CN104503434A - Fault diagnosis method based on active fault symptom pushing - Google Patents

Fault diagnosis method based on active fault symptom pushing Download PDF

Info

Publication number
CN104503434A
CN104503434A CN201410720508.9A CN201410720508A CN104503434A CN 104503434 A CN104503434 A CN 104503434A CN 201410720508 A CN201410720508 A CN 201410720508A CN 104503434 A CN104503434 A CN 104503434A
Authority
CN
China
Prior art keywords
fault
parameter
failure symptom
symbol
symptom
Prior art date
Application number
CN201410720508.9A
Other languages
Chinese (zh)
Other versions
CN104503434B (en
Inventor
周磊
宋凯
朱子环
蔡睿
周文怡
张登攀
宋振龙
耿卫国
王祁
Original Assignee
北京航天试验技术研究所
哈尔滨工业大学
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 北京航天试验技术研究所, 哈尔滨工业大学 filed Critical 北京航天试验技术研究所
Priority to CN201410720508.9A priority Critical patent/CN104503434B/en
Publication of CN104503434A publication Critical patent/CN104503434A/en
Application granted granted Critical
Publication of CN104503434B publication Critical patent/CN104503434B/en

Links

Classifications

    • GPHYSICS
    • G05CONTROLLING; REGULATING
    • G05BCONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
    • G05B23/00Testing or monitoring of control systems or parts thereof
    • G05B23/02Electric testing or monitoring

Abstract

The invention provides a fault diagnosis method based on active fault symptom pushing, comprising the following steps: symbols of all fault mode parameter representation forms are generated; a system fault symptom set is obtained; a system parameter is acquired, and the symbol of the representation form of the acquired parameter is generated; whether the generated symbol is matched with the symbol corresponding to at least one fault symptom in the fault symptom set is judged, the corresponding fault symptom is output and the method ends if the generated symbol is matched with the symbol corresponding to at least one fault symptom in the fault symptom set, or a fault symptom is determined based on the generated symbol; and the method goes to a step of active fault symptom pushing, a fault symptom already existing shows that an original knowledge base has a fault symptom fully consistent with the newly-emerging fault symptom, and the original knowledge base is not treated under the situation, and if a fault symptom is completely new, the corresponding relationship between the determined fault symptom and the generated symbol is added to the fault symptom set. The fault diagnosis method is based on active pushing instead of passive identification, so that complicated information search is reduced, and quick and correct fault diagnosis is realized.

Description

A kind of method for diagnosing faults based on failure symptom active push

Technical field

The present invention relates to a kind of method for diagnosing faults based on failure symptom active push, be applicable to complication system fault self-diagnosis and state confirmation certainly, belong to fault diagnosis technology field.

Background technology

Due in large scale, complex structure, part kind is many, influence factor is many feature, make the fault rate of complication system comparatively large, and do not have fault self-diagnosis function, whether the duty cannot understanding system is normal, once system jam, mistake can not be detected in real time.Existing diagnosis method for system fault, mostly adopts the method based on mathematical model, and this Method Modeling is complicated, and calculated amount is large, and for the nonlinear system of complexity, cannot set up the accurate dynamic mathematical models for fault diagnosis.

For the fault diagnosis system of complication system, the initial stage is all set up knowledge base according to expertise and historical data analysis result.Due to some non-intellectual of system factor and the uncertainty of emergency case, make set up knowledge base there is some incompleteness unavoidably, this just needs after new barrier pattern occurs, and upgrades existing knowledge base, supplements new fault mode knowledge.Because fault diagnosis compares the knowledge base depending on fault mode, so the pattern knowledge newly increased can cause larger impact to existing knowledge base.

In conventional fault diagnosis method (as methods such as pivot analysis, artificial neural network or support vector machine) when new fault mode, new fault mode knowledge and original training set are formed new training set by people usually, train the new diagnostic rule comprising new fault mode.Along with new fault mode increases, training data gets more and more, and occur that new model all will recalculate, diagnostic system adaptability is very poor at every turn, needs the limitation overcoming traditional fault diagnosis.

Summary of the invention

In view of this, the object of the invention is the problem can not carrying out self diagnosis in order to solve complication system to oneself state, and when breaking down, can not the problem of failure judgement type, and introduce failure symptom active push link renewal fault mode knowledge, propose a kind of method for diagnosing faults based on failure symptom active push, accelerate diagnostic reasoning speed and correctness.

Realizing technical scheme of the present invention is:

Based on a method for diagnosing faults for failure symptom active push, detailed process is:

Step one, system for required diagnosis, carry out symbol generation by all fault mode Parametric Representation forms;

Step 2, set up the corresponding relation of each failure symptom and fault mode parameter, obtain system failure symptom set;

Step 3, acquisition system parameter, and gathered Parametric Representation form is carried out symbol generation;

Whether the symbol generated in step 4, determining step three mates failure symptom is concentrated the symbol that at least one failure symptom is corresponding, if so, the method is terminated after then being exported by corresponding failure symptom, otherwise, based on the symbol determination failure symptom generated in step 3, after output, enter step 5;

Step 5, enter failure symptom active push link, for already present failure symptom, show that original knowledge stock is in situation on all four with emerging failure symptom, this situation does not do any process to original knowledge storehouse; For completely new failure symptom, add the corresponding relation of the symbol generated in the failure symptom determined and step 3 to failure symptom and concentrate, terminate the method.

Further, symbol of the present invention is generated as: parameter attribute is defined as four classes:

The first, whether threshold range is exceeded,

Parameter value Normally Higher than threshold value Lower than threshold value Unusual fluctuations Symbolic representation A1 A2 A3 A4

The second, Parameter Variation,

Three, parameter zero-bit characteristic variable,

Parameter zero-bit Be not classified as zero-bit Be classified as zero-bit Symbolic representation C1 C2

Four, whether there is sensor fault,

Sensor fault Not there is sensor fault feature There is sensor fault feature Symbolic representation D1 D2

For each parameter, carry out symbol generation according to above-mentioned four attribute.

Further, the present invention adopts binary value to represent parameter attribute symbol:

The first, whether threshold range is exceeded,

Parameter value Normally Higher than threshold value Lower than threshold value Unusual fluctuations Scale-of-two characterizes 00 01 10 11

The second, Parameter Variation,

Three, parameter zero-bit characteristic variable,

Parameter zero-bit Be not classified as zero-bit Be classified as zero-bit Scale-of-two characterizes 0 1

Four, whether there is sensor fault,

Sensor fault Not there is sensor fault feature There is sensor fault feature Scale-of-two characterizes 0 1

Beneficial effect:

The first, the present invention is based on the method for diagnosing faults of failure symptom active push, by mating the detected parameters of symbolism, changing active push into by passive discerning, decrease numerous and diverse information search work, realize fault quick, correctly diagnose.

Second, the present invention characterizes the measuring point parametic fault pattern of complication system by multidigit binary number, and information is clear, only can realize the sign of a large amount of fault characteristic information with less storage space, and be convenient to retrieval, improve the efficiency that failure symptom characteristic information stores and identifies.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of method for diagnosing faults of the present invention;

Fig. 2 is based on the inference machine schematic diagram of failure symptom active push;

The corresponding relation of Fig. 3 parameter and attribute and sign.

Embodiment

Below in conjunction with the drawings and specific embodiments, the present invention is described in detail.

As shown in Figure 1, a kind of method for diagnosing faults based on failure symptom active push of the present invention, detailed process is:

Step one, system for required diagnosis, carry out symbol generation by all fault mode Parametric Representation forms;

Step 2, set up the corresponding relation of each failure symptom and fault mode parameter, obtain system failure symptom set;

Step 3, acquisition system parameter, and gathered Parametric Representation form is carried out symbol generation;

Whether the symbol generated in step 4, determining step three mates failure symptom is concentrated the symbol that at least one failure symptom is corresponding, if, the method is terminated after then being exported by corresponding failure symptom, otherwise, adopting the method determination failure symptom of classical reasoning (as determined to adopt the concrete grammars such as pivot analysis, artificial neural network or support vector machine) based on the symbol generated in step 3, after output, entering step 5;

Step 5, enter failure symptom active push link, for already present failure symptom, show that original knowledge stock is in situation on all four with emerging failure symptom, this situation does not do any process to original knowledge storehouse; For completely new failure symptom, the corresponding relation of the symbol generated in the failure symptom determined and step 3 is added in fault knowledge storehouse and upgrades.

Failure symptom characteristic information, based on the method for diagnosing faults based on failure symptom active push, initiatively can not be passed to the shortcoming of decision maker, propose the failure symptom active push mode with data flow driven by the present invention for existing method.By structure and the principle of work of multianalysis complication system, failure mode analysis (FMA) is carried out to system, often kind of fault mode is carried out standard symbol, set up the failure symptom collection of system.According to the analysis of systematic parameter feature gathered and description, the failure symptom of coupling is delivered in decision maker's hand timely and accurately, reduces various numerous and diverse search work, realize fault quick, correctly diagnose.The fault diagnosis result do not appeared on symptom set is preserved by the present invention simultaneously, and then utilize it to upgrade failure symptom collection, if there is identical failure symptom after this, immediately diagnostic result is pushed to user, its process as shown in Figure 2.

Symbol of the present invention is generated as: parameter attribute is defined as four classes:

The first, whether threshold range is exceeded,

Parameter value Normally Higher than threshold value Lower than threshold value Abnormal (fluctuation) Symbolic representation A1 A2 A3 A4

The second, Parameter Variation,

Three, parameter zero-bit characteristic variable,

Parameter zero-bit Be not classified as zero-bit Be classified as zero-bit Symbolic representation C1 C2

Four, whether there is sensor fault,

Sensor fault Not there is sensor fault feature There is sensor fault feature Symbolic representation D1 D2

For each parameter, carry out symbol generation according to above-mentioned four attribute.

Such as, when collecting the parameter of sensor higher than threshold value, and be the form of Spline smoothing, simultaneously for not being classified as the state of zero-bit, then the representation after its symbolism is A2B3C1D2.

The present invention is by carrying out symbolism to the fault mode parameter that may exist of required diagnostic system, and each fault mode parameter is a corresponding symbol uniquely; Then set up the corresponding relation of each failure symptom and fault mode parameter, obtain system failure symptom set.

Such as, when there is filter clog fault sign, corresponding to it fault mode be that pressure and flow parameter decline all suddenly, the fault mode therefore pressure and flow parameter declined suddenly respectively symbol turns to corresponding symbol representation; Concentrate at failure symptom, the pressure that filter clog fault is corresponding and the symbol that flow parameter declines suddenly represent.Because the corresponding relation of each failure symptom and fault mode parameter is prior art, therefore do not enumerate at this.

Complication system in the present invention, in the course of the work, by sensor or Network Capture system measuring point parameter, utilizes above-mentioned four attributes provided to analyze, realizes the symbolism of measuring point parameter.If the measuring point parameter after symbolism is the fault mode match parameters concentrating at least one fault million corresponding with failure symptom, then represent that system exists this fault, if there is not situation about matching, then represent that the fault that system occurs does not represent in symptom set, now need to analyze parameter, redefine failure symptom, then utilize the failure symptom determined to upgrade failure symptom collection; Therefore the present invention is judging out of order while, and what can also realize symptom set is constantly perfect, so that the quick and precisely judgement of follow-up system fault.

Symptom set be designed to important component part of the present invention, failure symptom characteristic information correct, effectively and abundant be the prerequisite and guarantee of effectively dealing with problems based on the fault diagnosis system of failure symptom active push, to a certain extent, the level of failure symptom characteristic information determines the validity of this method.Originally be that in embodiment, the result of failure mode analysis (FMA) adopts binary mode to encode by symptom set, the corresponding parameter sets of each fault mode, the corresponding one group of parameter attribute of each parameter in parameter sets.Preserved in a database by these binary codes, make the memory capacity of symptom set little, retrieval rate is fast, calls conveniently.

The present invention is according to the concrete actual conditions of complication system, and the fault mode parameter form of expression includes but not limited to:

Parameter value exceeds threshold range: high and low.

Parameter Variation: gradual, step, impulse.

The specific performance characteristic of parameter, such as: the special value feature of fault parameter characteristic sum of measuring point or sensor.

As follows for each parameter performance formal definition parameter attribute variate-value:

L1 parameter values variable, span and implication as shown in table 1.

Table 1 parameter value scale-of-two characterizes

Parameter value Normally Higher than threshold value Lower than threshold value Abnormal (fluctuation) Scale-of-two characterizes 00 01 10 11

L2 parameter regularity variable, span and implication as shown in table 2.

The scale-of-two of table 2 Parameter Variation characterizes

L3 parameter zero-bit characteristic variable, span and implication as shown in table 3.

The scale-of-two of table 3 parameter zero-bit feature characterizes

Parameter zero-bit Be not classified as zero-bit Be classified as zero-bit Scale-of-two characterizes 0 1

L4 sensor characteristics variable, span and implication as shown in table 4.

The scale-of-two of table 4 sensor fault characterizes

Sensor fault Not there is sensor fault feature There is sensor fault feature Scale-of-two characterizes 0 1

For the value L4L3L2L1 of corresponding one group of this logical variable of each parameter, the corresponding matching condition of each one group of value of this logical variable set, namely as the sign of matching condition, as shown in Figure 3.

The present invention transforms fault diagnosis system, after carrying out sign symbolism, enter failure symptom active push link, the sign of symbolism is mated with symptom set, if the match is successful, directly export corresponding failure symptom result, if it fails to match, enter classical reasoning process, as adopted the concrete grammar such as pivot analysis, artificial neural network or support vector machine, and by the result feedback determined in symptom set.The result of each for system fault diagnosis is preserved, if there is identical failure symptom after which, immediately diagnostic result is pushed to user, accelerate the inference speed of inference machine.

Example 1:

For a specific process system, P represents pressure elements, and T represents temperature element (TE), and Q represents flow element, then typical fault feature coding is as follows:

(1) filter clog fault is characterized as pressure and flow parameter declines all suddenly.According to coding rule, the corresponding binary coding of its pressure versus flow is L4=0, L3=0, L2=010, L1=10.Then its pressure fault is encoded to Binary Zero 001010; Flow Fault is encoded to Binary Zero 001010.

(2) pressure-measuring-point fracture defect is characterized as the unexpected decreasing value atmospheric pressure of pressure parameter.According to coding rule, the corresponding binary coding of its pressure is L4=0, L3=1, L2=010, L1=10, and the corresponding binary coding of flow is L4=0, L3=0, L2=000, L1=00.Then its pressure fault is encoded to Binary Zero 101010; Flow Fault is encoded to Binary Zero 000000.

For the value of corresponding one group of this logical variable of each parameter, the corresponding Inference Conditions of each group value of this logical variable set, namely as the sign of Inference Conditions.The failure symptom collection coding of the present embodiment is as follows:

Fault type P Q Plugged filter 0001010 0001010 Pressure-measuring-point ruptures 0101010 0000000

This binary coding stored in failure symptom concentrate, when there is P=0001010 and Q=0001010 sign time, namely push " plugged filter " sign; When there is P=0101010 and Q=0000000 sign time, namely push " pressure-measuring-point fracture " sign.

In sum, these are only preferred embodiment of the present invention, be not intended to limit protection scope of the present invention.Within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.

Claims (3)

1. based on a method for diagnosing faults for failure symptom active push, it is characterized in that, detailed process is:
Step one, system for required diagnosis, carry out symbol generation by all fault mode Parametric Representation forms;
Step 2, set up the corresponding relation of each failure symptom and fault mode parameter, obtain system failure symptom set;
Step 3, acquisition system parameter, and gathered Parametric Representation form is carried out symbol generation;
Whether the symbol generated in step 4, determining step three mates failure symptom is concentrated the symbol that at least one failure symptom is corresponding, if so, the method is terminated after then being exported by corresponding failure symptom, otherwise, based on the symbol determination failure symptom generated in step 3, after output, enter step 5;
Step 5, add the corresponding relation of the symbol generated in the failure symptom determined and step 3 to failure symptom and concentrate, terminate the method.
2. according to claim 1 based on the method for diagnosing faults of failure symptom active push, it is characterized in that, described symbol is generated as: parameter attribute is defined as four classes:
The first, whether threshold range is exceeded,
Parameter value Normally Higher than threshold value Lower than threshold value Unusual fluctuations Symbolic representation A1 A2 A3 A4
The second, Parameter Variation,
Three, parameter zero-bit characteristic variable,
Parameter zero-bit Be not classified as zero-bit Be classified as zero-bit
Symbolic representation C1 C2
Four, whether there is sensor fault,
Sensor fault Not there is sensor fault feature There is sensor fault feature Symbolic representation D1 D2
For each parameter, carry out symbol generation according to above-mentioned four attribute.
3. according to claim 2 based on the method for diagnosing faults of failure symptom active push, it is characterized in that, adopt binary value to represent parameter attribute symbol:
The first, whether threshold range is exceeded,
Parameter value Normally Higher than threshold value Lower than threshold value Unusual fluctuations Scale-of-two characterizes 00 01 10 11
The second, Parameter Variation,
Three, parameter zero-bit characteristic variable,
Parameter zero-bit Be not classified as zero-bit Be classified as zero-bit Scale-of-two characterizes 0 1
Four, whether there is sensor fault,
Sensor fault Not there is sensor fault feature There is sensor fault feature Scale-of-two characterizes 0 1
CN201410720508.9A 2014-12-01 2014-12-01 Fault diagnosis method based on active fault symptom pushing CN104503434B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410720508.9A CN104503434B (en) 2014-12-01 2014-12-01 Fault diagnosis method based on active fault symptom pushing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410720508.9A CN104503434B (en) 2014-12-01 2014-12-01 Fault diagnosis method based on active fault symptom pushing

Publications (2)

Publication Number Publication Date
CN104503434A true CN104503434A (en) 2015-04-08
CN104503434B CN104503434B (en) 2017-05-03

Family

ID=52944838

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410720508.9A CN104503434B (en) 2014-12-01 2014-12-01 Fault diagnosis method based on active fault symptom pushing

Country Status (1)

Country Link
CN (1) CN104503434B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105096053A (en) * 2015-08-14 2015-11-25 哈尔滨工业大学 Health management decision-making method suitable for complex process system
CN105160081A (en) * 2015-08-14 2015-12-16 北京航天试验技术研究所 Fault diagnosis method based on three-dimensional virtual cognitive environment
CN107316158A (en) * 2017-07-05 2017-11-03 云能服(北京)科技有限公司 A kind of multidimensional work data treating method and apparatus
CN109632315A (en) * 2019-01-11 2019-04-16 浙江浙能技术研究院有限公司 A kind of Steam Turbine Vibration fault reasoning diagnostic method based on two-parameter rule match

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090037772A1 (en) * 2007-08-03 2009-02-05 Wegerich Stephan W Fuzzy classification approach to fault pattern matching
CN101634851A (en) * 2009-08-25 2010-01-27 西安交通大学 Method based on cause-and-effect relation of variables for diagnosing failures in process industry
CN102004486A (en) * 2010-09-26 2011-04-06 中国石油化工股份有限公司 Hybrid fault diagnosis method based on qualitative signed directed graph in petrochemical process
CN102521613A (en) * 2011-12-17 2012-06-27 山东省科学院自动化研究所 Method for fault diagnosis of automobile electronic system
CN103593336A (en) * 2013-10-30 2014-02-19 中国运载火箭技术研究院 Knowledge pushing system and method based on semantic analysis

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090037772A1 (en) * 2007-08-03 2009-02-05 Wegerich Stephan W Fuzzy classification approach to fault pattern matching
CN101634851A (en) * 2009-08-25 2010-01-27 西安交通大学 Method based on cause-and-effect relation of variables for diagnosing failures in process industry
CN102004486A (en) * 2010-09-26 2011-04-06 中国石油化工股份有限公司 Hybrid fault diagnosis method based on qualitative signed directed graph in petrochemical process
CN102521613A (en) * 2011-12-17 2012-06-27 山东省科学院自动化研究所 Method for fault diagnosis of automobile electronic system
CN103593336A (en) * 2013-10-30 2014-02-19 中国运载火箭技术研究院 Knowledge pushing system and method based on semantic analysis

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
杨思锋 等: "一种基于知识推送的故障诊断系统", 《航空动力学报》 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105096053A (en) * 2015-08-14 2015-11-25 哈尔滨工业大学 Health management decision-making method suitable for complex process system
CN105160081A (en) * 2015-08-14 2015-12-16 北京航天试验技术研究所 Fault diagnosis method based on three-dimensional virtual cognitive environment
CN105096053B (en) * 2015-08-14 2018-11-09 哈尔滨工业大学 A kind of health control decision-making technique suitable for complicated technology system
CN105160081B (en) * 2015-08-14 2018-11-09 北京航天试验技术研究所 A kind of method for diagnosing faults based on three-dimensional cognitive environment
CN107316158A (en) * 2017-07-05 2017-11-03 云能服(北京)科技有限公司 A kind of multidimensional work data treating method and apparatus
CN109632315A (en) * 2019-01-11 2019-04-16 浙江浙能技术研究院有限公司 A kind of Steam Turbine Vibration fault reasoning diagnostic method based on two-parameter rule match

Also Published As

Publication number Publication date
CN104503434B (en) 2017-05-03

Similar Documents

Publication Publication Date Title
Cai et al. Bayesian networks in fault diagnosis
US10192170B2 (en) System and methods for automated plant asset failure detection
CN104866632B (en) Fault data acquisition methods, device and the terminal of home appliance
Ahmed et al. Similarity analysis of industrial alarm flood data
US20160203036A1 (en) Machine learning-based fault detection system
Wu et al. Fuzzy filter design for Itô stochastic systems with application to sensor fault detection
Demetriou et al. Incipient fault diagnosis of dynamical systems using online approximators
US9734001B2 (en) Efficient health management, diagnosis and prognosis of a machine
US7496798B2 (en) Data-centric monitoring method
US5351247A (en) Adaptive fault identification system
US8036999B2 (en) Method for analyzing and classifying process data that operates a knowledge base in an open-book mode before defining any clusters
CN102779230B (en) State analysis and maintenance decision judging method of power transformer system
JP4762088B2 (en) Process abnormality diagnosis device
Bai et al. Generalized shock models based on a cluster point process
Wu et al. Inferring topologies of complex networks with hidden variables
CN105424395A (en) Method and device for determining equipment fault
Junejo et al. Behaviour-based attack detection and classification in cyber physical systems using machine learning
CN102208028A (en) Fault predicting and diagnosing method suitable for dynamic complex system
CN103914735A (en) Failure recognition method and system based on neural network self-learning
CN103699489B (en) A kind of remote software fault diagnosis and restorative procedure based on knowledge base
CN105391579B (en) Power communication network fault positioning method based on crucial alarm collection and supervised classification
US20150199224A1 (en) Method and Apparatus for Detection of Anomalies in Integrated Parameter Systems
CN102291392B (en) Hybrid intrusion detection method based on Bagging algorithm
US20120290261A1 (en) Valve actuator fault analysis system
CN102282552A (en) System, method and computer program for pattern based intelligent control, monitoring and automation

Legal Events

Date Code Title Description
PB01 Publication
C06 Publication
SE01 Entry into force of request for substantive examination
C10 Entry into substantive examination
GR01 Patent grant
GR01 Patent grant