CN116821833A - Data anomaly detection method for machine filter cloth adhesion - Google Patents
Data anomaly detection method for machine filter cloth adhesion Download PDFInfo
- Publication number
- CN116821833A CN116821833A CN202311090952.2A CN202311090952A CN116821833A CN 116821833 A CN116821833 A CN 116821833A CN 202311090952 A CN202311090952 A CN 202311090952A CN 116821833 A CN116821833 A CN 116821833A
- Authority
- CN
- China
- Prior art keywords
- data
- sequence
- monocycle
- cluster
- suspected
- 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.)
- Granted
Links
- 239000004744 fabric Substances 0.000 title claims abstract description 42
- 238000001514 detection method Methods 0.000 title claims abstract description 17
- 230000002159 abnormal effect Effects 0.000 claims abstract description 219
- 230000005856 abnormality Effects 0.000 claims abstract description 56
- 238000000034 method Methods 0.000 claims abstract description 39
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000010606 normalization Methods 0.000 claims description 9
- 230000011218 segmentation Effects 0.000 claims description 9
- 238000012545 processing Methods 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 abstract description 2
- 230000000875 corresponding effect Effects 0.000 description 114
- 239000012065 filter cake Substances 0.000 description 8
- 239000000047 product Substances 0.000 description 7
- 239000007787 solid Substances 0.000 description 6
- 238000001914 filtration Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 239000000126 substance Substances 0.000 description 4
- 238000009826 distribution Methods 0.000 description 3
- 239000002245 particle Substances 0.000 description 3
- 238000003825 pressing Methods 0.000 description 3
- 238000009825 accumulation Methods 0.000 description 2
- 238000010009 beating Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000005260 corrosion Methods 0.000 description 2
- 230000007797 corrosion Effects 0.000 description 2
- 239000003112 inhibitor Substances 0.000 description 2
- 239000007788 liquid Substances 0.000 description 2
- 150000007524 organic acids Chemical class 0.000 description 2
- 230000001105 regulatory effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000000903 blocking effect Effects 0.000 description 1
- 238000012824 chemical production Methods 0.000 description 1
- 238000004140 cleaning Methods 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 239000000706 filtrate Substances 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 239000011148 porous material Substances 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000010561 standard procedure Methods 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 238000004065 wastewater treatment Methods 0.000 description 1
Landscapes
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention relates to the technical field of data processing, in particular to a data anomaly detection method for machine filter cloth adhesion, which comprises the following steps: dividing the amplitude time sequence data sequence into a plurality of monocycle sequences, and determining the suspected abnormality degree of the monocycle sequences according to the data quantity in the monocycle sequences and the analysis of the data difference between the monocycle sequences. The method comprises the steps of obtaining a suspected abnormal degree sequence, dividing the suspected abnormal degree sequence into a plurality of clusters, distinguishing a normal cluster from a suspected abnormal cluster, determining the abnormal degree of the suspected abnormal cluster according to data in the suspected abnormal cluster, the data quantity in a corresponding monocycle sequence and the corresponding noise interference degree, determining the K distance of the suspected abnormal cluster, combining the K distance of the normal cluster, determining the K distance of each data in the amplitude time sequence data sequence, and obtaining abnormal data by using an LOF algorithm. According to the invention, noise data is filtered through the self-adaptive K distance, and the accuracy of abnormal data detection is ensured.
Description
Technical Field
The invention relates to the technical field of data processing, in particular to a data anomaly detection method for machine filter cloth adhesion.
Background
The solid is separated from the liquid by filtering the suspension containing the solid particles through a filter press, and can be used for removing the solid particles from the processes of wastewater treatment, mineral processing, chemical production, food processing and the like to obtain clean liquid or recycle the solid matters. The solid organic acid corrosion inhibitor in the form of filter cake is prepared through chemical reaction, compression and filtration, and then the solid organic acid corrosion inhibitor is granulated, dried and packaged into a product. When the separation reaches a certain degree, the filter cake is filled in the cavity of the filter press, the operation cannot be continued along with the increase of the filtering pressure, the filter cake on the filter cloth of the filter pressing plate is required to be removed by the automatic dismounting device through vibration beating, but the filter cloth is possibly affected by the uneven distribution of filtrate, the pressure distribution and the vibration force, so that the filter cake is formed and the quality is different, the vibration parameters of the intelligent adjusting device are required, and after the vibration beating is prevented, the filter cake particles still remain on the filter cloth, so that the filtration is insufficient or the next filtering effect is affected. Therefore, a data abnormal machine detection method for machine filter cloth adhesion needs to be set, vibration parameters are adjusted by analyzing abnormal data, and filter cake cleaning is guaranteed.
The vibration sensor can be arranged on the filter cloth to detect abnormal data in the vibration signal and analyze the adhesion condition of the filter press cloth, so that the automatic dismounting device is accurately regulated and controlled, and the filter cake is ensured to be cleaned. The LOF algorithm is a commonly used data anomaly detection algorithm, and determines the degree of anomaly by comparing the local density of each data point with the local densities of adjacent points, which gives an anomaly score for each data point, so that the result is more interpretable.
The existing problems are as follows: because the operation of equipment, the vibration of filter cloth, the flow of fluid and data transmission and the sensor itself are influenced by factors such as electromagnetic interference, noise exists in the vibration signal on the filter cloth of the collected machine, and the accuracy of detecting abnormal data by the LOF algorithm is influenced. And when the K distance in the LOF algorithm is selected to be too small, the sensitivity is excessive, noise data is easy to be misjudged as abnormal data, and when the K distance is selected to be too large, the detection accuracy of the abnormal data is low.
Disclosure of Invention
The invention provides a data anomaly detection method for machine filter cloth adhesion, which aims to solve the existing problems.
The invention relates to a data anomaly detection method for machine filter cloth adhesion, which adopts the following technical scheme:
the embodiment of the invention provides a data anomaly detection method for machine filter cloth adhesion, which comprises the following steps:
collecting vibration signals of the filter cloth by using a vibration sensor to obtain an amplitude time sequence data sequence; dividing the amplitude time sequence data sequence into a plurality of single-period sequences, and obtaining a duration set according to the data quantity in the single-period sequences;
obtaining a similarity set according to the data difference between the monocycle sequences; determining a standard single-period duration according to all data values in the duration set and the occurrence probability of the corresponding duration; determining the suspected abnormality degree of the monocycle sequence according to the difference between the monocycle sequence duration and the standard monocycle duration and all data values in the similarity set;
constructing a suspected abnormal degree sequence according to the suspected abnormal degrees of all the monocycle sequences, and dividing the suspected abnormal degree sequence into a plurality of cluster clusters; dividing the cluster into a normal cluster and a suspected abnormal cluster;
respectively constructing a reference amplitude sequence and a reference duration sequence according to the data average value and the data quantity in the monocycle sequence corresponding to all data in the suspected abnormal cluster; according to the correlation between the reference amplitude sequence and the reference time length sequence and the number of local extreme points in the reference amplitude sequence and the reference time length sequence, determining the noise interference degree corresponding to the suspected abnormal cluster; determining the abnormality degree corresponding to the suspected abnormal cluster according to the average value of the data in the suspected abnormal cluster, the sum of the data quantity in the monocycle sequence corresponding to all the data and the corresponding noise interference degree;
determining the K distance corresponding to the suspected abnormal cluster according to the degree of abnormality corresponding to the suspected abnormal cluster, determining the K distance corresponding to each piece of data in the amplitude time sequence data according to the K distance corresponding to the normal cluster and the K distance corresponding to the suspected abnormal cluster, and obtaining the abnormal data in the amplitude time sequence data by using an LOF algorithm.
Further, a similarity set is obtained according to the data difference between the monocycle sequences; according to all data values in the duration set and the occurrence probability of the corresponding duration, determining the standard single-period duration comprises the following specific steps:
calculating the DTW distances between any one monocycle sequence and other monocycle sequences respectively by using a DTW algorithm, and constructing a similarity set according to the DTW distances between the monocycle sequences and the other monocycle sequences respectively, wherein the DTW distances represent the data difference between the monocycle sequences and the other monocycle sequences;
dividing the number of times of occurrence of the value of the data quantity in the monocycle sequence in the time length set by the quotient of the data quantity in the time length set, and recording the quotient as the time length occurrence probability of the monocycle sequence;
and (3) recording the sum of products of the normalized value of the time duration occurrence probability of all the monocycle sequences and the data quantity in all the monocycle sequences as the standard monocycle time duration.
Further, the determining the suspected abnormal degree of the monocycle sequence according to the difference between the monocycle sequence duration and the standard monocycle duration and all the data values in the similarity set comprises the following specific steps:
recording the absolute value of the difference value between the data quantity in the monocycle sequence and the standard monocycle time length as the difference between the monocycle sequence time length and the standard monocycle time length;
the difference value of the normalized value of the data in the similarity set is marked as the weight value of the data in the similarity set;
according to the weight values of all the data in the similarity set and the normalized values of all the data in the similarity set, determining dissimilarity between the monocycle sequence and other monocycle sequences;
and determining the suspected abnormal degree of the monocycle sequence according to the difference between the monocycle sequence duration and the standard monocycle duration and the dissimilarity between the monocycle sequence and other monocycle sequences.
Further, according to the difference between the monocycle sequence duration and the standard monocycle duration and the dissimilarity between the monocycle sequence and other monocycle sequences, the specific calculation formula corresponding to the suspected abnormality degree of the monocycle sequence is determined as follows:
wherein the method comprises the steps ofFor the suspected abnormality degree of the z-th monocycle sequence, -/-, for example>Normalized value for j-th data in the similarity set,/>For the dissimilarity of the z-th monocycle sequence with other monocycle sequences, +.>For the number of data in the z-th monocycle sequence, is->For the number of data in the ith monocycle sequence, etc.>For the time-duration occurrence probability of the ith monocycle sequence, n is the number of monocycle sequences divided by the amplitude time-sequence data sequence,/->Is a linear normalization function.
Further, the suspected abnormality degree sequence is divided into a plurality of cluster clusters; the clustering clusters are divided into normal clustering clusters and suspected abnormal clustering clusters, and the method comprises the following specific steps of:
classifying the suspected anomaly degree sequence by using a DBSCAN clustering algorithm according to a preset neighborhood radius and the minimum sample number to obtain a plurality of clusters;
and (3) marking the average value of all data in the suspected abnormal degree sequence as a segmentation threshold value, enabling the cluster with the average value of the data smaller than the segmentation threshold value in the cluster to be a normal cluster, and enabling the cluster with the average value of the data larger than or equal to the segmentation threshold value in the cluster to be a suspected abnormal cluster.
Further, the determining the noise interference degree corresponding to the suspected abnormal cluster according to the correlation between the reference amplitude sequence and the reference time length sequence and the number of the local extreme points in the reference amplitude sequence and the reference time length sequence comprises the following specific steps:
if the number of data in the suspected abnormal cluster is larger than a preset number threshold, determining the noise interference degree corresponding to the suspected abnormal cluster according to the data covariance between the reference amplitude sequence and the reference time length sequence and the sum of the number of local extremum points in the reference amplitude sequence and the reference time length sequence;
if the number of data in the suspected abnormal cluster is smaller than or equal to a preset number threshold, setting the noise interference degree corresponding to the suspected abnormal cluster as the preset noise interference degree.
Further, the specific calculation formula corresponding to the noise interference degree corresponding to the suspected abnormal cluster is determined according to the data covariance between the reference amplitude sequence and the reference time length sequence and the sum of the local extreme point numbers in the reference amplitude sequence and the reference time length sequence, wherein the specific calculation formula is as follows:
wherein G is suspiciousThe noise interference degree corresponding to the quasi-abnormal cluster is m, wherein m is the data quantity in the suspected abnormal cluster, and m is equal to the data quantity in the reference amplitude sequence, m is also equal to the data quantity in the reference duration sequence,for the x-th data value in the reference amplitude sequence, is->For the mean value of all data in the reference amplitude sequence, +.>For the x-th data value in the sequence of reference durations, and (2)>For the mean value of all data in the sequence of reference duration, +.>For the data covariance between the reference amplitude sequence and the reference duration sequence, +.>And->The number of local extreme points in the reference amplitude sequence and the reference duration sequence respectively, < >>As a linear normalization function>K is a preset exponential function adjustment value for an exponential function based on a natural constant.
Further, the determining the abnormality degree corresponding to the suspected abnormal cluster according to the sum of the data average value in the suspected abnormal cluster and the data quantity in the monocycle sequence corresponding to all the data and the corresponding noise interference degree comprises the following specific steps:
determining the abnormal degree of the suspected abnormal cluster after removing the noise interference according to the noise interference degree corresponding to the suspected abnormal cluster and the average value of all data in the suspected abnormal cluster;
and determining the abnormality degree corresponding to the suspected abnormality cluster according to the sum of the abnormality degree corresponding to the suspected abnormality cluster after noise interference is removed and the data quantity in the monocycle sequence corresponding to all the data in the suspected abnormality cluster.
Further, the specific calculation formula corresponding to the abnormality degree corresponding to the suspected abnormal cluster is determined according to the sum of the abnormality degree corresponding to the suspected abnormal cluster after noise interference removal and the data quantity in the monocycle sequence corresponding to all the data in the suspected abnormal cluster, wherein the specific calculation formula corresponds to the abnormality degree corresponding to the suspected abnormal cluster and is as follows:
wherein F is the degree of abnormality corresponding to the suspected abnormal cluster, G is the degree of noise interference corresponding to the suspected abnormal cluster,is the average value of all data in the suspected abnormal cluster, < >>For the degree of abnormality of the suspected abnormal cluster after noise interference removal, M is the sum of the data amounts in the monocycle sequence corresponding to all the data in the suspected abnormal cluster, N is the data amount in the amplitude time sequence data sequence, < >>And the normalized value is the sum of the data quantity in the monocycle sequence corresponding to all the data in the suspected abnormal cluster.
Further, determining a K distance corresponding to the suspected abnormal cluster according to the degree of abnormality corresponding to the suspected abnormal cluster, determining a K distance corresponding to each data in the amplitude time sequence according to the K distance corresponding to the normal cluster and the K distance corresponding to the suspected abnormal cluster, and obtaining abnormal data in the amplitude time sequence by using an LOF algorithm, including the specific steps as follows:
determining the K distance corresponding to the suspected abnormal cluster according to the difference between normalized values of the degree of abnormality corresponding to the suspected abnormal cluster and the K distance value range of preset abnormal data and noise data; the K distances corresponding to the suspected abnormal cluster are given to all data in the monocycle sequence corresponding to all data in the corresponding suspected abnormal cluster, so that the K distances corresponding to all data in the monocycle sequence corresponding to all data in the suspected abnormal cluster are obtained;
setting the K distance corresponding to the normal cluster as the K distance of preset normal data; the K distances corresponding to the normal cluster are given to all data in the monocycle sequences corresponding to all data in the corresponding normal cluster, so that the K distances corresponding to all data in the monocycle sequences corresponding to all data in the normal cluster are obtained;
obtaining the K distance corresponding to each data in the amplitude time sequence data sequence according to the K distances corresponding to all data in the monocycle sequences corresponding to all data in all suspected abnormal clusters and the K distances corresponding to all data in the monocycle sequences corresponding to all data in all normal clusters;
and obtaining abnormal data in the amplitude time sequence data by using an LOF algorithm according to the K distance corresponding to each data in the amplitude time sequence data.
The technical scheme of the invention has the beneficial effects that:
in the embodiment of the invention, an amplitude time sequence data sequence is divided into a plurality of monocycle sequences, a time length set is obtained according to the data quantity in the monocycle sequences, and a similarity set is obtained according to the data difference between the monocycle sequences. And determining the suspected abnormal degree of the monocycle sequence according to all the data values in the time length set, the corresponding time length occurrence probability and all the data values in the similarity set. In consideration of the influence of noise on abnormal data detection, a suspected abnormal degree sequence is constructed according to suspected abnormal degrees of all monocycle sequences, the suspected abnormal degree sequence is divided into a plurality of clusters, the clusters are divided into normal clusters and suspected abnormal clusters, the abnormal degree corresponding to the suspected abnormal clusters is determined according to the sum of the data mean value in the suspected abnormal clusters and the data quantity in the monocycle sequences corresponding to all data and the corresponding noise interference degree, the K distance corresponding to the suspected abnormal clusters is determined according to the abnormal degree corresponding to the suspected abnormal clusters, the K distance corresponding to each data in the amplitude time sequence is determined according to the K distance corresponding to the normal clusters and the K distance corresponding to the suspected abnormal clusters, and the abnormal data in the amplitude time sequence is obtained by using an LOF algorithm. The real abnormal data is endowed with smaller K distance, and the neighborhood range is limited, so that the LOF algorithm is more focused on local density change, and abnormal data with obvious differences from surrounding data points can be detected more sensitively. And for data anomalies caused by noise, a larger K distance is provided, the neighborhood range is enlarged, more data points are considered, more comprehensive context information is provided, the local reachable density of the data points is estimated more accurately, and noise data which are far away from surrounding data points and have no obvious relevance can be filtered.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart showing the steps of a method for detecting data anomalies due to sticking of filter cloth of a machine according to the present invention.
Detailed Description
In order to further describe the technical means and effects adopted by the invention to achieve the preset aim, the following is a detailed description of specific implementation, structure, characteristics and effects of a data anomaly detection method for machine filter cloth adhesion according to the invention with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" means that the embodiments are not necessarily the same. Furthermore, the particular features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.
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 specifically describes a specific scheme of the data anomaly detection method for machine filter cloth adhesion provided by the invention with reference to the accompanying drawings.
Referring to fig. 1, a flowchart of a method for detecting abnormal data adhered to a filter cloth of a machine according to an embodiment of the invention is shown, the method includes the following steps:
step S001: collecting vibration signals of the filter cloth by using a vibration sensor to obtain an amplitude time sequence data sequence; dividing the amplitude time sequence data sequence into a plurality of single-period sequences, and obtaining a duration set according to the data quantity in the single-period sequences.
Vibration signals of each filter cloth in the same running state of the machine are collected through the vibration sensor arranged on the filter cloth, and an amplitude time sequence data sequence of each filter cloth is obtained.
Taking an amplitude time sequence of a filter cloth in a machine as an example, obtaining a local minimum point in the amplitude time sequence by using a valley detection algorithm, dividing the amplitude time sequence into a plurality of time sequence data sequence segments according to the local minimum point in the amplitude time sequence, and recording the time sequence data as a single-period sequence. It should be noted that the first and last monocycle sequences in the amplitude-sequence may be incomplete, affecting the analysis between monocycle sequences, so that only the second to the last monocycle sequence in the amplitude-sequence is analyzed below.
Counting the data quantity in each monocycle sequence to obtain a time length setWherein n is the number of monocycle sequences divided by the amplitude time sequence data, +.>The number of data in the nth single period sequence divided for the amplitude timing data sequence. It should be noted that n is specifically the number of second to last one-cycle sequences in the amplitude timing data sequence.
Step S002: obtaining a similarity set according to the data difference between the monocycle sequences; determining a standard single-period duration according to all data values in the duration set and the occurrence probability of the corresponding duration; and determining the suspected abnormality degree of the monocycle sequence according to the difference between the monocycle sequence duration and the standard monocycle duration and all data values in the similarity set.
Taking the z-th monocycle sequence as an example, using a DTW algorithm to obtain the DTW distances between the monocycle sequence and other monocycle sequences respectively to obtain a similarity setWherein->For the DTW distance between the n-1 th monocycle sequence and other monocycle sequences, the minimum maximum standard method is used to make similarity set +.>Normalized to [0,1 ]]Within the interval.
The valley detection algorithm, the DTW algorithm, and the minimum maximum normalization method are known techniques, and specific methods are not described herein.
Thus, the suspected abnormality degree of the monocycle sequence can be knownThe calculation formula of (2) is as follows:
wherein the method comprises the steps ofFor the suspected abnormality degree of the z-th monocycle sequence, -/-, for example>For similarity set->Normalized value of j-th data in (a),>for the number of data in the z-th monocycle sequence, is->For the number of data in the ith monocycle sequence, etc.>Probability of occurrence for the duration of the ith monocycle sequence,/->The solving process of (1) is as follows: the size is +.>Is set of data values in duration +.>Dividing the number of occurrences by the set of durations +.>The quotient of the number of data n, recorded as the probability of occurrence of the length of the ith monocycle sequence +.>N is the number of monocycle sequences divided by the amplitude timing data sequence. />Normalizing the data values to [0,1 ] as a linear normalization function]Within the interval.
What needs to be described is: the vibration frequency of a normally operating machine filter cloth is generally relatively stable and will vibrate in a relatively stable mannerOperating in the amplitude range. The amount of data in each monocycle sequence should be similar under normal conditions, and thereforeThe bigger the->Corresponding->The more authentic, thereby normalized +.>Is->Weight, weighted average->For a standard monocycle duration, i.e.)>Is the difference between the length of the z-th monocycle sequence and the length of the standard monocycle. Further analyzing the variation of vibration amplitude, the amplitude data in the normal monocycle sequence is similar, and the variation of the amplitude data in the abnormal monocycle sequence is caused by the variation of the adhesion degree of the filter cloth and the variation of the blocking degree of the pore, the smaller the DTW is, the greater the similarity of the two time sequence data sequences is, thus the normalization is achieved>Is->Weight, weighted average->Is the dissimilarity of the z-th monocycle sequence with other monocycle sequences. To this end use->Is->The product of the two is the suspected abnormality degree of the z-th monocycle sequence.
Step S003: constructing a suspected abnormal degree sequence according to the suspected abnormal degrees of all the monocycle sequences, and dividing the suspected abnormal degree sequence into a plurality of cluster clusters; the clusters are divided into normal clusters and suspected abnormal clusters.
According to the mode, calculating the suspected abnormality degree of all monocycle sequences divided by the amplitude time sequence data to obtain a suspected abnormality degree sequenceWherein n is the number of monocycle sequences divided by the amplitude time sequence data, +.>The degree of suspected abnormality of the nth monocycle sequence divided for the amplitude time series data sequence.
Because the noise data and the abnormal data of the amplitude time sequence data sequence can cause larger suspected abnormal degree of the corresponding monocycle sequence, it is known that in the filter pressing process, along with accumulation and falling of adherends on filter cloth, uneven distribution of the adherends on the filter cloth can be caused, intermittent continuous occurrence of the abnormal data in the amplitude time sequence data can be caused, and the noise can be superimposed on normal data or abnormal data due to the size and the randomness of the occurrence time of the noise and the shorter occurrence time of each time. Therefore, it is necessary to further analyze the influence of noise on abnormal data monitoring and calculate the accurate degree of abnormality.
Use of DBSCAN clustering algorithm for suspected anomaly degree sequenceAnd performing classification processing to obtain a plurality of clustering clusters. The present embodiment is described by taking the neighborhood radius equal to 0.2 and the minimum number of samples equal to 3 as an example, and other values may be set in other embodiments, and the present embodiment is not limited thereto. Wherein, DBSCAN clustering algorithm is a known technique, and the specific method is as followsNot described.
Sequence of suspected abnormality degreeThe average value of all data in the cluster is recorded as a segmentation threshold value, the cluster with the average value of the data in the cluster smaller than the segmentation threshold value is a normal cluster, and the cluster with the average value of the data in the cluster larger than or equal to the segmentation threshold value is a suspected abnormal cluster.
Step S004: respectively constructing a reference amplitude sequence and a reference duration sequence according to the data average value and the data quantity in the monocycle sequence corresponding to all data in the suspected abnormal cluster; according to the correlation between the reference amplitude sequence and the reference time length sequence and the number of local extreme points in the reference amplitude sequence and the reference time length sequence, determining the noise interference degree corresponding to the suspected abnormal cluster; and determining the abnormality degree corresponding to the suspected abnormal cluster according to the average value of the data in the suspected abnormal cluster, the sum of the data quantity in the monocycle sequence corresponding to all the data and the corresponding noise interference degree.
Taking a suspected abnormal cluster as an example, sequentially calculating the data average value in a monocycle sequence corresponding to each data in the suspected abnormal cluster to obtain a reference amplitude sequenceSequentially calculating the data quantity in the monocycle sequence corresponding to each data in the suspected abnormal cluster to obtain a reference time length sequence +.>Wherein m is the number of data in the suspected abnormal cluster, and m is the number of data in the reference amplitude sequence, and m is the number of data in the reference duration sequence, and +.>For the data mean value of the monocycle sequence corresponding to the mth data in the suspected abnormal cluster,/in the cluster>And the data quantity in the monocycle sequence corresponding to the mth data in the suspected abnormal cluster is obtained.
The reference amplitude sequences are obtained respectively by using a first derivative methodAnd the reference duration sequence->When the first derivative method is used for solving the local extreme points, at least three data exist in the sequence, so that if the number of the data in the sequence is less than three, the number of the local extreme points in the sequence is directly made to be zero. The first derivative method is a known technique, and the specific method is not described here.
From this, the calculation formula of the abnormality degree F corresponding to the suspected abnormal cluster is shown as follows:
when (when)In the process, the acquisition mode of G is as follows:
when (when)In the process, the acquisition mode of G is as follows:
wherein F is the degree of abnormality corresponding to the suspected abnormal cluster, G is the degree of noise interference corresponding to the suspected abnormal cluster,m is all numbers in the suspected abnormal cluster for the set noise interference degreeAccording to the sum of the data quantity in the corresponding monocycle sequence, m is the data quantity in the suspected abnormal cluster, and m is equal to the data quantity in the reference amplitude sequence, m is also equal to the data quantity in the reference duration sequence,/A>For the set number threshold, N is the number of data in the amplitude timing data sequence, +.>For the mean value of all data in the suspected abnormal cluster, +.>For the x-th data value in the reference amplitude sequence D, and (2)>For the mean value of all data in the reference amplitude sequence D, +.>For the reference duration sequence->Is selected from the group consisting of the x-th data value,for the reference duration sequence->Mean value of all data in>And->Respectively reference amplitude sequences->And a reference duration sequenceThe number of local extreme points in the model. />Normalizing the data values to [0,1 ] as a linear normalization function]Within the interval. />K is a set exponential function adjustment value for an exponential function based on a natural constant. In this embodiment k is set to 0.05, < >>The value range of G is 1 and G is 0,1]Setting +.>For example, 1 is described as an example, and other values may be set in other embodiments, and the present example is not limited thereto.
What needs to be described is: in the filter pressing process, as the adhesion substances are accumulated on the filter cloth, the amplitude and the frequency of vibration data are gradually increased, and as the adhesion substances are partially or completely dropped, the amplitude and the frequency of the vibration data are gradually reduced, so that the change of the amplitude and the frequency of the vibration data caused by the adhesion substances is positively correlated, the accumulation and the dropping of the adhesion substances are continuous, the occurrence time of noise is shorter each time, and when the data quantity m in the suspected abnormal cluster is less than or equal to a set quantity threshold valueWhen the suspected abnormality degree is caused by noise, the noise interference degree G corresponding to the suspected abnormality cluster is set to 1. When the data quantity m in the suspected abnormal cluster is larger than the set quantity threshold value +.>When the noise is superimposed on the abnormal data, the noise causes irregular change of the amplitude and frequency of the vibration data, so that the noise causes multiple trend conversion in the process of increasing and decreasing the amplitude and frequency of the vibration data caused by the adhesiveTo transform into->The larger the number of the explanatory noise is, the more +.>For the reference amplitude sequence->And the reference duration sequence->The smaller the value, the greater the positive correlation effect of the noise data value on the amplitude and frequency variation of the vibration data caused by the adherend, thus +.>And covariance->And the normalized value of the product of the two is the noise interference degree G corresponding to the suspected abnormal cluster. G is larger, say->Is due to the fact that noise interference is large, therefore +.>Is->The product of the two is the degree of abnormality after noise interference is removed, and the larger M is, the longer and more important the time of vibration abnormality data caused by the adhesion is, so the normalized +.>Is->The product of the two is the degree of abnormality F corresponding to the suspected abnormal cluster.
Obtaining suspected abnormality degree sequences according to the above modeThe abnormal degree corresponding to each suspected abnormal cluster divided in the process is normalized by using a minimum maximum normalization method to the abnormal degree corresponding to all suspected abnormal clusters until the abnormal degree is 0,1]And obtaining a normalized value of the abnormality degree corresponding to each suspected abnormal cluster in the interval.
Step S005: determining the K distance corresponding to the suspected abnormal cluster according to the degree of abnormality corresponding to the suspected abnormal cluster, determining the K distance corresponding to each piece of data in the amplitude time sequence data according to the K distance corresponding to the normal cluster and the K distance corresponding to the suspected abnormal cluster, and obtaining the abnormal data in the amplitude time sequence data by using an LOF algorithm.
In the LOF algorithm, a K distance parameter needs to be set for each data, and in this embodiment, the K distance of the normal data is set to 10, and the K distance range of the abnormal data and the noise data is set to [5,15 ]]In the description of this example, other values may be set in other embodiments, and the present example is not limited thereto. Thereby making the suspected abnormality degree sequenceAll data in the monocycle sequence corresponding to all data in all normal clusters divided in the (a) and K distances corresponding to all data in the first and last monocycle sequences in the amplitude time sequence are 10.
Suspected abnormality degree sequenceK distance +.>The calculation formula of (2) is as follows:
wherein the method comprises the steps ofK distance corresponding to the y suspected abnormal cluster,>the normalized value of the degree of abnormality corresponding to the y-th suspected abnormal cluster, z is the suspected abnormal degree sequence +.>The number of the suspected abnormal cluster divided in (a). And the K distance corresponding to each suspected abnormal cluster is obtained, the K distance corresponding to each suspected abnormal cluster is endowed to all data in the monocycle sequence corresponding to all data in the corresponding suspected abnormal cluster, and the K distance corresponding to all data in the monocycle sequence corresponding to all data in the suspected abnormal cluster is obtained.
What needs to be described is: when (when)And when the data is larger, the corresponding data is abnormal, the smaller K distance is needed, the neighborhood range is limited, so that the LOF algorithm is more focused on the local density change, and abnormal data with obvious differences from surrounding data points can be detected more sensitively. When->The smaller the data anomaly caused by noise, the larger the K distance is needed, the larger the neighborhood range is enlarged, more data points are considered, more comprehensive context information is provided, the local reachable density of the data points is estimated more accurately, and noise data which are far away from surrounding data points and have no obvious relevance can be filtered out.
So far, the K distance corresponding to each data of the amplitude time sequence data is obtained. And obtaining abnormal data in the amplitude time sequence data sequence by using an LOF algorithm. Further analyzing and excavating the detected abnormal data, judging the adhesion condition of the filter cloth, and accurately regulating and controlling the automatic dismounting device, ensuring that the filter cake is cleaned, and improving the product quality and the process efficiency. The LOF algorithm is a well known technique, and a specific method is not described herein.
The present invention has been completed.
In summary, in the embodiment of the present invention, the amplitude time sequence data sequence is divided into a plurality of monocycle sequences, a duration set is obtained according to the data amount in the monocycle sequences, a similarity set is obtained according to the data difference between the monocycle sequences, and the suspected abnormal degree of the monocycle sequences is determined according to the duration set and the data analysis in the similarity set. And constructing a suspected abnormal degree sequence according to the suspected abnormal degree of all the monocycle sequences, dividing the suspected abnormal degree sequence into a plurality of clusters, dividing the clusters into a normal cluster and a suspected abnormal cluster, determining the abnormal degree corresponding to the suspected abnormal cluster according to the sum of the data average value in the suspected abnormal cluster and the data quantity in the monocycle sequences corresponding to all the data and the corresponding noise interference degree, determining the K distance corresponding to the suspected abnormal cluster according to the abnormal degree corresponding to the suspected abnormal cluster, determining the K distance corresponding to each data in the amplitude time sequence according to the K distance corresponding to the normal cluster and the K distance corresponding to the suspected abnormal cluster, and obtaining the abnormal data in the amplitude time sequence by using an LOF algorithm. The real abnormal data is endowed with smaller K distance, and the neighborhood range is limited, so that the LOF algorithm is more focused on local density change, and abnormal data with obvious differences from surrounding data points can be detected more sensitively. And for data anomalies caused by noise, a larger K distance is provided, the neighborhood range is enlarged, more data points are considered, more comprehensive context information is provided, the local reachable density of the data points is estimated more accurately, and noise data which are far away from surrounding data points and have no obvious relevance can be filtered.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, alternatives, and improvements that fall within the spirit and scope of the invention.
Claims (10)
1. The data anomaly detection method for the adhesion of the filter cloth of the machine is characterized by comprising the following steps of:
collecting vibration signals of the filter cloth by using a vibration sensor to obtain an amplitude time sequence data sequence; dividing the amplitude time sequence data sequence into a plurality of single-period sequences, and obtaining a duration set according to the data quantity in the single-period sequences;
obtaining a similarity set according to the data difference between the monocycle sequences; determining a standard single-period duration according to all data values in the duration set and the occurrence probability of the corresponding duration; determining the suspected abnormality degree of the monocycle sequence according to the difference between the monocycle sequence duration and the standard monocycle duration and all data values in the similarity set;
constructing a suspected abnormal degree sequence according to the suspected abnormal degrees of all the monocycle sequences, and dividing the suspected abnormal degree sequence into a plurality of cluster clusters; dividing the cluster into a normal cluster and a suspected abnormal cluster;
respectively constructing a reference amplitude sequence and a reference duration sequence according to the data average value and the data quantity in the monocycle sequence corresponding to all data in the suspected abnormal cluster; according to the correlation between the reference amplitude sequence and the reference time length sequence and the number of local extreme points in the reference amplitude sequence and the reference time length sequence, determining the noise interference degree corresponding to the suspected abnormal cluster; determining the abnormality degree corresponding to the suspected abnormal cluster according to the average value of the data in the suspected abnormal cluster, the sum of the data quantity in the monocycle sequence corresponding to all the data and the corresponding noise interference degree;
determining the K distance corresponding to the suspected abnormal cluster according to the degree of abnormality corresponding to the suspected abnormal cluster, determining the K distance corresponding to each piece of data in the amplitude time sequence data according to the K distance corresponding to the normal cluster and the K distance corresponding to the suspected abnormal cluster, and obtaining the abnormal data in the amplitude time sequence data by using an LOF algorithm.
2. The method for detecting abnormal data adhered to filter cloth of a machine according to claim 1, wherein a similarity set is obtained according to data differences among single-period sequences; according to all data values in the duration set and the occurrence probability of the corresponding duration, determining the standard single-period duration comprises the following specific steps:
calculating the DTW distances between any one monocycle sequence and other monocycle sequences respectively by using a DTW algorithm, and constructing a similarity set according to the DTW distances between the monocycle sequences and the other monocycle sequences respectively, wherein the DTW distances represent the data difference between the monocycle sequences and the other monocycle sequences;
dividing the number of times of occurrence of the value of the data quantity in the monocycle sequence in the time length set by the quotient of the data quantity in the time length set, and recording the quotient as the time length occurrence probability of the monocycle sequence;
and (3) recording the sum of products of the normalized value of the time duration occurrence probability of all the monocycle sequences and the data quantity in all the monocycle sequences as the standard monocycle time duration.
3. The method for detecting abnormal data adhesion of filter cloth of a machine according to claim 1, wherein the determining the suspected abnormal degree of the monocycle sequence according to the difference between the monocycle sequence duration and the standard monocycle duration and all data values in the similarity set comprises the following specific steps:
recording the absolute value of the difference value between the data quantity in the monocycle sequence and the standard monocycle time length as the difference between the monocycle sequence time length and the standard monocycle time length;
the difference value of the normalized value of the data in the similarity set is marked as the weight value of the data in the similarity set;
according to the weight values of all the data in the similarity set and the normalized values of all the data in the similarity set, determining dissimilarity between the monocycle sequence and other monocycle sequences;
and determining the suspected abnormal degree of the monocycle sequence according to the difference between the monocycle sequence duration and the standard monocycle duration and the dissimilarity between the monocycle sequence and other monocycle sequences.
4. The method for detecting abnormal data adhesion between filter cloth of a machine according to claim 3, wherein the specific calculation formula corresponding to the suspected abnormal degree of the monocycle sequence is determined according to the difference between the monocycle sequence duration and the standard monocycle duration and the dissimilarity between the monocycle sequence and other monocycle sequences, and is as follows:
wherein the method comprises the steps ofFor the suspected abnormality degree of the z-th monocycle sequence, -/-, for example>Is the normalized value of the j-th data in the similarity set,for the dissimilarity of the z-th monocycle sequence with other monocycle sequences, +.>For the number of data in the z-th monocycle sequence, is->For the number of data in the ith monocycle sequence, etc.>For the time-duration occurrence probability of the ith monocycle sequence, n is the number of monocycle sequences divided by the amplitude time-sequence data sequence,/->Is a linear normalization function.
5. The method for detecting abnormal data adhered to filter cloth of a machine according to claim 1, wherein the suspected abnormal degree sequence is divided into a plurality of clusters; the clustering clusters are divided into normal clustering clusters and suspected abnormal clustering clusters, and the method comprises the following specific steps of:
classifying the suspected anomaly degree sequence by using a DBSCAN clustering algorithm according to a preset neighborhood radius and the minimum sample number to obtain a plurality of clusters;
and (3) marking the average value of all data in the suspected abnormal degree sequence as a segmentation threshold value, enabling the cluster with the average value of the data smaller than the segmentation threshold value in the cluster to be a normal cluster, and enabling the cluster with the average value of the data larger than or equal to the segmentation threshold value in the cluster to be a suspected abnormal cluster.
6. The method for detecting abnormal data adhesion of filter cloth of a machine according to claim 1, wherein the determining the noise interference degree corresponding to the suspected abnormal cluster according to the correlation between the reference amplitude sequence and the reference time length sequence and the number of local extremum points in the reference amplitude sequence and the reference time length sequence comprises the following specific steps:
if the number of data in the suspected abnormal cluster is larger than a preset number threshold, determining the noise interference degree corresponding to the suspected abnormal cluster according to the data covariance between the reference amplitude sequence and the reference time length sequence and the sum of the number of local extremum points in the reference amplitude sequence and the reference time length sequence;
if the number of data in the suspected abnormal cluster is smaller than or equal to a preset number threshold, setting the noise interference degree corresponding to the suspected abnormal cluster as the preset noise interference degree.
7. The method for detecting abnormal data adhesion between filter cloth of a machine according to claim 6, wherein the specific calculation formula corresponding to the noise interference degree corresponding to the suspected abnormal cluster is determined according to the data covariance between the reference amplitude sequence and the reference time length sequence and the sum of the local extreme point numbers in the reference amplitude sequence and the reference time length sequence, and is as follows:
wherein G is the noise interference degree corresponding to the suspected abnormal cluster, m is the data quantity in the reference amplitude sequence, m is the data quantity in the reference duration sequence,for the x-th data value in the reference amplitude sequence, is->For the mean value of all data in the reference amplitude sequence, +.>For the x-th data value in the sequence of reference durations, and (2)>For the mean value of all data in the sequence of reference duration, +.>For the data covariance between the reference amplitude sequence and the reference duration sequence, +.>And->The number of local extreme points in the reference amplitude sequence and the reference duration sequence respectively, < >>As a linear normalization function>K is a preset exponential function adjustment value for an exponential function based on a natural constant.
8. The method for detecting abnormal data adhered to filter cloth of a machine according to claim 1, wherein the determining the abnormal degree corresponding to the suspected abnormal cluster according to the sum of the average data value of the suspected abnormal cluster and the data quantity in the monocycle sequence corresponding to all data and the corresponding noise interference degree comprises the following specific steps:
determining the abnormal degree of the suspected abnormal cluster after removing the noise interference according to the noise interference degree corresponding to the suspected abnormal cluster and the average value of all data in the suspected abnormal cluster;
and determining the abnormality degree corresponding to the suspected abnormality cluster according to the sum of the abnormality degree corresponding to the suspected abnormality cluster after noise interference is removed and the data quantity in the monocycle sequence corresponding to all the data in the suspected abnormality cluster.
9. The method for detecting abnormal data adhered to a filter cloth of a machine according to claim 8, wherein the specific calculation formula for determining the abnormal degree corresponding to the suspected abnormal cluster according to the sum of the abnormal degree corresponding to the suspected abnormal cluster after noise interference removal and the data quantity in the monocycle sequence corresponding to all the data in the suspected abnormal cluster is:
wherein F is the degree of abnormality corresponding to the suspected abnormal cluster, G is the degree of noise interference corresponding to the suspected abnormal cluster,is the average value of all data in the suspected abnormal cluster, < >>For the degree of abnormality of the suspected abnormal cluster after noise interference is removed, M is all data in the suspected abnormal clusterThe sum of the data amounts in the corresponding monocycle sequence, N is the data quantity in the amplitude time sequence, < >>And the normalized value is the sum of the data quantity in the monocycle sequence corresponding to all the data in the suspected abnormal cluster.
10. The method for detecting abnormal data adhered to filter cloth of a machine according to claim 1, wherein the determining the K distance corresponding to the suspected abnormal cluster according to the degree of abnormality corresponding to the suspected abnormal cluster, determining the K distance corresponding to each data in the amplitude time series data sequence according to the K distance corresponding to the normal cluster and the K distance corresponding to the suspected abnormal cluster, and obtaining the abnormal data in the amplitude time series data sequence by using the LOF algorithm comprises the following specific steps:
determining the K distance corresponding to the suspected abnormal cluster according to the difference between normalized values of the degree of abnormality corresponding to the suspected abnormal cluster and the K distance value range of preset abnormal data and noise data; the K distances corresponding to the suspected abnormal cluster are given to all data in the monocycle sequence corresponding to all data in the corresponding suspected abnormal cluster, so that the K distances corresponding to all data in the monocycle sequence corresponding to all data in the suspected abnormal cluster are obtained;
setting the K distance corresponding to the normal cluster as the K distance of preset normal data; the K distances corresponding to the normal cluster are given to all data in the monocycle sequences corresponding to all data in the corresponding normal cluster, so that the K distances corresponding to all data in the monocycle sequences corresponding to all data in the normal cluster are obtained;
obtaining the K distance corresponding to each data in the amplitude time sequence data sequence according to the K distances corresponding to all data in the monocycle sequences corresponding to all data in all suspected abnormal clusters and the K distances corresponding to all data in the monocycle sequences corresponding to all data in all normal clusters;
and obtaining abnormal data in the amplitude time sequence data by using an LOF algorithm according to the K distance corresponding to each data in the amplitude time sequence data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311090952.2A CN116821833B (en) | 2023-08-29 | 2023-08-29 | Data anomaly detection method for machine filter cloth adhesion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311090952.2A CN116821833B (en) | 2023-08-29 | 2023-08-29 | Data anomaly detection method for machine filter cloth adhesion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116821833A true CN116821833A (en) | 2023-09-29 |
| CN116821833B CN116821833B (en) | 2023-11-14 |
Family
ID=88127716
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311090952.2A Active CN116821833B (en) | 2023-08-29 | 2023-08-29 | Data anomaly detection method for machine filter cloth adhesion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116821833B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117009910A (en) * | 2023-10-08 | 2023-11-07 | 湖南工程学院 | Intelligent monitoring method for abnormal change of ambient temperature |
| CN117056913A (en) * | 2023-10-10 | 2023-11-14 | 北京点聚信息技术有限公司 | Intelligent management method for electronic seal authorization |
| CN117093828A (en) * | 2023-10-16 | 2023-11-21 | 张家港市友明机械制造有限公司 | Method and system for detecting tightness of hydraulic cylinder body |
| CN117131530A (en) * | 2023-10-20 | 2023-11-28 | 合肥亚明汽车部件有限公司 | Intelligent factory sensitive data encryption protection method |
| CN117235557A (en) * | 2023-11-14 | 2023-12-15 | 山东贺铭电气有限公司 | Electrical equipment fault rapid diagnosis method based on big data analysis |
| CN117370898A (en) * | 2023-12-08 | 2024-01-09 | 钛合联(深圳)科技有限公司 | Electronic data safety control system |
| CN117473345A (en) * | 2023-12-27 | 2024-01-30 | 山东顺发重工有限公司 | Flange forging defect detection method based on data analysis |
| CN117540238A (en) * | 2024-01-05 | 2024-02-09 | 长春同泰企业管理服务有限责任公司 | Data security management method for industrial digital information acquisition device |
| CN117665935A (en) * | 2024-01-30 | 2024-03-08 | 山东鑫国矿业技术开发有限公司 | Monitoring data processing method for broken rock mass supporting construction process |
| CN117782364A (en) * | 2024-02-28 | 2024-03-29 | 西瑞思创(西安)实验仪器有限公司 | Real-time monitoring method and system for high-flux automatic material rapid screening reactor |
Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160219067A1 (en) * | 2015-01-28 | 2016-07-28 | Korea Internet & Security Agency | Method of detecting anomalies suspected of attack, based on time series statistics |
| US20190295408A1 (en) * | 2018-03-26 | 2019-09-26 | International Business Machines Corporation | Real-time service level monitor |
| CN110286656A (en) * | 2019-05-07 | 2019-09-27 | 清华大学 | A kind of the false-alarm filter method and device of wrong data tolerance |
| CN110826648A (en) * | 2020-01-09 | 2020-02-21 | 浙江鹏信信息科技股份有限公司 | Method for realizing fault detection by utilizing time sequence clustering algorithm |
| JP2020107016A (en) * | 2018-12-27 | 2020-07-09 | 三菱電機株式会社 | Similarity determination device and abnormality detection device |
| CN114861788A (en) * | 2022-04-28 | 2022-08-05 | 国网江苏省电力有限公司营销服务中心 | Load abnormity detection method and system based on DBSCAN clustering |
| US20220279045A1 (en) * | 2021-02-26 | 2022-09-01 | Trackerdetect Ltd | Global iterative clustering algorithm to model entities' behaviors and detect anomalies |
| CN114997276A (en) * | 2022-05-07 | 2022-09-02 | 北京航空航天大学 | Heterogeneous multi-source time sequence data abnormity identification method for compression molding equipment |
| CN115147052A (en) * | 2022-09-02 | 2022-10-04 | 泗水县信诺线缆厂(普通合伙) | Intelligent management system for HDMI cable reservoir area |
| KR102511137B1 (en) * | 2022-06-20 | 2023-03-17 | 대한민국 | Method and method for analyzing similarity of chromatogram image |
| CN116127344A (en) * | 2022-12-23 | 2023-05-16 | 中国建设银行股份有限公司 | Abnormality detection optimization method and related equipment |
| CN116226700A (en) * | 2023-03-15 | 2023-06-06 | 重庆邮电大学 | Flow anomaly detection method based on time sequence clustering |
| CN116340795A (en) * | 2023-05-29 | 2023-06-27 | 山东一然环保科技有限公司 | Operation data processing method for pure oxygen combustion heating furnace |
| CN116502169A (en) * | 2023-06-28 | 2023-07-28 | 深圳特力自动化工程有限公司 | Centrifugal dehydrator working state detection method based on data detection |
| CN116628529A (en) * | 2023-07-21 | 2023-08-22 | 山东科华电力技术有限公司 | Data anomaly detection method for intelligent load control system at user side |
| CN116628617A (en) * | 2023-07-25 | 2023-08-22 | 值数科技(北京)有限公司 | Method for realizing miniature strain monitoring based on nanocomposite |
-
2023
- 2023-08-29 CN CN202311090952.2A patent/CN116821833B/en active Active
Patent Citations (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160219067A1 (en) * | 2015-01-28 | 2016-07-28 | Korea Internet & Security Agency | Method of detecting anomalies suspected of attack, based on time series statistics |
| US20190295408A1 (en) * | 2018-03-26 | 2019-09-26 | International Business Machines Corporation | Real-time service level monitor |
| JP2020107016A (en) * | 2018-12-27 | 2020-07-09 | 三菱電機株式会社 | Similarity determination device and abnormality detection device |
| CN110286656A (en) * | 2019-05-07 | 2019-09-27 | 清华大学 | A kind of the false-alarm filter method and device of wrong data tolerance |
| CN110826648A (en) * | 2020-01-09 | 2020-02-21 | 浙江鹏信信息科技股份有限公司 | Method for realizing fault detection by utilizing time sequence clustering algorithm |
| US20220279045A1 (en) * | 2021-02-26 | 2022-09-01 | Trackerdetect Ltd | Global iterative clustering algorithm to model entities' behaviors and detect anomalies |
| CN114861788A (en) * | 2022-04-28 | 2022-08-05 | 国网江苏省电力有限公司营销服务中心 | Load abnormity detection method and system based on DBSCAN clustering |
| CN114997276A (en) * | 2022-05-07 | 2022-09-02 | 北京航空航天大学 | Heterogeneous multi-source time sequence data abnormity identification method for compression molding equipment |
| KR102511137B1 (en) * | 2022-06-20 | 2023-03-17 | 대한민국 | Method and method for analyzing similarity of chromatogram image |
| CN115147052A (en) * | 2022-09-02 | 2022-10-04 | 泗水县信诺线缆厂(普通合伙) | Intelligent management system for HDMI cable reservoir area |
| CN116127344A (en) * | 2022-12-23 | 2023-05-16 | 中国建设银行股份有限公司 | Abnormality detection optimization method and related equipment |
| CN116226700A (en) * | 2023-03-15 | 2023-06-06 | 重庆邮电大学 | Flow anomaly detection method based on time sequence clustering |
| CN116340795A (en) * | 2023-05-29 | 2023-06-27 | 山东一然环保科技有限公司 | Operation data processing method for pure oxygen combustion heating furnace |
| CN116502169A (en) * | 2023-06-28 | 2023-07-28 | 深圳特力自动化工程有限公司 | Centrifugal dehydrator working state detection method based on data detection |
| CN116628529A (en) * | 2023-07-21 | 2023-08-22 | 山东科华电力技术有限公司 | Data anomaly detection method for intelligent load control system at user side |
| CN116628617A (en) * | 2023-07-25 | 2023-08-22 | 值数科技(北京)有限公司 | Method for realizing miniature strain monitoring based on nanocomposite |
Non-Patent Citations (6)
| Title |
|---|
| TIANYU LI: "Anomaly Scoring for Prediction-Based Anomaly Detection in Time Series", 2020 IEEE AEROSPACE CONFERENCE * |
| 司方远;韩英华;赵强;汪晋宽;: "基于AP-LOF离群组检测的配电网连接验证", 东北大学学报(自然科学版), no. 08 * |
| 张晓;: "基于聚类和LOF算法的异常数据检测方法", 伊犁师范学院学报(自然科学版), no. 02 * |
| 文必龙;马强;李菲;: "储油罐液位时序数据模式发现", 重庆大学学报, no. 03 * |
| 武丽芬;罗永莲;王秀华;: "高维大数据流连续异常点检测方法仿真", 计算机仿真, no. 10 * |
| 陶抒青;刘晓强;李柏岩;SHEN JIE;: "基于密度聚类和投票判别的三维数据去噪方法", 计算机应用研究, no. 02 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117009910B (en) * | 2023-10-08 | 2023-12-15 | 湖南工程学院 | Intelligent monitoring method for abnormal change of ambient temperature |
| CN117009910A (en) * | 2023-10-08 | 2023-11-07 | 湖南工程学院 | Intelligent monitoring method for abnormal change of ambient temperature |
| CN117056913A (en) * | 2023-10-10 | 2023-11-14 | 北京点聚信息技术有限公司 | Intelligent management method for electronic seal authorization |
| CN117056913B (en) * | 2023-10-10 | 2023-12-22 | 北京点聚信息技术有限公司 | Intelligent management method for electronic seal authorization |
| CN117093828A (en) * | 2023-10-16 | 2023-11-21 | 张家港市友明机械制造有限公司 | Method and system for detecting tightness of hydraulic cylinder body |
| CN117093828B (en) * | 2023-10-16 | 2023-12-26 | 张家港市友明机械制造有限公司 | Method and system for detecting tightness of hydraulic cylinder body |
| CN117131530A (en) * | 2023-10-20 | 2023-11-28 | 合肥亚明汽车部件有限公司 | Intelligent factory sensitive data encryption protection method |
| CN117131530B (en) * | 2023-10-20 | 2024-01-30 | 合肥亚明汽车部件有限公司 | Intelligent factory sensitive data encryption protection method |
| CN117235557A (en) * | 2023-11-14 | 2023-12-15 | 山东贺铭电气有限公司 | Electrical equipment fault rapid diagnosis method based on big data analysis |
| CN117235557B (en) * | 2023-11-14 | 2024-02-02 | 山东贺铭电气有限公司 | Electrical equipment fault rapid diagnosis method based on big data analysis |
| CN117370898B (en) * | 2023-12-08 | 2024-03-12 | 钛合联(深圳)科技有限公司 | Electronic data safety control system |
| CN117370898A (en) * | 2023-12-08 | 2024-01-09 | 钛合联(深圳)科技有限公司 | Electronic data safety control system |
| CN117473345A (en) * | 2023-12-27 | 2024-01-30 | 山东顺发重工有限公司 | Flange forging defect detection method based on data analysis |
| CN117473345B (en) * | 2023-12-27 | 2024-03-22 | 山东顺发重工有限公司 | Flange forging defect detection method based on data analysis |
| CN117540238A (en) * | 2024-01-05 | 2024-02-09 | 长春同泰企业管理服务有限责任公司 | Data security management method for industrial digital information acquisition device |
| CN117540238B (en) * | 2024-01-05 | 2024-03-22 | 长春同泰企业管理服务有限责任公司 | Data security management method for industrial digital information acquisition device |
| CN117665935A (en) * | 2024-01-30 | 2024-03-08 | 山东鑫国矿业技术开发有限公司 | Monitoring data processing method for broken rock mass supporting construction process |
| CN117665935B (en) * | 2024-01-30 | 2024-04-19 | 山东鑫国矿业技术开发有限公司 | Monitoring data processing method for broken rock mass supporting construction process |
| CN117782364A (en) * | 2024-02-28 | 2024-03-29 | 西瑞思创(西安)实验仪器有限公司 | Real-time monitoring method and system for high-flux automatic material rapid screening reactor |
| CN117782364B (en) * | 2024-02-28 | 2024-05-03 | 西瑞思创(西安)实验仪器有限公司 | Real-time monitoring method and system for high-flux automatic material rapid screening reactor |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116821833B (en) | 2023-11-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN116821833B (en) | Data anomaly detection method for machine filter cloth adhesion | |
| CN110263846B (en) | Fault diagnosis method based on fault data deep mining and learning | |
| CN110738255A (en) | device state monitoring method based on clustering algorithm | |
| CN114781467B (en) | Fault detection method based on vibration similarity | |
| CN111898443B (en) | Flow monitoring method for wire feeding mechanism of FDM type 3D printer | |
| CN112380992B (en) | Method and device for evaluating and optimizing accuracy of monitoring data in machining process | |
| CN112414694A (en) | Equipment multistage abnormal state identification method and device based on multivariate state estimation technology | |
| CN117191305B (en) | State evaluation method and system for highway bridge | |
| CN111538311A (en) | Flexible multi-state self-adaptive early warning method and device for mechanical equipment based on data mining | |
| CN112068120A (en) | micro-Doppler time-frequency plane individual soldier squad identification method based on two-dimensional Fourier transform | |
| CN114817228A (en) | Abnormal value removing method for long-time sequence monitoring data of lake water level | |
| CN114118219A (en) | Data-driven real-time abnormal detection method for health state of long-term power-on equipment | |
| CN115220396A (en) | Intelligent monitoring method and system for numerical control machine tool | |
| CN107422373B (en) | A kind of coal mine rock burst disaster on-line early warning method of feature based drift | |
| CN115950666B (en) | Intelligent fault detection method for starch glucose production equipment | |
| CN115310497A (en) | Abnormity identification method for bearing of numerical control machine tool | |
| CN117271980B (en) | Drilling data processing method and device, electronic equipment and storage medium | |
| CN110956340A (en) | Engineering test detection data management early warning decision method | |
| CN116626454B (en) | Oil paper insulation UHF partial discharge signal anti-interference identification and positioning method and device based on correction time-frequency clustering | |
| US20230211451A1 (en) | Intelligent analysis system for measuring signals of polishing pad surface, method and computer readable medium thereof | |
| CN117439827B (en) | Network flow big data analysis method | |
| CN112464146B (en) | Key subsystem based on historical telemetering data and single-machine correlation health baseline construction method | |
| CN112087272B (en) | Automatic detection method for electromagnetic spectrum monitoring receiver signal | |
| CN116329197B (en) | Production cleaning control method of filter antenna | |
| CN117148812A (en) | Fault feature parameter extraction method based on high-dimensional time sequence data |
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 |