CN115293274A - Time sequence mutation point detection method based on antagonistic alternative sliding window - Google Patents
Time sequence mutation point detection method based on antagonistic alternative sliding window Download PDFInfo
- Publication number
- CN115293274A CN115293274A CN202210967432.4A CN202210967432A CN115293274A CN 115293274 A CN115293274 A CN 115293274A CN 202210967432 A CN202210967432 A CN 202210967432A CN 115293274 A CN115293274 A CN 115293274A
- Authority
- CN
- China
- Prior art keywords
- subsequence
- mutation
- mutation point
- detection
- sliding window
- 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
Landscapes
- Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)
Abstract
The invention discloses a time series mutation point detection method based on an antagonistic alternative sliding window. Aiming at an industrial process variable time sequence collected from a sensor, a confrontation window intercepting subsequence is designed and divided into a plurality of sections of equal-length subsequences; constructing by using two adjacent subsequences as positive and negative sample setsSVMCarrying out countermeasure data distribution inspection by the second classification model, and carrying out abnormal sequence location; according toAUCThe detection threshold value transfers the subsequences possibly having the mutation points into an alternative sliding window for detection; carrying out mutation point judgment through a sliding mean value and a sliding variance of an alternating window where the time sequence data point to be detected is located; then, smoothing the mutation points, adaptively adjusting the length of an anti-type window of the next wheel according to a local detection result after the subsequence is detected, and intercepting a new subsequence to perform the next round of mutation point detection; the invention effectively improves the real-time performance and the accuracy of the mutation point detection.
Description
Technical Field
The invention relates to a time sequence mutation point detection method based on an anti-alternant sliding window, belonging to the technical field of artificial intelligence data mining.
Background
With the rapid development of scientific technology and modern industry, mechanical electronic equipment in industrial production is developing towards the directions of intellectualization, high speed, complication, digitalization and high power, and mass production data and equipment operation data are generated every day. The time sequence data is the most widely existing data type, is a sequence formed by arranging numerical values of the same statistical index according to the occurrence time sequence, and can explain characteristic information closely related to production activities, such as environment change, equipment running state and the like. In actual production, the industrial process variable data are collected through a multi-sensor technology, and the correlation and difference inside the data are deeply analyzed and mined, so that the running state of the current equipment is obtained, and the method has important significance for promoting high-quality development of industrial production.
The detection of the time series mutation point is an important research branch of the data mining technology, and means that a change point is found in the time series, and the distribution of two sections of data before and after the change point is not the same. In the classification problem, data in two sections before and after the mutation point can be divided into two categories. When the data distribution of the time sequence changes, how to quickly and accurately find the position of the mutation point is the basis for further fault prediction, abnormal positioning and fault root cause analysis. The method has important significance for constructing a safe and reliable industrial production system and reducing the occurrence of faults to the maximum extent in time. Therefore, the problem to be solved in the prior art is to improve the detection accuracy and efficiency of the time-series mutation point.
The sliding window model is one of the key techniques for mutation point detection. The data to be detected can be divided into a plurality of subsequences by a sliding window technology, each subsequence is detected according to a window sequence, rapid detection of multiple mutation points of time sequence data is achieved, the real-time requirement of online detection is met, and the window size has a large influence on the mutation point detection accuracy rate in a sliding window model. If the window is too large, the data fluctuation in the window can be covered, and the detection accuracy of the mutation point is reduced. The window is too small, the carried information data is less, and the detection precision of the mutation point is reduced. Therefore, how to select a proper window size and a detection method is a key for improving the mutation point detection technology. How to select the window length in a self-adaptive manner becomes a key for improving the detection precision.
Disclosure of Invention
In order to solve the technical problem, the invention provides a time sequence mutation point detection method based on an anti-alternant self-adaptive sliding window.
The specific technical scheme of the invention is as follows:
s1: preprocessing time series data of the process variable to obtain a subsequence set;
s2: adaptively adjusting the size of a confrontation type sliding window of the interception subsequence;
s3: positioning the abnormal subsequence by using a confrontation type sliding window;
s4: mutation point detection is performed using an alternating sliding window.
Further, the step S1 of preprocessing time series data specifically includes:
(1) A process variable time series of length T acquired in real time from an industrial process is represented as: x = { X 1 ,...,x t ,...,x T In which x t Is the value of sequence X at time T, and T ∈ {1, 2.., T };
(2) For the time sequence X, after the antagonistic type self-adaptive sliding window segmentation, truncated into a set of subsequences and denoted X = { S = 1 ,...,S n ,...,S N In which S is n An nth subsequence representing X; for subsequence S n Data x in (1) t Mapping to [0,1 ] based on max-min windowing normalization method in equation (1)]Interval(s)
In the formula, max (S) n ) Is a subsequence S n Maximum value of (1), min (S) n ) Is a subsequence S n The minimum value of (a) is greater than (b),the normalized data is obtained; then, the normalized time series can be recorded asAnd is provided with
Further, the step S2 is based on a window size adaptive adjustment strategy of the local detection result, which specifically includes:
(1) For subsequencesIt is divided into 4 subsequences of the same length from left to right in time order, and is represented as The mutation point detection result is defined as a local detection result expressed as Representing subsequencesThe ith subsequence ofResult of mutation point detection of (1), initialIndicating that the mutation point is not detected, if the mutation point determination is performedWhen a mutation point is detected, then
(2) Defining subsequencesThe ith subsequence ofHas a detection weight ofSubsequence(s)Are cumulatively weighted byCalculated from equation (2):
s2-3: setting a truncated subsequence S 1 Has an initial size of w 1 ,Truncating the subsequence S with increasing value of n n Of antagonistic window size w n The adaptive adjustment can be performed according to formula (3):
further, the step S3 is based on the abnormal sequence localization of the countermeasure type sliding window, specifically as follows:
(1) For subsequencesPositive and negative confrontation training sample set constructed by two adjacent subsequencesWherein the superscripts "+" indicate that the data in the subsequence are each given a "positive" label, wherein the superscripts "-" indicate that the data in the subsequence are each given a "negative" label;
(2) For thePositive and negative confrontation sample set in (1)Constructing a two-classification model based on a Support Vector Machine (SVM) as an abnormal distribution detection model;
(3) According to the training sample setConstructing a two-classification model based on a Support Vector Machine (SVM) as an abnormal distribution detection model;
(4) The area under the ROC curve (AUC) value is used as the evaluation index of the SVM two-classification model, and is shown in formula (4)
In the formula, M and N respectively represent the number of positive samples and negative samples, and Tol represents the number of positive samples with the prediction probability higher than that of negative samples in M multiplied by N pairs of samples;the value range is [0,1 ]]The larger the value is, the higher the classification accuracy of the SVM model is; if it isThen determine the adjacent bipartite sequenceAndthe distribution is the same, and mutation point detection is not needed; otherwise, switching to the step S4 to detect the mutation point by using an alternative sliding window;
further, the step S4 is based on mutation point detection of the alternative sliding window, and specifically includes the following steps:
(1) ForAny one data point ofBy a length ofWith respect to alternative window interceptionIs a data set ofDefinition of d t The abnormality determination threshold value of (2) is g (d) t )
g(d t )=min(max(d t )-mean(d t ),mean(d t )-min(d t )) (5)
In the formula, max (d) t )、min(d t ) And mean (d) t ) Respectively represents d t Maximum, minimum and mean values of;
(2) Mutation point determination according to equations (6) and (7)
In the formula, std (d) T ) Denotes d t Alpha is a regulating parameter;
in the formula, g 0 Indicating an initial anomaly number threshold, signRepresents rounding down;representing the number of judged abnormal points in the past 2d data points;
when the temperature is higher than the set temperatureWhen the formulas (6) and (7) are satisfied at the same time, it is judgedFor mutation points, markingLTR =1 of the located subsequences and storing the LTR in a global mutation point result set Test;
(3) If the step (2) is performed, theIf the mutation point is determined, smoothing is performed according to the formula (8)
Wherein k represents a discontinuity smoothing coefficient;
(4) Using pairs of step S3 to step S4After mutation point detection, sequencingRepeating the steps S3 to S4 as a next wheel pair resistance sample set to perform next wheel mutation point detection; to a subsequenceAfter the detection is finished, according to the local detection resultUtilizing the step S2 to obtain the width w of the next wheel to the resisting window n+1 Adaptive truncation of subsequence S n+1 And after the pretreatment of the step S1, repeating the steps S3 to S4 to carry out a new round of mutation point detection.
The invention has the beneficial effects that: the invention discloses a time sequence anomaly detection method which is a data driving method based on a sliding window model, and utilizes time sequence data in two adjacent sliding windows as training samples to construct a binary model for resistance distribution detection so as to realize anomaly sequence positioning. Then, a dynamic judgment threshold value is set based on the alternative window, mutation point detection is carried out, the size of the sliding window is dynamically adjusted according to the local detection result, and the real-time performance and the accuracy of mutation point detection are effectively improved.
Drawings
In order to illustrate embodiments of the present invention or technical solutions in the prior art more clearly, the drawings which are needed in the embodiments will be briefly described below, so that the features and advantages of the present invention can be understood more clearly by referring to the drawings, which are schematic and should not be construed as limiting the present invention in any way, and for a person skilled in the art, other drawings can be obtained on the basis of these drawings without any inventive effort. Wherein:
FIG. 1 is a block flow diagram of the method of the present invention;
FIG. 2 is a view of the turntable with the counterweight screws installed thereon;
FIGS. 3 (a) - (c) are time series diagrams of different states of a rotating mechanical device as captured by a sensor in a method embodiment of the present invention;
FIG. 4 is a window adjustment strategy flow diagram;
FIG. 5 is a flow chart of the resistance distribution test performed by the present invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced otherwise than as specifically described herein and, therefore, the scope of the present invention is not limited by the specific embodiments disclosed below.
As shown in FIG. 1, the invention is based on collecting vibration acceleration signals from the normal state of the motor rotor and four rotor unbalance fault states, and obtaining the time sequence of transition from the normal state to any fault state; intercepting the subsequence according to the confrontation window, dividing the subsequence into 4 equal-length subsequences, and performing windowing normalization processing; secondly, constructing an SVM two-classification model by taking two adjacent subsequences as positive and negative training sample sets to perform antagonism data distribution inspection; transferring the subsequences possibly having mutation points into an alternative sliding window for detection according to a model AUC detection threshold value; carrying out catastrophe point confidence test by combining an experience threshold through a sliding mean value and a sliding variance of an alternating window where the time sequence data point to be detected is located; carrying out mutation point judgment on data points exceeding the upper and lower confidence limits based on an iterative mode; and finally, storing the mutation points into a global mutation point result set, smoothing the mutation points, adaptively updating the length of an anti-window of the next wheel according to a local detection result after the detection of the subsequence is finished, and intercepting a new subsequence to perform the next round of mutation point detection.
For the convenience of understanding the technical solutions of the present invention, the technical solutions of the present invention will be described in detail by specific examples.
Examples
The embodiment comprises the following steps:
s1: the invention takes the process that the motor rotor in the rotating machinery fault is changed into four rotor unbalance fault grades from a normal state as an example; the generation of rotor faults generally can cause abnormal vibration of the generator rotor, and the abnormal vibration information source (sensor) can cause the amplitude of a specific frequency component to increase or decrease; the present invention sets 4 types of fault conditions in this embodiment: normal (D) 1 ) Rotor imbalance fault (D) 2 ) Rotor imbalance fault (D) 3 ) Rotor imbalance fault (D) 4 ) Rotor imbalance fault (D) 5 ) A larger value of the recorded signal indicates a higher degree of imbalance failure of the rotor.
As shown in fig. 2, for the turntable additionally mounted on the rotor test bed, 1 to 4 counter weight screws are respectively additionally mounted in the same direction to simulate four levels of unbalance in a certain direction, and the more screws are mounted, the more obvious the unbalance effect is; in the example, the experimental equipment is a flexible rotor ZHS-2 multifunctional experimental platform of the motor, the rotating speed of the motor rotor is set to be 1500r/min, the sampling frequency is 1280Hz, and the fundamental frequency is 1X =25H; according to the above experimental conditions, three vibration acceleration sensors are equidistantly placed in the horizontal direction of the middle base of the rotor test bed, four unbalance faults are alternately operated on the experimental platform, and the HG8902 data collector is used for data acquisition, so as to obtain the time sequence X = { X = of the process variables monitored by the three sensors as shown in fig. 3 (a) - (c) 1 ,x 2 ,...,x T },x t ={x t,1 ,x t,2 ,x t,3 } T 。
X = { X) for time series 1 ,x 2 ,...,x T After adaptive countersliding window segmentation, it is truncated into a subsequence set and denoted as X = { S = 1 ,S 2 ,...,S N },S n The nth subsequence of X, and each subsequence has a length corresponding to the antagonism windowDetermining the size; for subsequence S n Normalizing the time series within the window using the maximum and minimum values of the series within the window according to equation (1) to map the raw data to [0,1]The interval solves the problem of sequence amplitude difference under different measurement conditions when the time span is large; the normalized time series is recorded as
Where, max (S) n ) Is a subsequence S n Maximum value of (1), min (S) n ) Is a subsequence S n The minimum value of (a) is greater than (b),is normalized data.
S2: in this embodiment the initial confrontation window size w 1 Selecting based on empirical parameter table 1, and intercepting the obtained subsequence according to countermeasure windowInitial sequence detection weight ofThe local detection result isInitialIf in the sub-sequenceWhen a mutation point is detected, then
The closer the subsequence detection result to the current moment is specified to have higher weight, so that the mutation point detection condition at the current moment can be reflected; defining subsequencesThe ith subsequence ofHas a detection weight ofSubsequence(s)Are cumulatively weighted byCalculated from equation (2):
initial size w 1 The countermeasure window sliding process is adaptively adjusted according to the formula (3), and a window adjustment strategy flow chart is shown in fig. 4:
when the temperature is higher than the set temperatureIs represented in a subsequenceThe probability that the mutation point is not detected is high, the current data distribution is smooth, the window size is required to be enlarged, and the mutation point detection efficiency is improved; when the temperature is higher than the set temperatureWhen the window size w is not changed, the probability that the mutation point is detected is close to the probability that the mutation point is not detected; when in useWhen the detection is carried out, the probability that the mutation point is not detected is higher, the current data distribution is more oscillatory, the size of a window is reduced, and the condition that the detection is missed due to the fact that the window is too large is prevented;
s3: for subsequencesPositive and negative confrontation training sample set constructed by two adjacent subsequencesThe countermeasure testing flow is shown in fig. 5;
to be provided withFor example, the SVM model obtained by training the SVM model is further describedClassifying as a test set; if it isThe poor classification of the model, the inability to distinguish between positive and negative samples, indicatesAndbelonging to the same data distribution and not containing mutation points; if it isOrThe model is shown to have the capability of separating positive and negative samples in adjacent bipartite sequencesMutation points exist at a large probability, and then mutation point detection is carried out by switching to an alternating window;
s4: finding out the two classification models according to the classification result of the SVM two classification modelsCarrying out mutation point detection by using the positive and negative mutation boundaries as reference front and back iteration time sequence data points;
(1) To be provided withAs the mutation boundary, a length ofWith respect to alternative window interceptionIs a data set ofd t The abnormality determination threshold value of (2) is g (d) t )
g(d t )=min(max(d t )-mean(d t ),mean(d t )-min(d t )) (5)
In the formula, max (d) t )、min(d t ) And mean (d) t ) Respectively represent d t Maximum, minimum and mean values of;
(2) Mutation point determination according to equations (6) and (7)
In the formula, std (d) T ) Denotes d t Alpha is a regulating parameter;
in the formula, g 0 Indicating an initial anomaly number threshold, signRepresents rounding down;representing the number of abnormal points judged in the past 2d data points;
when in useWhen the formulas (6) and (7) are satisfied at the same time, it is judged thatFor mutation points, markingLTR =1 of the located subsequences and storing the LTR in a global mutation point result set Test;
(3) Setting an abnormal value smoothing coefficient k, and carrying out mutation point x according to a formula (8) T Carrying out smoothing treatment;
(4) Using the pairs of steps S3 to S4After the data points in the sequence are subjected to mutation point detection, the data points are sequenced in sequenceRepeating the steps S3-S4 to detect the next round of mutation points as a next round of anti-sample set; wait to subsequenceAfter the detection is finished, according to the local detection resultObtaining the width w of the next pair of reactance windows by utilizing the step S2 n+1 Adaptive truncation of subsequence S n+1 And after the pretreatment of the step S1, repeating the steps S3-1 to S4-3 to perform a new round of mutation point detection.
Finally, wait time sequence X = { X = 1 ,x 2 ,...,x T Finishing detection of all subsequences, and evaluating; if n actual mutation points exist in the time sequence X, c mutation points are detected by the method, wherein the actual mutation points include TP actual mutation points and FP false mutation points detected by mistake; the remaining undetected data comprise FN undetected actual mutation points and TN actual non-mutation points; from this, the following equation can be obtained:
based on the above-mentioned variable assumptions, evaluation is performed according to the evaluation indexes of formula (10) and formula (11),
in the formula, hit is a Hit rate and represents the ratio of the number of correctly identified mutation points to the number of actual mutation points; maE is an average absolute error and represents the ratio of the sum of the distances between each detected mutation point and the actual mutation point to the number of the mutation points; test represents the detected mutation point set, and Actual represents the set of Actual mutation points;
in this embodiment, the parameter w in Table 1 is set 1 D, alpha and g 0 Corresponding values are randomly combined for 200 times, and a parameter set { w after combination is carried out 1 ,d,α,g 0 Detecting mutation points of the time sequences in the graphs 3 (a) to 3 (c), and then detecting the mutation points according to the actual mutation point positions marked in the graphs and the actual mutation point positionsThe positions of the mutation points detected by the invention are evaluated by an algorithm according to the formula (9) to the formula (11), and the evaluation result is shown in table 2.
Table 1 table of empirical parameters referred to under the method of the invention
TABLE 2 statistical table of mutation point 200 times detection results under the method of the present invention
Claims (5)
1. A time series mutation point detection method based on an antagonistic alternative sliding window is characterized by comprising the following steps:
s1: preprocessing time series data of process variables to obtain a subsequence set;
s2: adaptively adjusting the size of a confrontation type sliding window of the interception subsequence;
s3: positioning the abnormal subsequence by using a confrontation type sliding window;
s4: mutation point detection is performed using an alternating sliding window.
2. The method for detecting time-series mutation points based on the sliding window against alternans according to claim 1, wherein the step S1 comprises:
s1-1 a process variable time series of length T acquired in real time from an industrial process is represented as: x = { X 1 ,...,x t ,...,x T In which x t Is the value of the process variable time series X at time T, and T ∈ {1, 2.
S1-2: for time series X, after the antagonistic adaptive sliding window is divided, the time series X is cut into a subsequence set and is expressed as X = { S = { (S) } 1 ,...,S n ,...,S N In which S is n Represents XThe nth subsequence of (1);
for subsequence S n Data x in (1) t It is mapped to [0,1 ] based on the max-min windowing normalization method in equation (1)]Interval(s)
3. The method for detecting time-series mutation points based on the sliding window against alternans according to claim 2, wherein the step S2 is specifically as follows:
s2-1: for subsequencesIt is divided into four subsequences of the same length from left to right in time order, and is shown as The mutation point detection result of (2) is defined as a local detection result expressed as Representing subsequencesThe ith subsequence ofDetection of mutation points of (1), initiationIndicating that the mutation point is not detected, if the mutation point determination is performedIn detection of a mutation point, then
S2-2: defining subsequencesThe ith subsequence ofHas a detection weight ofSubsequence(s)Are cumulatively weighted byCalculated from equation (2):
s2-3: setting a truncated subsequence S 1 Has an initial size of w 1 ,Truncating the subsequence S with increasing value of n n Confrontation type window size w n Performing adaptive adjustment according to formula (3):
4. the method for detecting time-series mutation points based on the sliding window against alternans according to claim 3, wherein the step S3 is specifically as follows:
s3-1: for subsequencesConstructing positive and negative confrontation training sample set by two adjacent subsequencesWherein the superscripts "+" indicate that the data in the subsequence are each given a "positive" label, wherein the superscripts "-" indicate that the data in the subsequence are each given a "negative" label;
s3-2: for thePositive and negative confrontation sample set in (1)Establishing a second classification model based on a support vector mechanism to serve as an abnormal distribution detection model;
s3-3: the area value under the ROC curve is used as the evaluation index of the SVM two-classification model, and the evaluation index is shown in a formula (4)
In the formula, M and N respectively represent the number of positive samples and negative samples, and Tol represents the number of positive samples with the prediction probability higher than that of negative samples in M multiplied by N pairs of samples;the value range is [0,1 ]]The larger the value is, the higher the classification accuracy of the SVM model is; if it isThen determine the adjacent bipartite sequenceAndthe distribution is the same, and mutation point detection is not needed; otherwise, the step S4 is carried out, and mutation point detection is carried out by using an alternative sliding window.
5. The method for detecting time-series mutation points based on the sliding window against alternans according to claim 4, wherein the step S4 comprises:
s4-1: for theAny one data point ofIntercepting with alternate windows of length dIs a data set ofDefinition of d t The abnormality determination threshold value of (2) is g (d) t )
g(d t )=min(max(d t )-mean(d t ),mean(d t )-min(d t )) (5)
In the formula, max (d) t )、min(d t ) And mean (d) t ) Respectively represents d t Maximum, minimum and mean values of;
s4-2: mutation point determination according to equations (6) and (7)
In the formula, std (d) T ) Denotes d t Alpha is a regulating parameter;
in the formula, g 0 Indicating an initial anomaly number threshold, signRepresents rounding down;representing the number of judged abnormal points in the past 2d data points;
when the temperature is higher than the set temperatureWhen the formulas (6) and (7) are satisfied at the same time, it is judgedFor mutation points, markingLTR =1 of the located subsequences and is stored in a global mutation point result set Test;
s4-3 if going through step S4-2, willIf it is judged as a mutation point, smoothing is performed according to the formula (8)
In the formula, k represents a mutation point smoothing coefficient;
s4-4 uses the pairs from step S3-1 to step S4-4After mutation point detection, the sequence is divided into sequencesRepeating the steps S3-1 to S4-4 to detect the next round of mutation points as a next round of anti-sample set;
to a subsequenceAfter the detection is finished, according to the local detection resultObtaining the width w of the next counter reactance window by using the formula (3) in the step S2 n+1 Adaptive truncation of subsequence S n+1 And after the pretreatment of the step S1, repeating the steps S3-1 to S4-3 to perform a new round of mutation point detection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210967432.4A CN115293274A (en) | 2022-08-12 | 2022-08-12 | Time sequence mutation point detection method based on antagonistic alternative sliding window |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210967432.4A CN115293274A (en) | 2022-08-12 | 2022-08-12 | Time sequence mutation point detection method based on antagonistic alternative sliding window |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115293274A true CN115293274A (en) | 2022-11-04 |
Family
ID=83828223
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210967432.4A Pending CN115293274A (en) | 2022-08-12 | 2022-08-12 | Time sequence mutation point detection method based on antagonistic alternative sliding window |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115293274A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117088071A (en) * | 2023-10-19 | 2023-11-21 | 山西戴德测控技术股份有限公司 | System, server and method for positioning damaged position of conveyor belt |
CN117591987A (en) * | 2024-01-18 | 2024-02-23 | 北京国旺盛源智能终端科技有限公司 | Electric equipment monitoring method and system based on artificial intelligence |
-
2022
- 2022-08-12 CN CN202210967432.4A patent/CN115293274A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117088071A (en) * | 2023-10-19 | 2023-11-21 | 山西戴德测控技术股份有限公司 | System, server and method for positioning damaged position of conveyor belt |
CN117088071B (en) * | 2023-10-19 | 2024-01-23 | 山西戴德测控技术股份有限公司 | System, server and method for positioning damaged position of conveyor belt |
CN117591987A (en) * | 2024-01-18 | 2024-02-23 | 北京国旺盛源智能终端科技有限公司 | Electric equipment monitoring method and system based on artificial intelligence |
CN117591987B (en) * | 2024-01-18 | 2024-04-26 | 北京国旺盛源智能终端科技有限公司 | Electric equipment monitoring method and system based on artificial intelligence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Huo et al. | Incipient fault diagnosis of roller bearing using optimized wavelet transform based multi-speed vibration signatures | |
CN115293274A (en) | Time sequence mutation point detection method based on antagonistic alternative sliding window | |
CN110276416B (en) | Rolling bearing fault prediction method | |
CN110070060B (en) | Fault diagnosis method for bearing equipment | |
Kang et al. | A hybrid feature selection scheme for reducing diagnostic performance deterioration caused by outliers in data-driven diagnostics | |
CN108760305B (en) | Bearing fault detection method, device and equipment | |
CN112746934B (en) | Method for diagnosing fan fault through self-association neural network | |
CN116413395B (en) | Intelligent detection method for environmental abnormality | |
CN110160765A (en) | A kind of shock characteristic recognition methods and system based on sound or vibration signal | |
CN116380445B (en) | Equipment state diagnosis method and related device based on vibration waveform | |
CN116448419A (en) | Zero sample bearing fault diagnosis method based on depth model high-dimensional parameter multi-target efficient optimization | |
CN113947017A (en) | Method for predicting residual service life of rolling bearing | |
Song et al. | Data and decision level fusion-based crack detection for compressor blade using acoustic and vibration signal | |
CN115238753A (en) | Self-adaptive SHM data cleaning method based on local outlier factor | |
CN115890366A (en) | Online monitoring method for early flutter of inner circle of bearing ring during plunge grinding | |
CN115273791A (en) | Industrial equipment fault detection method and device based on acoustic feature driving | |
CN113743010B (en) | Rolling bearing running state evaluation method based on EEMD energy entropy | |
CN114611548A (en) | State classification method and system based on vibration analysis | |
CN113237619B (en) | Fault early warning method, device, equipment and storage medium for variable-speed rotating machinery vibration | |
CN115795999B (en) | Early warning method for abnormal performance of long-term service pumped storage unit | |
US11339763B2 (en) | Method for windmill farm monitoring | |
CN113889147A (en) | Mechanical system voiceprint signal matching method based on dynamic time warping algorithm | |
CN111695634A (en) | Data abnormal mutation point detection algorithm based on limited accompanying censoring mechanism | |
CN111044176A (en) | Method for monitoring temperature abnormity of generator | |
Zhao et al. | Fault Diagnosis of Rolling Bearings based on GA-SVM model |
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 |