CN110008350A - A kind of pump Ankang knowledge base lookup method based on Bayesian inference - Google Patents
A kind of pump Ankang knowledge base lookup method based on Bayesian inference Download PDFInfo
- Publication number
- CN110008350A CN110008350A CN201910169761.2A CN201910169761A CN110008350A CN 110008350 A CN110008350 A CN 110008350A CN 201910169761 A CN201910169761 A CN 201910169761A CN 110008350 A CN110008350 A CN 110008350A
- Authority
- CN
- China
- Prior art keywords
- pump
- bayesian network
- network
- data
- ankang
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Databases & Information Systems (AREA)
- Animal Behavior & Ethology (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Control Of Non-Positive-Displacement Pumps (AREA)
Abstract
The present invention relates to pump monitoring and fault diagnosis technologies, it is desirable to provide a kind of pump Ankang knowledge base lookup method based on Bayesian inference.It include: using pump critical component, specific fault message and the vibration signal characteristics in pump Ankang knowledge base as priori knowledge, establish the study of Bayesian network search model, new database is established using the score that score function calculates Bayesian network, using obtained Bayesian network, the probability of the nodes break down is inferred according to the real-time status of the vibration signal characteristics obtained in pump operational process.The present invention can effectively adjust penalty term when can be for the training of certain data, it prevents from being excessively fitted, the topological structure of Bayesian network can be effectively determined, pass through priori knowledge, the conditional probability density at each node of Bayesian network model is obtained, to obtain safe and comfortable state and the failure mode in pump operational process；Safe and comfortable degree and fault diagnosis in pump operational process are analyzed and diagnosed, realizes pump safe and highly efficient operation.
Description
Technical field
The present invention relates to pump monitoring and fault diagnosis technology, in particular to a kind of pump based on Bayesian inference is safe and comfortable
Knowledge base lookup method, belongs to field of intelligent control.
Background technique
The pumps equipment such as blower, water pump, compressor, steam turbine is the modern enterprise such as petroleum, chemical industry, metallurgy, steel and electric power
The crucial production equipment of industry.With intelligentized development, such equipment is constantly to high speed, heavy, efficient and intelligent development, together
When to the condition monitoring and fault diagnosis system of equipment, more stringent requirements are proposed.Pump operating status quality directly affects enterprise
The efficiency of industry production, even will cause safety problem when catastrophe failure occurs.So the safe and comfortable monitoring to pump equipment seems outstanding
It is important.
Existing monitoring technology establishes operational parameter data library primarily directed to pump, by monitoring device operation data and incites somebody to action
It is mutually matched inquiry between monitoring data and operation data library, confirms the operating status of monitoring device.But such monitoring skill
Art needs a large amount of manpower to go to carry out the data analysis of profession and inquiry, and is not previously predicted function, and accuracy is not high.
Summary of the invention
The technical problem to be solved by the present invention is to overcome deficiency in the prior art, provide a kind of based on Bayesian inference
Pump Ankang knowledge base lookup method.
In order to solve the technical problem, solution of the invention is:
A kind of pump Ankang knowledge base lookup method based on Bayesian inference is provided, is included the following steps:
(1) using pump critical component, specific fault message and the vibration signal characteristics in pump Ankang knowledge base as first
Knowledge is tested, Bayesian network search model is established；
(2) vibration signal characteristics traced into pump operational process in setting time length and specific fault message are made
For training data sample D, the study of Bayesian network is carried out；
(3) collected vibration signal characteristics are as network inputs using in pump actual moving process, with the specific of pump
Safe and comfortable condition information is exported as network；
(4) the searching times m of Bayesian network is set；
(5) score of Bayesian network is calculated using score function；
(6) new database is established, for saving the score of search Bayesian network every time；
(7) Bayesian network for finding out highest scoring realizes the structure of Bayesian network；
(8) parameter learning of network is realized using EM algorithm；
(9) using obtained Bayesian network, according to the real-time shape of the vibration signal characteristics obtained in pump operational process
State, and then infer the probability of the nodes break down.
In the present invention, vibration signal characteristics described in step (1) refer to the kurtosis index of vibration data, peak index,
Margin index, waveform index and a frequency multiplication, two frequencys multiplication, frequency tripling and high frequency multiplication relevant time domain and frequency-domain index.
In the present invention, pump key position described in step (1) refers to 5 kinds of pump components for easily sending out failure: rotor, sliding
Bearing, gear-box, lower margin and impeller；The specific fault message and vibration signal characteristics are used to construct pump Ankang knowledge base,
Its data source includes: the fail result of test simulation simulation, the daily fortune of Research Literature, pump of pump malfunction monitoring and analysis
Inspection failure logging when row.
In the present invention, pump described in step (3) specifically safe and comfortable situation include at least be in a good state of health and pump therefore
Hinder two kinds of situations, if pump failure, exports specific defect content.
In the present invention, score function described in step (5) refers to that Bayesian network conditional probability counts score function
BCPS, specific as shown in formula (1):
N, i, j, k ∈ N
Wherein, S is network structure, and D is data sample, and n, i, j, k are different natural numbers, and q is failure father node, and r is
Network node number, m is searching times, and α is the optimum structure parameter sought；
The penalty term of BCPS function isWhereinFor the complexity of network structure, λ is
Penalty coefficient,For the likelihood score of Bayesian network；The score value of BCPS function is bigger, the pass between explanatory variable
Connection intensity is bigger, and the fitness of data and network structure is better.
In the present invention, parameter learning described in step (8) refers to the parameter learning that Bayesian network is realized using EM algorithm,
For realizing the calculating of maximum likelihood value；It specifically includes:
For Bayesian network sample data D, its conditional probability P (x is calculatedi, πi|Dl, θ (t))；Given D, likelihood letter
Number are as follows:
I, j, k ∈ N
Wherein, 1 is likelihood function value, and i, j, k is different natural numbers, xi∈{xi 1xi 2..., xi k..., xi q1}；πiFor
The set of father node, putting in order is 1,2 ..., qi；θ is the optimized parameter sought, θjkTo work as xi=xi kAnd πi=j is
Optimized parameter；X is worked as in expressioni=k and πiValue when=j in data set, obtains following formula:
Starting to set an initial estimation θ(0), then constantly correct；From current estimation θ(t), arrive next estimation
θ(t+1), t is time value, carry out expectation calculating and maximization:
It is expected that calculating is calculated when determining D, the likelihood function of current θ it is expected:
To all θ, should meet l (θ | θ(t+1)≥l(θ|θ(t))), according to formula (2)
Maximizing calculating is to select next estimation θ by maximizing current expectation likelihood function value(t+1):
Compared with prior art, the present invention has following technical effect that
1, the present invention using pump Ankang knowledge base be used as the priori knowledge of Bayesian network, and utilization EM algorithm to network into
Row parameter learning combines improved BCPS score function, can effectively adjust penalty term when can train for certain data, prevent
Only excessively fitting, can effectively determine the topological structure of Bayesian network, by priori knowledge, it is each to obtain Bayesian network model
Conditional probability density at node, to obtain safe and comfortable state and the failure mode in pump operational process.
2, the present invention combines Bayesian network with pump Ankang knowledge base, can analyze and diagnose pump operation well
Safe and comfortable degree and fault diagnosis in the process can be avoided and reduce weight so as to find the early stage sign of elevator faults in time
The generation of mass production accident, effectively extend pump service life, reduce maintenance cost, realize pump Life cycle safety with
Efficiently.
Detailed description of the invention
Fig. 1 is the embodiment of the present invention flow chart.
Fig. 2 is the key position and fault signature schematic diagram of pump Ankang condition diagnosing.
Fig. 3 is the schematic illustration for realizing pump Ankang maintenance system of the invention.
Specific embodiment
In order to which the purpose of the present invention and technical solution and advantage is more clearly understood, make in conjunction with attached drawing further detailed
Explanation.It should be appreciated that specific embodiment described herein only possesses the explanation present invention, and does not have to and limit the present invention.
Firstly the need of explanation, realization of the invention can be related to signal detection technique, and system, computer technology is raw in industry
The application in production field.During concrete application of the invention, the application of software function module may relate to.Applicant recognizes
It is existing known combining such as after reading over application documents, accurate understanding realization principle and goal of the invention of the invention
In the case where technology, the software programming technical ability that those skilled in the art can grasp completely with it realizes the present invention.Fan Benfa
Category this scope that bright application documents refer to, applicant will not enumerate.
Pump Ankang knowledge base is a large scale knowledge base, can be made of multiple components.Such as: (1) each category is transported
Turn the product component of equipment different model vibration severity data in different usage situations and spectrum signature is summarized, returned
Class can be configured to running device component vibration data package；(2) with whole as defined in standard (such as ISO2372 vibration standard)
Body vibration velocity earthquake intensity rate range constructs vibration severity data package；(3) with standard (such as JB/T 5294-91 temperature standard)
Defined running temperature rate range constructs temperature data component；(4) failure logging covered with history big data, and integration
Magnanimity running device fault diagnosis example forms knowledge library component；The component can cover the failure that most of running device is likely to occur
And its detection method, be divided into mechanical failure, electric information failure, installation it is lack of standardization, maintenance it is not in place, product element fault, make
With several major class such as improper failure.The teaming method of pump Ankang knowledge base can be realized with customary technical means in the art, of the invention
It repeats no more.Using the pump Ankang knowledge base after arrangement as database module lookup for use in the present invention.
As shown in Fig. 1, the pump Ankang knowledge base lookup method provided by the invention based on Bayesian inference, including with
Lower step:
S1: using in pump Ankang knowledge base pump failure critical component and specific failure and vibration signal characteristics as
Priori knowledge establishes Bayesian network search model；Wherein, prior probability is mainly derived from the pump Ankang knowledge established in advance
Library, data source may include test simulation simulation fail result, pump malfunction monitoring and analysis research papers,
And the failure logging etc. that worker's inspection in pump operational process goes out.
S2: the vibration signal traced into setting time length (such as one longer-term time) interior pump operational process
Feature and its specific fault message carry out the study of Bayesian network as training data sample D；Pump peace is illustrated in Fig. 2
The key position and fault signature schematic diagram of health condition diagnosing, 5 critical components including pump, rotor, sliding bearing, gear
Case, lower margin and impeller, and the corresponding failure feature for Bayesian network training；
S3: using vibration signal characteristics collected in pump actual moving process as network inputs, with the specific of pump
Safe and comfortable condition information is exported as network；Specifically safe and comfortable situation is included at least and is in a good state of health and two kinds of feelings of pump failure pump
Condition (can also divide multiple grades), if pump failure, export specific defect content.
S4: the number m of the search of setting network；
S5: the score of Bayesian network is calculated using score function；
The present invention uses score function BCPS, specific as shown in formula (1):
N, i, j, k ∈ N
Wherein, S is network structure, and D is data sample, and n, i, j, k are different natural numbers, and q is failure father node, and r is network
Node serial number, m are searching times, and α is the optimum structure parameter sought.The penalty term of BCPS function isIts
InFor the complexity of network structure, λ is penalty coefficient,For the likelihood of Bayesian network
Degree；The score value of BCPS function is bigger, and the strength of association between explanatory variable is bigger, and the fitness of data and network structure is better.
S6: establishing new database, for saving the score of search Bayesian network every time；
S7: finding out the Bayesian network of highest scoring, realizes the structure of Bayesian network；
S8: the parameter learning of network is realized using EM algorithm；
The present invention realizes the parameter learning of Bayesian network using EM algorithm, for realizing the calculating of maximum likelihood value；Tool
Body includes:
For Bayesian network sample data D, its conditional probability P (x is calculatedi, πi|Dl, θ (t))；Given D, likelihood letter
Number are as follows:
I, j, k ∈ N
Wherein, 1 is likelihood function value, xi∈{xi 1, xi 2..., xi k..., xi qi}；πiFor the set of father node, arrangement
Sequence is 1,2 ..., qi；θ is the optimized parameter sought；X is worked as in expressioni=k and πiValue when=j in data set,
Obtain following formula:
Starting to set an initial estimation θ(0), then constantly correct；From current estimation θ(t), arrive next estimation
θ(t+1), t is time value, carry out expectation calculating and maximization:
It is expected that calculating is calculated when determining D, the likelihood function of current θ it is expected:
To all θ, should meet 1 (θ | θ(t+1)≥l(θ|θ(t))), according to formula (2)
Maximizing calculating is to select next estimation θ by maximizing current expectation likelihood function value(t+1):
S8: using obtained Bayesian network, according to the real-time shape of the vibration signal characteristics obtained in pump operational process
State, and then infer the probability of the nodes break down；Finally by priori knowledge, obtain at each node of Bayesian network model
Conditional probability density, to obtain safe and comfortable state and the failure mode in pump operational process.
As an example, being described below to using the pump Ankang maintenance system of lookup method of the present invention.
Pump Ankang maintenance system is as shown in figure 3, include for showing the host computer of monitored results, being mounted on pump
Vibration temperature sensor and the wireless base station NB-IOT, data storage server and data analytics server；The wireless base station Rola
Mode is connect by wireless communication between data storage server, data storage server, data analytics server and upper
It is connected between machine by signal wire；
Vibration temperature sensor includes the shell for encapsulation, built-in three-axial vibration sensing module, temperature in shell
Monitoring modular, filter module, data processing module, wireless communication module and battery module, three-axial vibration sensing module pass through
Signal wire is sequentially connected filter module and data processing module, and temperature monitoring module and wireless communication module pass through signal wire respectively
Data processing module is connected, battery module is used to power for each component.
Vibrating sensor one shares 2, is separately mounted to the sliding axle socket end and base end of pump.
The shell (such as 304 stainless steel casings) of integral packaging can be used in vibrating sensor, installs simple, easy to use, knot
Structure is compact, compact.It is based on wireless digital signal transmission mode and eliminates long cable transmission bring noise jamming, used
Filter meets 10816 vibration level standard of ISO, has anti-aliasing function.In vibration temperature sensor of the present invention, three axis
To vibrating sensing module for realizing the real-time monitoring of axial (X-axis, the Y-axis, Z axis) acceleration signal of running device three, can be selected
The tri- shaft vibration acceleration chip of MEAS 7131A-0050 of U.S. TE Connectivity production.The temperature of temperature sensor is believed
Number in conjunction with analysis of vibration signal, the accuracy and reliability of diagnostic analysis has further sufficiently been ensured.
The wireless base station NB-IOT, for realizing the data transmission between vibrating sensor and data storage server；Pass through
NB-IOT wireless communication mode carries out acquisition, storage and the analysis of the vibrational state and temperature data of pump, is a kind of low-power consumption
The wide area network communication modes of long range, possess superelevation data receiver sensitivity and superpower signal-to-noise ratio, be used for vibrating sensor
With the communication for being video camera, the safe and reliable transmission of data has been ensured, while also increasing the use of vibrating sensor
Service life.
Data storage server is mounted on Pump control center, receives the data acquired by vibrating sensor, is stored,
Data basis is provided for the accident analysis and diagnosis of pump, historical data can also be recalled at any time for checking.The system further includes
It is installed on the remote login service end software module of data storage server, it is whole mobile communication equipment can be connected by internet
End.
Data analytics server is also mountable at Pump control center, wherein built-in analyzing and diagnosing algorithm software module and machine
Pump safe and comfortable base module.Data perception is carried out by the mass data acquired to vibrating sensor, then to component each in pump
Vibration frequency and pump vibration severity carry out analysis and assessment, with pump Ankang knowledge base in all kinds of most common failures divided
Analysis and comparison realize the monitoring of pump all-around intelligent and management and right to carry out health analysis and security diagnostics to pump
The visualization of pump health status exports.
Pump Ankang knowledge base is a large scale knowledge base, can be made of multiple components.Such as: (1) by each category machine
It pumps the vibration frequency data of the product component of different model in different usage situations to be summarized, sorted out, can be configured to
Pump component vibration frequency data component；(2) the vibration severity rate range of pump is configured to vibration severity data package；
(3) record of the failure occurred according to corresponding to history big data in data storage server, and a large amount of pumps events of integration
Hinder knowledge library component composed by example；The component can cover the failure and its detection method that most of pump is likely to occur,
It is several to be divided into mechanical failure, electric information failure, installation lack of standardization, not in place, the product element fault of maintenance, improper use failure etc.
Major class.Data for constructing pump Ankang knowledge base include initial vibration frequency data and fault signature, are mainly derived from: examination
Test the inspection failure note when fail result of analogue simulation, the Research Literature of pump malfunction monitoring and analysis, pump day-to-day operation
Record, etc..
Host computer may be mounted in building master control room.Built-in visualization output software module in host computer, for connecing
The calculated result of data analytics server transmission, and the real-time display after analysis and processing are received, is issued according to preset failure condition
Real-time early warning and alarm.
Pump Ankang maintenance monitoring method based on the system, comprising the following steps:
(1) vibration temperature sensor obtains vibration frequency and acceleration from pump using its three-axial vibration sensing module
Degree detection signal；After filter eliminates clutter, signal will test by data processing module and be converted to the data format that can be transmitted,
And it is sent to data storage server through wireless internet of things module and the wireless base station NB-IOT, collected data by it and is deposited
Storage；
(2) built-in analyzing and diagnosing algorithm software module and pump Ankang base module in data analytics server；Analysis
Diagnosis algorithm software module is extracted the data of data storage server storage and is calculated, and the vibration of each component in pump is obtained
Frequency and body vibration speed data；After being analyzed and being compared with pump Ankang knowledge base, secure good health analysis and safety
Diagnostic result；
(3) built-in visualization output software module in the host computer, in the calculating for receiving data analytics server transmission
As a result its result is analyzed and is handled afterwards simultaneously real-time display, while real-time early warning and report are issued according to preset failure condition
It is alert.
It analysis in the step (2) and compares including following the description:
(2.1) for the analysis of each component vibration frequency in pump with compare
Each component of pump is formed because of the otherness of its material and shape, there is the fixation vibration frequency model being mutually distinguishable
It encloses；Display can be different spectrum signature in vibration frequency testing result, like the vocal print difference between different human individuals.
By before assembling to each component of pump (such as the components such as pedestal, rotor, sliding bearing, gear-box, impeller) vibration frequency
Measurement, can obtain corresponding data；By the vibration frequency of the component of each category pump different model product in different usage situations
Rate data are summarized, are sorted out, and a significant components of pump Ankang knowledge base can be configured to；It, can in the knowledge base
With qualitative for different fault levels according to different use states.
After analyzing and diagnosing algorithm software module extracts data, the synthesis oscillation frequency data of pump are obtained；Then basis is worked as
Synthesis oscillation frequency data are parsed, inquired and are matched, mentioned by the data of preceding pump product type and pump Ankang knowledge base
Take out the vibration frequency data of each component in pump；The vibration frequency data of each component and pump Ankang knowledge base are compared again
To and analysis, obtain the corresponding fault level of each component；Since data are extracted, parsing, inquiry, compare, matching, analysis etc. are specific
Realization process is that those skilled in the art skillfully grasp, and has many mature technology means can use, the present invention repeats no more.
(2.2) for pump body vibration speed analysis with compare
The body vibration situation of pump can reflect a possibility that its building block breaks down from side.It therefore, can be with
Use pump body vibration speed earthquake intensity as pump breakdown judge foundation, the data of pump body vibration speed earthquake intensity can be used as group
Part module is summarized in together in pump Ankang knowledge base.
After analyzing and diagnosing algorithm software module extracts data, to vibrating sensor vibration acceleration value measured directly, lead to
It crosses wavelet transformation and carries out data prediction, vibration velocity virtual value is then obtained by integral calculation；By be summarized in pump
Pump vibration severity rate range in safe and comfortable knowledge base is compared, and obtains the fault level of pump body vibration speed.
For example, carrying out integral operation to vibration acceleration signal a (t) obtains vibration speed value v (t), such as formula (1):
Wherein, t is the time of data acquisition；
Vibration velocity virtual value vrms is acquired further according to vibration speed value v (t), such as formula (2):
Wherein, T is the vibration period.
Since data are extracted, parsing, inquiry, compare, matching, analysis and the specific implementation such as wavelet transformation, integral calculation
Process is that those skilled in the art skillfully grasp, and has many mature technology means can use, the present invention repeats no more.
For example, difference of the present invention according to pump vibration velocity virtual value, is divided into A, B, C, D tetra- for pump vibration severity
Different rate ranges, the area A indicates that pump operation conditions is good, without any failure problems；The area B indicates there is micro- vibration in pump operation
It is dynamic, it is the permissible value in pump operational process, does not need to repair, but needs to continue to observe；The area C indicates have in pump operation
Apparent vibration, it is proposed that the analysis that inquiry failure cause is carried out by pump Ankang knowledge base carries out corresponding maintenance；The area D table
Show in pump operation there is violent vibration, the operation of pump must be stopped immediately, carries out emergency maintenance.
Based on above-mentioned pump Ankang maintenance system, the present invention can also be realized further more using the technology of mobile Internet
It is multi-functional, such as: (1) maintenance staff use mobile communication equipment terminal, connect remote login service end software by internet
Module checks that pump history maintenance records according to permission, fills in current maintenance record, and be stored in server；Alternatively, (2) pump
Operator uses mobile communication equipment terminal, connects remote login service end software module by internet, is looked into according to permission
It sees that pump history maintenance records, uploads failure and declare.
Pump Ankang knowledge base lookup method of the present invention based on Bayesian inference, always through in step (2)
Analysis and comparison process in.For example, before being analyzed and being compared, parameter learning and scoring by Bayesian network,
The Bayesian network of highest scoring is confirmed as analysis and compares Bayesian network used；Then in analysis and comparison process
In, using obtained Bayesian network, the real-time status of node is judged according to the vibration signal obtained in pump operational process, into
And infer the probability of the nodes break down, analysis result is pushed to host computer.
Claims (6)
1. a kind of pump Ankang knowledge base lookup method based on Bayesian inference, which comprises the steps of:
(1) known using pump critical component, specific fault message and the vibration signal characteristics in pump Ankang knowledge base as priori
Know, establishes Bayesian network search model；
(2) using the vibration signal characteristics and specific fault message traced into pump operational process in setting time length as instruction
Practice data sample D, carries out the study of Bayesian network；
(3) collected vibration signal characteristics are as network inputs using in pump actual moving process, with the specific Ankang of pump
Condition information is exported as network；
(4) the searching times m of Bayesian network is set；
(5) score of Bayesian network is calculated using score function；
(6) new database is established, for saving the score of search Bayesian network every time；
(7) Bayesian network for finding out highest scoring realizes the structure of Bayesian network；
(8) parameter learning of network is realized using EM algorithm；
(9) using obtained Bayesian network, according to the real-time status of the vibration signal characteristics obtained in pump operational process, into
And infer the probability of the nodes break down.
2. the method according to claim 1, wherein vibration signal characteristics described in step (1) refer to vibration
Kurtosis index, peak index, margin index, waveform index and the frequency multiplication of data, two frequencys multiplication, frequency tripling and high frequency multiplication
Relevant time domain and frequency-domain index.
3. the method according to claim 1, wherein pump key position described in step (1) refers to 5 kinds of easily hairs
The pump component of failure: rotor, sliding bearing, gear-box, lower margin and impeller；The specific fault message and vibration signal characteristics
For constructing pump Ankang knowledge base, data source includes: the fail result of test simulation simulation, pump malfunction monitoring and divides
Inspection failure logging when the Research Literature of analysis, pump day-to-day operation.
4. the method according to claim 1, wherein specifically safe and comfortable situation is at least for pump described in step (3)
Specific defect content is exported if pump failure with two kinds of situations of pump failure including being in a good state of health.
5. the method according to claim 1, wherein score function described in step (5) refers to Bayesian network
Network conditional probability counts score function BCPS, specific as shown in formula (1):
Wherein, S is network structure, and D is data sample, and n, i, j, k are different natural numbers, and q is failure father node, and r is network
Node serial number, m are searching times, and α is the optimum structure parameter sought；
The penalty term of BCPS function isWhereinFor the complexity of network structure, λ is punishment
Coefficient,For the likelihood score of Bayesian network；The score value of BCPS function is bigger, and the association between explanatory variable is strong
Degree is bigger, and the fitness of data and network structure is better.
6. the method according to claim 1, wherein parameter learning described in step (8), which refers to, utilizes EM algorithm
The parameter learning of Bayesian network is realized, for realizing the calculating of maximum likelihood value；It specifically includes:
For Bayesian network sample data D, its conditional probability P (x is calculatedi, πi|D1, θ(t))；Given D, likelihood function are as follows:
Wherein, 1 is likelihood function value, and i, j, k is different natural numbers, xi∈{xi 1, xi 2..., xi k..., xi qi}；πiFor father
The set of node, putting in order is 1,2 ..., qi；θ is the optimized parameter sought, θjkTo work as xi=xi kAnd ` πi=j is yes
Optimized parameter；X is worked as in expressioni=k and πiValue when=j in data set, obtains following formula:
Starting to set an initial estimation θ(0), then constantly correct；From current estimation θ(t), arrive next estimation
θ(t+1), t is time value, carry out expectation calculating and maximization:
It is expected that calculating is calculated when determining D, the likelihood function of current θ it is expected:
l(θ|θ(t))=∑l∑xilnp(Dl, xl|θ)p(xl|Dl, θ(t)) (4)
To all θ, should meet l (θ | θ(t+1)≥l(θ|θ(t))), according to formula (2)
Maximizing calculating is to select next estimation θ by maximizing current expectation likelihood function value(t+1):
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910169761.2A CN110008350A (en) | 2019-03-06 | 2019-03-06 | A kind of pump Ankang knowledge base lookup method based on Bayesian inference |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910169761.2A CN110008350A (en) | 2019-03-06 | 2019-03-06 | A kind of pump Ankang knowledge base lookup method based on Bayesian inference |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110008350A true CN110008350A (en) | 2019-07-12 |
Family
ID=67166502
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910169761.2A Pending CN110008350A (en) | 2019-03-06 | 2019-03-06 | A kind of pump Ankang knowledge base lookup method based on Bayesian inference |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110008350A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020134890A1 (en) * | 2018-07-30 | 2020-07-02 | 杭州哲达科技股份有限公司 | Vibration and temperature sensor for monitoring data of safe and healthy operation of operating and rotating device |
| CN111664083A (en) * | 2020-05-29 | 2020-09-15 | 中核武汉核电运行技术股份有限公司 | Nuclear power main pump fault diagnosis method based on Bayesian network |
| CN111708815A (en) * | 2020-05-11 | 2020-09-25 | 中国石油集团工程股份有限公司 | Pump model selection and analysis technology based on big data algorithm |
| CN111897945A (en) * | 2020-07-07 | 2020-11-06 | 广州视源电子科技股份有限公司 | Method, device, equipment and medium for marking and item pushing of preposed knowledge points |
| CN112906775A (en) * | 2021-02-05 | 2021-06-04 | 深圳市芯聚智科技有限公司 | Equipment fault prediction method and system |
| CN114962239A (en) * | 2022-06-01 | 2022-08-30 | 黄河科技集团创新有限公司 | Equipment fault detection method based on intelligent Internet of things |
| CN116739345A (en) * | 2023-06-08 | 2023-09-12 | 南京工业大学 | Real-time evaluation method for possibility of dangerous chemical road transportation accident |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060136403A1 (en) * | 2004-12-22 | 2006-06-22 | Koo Charles C | System and method for digital content searching based on determined intent |
| CN103198217A (en) * | 2013-03-26 | 2013-07-10 | X·Q·李 | Fault detection method and system |
| EP2656259A1 (en) * | 2010-12-21 | 2013-10-30 | Koninklijke Philips N.V. | Patient condition detection and mortality |
| CN104462842A (en) * | 2014-12-22 | 2015-03-25 | 厦门大学 | Excavating diagnosis method of failure data based on bayesian network |
| CN107679566A (en) * | 2017-09-22 | 2018-02-09 | 西安电子科技大学 | A kind of Bayesian network parameters learning method for merging expert's priori |
-
2019
- 2019-03-06 CN CN201910169761.2A patent/CN110008350A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20060136403A1 (en) * | 2004-12-22 | 2006-06-22 | Koo Charles C | System and method for digital content searching based on determined intent |
| EP2656259A1 (en) * | 2010-12-21 | 2013-10-30 | Koninklijke Philips N.V. | Patient condition detection and mortality |
| CN103198217A (en) * | 2013-03-26 | 2013-07-10 | X·Q·李 | Fault detection method and system |
| CN104462842A (en) * | 2014-12-22 | 2015-03-25 | 厦门大学 | Excavating diagnosis method of failure data based on bayesian network |
| CN107679566A (en) * | 2017-09-22 | 2018-02-09 | 西安电子科技大学 | A kind of Bayesian network parameters learning method for merging expert's priori |
Non-Patent Citations (2)
| Title |
|---|
| 戚梦成: ""白车身焊接自动化生产线的故障诊断专家系统的设计与实现"", 《中国优秀硕士学位论文全文数据库库工程科技II级》 * |
| 李淑智等: ""贝叶斯网络结构评分函数研究"", 《中国学术会议》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020134890A1 (en) * | 2018-07-30 | 2020-07-02 | 杭州哲达科技股份有限公司 | Vibration and temperature sensor for monitoring data of safe and healthy operation of operating and rotating device |
| CN111708815A (en) * | 2020-05-11 | 2020-09-25 | 中国石油集团工程股份有限公司 | Pump model selection and analysis technology based on big data algorithm |
| CN111708815B (en) * | 2020-05-11 | 2023-04-18 | 中国石油集团工程股份有限公司 | Pump model selection and analysis method based on big data algorithm |
| CN111664083A (en) * | 2020-05-29 | 2020-09-15 | 中核武汉核电运行技术股份有限公司 | Nuclear power main pump fault diagnosis method based on Bayesian network |
| CN111664083B (en) * | 2020-05-29 | 2024-06-11 | 中核武汉核电运行技术股份有限公司 | Nuclear power main pump fault diagnosis method based on Bayesian network |
| CN111897945A (en) * | 2020-07-07 | 2020-11-06 | 广州视源电子科技股份有限公司 | Method, device, equipment and medium for marking and item pushing of preposed knowledge points |
| CN111897945B (en) * | 2020-07-07 | 2024-07-02 | 广州视源电子科技股份有限公司 | Method, device, equipment and medium for labeling prepositioned knowledge points and pushing topics |
| CN112906775A (en) * | 2021-02-05 | 2021-06-04 | 深圳市芯聚智科技有限公司 | Equipment fault prediction method and system |
| CN112906775B (en) * | 2021-02-05 | 2023-12-01 | 深圳市芯聚智科技有限公司 | Equipment fault prediction method and system |
| CN114962239A (en) * | 2022-06-01 | 2022-08-30 | 黄河科技集团创新有限公司 | Equipment fault detection method based on intelligent Internet of things |
| CN116739345A (en) * | 2023-06-08 | 2023-09-12 | 南京工业大学 | Real-time evaluation method for possibility of dangerous chemical road transportation accident |
| CN116739345B (en) * | 2023-06-08 | 2024-03-22 | 南京工业大学 | Real-time evaluation method for possibility of dangerous chemical road transportation accident |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110008350A (en) | A kind of pump Ankang knowledge base lookup method based on Bayesian inference | |
| CN102053016B (en) | System for monitoring vibration of rotating machinery rolling bearing in wireless mode | |
| CN110766277B (en) | Health assessment and diagnosis system and mobile terminal for nuclear industry field | |
| CN107271170B (en) | A kind of automatic diagnosis method and system of mechanical equipment fault type | |
| CN113562562A (en) | Elevator safety early warning monitoring system and judgment method thereof | |
| CN111428386B (en) | Elevator traction machine rotor fault diagnosis information fusion method based on complex network | |
| CN115159285B (en) | Elevator state monitoring method and elevator state monitoring device | |
| CN112880750A (en) | Transformer multidimensional comprehensive online monitoring intelligent diagnosis system | |
| CN116720112A (en) | Machine pump fault diagnosis method and system based on machine learning | |
| CN110133500A (en) | Motor on-line monitoring and failure omen diagnostic system and method based on multi-layer framework | |
| CN110837045A (en) | Method for diagnosing potential fault of pump system and detection system | |
| KR102545672B1 (en) | Method and apparatus for machine fault diagnosis | |
| CN116540015A (en) | Power distribution network fault early warning method and system based on transient waveform signals | |
| CN113697424B (en) | Belt conveyor monitoring and fault diagnosis system and method based on cloud technology | |
| CN114553756A (en) | Equipment fault detection method based on joint generation countermeasure network and electronic equipment | |
| CN105865789A (en) | Detection system for bearing fault of decomposer gear reducer | |
| CN114019935A (en) | Real-time detection and diagnosis system based on industrial Internet of things equipment | |
| CN117272032B (en) | Air compressor fault diagnosis method and device based on vibration detection | |
| CN117252542B (en) | Digital operation and maintenance method and platform for electric power system based on Internet of things | |
| CN111456915A (en) | Fault diagnosis device and method for internal components of fan engine room | |
| CN118148898A (en) | Intelligent monitoring system and method applied to digital energy air compression station | |
| CN117668646A (en) | Speed reducer running state detection method, system, equipment and medium based on multi-feature fusion | |
| CN110057587A (en) | A kind of nuclear power pump bearing intelligent failure diagnosis method and system | |
| CN117150313A (en) | Power information communication method and system based on abnormal state identification | |
| CN114167282B (en) | Motor fault diagnosis and degradation trend prediction system |
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 |