CN114611633B - Health monitoring method of electromagnetic valve - Google Patents

Health monitoring method of electromagnetic valve Download PDF

Info

Publication number
CN114611633B
CN114611633B CN202210500346.2A CN202210500346A CN114611633B CN 114611633 B CN114611633 B CN 114611633B CN 202210500346 A CN202210500346 A CN 202210500346A CN 114611633 B CN114611633 B CN 114611633B
Authority
CN
China
Prior art keywords
health
sequence
characteristic
sequences
electromagnetic valve
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210500346.2A
Other languages
Chinese (zh)
Other versions
CN114611633A (en
Inventor
王强
胡俊
王平
鲁相
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Equipment Design and Testing Technology Research Institute of China Aerodynamics Research and Development Center
Original Assignee
Equipment Design and Testing Technology Research Institute of China Aerodynamics Research and Development Center
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Equipment Design and Testing Technology Research Institute of China Aerodynamics Research and Development Center filed Critical Equipment Design and Testing Technology Research Institute of China Aerodynamics Research and Development Center
Priority to CN202210500346.2A priority Critical patent/CN114611633B/en
Publication of CN114611633A publication Critical patent/CN114611633A/en
Application granted granted Critical
Publication of CN114611633B publication Critical patent/CN114611633B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2474—Sequence data queries, e.g. querying versioned data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis

Abstract

The invention relates to the field of health monitoring of mechanical equipment, and particularly discloses a health monitoring method of an electromagnetic valve, which comprises the steps of acquiring a characteristic sequence, standardizing the characteristic sequence, screening the characteristic sequence, calculating a health monitoring index according to monotonicity of all the characteristic sequences and judging the health condition of the electromagnetic valve according to the health index; the monitoring method provided by the invention obtains a weighted value based on the weighted Wasserstein distance, namely, the weighted value is used as a health index for comprehensively representing the electromagnetic valve at the moment.

Description

Health monitoring method of electromagnetic valve
Technical Field
The invention relates to the field of health monitoring of mechanical equipment, in particular to a health monitoring method of an electromagnetic valve.
Background
In a liquid nitrogen supply system, an on-off electromagnetic valve is an important element, and the on-off electromagnetic valve has the advantages of accurate action, high automation degree and stable and reliable work. Various problems arise inevitably during use. Although the cost of the electromagnetic valve is low, the maintenance is not complicated only by replacing the electromagnetic valve, but the fault is not obvious and is difficult to detect at the initial stage of the fault, which brings serious problems, the product quality is influenced if the fault is light, and accidents may occur if the fault is heavy; solenoid valve monitoring is therefore essential. In the existing monitoring method, two indexes for measuring the similarity degree of distribution are adopted, namely KL Divergence (Kullback-Leibler Divergence) and JS Divergence (Jensen-Shannon Divergence); however, if the two distributions are far apart, with no overlap at all, then the KL divergence value is meaningless, while the JS divergence value is a constant in this case. This is fatal in the learning algorithm, which means that the gradient of this point is 0, i.e., the gradient disappears.
Disclosure of Invention
Therefore, in order to solve the above-mentioned shortcomings, the present invention proposes a solenoid health monitoring method based on weighted Wasserstein distance, which is used for monitoring the health status of a solenoid, by using sensor data of the solenoid.
Specifically, the health monitoring method of the electromagnetic valve comprises the following steps:
step one, acquiring a characteristic sequence, and directly taking a data sequence of sensor measurement data directly representing the electromagnetic valve as the characteristic sequence to obtain a characteristic sequence set
Figure DEST_PATH_IMAGE002A
WhereinNIs composed of
Figure DEST_PATH_IMAGE004AA
The number of medium characteristic sequences;
step two, standardizing the characteristic sequences, eliminating the difference of the characteristic sequences, wherein the standardized formula is as follows:
Figure DEST_PATH_IMAGE006A
，
wherein
Figure DEST_PATH_IMAGE008A
And
Figure DEST_PATH_IMAGE010AA
respectively before and after standardizationiThe characteristic sequences are arranged in a sequence of the characteristic,
Figure DEST_PATH_IMAGE012A
and
Figure DEST_PATH_IMAGE014AA
are respectively the firstiMean and standard deviation of individual signature sequences;
step three, screening the characteristic sequences, and calculating the second step by adopting a monotonicity formulaiGlobal monotonicity of individual feature sequences
Figure DEST_PATH_IMAGE016A
Removing the characteristic sequence of the sensor with little or no degradation tendency, wherein the monotonicity formula is as follows:
Figure DEST_PATH_IMAGE018A
，
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE020AA
，
Figure DEST_PATH_IMAGE022A
and
Figure DEST_PATH_IMAGE024AA
are respectively the firstiA first of the characteristic sequencesjNumerical value and
Figure DEST_PATH_IMAGE026AA
the number of the individual values is,
Figure DEST_PATH_IMAGE028A
is the length of the sample;
step four, according to the monotonicity of all the characteristic sequences, dividing the characteristic sequences into two types through K-means, selecting the characteristic sequences with larger monotonicity, and supposing that the screened characteristic sequences are the characteristic sequences
Figure DEST_PATH_IMAGE030A
Wherein
Figure DEST_PATH_IMAGE032A
，RIs composed of
Figure DEST_PATH_IMAGE034AA
The number of the characteristic sequences in the medium,
Figure DEST_PATH_IMAGE036A
the characteristic sequence with larger monotonicity is adopted;
step five, calculating health monitoring indexes and adding WasserUse of stein distance for health indicators
Figure DEST_PATH_IMAGE038AA
；
Step six, according to the health indexes
Figure DEST_PATH_IMAGE039
And judging the health condition of the electromagnetic valve.
Preferably, the specific method of the fifth step is to perform characteristic sequence
Figure DEST_PATH_IMAGE041
Respectively selecting the length of the sequence from each characteristic sequence from the initial time
Figure DEST_PATH_IMAGE043
A sequence of (2) as a reference sequence, denoted
Figure DEST_PATH_IMAGE045
The characteristic sequence at the t-th moment is recorded as
Figure DEST_PATH_IMAGE047
，
Respectively obtaining probability density functions of the reference sequence and the characteristic sequence at the t-th moment through kernel density estimation, and obtaining the probability density functions through a formula
Figure DEST_PATH_IMAGE049A
Calculating a reference sequence
Figure DEST_PATH_IMAGE050A
Probability density function corresponding to reference sequence and characteristic sequence at t-th moment
Figure DEST_PATH_IMAGE051
Wasserstein distance between corresponding probability density functions
Figure DEST_PATH_IMAGE053
；
By the formula
Figure DEST_PATH_IMAGE055
Weighting the Wasserstein distances calculated by all the characteristic sequences, namely, the weighted Wasserstein distances are used as the health indexes for comprehensively representing the electromagnetic valve at the moment
Figure DEST_PATH_IMAGE057
Wherein
Figure DEST_PATH_IMAGE059
Is the firstjA characteristic sequence oftThe weight of the time of day.
In this invention, the Wasserstein distance is a measure of the distance between two probability distributions, and is defined as:
Figure DEST_PATH_IMAGE061
in the formula: wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE063
and
Figure DEST_PATH_IMAGE065
is the first of two probability density functionskA value;
Figure DEST_PATH_IMAGE067
is a distance measure;r is the order of the figure. The distance between Wasserstein and the two probability density functions is in the gaprIntegral at order index.
The invention has the following advantages:
the monitoring method provided by the invention obtains a weighted value based on the weighted Wasserstein distance, namely the weighted value is used as a health index for comprehensively representing the electromagnetic valve at the moment
Figure DEST_PATH_IMAGE069A
The monitoring method provided by the invention is less influenced by the number of the sensors, the range of the health evaluation value is not influenced by data difference in the system, and the accurate health condition of the electromagnetic valve is provided for the working personnel so as to provide better electromagnetic valve health conditionThe maintenance schedule of (1).
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention.
Detailed Description
The present invention will be described in detail with reference to the accompanying drawing 1, and the technical solutions in the embodiments of the present invention will be clearly and completely described.
A health monitoring method of an electromagnetic valve comprises the following steps:
step one, obtaining a characteristic sequence, extracting characteristics of time domain, frequency domain, energy, entropy and the like of sensor data of the electromagnetic valve and using the characteristics as the characteristic sequence to obtain a characteristic sequence set
Figure DEST_PATH_IMAGE002AA
WhereinNIs composed of
Figure DEST_PATH_IMAGE004AAA
Number of medium signature sequences.
Step two, standardizing the characteristic sequence data,
because the data of each sensor in the electromagnetic valve has certain difference, the invention carries out standardization processing on the characteristic sequence through a standardization formula to eliminate the data difference, wherein the standardization formula is as follows:
Figure DEST_PATH_IMAGE070A
wherein
Figure DEST_PATH_IMAGE071A
And
Figure DEST_PATH_IMAGE010AAA
respectively before and after standardizationiThe characteristic sequences are arranged in a sequence of the characteristic,
Figure DEST_PATH_IMAGE072A
and
Figure DEST_PATH_IMAGE073AA
are respectively the firstiMean and standard deviation of individual signature sequences.
Step three, screening the characteristic sequences
In order to reduce the amount of calculation, the sensor parameter signature sequence with little or no degradation tendency is removed. The common method is to select useful characteristic sequence manually by directly observing the variation trend of the characteristic sequence, and the method has certain subjectivity. In this respect, the present invention uses a monotonicity formula to calculateiGlobal monotonicity of individual feature sequences
Figure DEST_PATH_IMAGE074A
The formula for monotonicity is as follows:
Figure DEST_PATH_IMAGE075A
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE076A
，
Figure DEST_PATH_IMAGE077A
and
Figure DEST_PATH_IMAGE024AAA
are respectively the firstiA first of the characteristic sequencesjNumerical value and
Figure DEST_PATH_IMAGE078A
the number of the individual values is,
Figure DEST_PATH_IMAGE079A
is the length of the sample.
Figure DEST_PATH_IMAGE080A
The larger the monotonicity, the stronger the monotonicity, and conversely, the weaker the monotonicity. If it is firstiThe sequence of individual features is completely monotonic,
Figure DEST_PATH_IMAGE081A
is 1; if it is firstiThe sequence of individual features is non-monotonic,
Figure DEST_PATH_IMAGE082A
is 0.
And step four, dividing the characteristic sequences into two types through K-means according to the monotonicity of all the characteristic sequences, and selecting the characteristic sequence with larger monotonicity. The sequence of the screened features is assumed to be:
Figure DEST_PATH_IMAGE083AA
wherein
Figure DEST_PATH_IMAGE084AA
，RIs composed of
Figure DEST_PATH_IMAGE085
The number of the characteristic sequences in the medium,
Figure DEST_PATH_IMAGE086A
the characteristic sequence with larger monotonicity is obtained. At this time, the characteristic sequences obtained after screening all have a certain tendency of degeneration.
Step five, calculating health monitoring indexes
In the invention, the Wasserstein distance is used for calculating the health index, and the Wasserstein distance is used in the characteristic sequence
Figure DEST_PATH_IMAGE087
Respectively selecting the length of the sequence from each characteristic sequence from the initial time
Figure DEST_PATH_IMAGE088A
As a reference sequence, is noted
Figure DEST_PATH_IMAGE089
；
In order to calculate the Wasserstein distance at time t relative to the initial time, the length of the Wasserstein distance at time t needs to be selected
Figure DEST_PATH_IMAGE088AA
And a reference sequence
Figure DEST_PATH_IMAGE090A
The Wasserstein distance is calculated. Wherein the characteristic sequence at the t-th time is expressed as
Figure DEST_PATH_IMAGE092A
；
Respectively obtaining probability density functions of the reference sequence and the characteristic sequence at the t-th moment through kernel density estimation, and obtaining the probability density functions through a formula
Figure DEST_PATH_IMAGE094A
Calculating a reference sequence
Figure DEST_PATH_IMAGE096A
Probability density function corresponding to reference sequence and characteristic sequence at t-th moment
Figure DEST_PATH_IMAGE098A
The Wasserstein distance between the probability density functions corresponding to the characteristic sequences at the t-th time
Figure DEST_PATH_IMAGE100A
。
Because each characteristic sequence can represent the health degree of the system at a certain moment, the Wasserstein distance calculated by all the characteristic sequences is formulated
Figure DEST_PATH_IMAGE102A
Weighting to obtain a weighted value as a health index of the solenoid valve
Figure DEST_PATH_IMAGE104
(ii) a Wherein, the first and the second end of the pipe are connected with each other,
Figure DEST_PATH_IMAGE106
is the firstjA characteristic sequence oftThe weight of the time of day.
Step six, according to the health indexes
Figure DEST_PATH_IMAGE108
The method for judging the health condition of the electromagnetic valve comprises the following steps:
the idea of confidence interval in statistics is adopted to design self-adaptive threshold value and health index
Figure DEST_PATH_IMAGE109A
The mean and variance of (a) are:
Figure DEST_PATH_IMAGE111A
and
Figure DEST_PATH_IMAGE113
(ii) a According to health index
Figure DEST_PATH_IMAGE114A
The distribution condition of (2) needs to set an upper limit value and a lower limit value; meanwhile, the health index can be under the condition of no fault due to noise disturbance in the actual operation process
Figure DEST_PATH_IMAGE116A
Is not 0, so a constant is introduced
Figure DEST_PATH_IMAGE118A
To improve the robustness of the threshold; therefore, according to the health index
Figure DEST_PATH_IMAGE119
The mean and variance of (c) determine the following thresholds:
Figure DEST_PATH_IMAGE121
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE123
is the upper limit value of the number of bits,
Figure DEST_PATH_IMAGE125
is the lower limit value;
Figure DEST_PATH_IMAGE127
is a constant.
The logic for judging whether the electromagnetic valve is healthy is as follows:
Figure DEST_PATH_IMAGE129
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE131
is a health indicator obtained from the latest sequence.
The method monitors the health of the electromagnetic valve by utilizing the sensor data of the electromagnetic valve and based on the weighted Wasserstein distance, is less influenced by the number of the sensors, and the range of the health evaluation value is not influenced by the data difference in the system.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (2)

1. A health monitoring method of an electromagnetic valve is characterized by comprising the following steps:
s1, acquiring a characteristic sequence, and directly taking a data sequence of sensor measurement data directly representing the electromagnetic valve as the characteristic sequence to obtain a characteristic sequence set;
s2, standardizing the characteristic sequences and eliminating the difference of the characteristic sequences;
s3, screening the characteristic sequences, calculating the integral monotonicity of the characteristic sequences by adopting a monotonicity formula, and removing the characteristic sequences which are not degraded;
s4, dividing the characteristic sequences into two types through K-means according to the monotonicity of all the characteristic sequences, and selecting the characteristic sequence with larger monotonicity;
s5, calculating a health monitoring index, and using the Wasserstein distance as the health index;
s6, judging the health condition of the electromagnetic valve according to the health index;
the specific method for calculating the health monitoring index comprises the following steps:
in the feature sequences, a sequence with a specified length from an initial time is selected from each feature sequence as a reference sequence, probability density functions of the reference sequence and the feature sequences at corresponding times are obtained through kernel density estimation, Wassertein distances between the probability density functions corresponding to the reference sequences in the reference sequence and the probability density functions corresponding to the feature sequences at corresponding times are calculated, and the Wassertein distances are used for health indexes.
2. The health monitoring method of the electromagnetic valve according to claim 1, wherein the specific method for judging the health condition of the electromagnetic valve according to the health index is as follows: designing an adaptive threshold by adopting the idea of confidence interval in statistics; setting an upper limit value and a lower limit value by combining the mean value and the variance of the health indexes according to the distribution condition of the health indexes;
when the health index obtained by the latest sequence is greater than the upper limit value or less than the lower limit value, the fault is detected;
and when the health index obtained by the latest sequence is greater than or equal to the lower limit value and less than or equal to the upper limit value, determining that the health index is not in fault.
CN202210500346.2A 2022-05-10 2022-05-10 Health monitoring method of electromagnetic valve Active CN114611633B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210500346.2A CN114611633B (en) 2022-05-10 2022-05-10 Health monitoring method of electromagnetic valve

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210500346.2A CN114611633B (en) 2022-05-10 2022-05-10 Health monitoring method of electromagnetic valve

Publications (2)

Publication Number Publication Date
CN114611633A CN114611633A (en) 2022-06-10
CN114611633B true CN114611633B (en) 2022-07-29

Family

ID=81869064

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210500346.2A Active CN114611633B (en) 2022-05-10 2022-05-10 Health monitoring method of electromagnetic valve

Country Status (1)

Country Link
CN (1) CN114611633B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115310561B (en) * 2022-09-29 2022-12-20 中国空气动力研究与发展中心设备设计与测试技术研究所 Electromagnetic valve fault monitoring method based on integrated instant learning

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106446540A (en) * 2016-09-20 2017-02-22 华北电力大学（保定） Real-time evaluation method for health state of wind turbine unit
CN109000924A (en) * 2018-10-24 2018-12-14 哈工大机器人（山东）智能装备研究院 A kind of ball screw assembly, state monitoring method based on K mean value
CN112434636A (en) * 2020-12-03 2021-03-02 西安交通大学 Machine tool part health state monitoring method and system
CN113705738A (en) * 2021-08-31 2021-11-26 长安大学 Engineering equipment bearing degradation assessment method

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11310141B2 (en) * 2019-12-11 2022-04-19 Cisco Technology, Inc. Anomaly detection of model performance in an MLOps platform
CN111090050A (en) * 2020-01-21 2020-05-01 合肥工业大学 Lithium battery fault diagnosis method based on support vector machine and K mean value
CN113569903B (en) * 2021-06-09 2024-04-09 西安电子科技大学 Method, system, equipment, medium and terminal for predicting cutter abrasion of numerical control machine tool

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106446540A (en) * 2016-09-20 2017-02-22 华北电力大学（保定） Real-time evaluation method for health state of wind turbine unit
CN109000924A (en) * 2018-10-24 2018-12-14 哈工大机器人（山东）智能装备研究院 A kind of ball screw assembly, state monitoring method based on K mean value
CN112434636A (en) * 2020-12-03 2021-03-02 西安交通大学 Machine tool part health state monitoring method and system
CN113705738A (en) * 2021-08-31 2021-11-26 长安大学 Engineering equipment bearing degradation assessment method

Also Published As

Publication number Publication date
CN114611633A (en) 2022-06-10

Similar Documents

Publication Publication Date Title
CN111508216B (en) Intelligent early warning method for dam safety monitoring data
CN109186813B (en) Temperature sensor self-checking device and method
US8892478B1 (en) Adaptive model training system and method
US7496798B2 (en) Data-centric monitoring method
CN112348237B (en) Abnormal trend detection method for dynamic drilling data
Khaw et al. Optimal design of the synthetic control chart for monitoring the multivariate coefficient of variation
KR102116711B1 (en) Early warning system and method of power plants using multiple prediction model
CN102901529B (en) Calibration steps
CN109325692A (en) The data real-time analysis method and device of pipe network
CN110375983B (en) Valve fault real-time diagnosis system and method based on time series analysis
CN114611633B (en) Health monitoring method of electromagnetic valve
CN111324105B (en) Small sensor fault detection and separation method based on statistic Mahalanobis distance
CN116881745B (en) Pressure transmitter abnormality monitoring method based on big data
US7552035B2 (en) Method to use a receiver operator characteristics curve for model comparison in machine condition monitoring
CN107851294A (en) The preventive maintenance device and method based on state of large-scale runtime
CN117314020A (en) Wetland carbon sink data monitoring system of plankton
CN116627116A (en) Process industry fault positioning method and system and electronic equipment
JP4782734B2 (en) Plant measuring instrument calibration support apparatus and plant measuring instrument calibration support method
CN106649204B (en) Method for identifying abnormity of dissolved gas in transformer oil
CN111966966B (en) Method and system for analyzing feasible domain of sensor measurement error model parameters
WO2020183781A1 (en) Abnormality diagnosis device
JP5451927B1 (en) Abnormality cause identification system, abnormality cause identification method, and abnormality cause identification program
CN114001887A (en) Bridge damage assessment method based on deflection monitoring
JP2018159111A (en) Blast furnace status condition determining device and method of operating blast furnace
Hao et al. A KPI-related multiplicative fault diagnosis scheme for industrial processes

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant