CN115935286A - Abnormal point detection method, device and terminal for railway bearing state monitoring data - Google Patents
Abnormal point detection method, device and terminal for railway bearing state monitoring data Download PDFInfo
- Publication number
- CN115935286A CN115935286A CN202211511727.7A CN202211511727A CN115935286A CN 115935286 A CN115935286 A CN 115935286A CN 202211511727 A CN202211511727 A CN 202211511727A CN 115935286 A CN115935286 A CN 115935286A
- Authority
- CN
- China
- Prior art keywords
- abnormal
- feature
- sample subset
- frequency domain
- segment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The invention provides a method, a device and a terminal for detecting abnormal points of railway bearing state monitoring data. The method comprises the following steps: acquiring a vibration acceleration signal of a railway bearing as an original sample; carrying out non-delay cutting on an original sample to obtain a sample subset; the sample subset comprises n segments; respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain a characteristic set of the sample subset; performing feature selection on the feature set of the sample subset to obtain a feature matrix of the sample subset; based on an isolated forest algorithm, carrying out anomaly detection according to the characteristic matrix of the sample subset to obtain an anomaly score of each segment; and based on a preset abnormal score threshold value, positioning the abnormal segments according to the abnormal scores of the segments. The method can effectively position the abnormal segment, monitor the data quality of the mechanical equipment, provide good data quality for fault identification of subsequent fault diagnosis, service life prediction, migration diagnosis and the like, and enhance the decision accuracy.
Description
Technical Field
The invention relates to the technical field of bearing state monitoring, in particular to a method, a device and a terminal for detecting abnormal points of railway bearing state monitoring data.
Background
With the vigorous development of various sensor technologies, the monitoring of the railway bearing state is brought to the industrial big data era. In the current railway bearing state monitoring process, no matter fault diagnosis or service life prediction, signals collected in the operation process are often relied on. However, while the abundant signal data brings opportunities for realizing a series of applications such as industrial modeling, prediction, control, decision, optimization, fault diagnosis and the like of the railway bearing, a new challenge is brought, namely how to ensure the quality of the acquired big data.
The railway bearing signal data often causes abnormal conditions such as deviation values, missing values, drift values, inconsistent data and the like due to factors such as sensor faults, manual operation factors, system errors, the influence of severe working environments, acquisition line faults and acquisition instrument faults. The abnormal data is directly applied to diagnosis or analysis, and most of the abnormal data can cause interference or influence the accuracy of the abnormal data. Therefore, the concept of "data cleaning" is developed, and the first task of data cleaning is to find the abnormal points in the complex railway bearing signals and then delete or repair the abnormal points according to the analysis of the abnormal points, so as to improve the data quality.
In summary, how to accurately and timely detect the abnormality in the data becomes a key problem to be urgently solved for the health monitoring of the mechanical equipment under the current big data.
Disclosure of Invention
The embodiment of the invention provides a method, a device and a terminal for detecting abnormal points of railway bearing state monitoring data, which are used for accurately and timely detecting the abnormal points in the railway bearing state monitoring data.
In a first aspect, an embodiment of the present invention provides a method for detecting an abnormal point of railway bearing state monitoring data, including:
acquiring a vibration acceleration signal of a railway bearing as an original sample;
carrying out non-delay cutting on an original sample to obtain a sample subset; the sample subset comprises n segments;
respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain a characteristic set of the sample subset;
performing feature selection on the feature set of the sample subset to obtain a feature matrix of the sample subset;
based on an isolated forest algorithm, carrying out anomaly detection according to the characteristic matrix of the sample subset to obtain an anomaly score of each segment;
and based on a preset abnormal score threshold value, positioning the abnormal segments according to the abnormal scores of the segments.
In one possible implementation, performing non-delayed slicing on the original samples to obtain a sample subset includes:
and performing non-delay cutting on the original sample by adopting a sliding window technology to obtain a sample subset.
In a possible implementation manner, the time domain, the frequency domain, and the time-frequency domain feature extraction are performed on each segment of the sample subset, so as to obtain a feature set of the sample subset, where the feature set includes:
respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment;
performing parallel fusion on the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment to obtain an initial characteristic set;
and carrying out normalization processing on the initial feature set to obtain a feature set of the sample subset.
In one possible implementation, the temporal features include 16-dimensional temporal features; the frequency domain features comprise 13-dimensional frequency domain features; the time-frequency domain features include 8-dimensional wavelet packet energy features.
In a possible implementation manner, the performing feature selection on the feature set of the sample subset to obtain a feature matrix of the sample subset includes:
based on a filtering type feature selection method Relief, carrying out feature importance analysis on related feature quantity on the features of the feature set of the sample subset to obtain feature weights of all the features of the feature set of the sample subset, and screening out the features lower than a preset weight threshold to obtain the feature set after the first feature selection;
based on an embedded random forest selection method, screening is carried out according to the contribution of each feature in the feature set after the first feature selection on each tree in the random forest, and a feature matrix of the sample subset is obtained.
In a possible implementation manner, based on an isolated forest algorithm, performing anomaly detection according to a feature matrix of a sample subset to obtain an anomaly score of each segment, including:
constructing an isolated forest model according to the characteristic matrix of the sample subset;
and determining the abnormal score of each segment according to the isolated forest model.
In a possible implementation manner, based on a preset abnormality score threshold, the locating an abnormal segment according to the abnormality score of each segment includes:
and taking the section with the abnormal score larger than a preset abnormal score threshold value as an abnormal section.
In a second aspect, an embodiment of the present invention provides an abnormal point detection apparatus for railway bearing state monitoring data, including:
the acquisition module is used for acquiring a vibration acceleration signal of the railway bearing as an original sample;
the cutting module is used for carrying out non-delay cutting on the original sample to obtain a sample subset; the sample subset comprises n segments;
the characteristic extraction module is used for respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain a characteristic set of the sample subset;
the characteristic selection module is used for carrying out characteristic selection on the characteristic set of the sample subset to obtain a characteristic matrix of the sample subset;
the anomaly detection module is used for carrying out anomaly detection according to the characteristic matrix of the sample subset based on an isolated forest algorithm to obtain an anomaly score of each segment;
and the abnormal positioning module is used for positioning the abnormal segments according to the abnormal scores of the segments on the basis of a preset abnormal score threshold value.
In a third aspect, an embodiment of the present invention provides a terminal, which includes a processor and a memory, where the memory is configured to store a computer program, and the processor is configured to call and run the computer program stored in the memory, and execute the method for detecting an abnormal point of railway bearing state monitoring data according to the first aspect or any one of the possible implementation manners of the first aspect.
In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, where a computer program is stored, and the computer program, when executed by a processor, implements the steps of the method for detecting an abnormal point in railway bearing condition monitoring data according to the first aspect or any one of the possible implementations of the first aspect.
The embodiment of the invention provides a method, a device and a terminal for detecting abnormal points of railway bearing state monitoring data, wherein the method can effectively monitor abnormal signals generated in the railway bearing signal acquisition process, the abnormal signals are different from fault signals generated due to faults of a bearing or equipment, and the abnormal signals can be detected under the interference of the fault signals, so that the embodiment of the invention can effectively locate abnormal segments, monitor the data quality of mechanical equipment, provide good data quality for fault identification such as follow-up fault diagnosis, service life prediction, migration diagnosis and the like, enhance decision accuracy, and provide a new idea for data monitoring of unstable nonlinear vibration acceleration of rotating machinery.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive exercise.
FIG. 1 is a schematic flow chart of a method for detecting an abnormal point of railway bearing condition monitoring data according to an embodiment of the present invention;
FIG. 2 is a diagram of a vibration acceleration signal with an abnormal signal in a time series according to an embodiment of the present invention;
FIG. 3 is a schematic illustration of segment anomaly scores provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram of an anomaly detection confusion matrix provided by an embodiment of the present invention;
FIG. 5 is a schematic structural diagram of an abnormal point detecting device for monitoring data of a railway bearing state according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a terminal according to an embodiment of the present invention.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth, such as particular system structures, techniques, etc. in order to provide a thorough understanding of the embodiments of the invention. It will be apparent, however, to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present invention with unnecessary detail.
Referring to fig. 1, it shows a flowchart of an implementation of the abnormal point detection method for monitoring data of a railway bearing state according to an embodiment of the present invention. The main body of the method for detecting the abnormal point of the railway bearing condition monitoring data may be a terminal.
Referring to fig. 1, the method for detecting an abnormal point of the railway bearing condition monitoring data includes:
in S101, a railway bearing vibration acceleration signal is acquired as a raw sample.
The vibration acceleration signal of the railway bearing is collected as a raw sample by arranging an acceleration sensor on the railway bearing.
Setting acquisition time t, acquiring signals of the vibration acceleration signals of the railway bearing in a time period t, and acquiring an original sample x (t) = [ x = [ [ x ]) 1 ,x 2 ，…x N ]Wherein x (t) represents the vibration acceleration signal collected in the time period t, and N represents the time length of the vibration acceleration signal, and the collection frequency and the time t are jointly determined.
In S102, carrying out non-delay cutting on the original sample to obtain a sample subset; the sample subset includes n fragments.
In this embodiment, the original sample may be cut into n pieces by performing non-delay cutting on the original sample, so as to obtain a sample subset.
In S103, time domain, frequency domain, and time-frequency domain feature extraction are performed on each segment of the sample subset, so as to obtain a feature set of the sample subset.
The time domain features, the frequency domain features and the time-frequency domain features corresponding to the segments of the sample subset respectively form a feature set of the sample subset.
In S104, feature selection is performed on the feature set of the sample subset to obtain a feature matrix of the sample subset.
Based on the feature selection technology, feature selection can be performed on the feature set of the sample subset, some features with lower weight or relevance are screened out, and after the feature selection is performed, the retained features form a feature matrix of the sample subset.
In S105, based on the isolated forest algorithm, anomaly detection is performed according to the feature matrix of the sample subset, and anomaly scores of all the segments are obtained.
According to the embodiment, through an isolated forest algorithm, abnormality detection can be performed on each segment according to the characteristic matrix of the sample subset, and the abnormality score of each segment is obtained.
In S106, based on a preset abnormality score threshold, an abnormal segment is located according to the abnormality score of each segment.
The preset abnormal score threshold may be set according to actual requirements, and is not limited specifically herein.
The method provided by the embodiment can effectively monitor the abnormal signal generated in the railway bearing signal acquisition process, the abnormal signal is different from the fault signal generated due to the fault of the bearing or equipment, and the abnormal signal can be detected under the interference of the fault signal, so that the embodiment of the invention can effectively locate the abnormal segment, monitor the data quality of the mechanical equipment, provide good data quality for fault identification such as follow-up fault diagnosis, service life prediction, migration diagnosis and the like, enhance decision accuracy, and provide a new idea for data monitoring of the non-stable non-linear vibration acceleration of the rotary machine.
In some embodiments, the S102 may include:
and performing non-delay cutting on the original sample by adopting a sliding window technology to obtain a sample subset.
The original vibration acceleration signal, namely the original sample, is cut by adopting a sliding window technology, and the main principle can be expressed as follows: the window length w is set and represents the length of the data point included in each repeated cut of the window. Setting a delay time tau, wherein tau represents whether the starting point of the next time is delayed after the cutting is finished each time, and if tau is selected to be 0, the starting point of the first cutting is x 1 End point is x w The second cut starts at x w+1 Therefore, setting the delay to 0 mainly aims to cut the original vibration acceleration signal into N/w segments without repetition, thereby dividing the x (t) segment into N/w new segments, i.e. N = N/w.
The subset of samples obtained after cutting comprises the following fragments:
a first stage: x 1 ＝[x 1 ,x 2 ……x w ]
And a second stage: x 2 ＝[x w+1 ,x w+2 ，……x 2*w ]
……
An nth segment: x n ＝[x (n-1)w+1 ,……x nw ]
In some embodiments, the S103 may include:
respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment;
performing parallel fusion on the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment to obtain an initial characteristic set;
and carrying out normalization processing on the initial feature set to obtain a feature set of the sample subset.
In some embodiments, the temporal features comprise 16-dimensional temporal features; the frequency domain features comprise 13-dimensional frequency domain features; the time-frequency domain features include 8-dimensional wavelet packet energy features.
And respectively extracting the features of each segment, wherein the modes of extracting the features comprise time domain feature index extraction, frequency domain feature index extraction and time-frequency domain feature index extraction. And obtaining m characteristic indexes from each segment, performing parallel fusion on the characteristic indexes obtained from each segment to obtain an n-m initial characteristic set, and performing normalization processing on the initial characteristic set by using a normalization method to obtain a characteristic set of the sample subset. The feature set characterizes the original sample, and is input in place of the original sample.
The time domain characteristics can comprise 10-dimensional indexes and 6-dimensional dimensionless indexes of time domain analysis in signal analysis, and 16-dimensional time domain characteristic indexes are formed together; the frequency domain characteristics comprise 13-dimensional frequency domain characteristic indexes of dimension and dimensionless; the time-frequency domain indexes comprise 8-dimensional wavelet packet energy characteristics and entropy indexes.
The time-frequency domain feature extraction selection index adopts db10 wavelet as wavelet basis function according to Continuous Wavelet Transform (CWT) theory to perform continuous wavelet transform on signals, and the result is as follows:
wherein the content of the first and second substances,
is->
A is a translation factor and b is a scaling factor.
According to the energy conservation principle in the wavelet transformation process, the following are obtained by calculation:
further derivation yields:
calculating to obtain a time wavelet energy spectrum sequence E:
the wavelet transform has the characteristic of multi-resolution analysis, has the capability of representing the local characteristics of signals in both time domain and frequency domain, and is very suitable for analyzing non-stationary signals and extracting the characteristics of the signals.
In some embodiments, the S104 may include:
based on a filtering type feature selection method Relief, carrying out feature importance analysis on related feature quantity on the features of the feature set of the sample subset to obtain feature weights of all the features of the feature set of the sample subset, and screening out the features lower than a preset weight threshold to obtain the feature set after the first feature selection;
based on an embedded random forest selection method, screening is carried out according to the contribution of each feature in the feature set after the first feature selection on each tree in the random forest, and a feature matrix of the sample subset is obtained.
In the present embodiment, feature selection is first performed using a filtering-type feature selection method Relief. The Relief algorithm is a characteristic weight algorithm, the basic principle of which is that a class of samples R are randomly selected from a training set X, then a nearest leading sample Near Hit is searched from samples similar to R, a nearest neighbor sample Near Miss is searched from samples different from R, if the distance between R and Near Hit on a certain characteristic is smaller than that between R and Near Miss, the characteristic is beneficial to distinguishing the nearest neighbors of the same class and different classes, the characteristic obtains larger weight, otherwise, the weight is reduced, and the process is repeated to obtain the average weight of each characteristic. And screening the lowest d-dimensional characteristic according to the weight to obtain a new characteristic matrix of m-d dimensions.
Specifically, the Relief designs a "correlation statistic" to measure the importance of the features, mainly by: given a training set { (x) 1 ,y 1 )，(x 2 ,y 2 ),…，(x m ,y m ) For each example x i Now Relief x i Find the nearest neighbor, called Near hit, from the same sample as x i The nearest neighbor miss of the heterogeneous sample is searched, so that the component of the corresponding feature of the relevant statistic is as follows:
wherein the content of the first and second substances,
representative sample x a Value in characteristic, in>
Depending on the type of feature.
As can be seen from the above formula, if x i If the distance to its guessing neighbor on the feature is smaller than its distance to its guessing neighbor, the feature is good for classification, and the statistical component becomes larger, and vice versa.
And setting a preset weight threshold, and removing the statistical components lower than the weight threshold to obtain a feature set after the first feature selection. The preset weight threshold may be set according to actual requirements, and is not limited specifically herein.
And after the Relief selection, further analyzing the feature importance by using an embedded selection method in the feature selection and random forests, and performing the feature selection. And screening by using the contribution of each feature on each tree in the random forest, and using the Kerniy index as an evaluation index to screen out a useful feature combination to form a new n x k (k < m) feature matrix which is used as the final characterization of the original signal and used as input.
The basic principle of random forests is as follows: when a tree is constructed, the contribution of each feature on each tree is determined by using a kini index as an index, the specific evaluation process is that in the process of constructing the tree, each node is divided into two branches of a decision tree by a selected feature, the Gini value of the node is obtained according to the splitting purity of the node, the Gini index variable quantities before and after the branches of the adjacent nodes are compared to obtain the importance of the feature in one tree, the average variable quantity of the splitting purity of the node in all the decision trees is traversed on average to obtain the VIM of the feature, the VIM of all the features is checked, the feature with lower importance is screened out according to a threshold value to obtain a final feature matrix, and the final feature matrix is used as original input representation original data.
Selecting Gini index as an evaluation index, adopting VIM to represent feature importance, assuming that j features, i decision trees and c categories exist, calculating the average change quantity of the node splitting purity of the j-th feature in all the RF decision trees as follows:
the Gini index of the ith tree node q is calculated as:
wherein c represents c categories, p qc Represents the proportion of the class c in q.
The importance of the feature j at the ith tree node q, namely the Gini index variation before and after the node q branches is as follows:
GI l (i) and GI r (i) And respectively representing Gini indexes of two nodes before and after the node.
If a feature appears in decision tree i as set Q, then the importance of the feature in the ith tree is:
assuming that there are I trees in the RF, then:
and finishing the feature importance presentation according to the construction of the random forest, screening out e features with lower importance, and obtaining a new feature matrix again.
In some embodiments, the S105 may include:
constructing an isolated forest model according to the characteristic matrix of the sample subset;
and determining the abnormal score of each segment according to the isolated forest model.
In this embodiment, the n × k feature matrix obtained after feature selection is used as input, parameters of the leaf and tree of the isolated forest are determined, and each parameter of the isolated forest tree is determined; selecting a random subset by using input as a training set, training a model by using determined parameters, recursively constructing a tree, and forming an isolated forest model through constructing the tree for multiple times; inputting the test data into the trained isolated forest model, enabling the test data to traverse the whole isolated forest, standardizing the recorded path length through the average path length of tree, and calculating the abnormal score of each subset signal through the introduced abnormal value function S.
When each isolated tree is constructed, a random value in the value range of the characteristic value is randomly selected according to different characteristic values of the data set, all data are divided into two parts according to the random value of the characteristic, all data can be gradually separated, and abnormal data generally occupy a small proportion and have a large difference with normal data, so that the abnormal data can be generally distinguished from the normal data by using few attributes.
Specifically, the process of anomaly detection using the isolated forest algorithm is detailed as follows:
the method comprises the following steps: psi points are randomly selected from the training data as subsamples and placed into the root node of an isolated tree.
Step two: randomly appointing a dimension, and randomly generating a cutting point q in the data range of the current node (q is based on the maximum value and the minimum value of the dimension data).
Step three: and generating a hyperplane by selecting the cutting point, cutting the current node data to form 2 subspaces, placing the part with the current dimensionality smaller than q into the left node, and placing the part with the current dimensionality larger than q into the right space.
Step four: repeating the second step and the third step in the left and right space recursion of the nodes, continuously constructing new leaf nodes until the maximum height is reached or only one data point on the node is finished, and completing the construction of a single tree; the process is repeated to build an isolated forest from a single tree.
Step five: putting the test data into the constructed isolated forest model, enabling the data to traverse the whole forest as much as possible, calculating the abnormal score S (x, n) of each data segment according to the path length of each data segment, and introducing an abnormal score formula as follows:
c(n)＝2H(n-1)-(2(n-1)/n)
H(k)＝ln(k)+ξ,ξ＝0.5772156649
wherein h (x) is the height of x in each tree; c (n) is the average of the path lengths for a given number of samples; ξ is the Euler constant.
In some embodiments, the S106 may include:
and taking the segment with the abnormal score larger than a preset abnormal score threshold value as an abnormal segment.
In the embodiment, the abnormal segment is positioned by setting the preset abnormal score threshold, so that the position of the abnormal signal is positioned.
The abnormal score of each segment can be obtained in an isolated forest abnormal detection algorithm, when the obtained abnormal score is close to 1, the path length is very small, a data point is easy to be isolated, and the segment is abnormal; if the anomaly score is close to 0.5, then the overall path may be relatively average and there may be no anomalies in the sample; if the abnormal score is close to 0, the path length is very large, and the data can be determined to be normal.
In some possible implementations, the abnormality score threshold may be set to 0.5, and when the abnormality score is greater than 0.5, the segment is determined to be an abnormal segment.
In a specific embodiment, the implantation of the experimentally constructed signal and the self-constructed anomaly signal can be described by a Matlab platform.
According to the acquired railway bearing signals, signals containing electromagnetic interference are selected as a basis, missing abnormal signals, sensor saturation abnormal signals and deviation abnormal signals are implanted into the data, fault bearing data containing 4 types of abnormalities in the signals are used as original signals, time series vibration acceleration abnormal signals are shown in figure 2, and the signal acquisition length is N =360000.
By adopting a sliding window technology, setting a window length w =1200, cutting an original signal without delay, and cutting the original signal into 300 segments, wherein abnormal signals exist in 3 rd, 4 th, 5 th, 14 th-27 th and 31 th segments, which shows that 18 segments containing abnormal signals are contained in the 300 segments of signals, obviously, the existence of the abnormal signals accords with the few and different principles proposed by the isolated forest algorithm.
Respectively extracting features of 300 segments of data by using a time domain, a frequency domain and a time-frequency domain as feature indexes, obtaining a 300-37-dimensional feature matrix after extraction is finished, carrying out normalization processing on the feature matrix, then calculating the size of 37 feature related statistics by using a Relief feature screening mode, obtaining features with extremely low 12-dimensional correlation degree by checking the correlation coefficient of each feature, indicating that the 12-dimensional features are useless for anomaly detection, setting a threshold value and deleting the features; and then, performing random forest feature importance analysis on the obtained new feature matrix, taking Gini index as an evaluation index, screening the features with low importance in classification again, selecting the 5-dimensional features with the most accumulated appearance by using a voting method through 10 cycles, and fusing the rest features to obtain a 300 × 20 feature matrix.
And constructing an isolated forest anomaly detection model, wherein the sub-sampling size of the isolated forest anomaly detection model is 256, the number of trees is set to be 100, and the maximum depth of the trees is set to be 8 by a parameter optimization mode. And (3) replacing the original signal with the feature matrix of 300-20, carrying out anomaly detection on the isolated forest model obtained by training, wherein the threshold value is obtained by t = anomaly sample/total sample +0.5, and the threshold value is 0.56, so that the segment with the anomaly score exceeding 0.56 is an anomaly segment, the experimental result is shown in fig. 3, the anomaly score starts to suddenly rise between 2 segments and 3 segments, suddenly falls between 5 segments and 6 segments, the anomaly score continuously stabilizes above 0.56 between 13 segments and 27 segments, and suddenly rises again between 30 segments and 31 segments. Through the confusion matrix in fig. 4, it can be clearly seen that the accuracy of the algorithm for detecting the abnormal data of the 18 sections reaches 100%, and the railway bearing vibration acceleration abnormal signal is quickly and accurately detected.
It should be understood that, the sequence numbers of the steps in the foregoing embodiments do not imply an execution sequence, and the execution sequence of each process should be determined by functions and internal logic of the process, and should not limit the implementation process of the embodiments of the present invention in any way.
The following are embodiments of the apparatus of the invention, reference being made to the corresponding method embodiments described above for details which are not described in detail therein.
Fig. 5 is a schematic structural diagram of an abnormal point detecting device for railway bearing condition monitoring data according to an embodiment of the present invention, and for convenience of description, only the parts related to the embodiment of the present invention are shown, which is detailed as follows:
as shown in fig. 5, the abnormal point detecting device 30 for railway bearing condition monitoring data may include: the system comprises an acquisition module 31, a cutting module 32, a feature extraction module 33, a feature selection module 34, an anomaly detection module 35 and an anomaly localization module 36.
The acquisition module 31 is used for acquiring a vibration acceleration signal of the railway bearing as an original sample;
a cutting module 32, configured to perform non-delay cutting on an original sample to obtain a sample subset; the sample subset comprises n segments;
the feature extraction module 33 is configured to perform time domain, frequency domain and time-frequency domain feature extraction on each segment of the sample subset respectively to obtain a feature set of the sample subset;
the feature selection module 34 is configured to perform feature selection on the feature set of the sample subset to obtain a feature matrix of the sample subset;
the anomaly detection module 35 is used for performing anomaly detection according to the feature matrix of the sample subset based on an isolated forest algorithm to obtain an anomaly score of each segment;
and an anomaly positioning module 36, configured to position an anomaly segment according to the anomaly score of each segment based on a preset anomaly score threshold.
In one possible implementation, the cutting module 32 is specifically configured to:
and (3) performing non-delay cutting on the original sample by adopting a sliding window technology to obtain a sample subset.
In a possible implementation manner, the feature extraction module 33 is specifically configured to:
respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment;
performing parallel fusion on the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment to obtain an initial characteristic set;
and carrying out normalization processing on the initial feature set to obtain a feature set of the sample subset.
In one possible implementation, the temporal features include 16-dimensional temporal features; the frequency domain features comprise 13-dimensional frequency domain features; the time-frequency domain features include 8-dimensional wavelet packet energy features.
In one possible implementation, the feature selection module 34 is specifically configured to:
based on a filtering type feature selection method Relief, carrying out feature importance analysis on related feature quantity on the features of the feature set of the sample subset to obtain feature weights of all the features of the feature set of the sample subset, and screening out the features lower than a preset weight threshold to obtain the feature set after the first feature selection;
based on an embedded random forest selection method, screening is carried out according to the contribution of each feature in the feature set after the first feature selection on each tree in the random forest, and a feature matrix of the sample subset is obtained.
In a possible implementation manner, the anomaly detection module 35 is specifically configured to:
constructing an isolated forest model according to the characteristic matrix of the sample subset;
and determining the abnormal score of each segment according to the isolated forest model.
In a possible implementation, the anomaly locating module 36 is specifically configured to:
and taking the section with the abnormal score larger than a preset abnormal score threshold value as an abnormal section.
Fig. 6 is a schematic diagram of a terminal according to an embodiment of the present invention. As shown in fig. 6, the terminal 4 of this embodiment includes: a processor 40 and a memory 41. The memory 41 is configured to store a computer program 42, and the processor 40 is configured to call and run the computer program 42 stored in the memory 41 to execute the steps in the above-mentioned embodiment of the method for detecting an abnormal point of monitoring data of a railway bearing status, for example, S101 to S106 shown in fig. 1. Alternatively, the processor 40 is configured to call and run the computer program 42 stored in the memory 41, so as to implement the functions of the modules/units in the above-mentioned device embodiments, for example, the functions of the modules/units 31 to 36 shown in fig. 5.
Illustratively, the computer program 42 may be partitioned into one or more modules/units that are stored in the memory 41 and executed by the processor 40 to implement the present invention. The one or more modules/units may be a series of computer program instruction segments capable of performing specific functions, which are used to describe the execution of the computer program 42 in the terminal 4. For example, the computer program 42 may be divided into the modules/units 31 to 36 shown in fig. 5.
The terminal 4 may be a desktop computer, a notebook, a palm computer, a cloud server, or other computing devices. The terminal 4 may include, but is not limited to, a processor 40, a memory 41. Those skilled in the art will appreciate that fig. 6 is merely an example of a terminal 4 and is not intended to be limiting of terminal 4, and may include more or fewer components than those shown, or some components in combination, or different components, e.g., the terminal may also include input-output devices, network access devices, buses, etc.
The Processor 40 may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, discrete Gate or transistor logic, discrete hardware components, etc. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 41 may be an internal storage unit of the terminal 4, such as a hard disk or a memory of the terminal 4. The memory 41 may also be an external storage device of the terminal 4, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card) and the like provided on the terminal 4. Further, the memory 41 may also include both an internal storage unit and an external storage device of the terminal 4. The memory 41 is used for storing the computer program and other programs and data required by the terminal. The memory 41 may also be used to temporarily store data that has been output or is to be output.
It should be clear to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional units and modules is only used for illustration, and in practical applications, the above function distribution may be performed by different functional units and modules as needed, that is, the internal structure of the apparatus may be divided into different functional units or modules to perform all or part of the above described functions. Each functional unit and module in the embodiments may be integrated in one processing unit, or each unit may exist alone physically, or two or more units are integrated in one unit, and the integrated unit may be implemented in a form of hardware, or in a form of software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working processes of the units and modules in the system may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the above embodiments, the description of each embodiment has its own emphasis, and reference may be made to the related description of other embodiments for parts that are not described or recited in any embodiment.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal and method may be implemented in other ways. For example, the above-described apparatus/terminal embodiments are merely illustrative, and for example, the division of the modules or units is only one logical division, and there may be other divisions when actually implemented, for example, a plurality of units or components may be combined or may be integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow in the method according to the above embodiments may be implemented by a computer program, which may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the steps of the above embodiments of the method for detecting an abnormal point in monitoring data of a railway bearing state may be implemented. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, usb disk, removable hard disk, magnetic disk, optical disk, computer Memory, read-Only Memory (ROM), random Access Memory (RAM), electrical carrier wave signals, telecommunications signals, software distribution medium, and the like. It should be noted that the computer readable medium may contain other components which may be suitably increased or decreased as required by legislation and patent practice in jurisdictions, for example, in some jurisdictions, computer readable media which may not include electrical carrier signals and telecommunications signals in accordance with legislation and patent practice.
The above-mentioned embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present invention, and are intended to be included within the scope of the present invention.
Claims (10)
1. An abnormal point detection method for railway bearing state monitoring data is characterized by comprising the following steps:
acquiring a vibration acceleration signal of a railway bearing as an original sample;
carrying out non-delay cutting on the original sample to obtain a sample subset; the subset of samples comprises n segments;
respectively extracting time domain, frequency domain and time-frequency domain features of each segment of the sample subset to obtain a feature set of the sample subset;
performing feature selection on the feature set of the sample subset to obtain a feature matrix of the sample subset;
based on an isolated forest algorithm, carrying out anomaly detection according to the characteristic matrix of the sample subset to obtain an anomaly score of each segment;
and based on a preset abnormal score threshold value, positioning abnormal segments according to the abnormal score of each segment.
2. The method for detecting abnormal points of railway bearing condition monitoring data according to claim 1, wherein the performing non-delayed cutting on the original sample to obtain a sample subset comprises:
and performing non-delay cutting on the original sample by adopting a sliding window technology to obtain a sample subset.
3. The method for detecting the abnormal point of the railway bearing state monitoring data according to claim 1, wherein the step of respectively extracting the time domain, the frequency domain and the time-frequency domain characteristics of each segment of the sample subset to obtain the characteristic set of the sample subset comprises the following steps:
respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment;
performing parallel fusion on the time domain characteristics, the frequency domain characteristics and the time-frequency domain characteristics of each segment to obtain an initial characteristic set;
and carrying out normalization processing on the initial feature set to obtain a feature set of the sample subset.
4. The method of claim 3, wherein the time domain feature comprises a 16-dimensional time domain feature; the frequency domain features comprise 13-dimensional frequency domain features; the time-frequency domain features include 8-dimensional wavelet packet energy features.
5. The method for detecting an abnormal point of railway bearing condition monitoring data according to claim 1, wherein the step of performing feature selection on the feature set of the sample subset to obtain the feature matrix of the sample subset comprises:
based on a filtering type feature selection method Relief, carrying out feature importance analysis on related feature quantity on the features of the feature set of the sample subset to obtain feature weights of all the features of the feature set of the sample subset, and screening out the features lower than a preset weight threshold value to obtain a feature set after first feature selection;
and based on an embedded random forest selection method, screening according to the contribution of each feature in the feature set after the first feature selection on each tree in the random forest to obtain a feature matrix of the sample subset.
6. The method for detecting the abnormal point of the railway bearing state monitoring data according to claim 1, wherein the abnormal detection is performed according to the feature matrix of the sample subset based on the isolated forest algorithm to obtain the abnormal score of each segment, and the method comprises the following steps:
constructing an isolated forest model according to the characteristic matrix of the sample subset;
and determining the abnormal score of each segment according to the isolated forest model.
7. The method for detecting abnormal points in railway bearing condition monitoring data according to any one of claims 1 to 6, wherein the locating abnormal segments according to the abnormal score of each segment based on a preset abnormal score threshold comprises:
and taking the segments with the abnormal score larger than the preset abnormal score threshold value as abnormal segments.
8. An abnormal point detecting device for railway bearing condition monitoring data, comprising:
the acquisition module is used for acquiring a vibration acceleration signal of the railway bearing as an original sample;
the cutting module is used for carrying out non-delay cutting on the original sample to obtain a sample subset; the subset of samples comprises n segments;
the characteristic extraction module is used for respectively extracting time domain, frequency domain and time-frequency domain characteristics of each segment of the sample subset to obtain a characteristic set of the sample subset;
the characteristic selection module is used for carrying out characteristic selection on the characteristic set of the sample subset to obtain a characteristic matrix of the sample subset;
the anomaly detection module is used for carrying out anomaly detection according to the characteristic matrix of the sample subset based on an isolated forest algorithm to obtain an anomaly score of each segment;
and the abnormal positioning module is used for positioning the abnormal segments according to the abnormal scores of the segments on the basis of a preset abnormal score threshold value.
9. A terminal comprising a processor and a memory, the memory storing a computer program, the processor calling and executing the computer program stored in the memory to execute the abnormal point detecting method of the railway bearing condition monitoring data according to any one of claims 1 to 7.
10. A computer-readable storage medium, in which a computer program is stored, which, when being executed by a processor, carries out the steps of the method for detecting an abnormal point in railway bearing condition monitoring data according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211511727.7A CN115935286A (en) | 2022-11-29 | 2022-11-29 | Abnormal point detection method, device and terminal for railway bearing state monitoring data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211511727.7A CN115935286A (en) | 2022-11-29 | 2022-11-29 | Abnormal point detection method, device and terminal for railway bearing state monitoring data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115935286A true CN115935286A (en) | 2023-04-07 |
Family
ID=86651874
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211511727.7A Pending CN115935286A (en) | 2022-11-29 | 2022-11-29 | Abnormal point detection method, device and terminal for railway bearing state monitoring data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115935286A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116992365A (en) * | 2023-08-02 | 2023-11-03 | 广东海洋大学 | Fault diagnosis method and system under random impact interference |
| CN117331921A (en) * | 2023-09-28 | 2024-01-02 | 石家庄铁道大学 | Bearing monitoring multisource data processing method |
| CN117743836A (en) * | 2024-02-21 | 2024-03-22 | 聊城市产品质量监督检验所 | Abnormal vibration monitoring method for bearing |
-
2022
- 2022-11-29 CN CN202211511727.7A patent/CN115935286A/en active Pending
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116992365A (en) * | 2023-08-02 | 2023-11-03 | 广东海洋大学 | Fault diagnosis method and system under random impact interference |
| CN116992365B (en) * | 2023-08-02 | 2024-03-08 | 广东海洋大学 | Fault diagnosis method and system under random impact interference |
| CN117331921A (en) * | 2023-09-28 | 2024-01-02 | 石家庄铁道大学 | Bearing monitoring multisource data processing method |
| CN117743836A (en) * | 2024-02-21 | 2024-03-22 | 聊城市产品质量监督检验所 | Abnormal vibration monitoring method for bearing |
| CN117743836B (en) * | 2024-02-21 | 2024-05-03 | 聊城市产品质量监督检验所 | Abnormal vibration monitoring method for bearing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115935286A (en) | Abnormal point detection method, device and terminal for railway bearing state monitoring data | |
| CN111475680A (en) | Method, device, equipment and storage medium for detecting abnormal high-density subgraph | |
| CN111368890A (en) | Fault detection method and device and information physical fusion system | |
| CN107714038A (en) | The feature extracting method and device of a kind of EEG signals | |
| CN116148679B (en) | Battery health state prediction method and related device | |
| CN111881594B (en) | Non-stationary signal state monitoring method and system for nuclear power equipment | |
| CN116366374B (en) | Security assessment method, system and medium for power grid network management based on big data | |
| CN112580780A (en) | Model training processing method, device, equipment and storage medium | |
| CN113420652A (en) | Method, system, medium and terminal for recognizing abnormity of time sequence signal fragment | |
| CN116306806A (en) | Fault diagnosis model determining method and device and nonvolatile storage medium | |
| CN114819315A (en) | Bearing degradation trend prediction method based on multi-parameter fusion health factor and time convolution neural network | |
| CN117454283A (en) | State evaluation method for wind turbine generator operation detection data | |
| CN115840676A (en) | Computer hardware fault early warning method and system based on multidimensional data | |
| CN114548280A (en) | Fault diagnosis model training method, fault diagnosis method and electronic equipment | |
| CN113726558A (en) | Network equipment flow prediction system based on random forest algorithm | |
| CN116660761B (en) | Lithium ion battery detection method and system | |
| CN116502155B (en) | Safety supervision system for numerical control electric screw press | |
| CN115994327B (en) | Equipment fault diagnosis method and device based on edge calculation | |
| CN114674511B (en) | Bridge modal anomaly early warning method for eliminating time-varying environmental factor influence | |
| CN115729761A (en) | Hard disk fault prediction method, system, device and medium | |
| CN115659271A (en) | Sensor abnormality detection method, model training method, system, device, and medium | |
| CN112580781A (en) | Processing method, device and equipment of deep learning model and storage medium | |
| Yuan et al. | Deep learning approach to multiple features sequence analysis in predictive maintenance | |
| CN117829002B (en) | Aging diagnosis monitoring method and system for power cable | |
| CN113609207B (en) | Data preprocessing method for slope deformation monitoring 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 |