CN115392408B (en)

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115829422B (en) * 2023-02-21 2024-01-02 北京瀚海蓝山智能科技有限公司 Industrial equipment operation abnormal state identification method based on big data
CN115858630B (en) * 2023-02-21 2023-05-09 新风光电子科技股份有限公司 Abnormality detection method for energy storage data of energy storage power station
CN115860579B (en) * 2023-02-27 2023-05-09 山东金利康面粉有限公司 Production quality monitoring system for flour processing
CN115982602B (en) * 2023-03-20 2023-06-30 济宁众达利电气设备有限公司 Photovoltaic transformer electrical fault detection method
CN116185306B (en) * 2023-04-24 2023-07-14 山东爱福地生物股份有限公司 Sewage treatment system data storage method using potamogeton crispus
CN116561535B (en) * 2023-07-11 2023-09-19 安徽建筑大学 Individualized building interaction design processing method

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113515678A (en) * 2021-05-13 2021-10-19 上海梯之星信息科技有限公司 Abnormal data screening method
CN115171125A (en) * 2022-06-06 2022-10-11 阿里巴巴（中国）有限公司 Data anomaly detection method

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009013252A (en) * 2007-07-03 2009-01-22 Napura:Kk Fluorescent assembly and fluorescent paste composition
US10296832B1 (en) * 2015-12-05 2019-05-21 ThetaRay Ltd. System and method for detecting an undesirable event
CN112348090A (en) * 2020-11-10 2021-02-09 北京航空航天大学 Neighbor anomaly detection system based on neighbor self-encoder
CN114841621A (en) * 2022-04-30 2022-08-02 宁波大学科学技术学院 Method for detecting abnormal operation state of synthetic ammonia process based on online neighbor embedding index

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113515678A (en) * 2021-05-13 2021-10-19 上海梯之星信息科技有限公司 Abnormal data screening method
CN115171125A (en) * 2022-06-06 2022-10-11 阿里巴巴（中国）有限公司 Data anomaly detection method

Also Published As

Publication number Publication date
CN115392408A (en) 2022-11-25

Similar Documents

Publication Publication Date Title
CN115392408B (en) Method and system for detecting abnormal operation of electronic tablet counting machine
JP2020035097A (en) Image identification device, image identification method, and image identification program
CN104216349B (en) Utilize the yield analysis system and method for the sensing data of manufacturing equipment
US8037533B2 (en) Detecting method for network intrusion
US8230272B2 (en) Methods and systems for detection of anomalies in digital data streams
US20020169735A1 (en) Automatic mapping from data to preprocessing algorithms
CN117349220B (en) Data processing method and system based on PCI bus
TW200818060A (en) Clustering system, and defect kind judging device
CN113723157B (en) Crop disease identification method and device, electronic equipment and storage medium
CN109257383B (en) BGP anomaly detection method and system
CN111612038A (en) Abnormal user detection method and device, storage medium and electronic equipment
WO2007020466A2 (en) Data classification apparatus and method
CN114707608A (en) Medical quality control data processing method, apparatus, device, medium, and program product
CN117197591B (en) Data classification method based on machine learning
CN117093947B (en) Power generation diesel engine operation abnormity monitoring method and system
CN113553319A (en) LOF outlier detection cleaning method, device and equipment based on information entropy weighting and storage medium
CN107895361A (en) A kind of hyperspectral abnormity detection method based on local density&#39;s purifying background
CN113971426A (en) Information acquisition method, device, equipment and storage medium
CN115204323B (en) Seed multi-feature based clustering and synthesis method, system, device and medium
CN108415958B (en) Weight processing method and device for index weight VLAD features
CN116434197A (en) Abnormal action detection closed loop iterative optimization method, system, medium and equipment
CN117789907B (en) Intelligent medical data intelligent management method based on multi-source data fusion
EP4187438A1 (en) Object sample selection for training of neural networks
CN118052540A (en) Intelligent management method and system for equipment maintenance scheme
US20230409422A1 (en) Systems and Methods for Anomaly Detection in Multi-Modal Data Streams

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
GR01 Patent grant
GR01 Patent grant