CN106250631A - A kind of method for diagnosing faults based on fault test correlation matrix - Google Patents
A kind of method for diagnosing faults based on fault test correlation matrix Download PDFInfo
- Publication number
- CN106250631A CN106250631A CN201610629637.6A CN201610629637A CN106250631A CN 106250631 A CN106250631 A CN 106250631A CN 201610629637 A CN201610629637 A CN 201610629637A CN 106250631 A CN106250631 A CN 106250631A
- Authority
- CN
- China
- Prior art keywords
- node
- fault
- matrix
- test
- bayesian network
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Analysis (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Computer Hardware Design (AREA)
- Computational Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Geometry (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
Abstract
The present invention is a kind of method for diagnosing faults based on fault test correlation matrix, comprises the following steps: 1) determine that the system failure tests correlation matrix D by the dependency graph representation model of system；2) Bayesian network node is set up；3) according to element value Connection Step 2 each in matrix D) in two-layer Bayesian network node；4) for representing that the Bayesian network root node of fault arranges conditional probability table；5) for representing that the Bayesian network leaf node of test item arranges conditional probability table；6) evidence variable is set according to the test result of each test item；7) diagnostic result is obtained；The method for diagnosing faults based on fault test correlation matrix that the present invention proposes, available succinct figure intuitively describes the logical relation between fault and test, the probability that effective integration system each fault occurs, overcome existing direct application correlation matrix and carry out the defect of method for diagnosing faults existence, make diagnostic analysis result more quantitative than correlation matrix.
Description
Technical field
The invention belongs to the Analysis on Fault Diagnosis field of device integration system, be specifically related to a kind of based on fault-test phase
Close the method for diagnosing faults of matrix.
Background technology
Along with the new and high technology extensive application in various kinds of equipment, the synthesization of device integration system, informationization, systematization
Degree is more and more higher.While improve systems technology performance, owing to complexity and unit volume comprise number of devices
Increasing, the properly functioning factor of the system that affects increases the most therewith, and the probability resulting in trouble or failure is increasing.Cause
This, a kind of method of fast and effeciently removal system fault is increasingly important.
At present, occur in that a large amount of system fault diagnosis method method, including correlation model, fuzzy theory, specialist system, artificial
Neutral net, Bayesian network etc..Wherein, system diagnosis method based on correlation model is widely applied in engineering,
Especially at aerospace field.Correlation models comes across the 1950's the earliest, by DSI company of U.S. founder De
Paul started first this theory to be applied to system fault diagnosis from the sixties in 20th century.Correlation models is made with Correlation Reasoning
Based on, application and trouble-test matrix (abbreviation matrix D) carries out fault detection and diagnosis to system or equipment.Therefore based on relevant
The method for diagnosing faults of property model, its core is the analyzing and processing to D matrix, thus obtains diagnosis.But, D matrix
In each element qualitative expression fault and the incidence relation of test, do not consider the probability that each fault occurs.When multiple events
When barrier can be carried out state-detection by same test item, do not consider each fault rate, only will be unable to judge have most by D matrix
Contingent fault, diagnostic method lost efficacy.Such as, train is made up of element A and B, and two elements all can be carried out by test T
Detection, i.e.Work as T=1, i.e. during test item T test result exception, can determine whether that A and B is likely to occur by matrix D
Fault.But do not merge the probability of malfunction information of A and B in view of matrix D, therefore cannot specifically judge which element of A and B more has can
Can break down.
Method for diagnosing faults based on Bayesian network is combined closely with theory of probability, it is possible to effective integration constituent system components
Probability of malfunction information, therefore can effectively make up direct application matrix D and carry out the deficiency of fault diagnosis.It is currently used for fault diagnosis
Bayesian network is made up of two-layer node.Bayesian network for fault diagnosis is also called fault Bayesian network, and it is first years old
Layer is failure cause layer, and this node layer have expressed the fault of constituent system components, and node state is that fault occurs or do not occurs two
Kind；The second layer is failure symptom layer, and each node have expressed the phenomenon of the failure caused by failure cause, and phenomenon of the failure is typically by surveying
Examination item measures, and this node layer state is that phenomenon of the failure occurs or do not occurs two kinds.In Bayesian network, each node is all with one
Open conditional probability table to be associated.Wherein, the conditional probability table that associates with ground floor node, quantitative expression constituent system components
Probability of malfunction；The conditional probability table associated with second layer node, the incidence relation of quantitative expression and ground floor connected node.So
And, the incidence relation (i.e. fault and test incidence relation) between two-layer node is difficult to directly give, and this modeling difficulty limits base
Method for diagnosing faults extensively applying in engineering in Bayesian network.
To sum up, traditional method for diagnosing faults based on correlation matrix D exists can not the event of effective integration system component units
The shortcoming of barrier probability, there is the shortcoming being difficult to directly give fault with test incidence relation in traditional fault Bayesian network.With
Upper two shortcomings all limit the extensive application in systems in practice of each self-diagnosing method, and therefore seeking one can effectively overcome
The new method for diagnosing faults of existing two kinds of method shortcomings is significant for system fault diagnosis.
Summary of the invention
It is difficult to merge probability of malfunction makes the not accurate enough problem of diagnostic result for solving existing fault-test matrix, with
Time in order to overcome malfunctioning node and test node incidence relation in existing fault Bayesian network to be difficult to the defect obtained, the present invention
A kind of method for diagnosing faults based on fault-test correlation matrix is proposed.
The technical solution of the present invention is: the present invention comprises three core: A by fault-test correlation matrix structure
The method building two-layer Bayesian Network Topology Structures；The method to set up of B two-layer Bayesian network each node condition probability tables；C is originally
The diagnostic method flow process based on fault-test correlation matrix that invention proposes.
The process that realizes of the present invention comprises the following steps: a kind of method for diagnosing faults based on fault-test correlation matrix,
It is characterized in that: said method comprising the steps of:
1) system failure-test correlation matrix D is determined by the dependency graph representation model of system；Wherein, in matrix D, often
The fault of the only element in a line correspondence system, the unique test item in every string correspondence system；That is, for being formed by m
Unit, the system of n test item, correlation matrix D has m row n column data；
2) Bayesian network node is set up；The fault being followed successively by system component units corresponding to each row in correlation matrix D is built
Vertical unique corresponding node, as the root node of Bayesian network；The test item being followed successively by correlation matrix D each row corresponding is set up only
One corresponding node, as the leaf node of Bayesian network；
3) according to element value Connection Step 2 each in matrix D) in two-layer Bayesian network node；The operation step of line
Rapid as follows:
The i-th row jth column element d for matrix Dij(1≤i≤m, 1≤j≤n), if dij=1, then by fault FiIn step
2) Bayesian network root node line corresponding in points to test item TjIn step 2) in corresponding Bayesian network leaf node；If
dij=0, do not carry out any line operation；
Traversal i and j, carries out described line operation successively to element each in matrix D；
4) for representing that the Bayesian network root node of fault arranges conditional probability table；
5) for representing that the Bayesian network leaf node of test item arranges conditional probability table；If leaf node Tj(1≤j≤n) has k
Individual father node, then father node combinations of states has 2kThe situation of kind；And when to have any node state in k father node be 1, leaf node
TjState be 1 probability be set to 1.0；When k father node state that and if only if is 0, leaf node TjState be 1 probability set
It is set to 0.0；
6) evidence variable is set according to the test result of each test item: for testing the leaf node T passed throughjState T is setj
=0, leaf node T unsanctioned for test resultjState T is setj=1,1≤j≤n；
7) diagnostic result is obtained；The method of described acquisition diagnostic result is as follows:
Calculate Bayesian network posterior probability Pr (Fi=1 | T=t) (1≤i≤m, Fi∈ F), calculate knot according to posterior probability
Fruit is from big to small to fault FiIt is ranked up, according to FiRanking results carries out troubleshooting successively to system.
Above-mentioned steps 1) determine that the concrete of the system failure-test correlation matrix D walks by the dependency graph representation model of system
Suddenly:
1.1] for having m unit, the system of n test item, set up two-dimensional correlation matrix D, matrix D size for m
Row n arranges, the fault of the unique component units in each row correspondence system dependency graph representation model, each row correspondence system correlation models
In unique test item；Initializing each element in D is 0；
1.2] i=1 is made, wherein, i representing matrix D the i-th row data, FiThe system group that representing matrix D the i-th row data are corresponding
Become the fault of unit；
1.3] in dependency graph representation model, all square frames and circle are regarded as node, all node states are set and are
0, i.e. put Fi=0 and Tj=0；
1.4] in dependency graph representation model, at numbered FiNode in place a token, put numbered FiNode
State is 1, i.e. Fi=1, and search for, by searching algorithm, whole nodes that this token can arrive along directed line；
1.5] in dependency graph representation model, if node FiThe node that can arrive is the circle representing test item, that
Determine respective column j in matrix D of the test item represented by this circle, and in matrix D, put dij=1；
1.6] to i from adding 1:i=i+1；If i < m+1, then forward step 1.3 to]；Otherwise, in matrix D, each element has determined
Finishing, algorithm terminates.
Step 1.4] searching algorithm specifically includes following steps:
1.4.1] an empty linear list L is set up；
1.4.2] with token place node as starting point, point to along directed line segment and determine whether the adjacent node of starting point meets two
Individual condition: the adjacent node state of starting point is 0, the adjacent node of starting point is not present in linear list L；For meeting two conditions
The adjacent joint of starting point is stored in linear list L；
1.4.3] if linear list L is empty, reject the token in dependency graph representation model, terminate searching algorithm；Otherwise, exist
In dependency graph representation model, token is moved in the node that linear list L end data represents, linear list L end data is represented
The state of node change 1 into from 0, delete linear list L end data, forward step 1.4.2 to]；
Above-mentioned steps 4) if fault FiProbability of happening be Pi, then it represents that fault FiNode condition probability tables parameter arrange
As follows:
Root node Fi Pr(Fi)
0 1.0-Pi
1 Pi。
The invention have the benefit that
The method for diagnosing faults based on fault-test correlation matrix that the present invention proposes, by building according to correlation matrix
Bayesian network, not only with succinct figure the most intuitively, the logical relation between fault and test can be described, it is often more important that effectively
Merge the probability that each fault of system occurs, overcome existing direct application correlation matrix and carry out method for diagnosing faults existence
Defect so that diagnostic analysis result is more quantitative than correlation matrix；
The method for diagnosing faults based on fault-test correlation matrix that the present invention proposes simultaneously, overcomes and is currently used for event
It is tired that the internodal incidence relation of two-layer Bayesian network (i.e. fault with test incidence relation) of barrier diagnosis is difficult to directly give
Difficulty, can promote the method for diagnosing faults based on Bayesian network extensive application in engineering.
Finally, method for diagnosing faults effective integration of based on fault-test correlation matrix the Correlation Moment that the present invention proposes
Battle array and Bayesian network each advantage, quantitative scoring can calculate the probability that each fault occurs in the case of known test result, and give
Go out system maximum possible and explain to support fault diagnosis, for carrying out the research of the strong failure diagnosis tool of maneuverability, there is reference
Meaning, can assist the intellectuality realizing fault diagnosis, improves diagnosis efficiency.
Accompanying drawing explanation
Fig. 1 is present invention method for diagnosing faults flow chart based on fault-test correlation matrix；
Fig. 2 is the dependency graph representation model of system in the present invention-example；
Fig. 3 is the Bayesian network of equal value of dependency graph representation model in the present invention-example.
Detailed description of the invention
The present invention is a kind of method for diagnosing faults based on fault-test correlation matrix.Below in conjunction with Figure of description pair
Technical scheme is described in further detail.The specific implementation process of the embodiment of the present invention is as follows:
Step 1, obtains the system failure-test correlation matrix D by the dependency graph representation model of system.
The fault of the only element in described matrix D every a line correspondence system, the unique test in every string correspondence system
?；That is, for there being m component units, the system of n test item, correlation matrix D has m row n column data.Each group of described system
Unit is become the most only to include normal and fault two states, the event of the system component units that the i-th row (1≤i≤m) is corresponding in matrix D
Barrier FiRepresent, the test item T that jth row (1≤j≤n) is correspondingjRepresenting, matrix D method for expressing is as follows:
I-th row jth column data d in matrix DijIndicate fault FiWith test item TjDependency.Work as TjFault can be measured
FiFault message time, then test item TjWith fault FiRelevant, dij=1；Work as TjFault F can not be measurediFault message time, then
Test item TjWith fault FiUnrelated, dij=0.
System refer to by interact and complementary some unit be combined in order to realize the organic of specific function
Overall.Wherein, unit can be element, equipment, subsystem and the module with specific function.Described correlation models is to express
Cell failure and the model of test correlation logic relation in system, including dependency graph representation model and dependency mathematical model two
The form of kind.Described dependency graph representation model can represent the relation between unit and test item intuitively；And dependency mathematical modulo
Type fault-test matrix D describes the dependency between unit and test, and matrix D is also known as correlation matrix.
Specifically, the method for expressing of described dependency graph representation model is the most each component units function and structure conjunction
After reason divides, on the basis of functional block diagram, clearly indicate function information flow path direction and each component units interconnected relationship,
And mark position and the numbering understanding test item, to show the correlative relationship of each component units and each test item.Dependency graph
Representation model includes three class graphic elements: square frame, circle, directed line segment.Wherein, each component units in box indicating system,
Circle illustrates the test item of each unit in test system, and directed line segment indicates the function information of each unit in system and passes
Pass direction.Additionally, in dependency graph representation model, each square frame is with the fault of its system component units represented for numbering, each circle
Enclose with its test item represented for numbering.
Dependency graph representation model is the basis setting up matrix D.Described by dependency graph representation model acquisition correlation matrix D's
Algorithm is as follows:
1. for having m unit, the system of n test item, set up two-dimensional correlation matrix D, matrix D size for m row n
Row, the fault of the unique component units in each row correspondence system dependency graph representation model, in each row correspondence system correlation models
Unique test item.Initializing each element in D is 0.
2. i=1 is made.Wherein, i representing matrix D the i-th row data, FiThe system composition that representing matrix D the i-th row data are corresponding
The fault of unit.
3. in dependency graph representation model, all square frames and circle are regarded as node, all node states are set and are 0,
I.e. put Fi=0 and Tj=0.
4. in dependency graph representation model, at numbered FiNode in place a token, put numbered FiNode shape
State is 1, i.e. Fi=1, and search for, along directed line segment, all nodes that this token can arrive.Described searching algorithm is relevant
Property figure representation model includes 3 steps, specific as follows:
I) an empty linear list L is set up；
Ii) with token place node as starting point, point to along directed line segment and determine whether the adjacent node of starting point meets two
Condition: the adjacent node state of starting point is 0, the adjacent node of starting point is not present in linear list L.Rise for meeting two conditions
The adjacent node of point is stored in linear list L.
Described adjacent node refers between two nodes, points to the oriented of another node if there is by a node
Line segment, then later node is the adjacent node of previous node；
Iii) if linear list L is empty, the token in dependency graph representation model is rejected, the search calculation that end step is 4. described
Method；Otherwise, in dependency graph representation model, token is moved in the node that linear list L end data represents, by linear list L end
The state of the node that end data represents changes 1 into from 0, deletes linear list L end data, forwards step ii to).
Token refers to a kind of labelling in dependency graph representation model, marked the node that searching algorithm is being searched for, logical
Often represent with stain.
Complete step i)～iii) after, state be 1 node be node FiAccessibility all nodes.
5. in dependency graph representation model, if node FiThe node that can arrive is the circle representing test item, then
Determine respective column j in matrix D of the test item represented by this circle, and in matrix D, put dij=1.
6. to i from adding 1:i=i+1.If i < m+1, then forward step to 3.；Otherwise, in matrix D, each element determines complete, calculates
Method terminates.
3.～5. step constitutes a circulation controlled by variable i: as i=1,3.～5. by the 1st time step performs, can
Complete the determination of matrix D the 1st row data；As i=2,3.～5. by the 2nd time step performs, and can complete matrix D the 2nd row data
Determination；The like, step is 3.～5. by the 3rd, 4 ... performs, is sequentially completed matrix D the 3rd, 4 ... m row data are really for m time
Fixed.Wherein, 3.～5. step 6. for being circulated the statement of operation by variable i rate-determining steps: when i < during m+1, to step 3.
～5. operate, i.e. enter and circulate next time；Otherwise, 3.～the most no longer step is performed, i.e. loop ends.
When, after loop ends, 1.～6. step terminates, so far complete the determination of fault-test correlation matrix D.
In the present embodiment, as shown in dependency graph representation model in Fig. 2, system includes 5 component units, 2 test items.
Each component units includes two states: normal and fault.The fault of 5 component units uses F respectively1、F2、F3、F4And F5Represent,
And F1～F5Probability of malfunction be followed successively by 0.001,0.002,0.003,0.004 and 0.005, test item uses variable respectively in addition
T1And T2Represent.T1It is directly used in test failure F1And F5, work as F1And F5Middle Arbitrary Fault occurs, test item T1Test result not
Pass through；T2It is directly used in test failure F3, work as F3Fault occurs, test item T2Test result do not pass through.
In the present embodiment, determine that the implementation process of correlation matrix D is as follows by dependency graph representation model shown in Fig. 2:
1. there are 5 component units and 2 test items in the dependency graph representation model of system shown in Figure 2, then m=5 and n=
2 set up.The correlation matrix D with 5 row 2 column data, matrix D the i-th row (1≤i≤m, m=5) is set up for dependency graph representation model
Fault F with component unitsiCorrespondence, jth row (1≤j≤n, n=2) and the test item T of matrix DjCorrespondence, the matrix D of foundation
Expression-form is as follows:
In matrix D, dijValue is 0 or 1.Initialize each element d in DijIt is 0 (1≤i≤m, 1≤j≤n):
2. i=1 is made.Wherein, i representing matrix D the i-th row data, FiThe system composition that representing matrix D the i-th row data are corresponding
The fault of unit.
3., in the dependency graph representation model shown in Fig. 2, F is set1=0, F2=0, F3=0, F4=0 and F5=0, arrange
T1=0 and T2=0；
4. in the dependency graph representation model shown in Fig. 2, at numbered FiNode in place a token, token is with black
Point represents, puts Fi=1, i=1.F is searched for along directed line segmentiWhole nodes that middle token can arrive.Step is as follows:
I) an empty linear list L is set up；
Ii) with token place node as starting point, point to along directed line segment and determine whether the adjacent node of starting point meets two
Condition: the adjacent node state of starting point is 0, the adjacent node of starting point is not present in linear list L.Rise for meeting two conditions
The adjacent node of point is stored in linear list L.
In fig. 2, node FiIt is placed with token, then with FiFor starting point, point to along directed line segment and determine FiAdjacent node be
No satisfied two condition: FiAdjacent node state be 0, FiAdjacent node be not present in linear list L.In fig. 2, meet
The node of two conditions is F2And T2, two nodes are stored in linear list L, then L={F2, T2}。
Iii) if linear list L is empty, reject the token in dependency graph representation model, terminate searching algorithm；Otherwise, in phase
In closing property figure representation model, token is moved in the node that linear list L end data represents, linear list L end data is represented
The state of node changes 1 into from 0, deletes linear list L end data, forwards step ii to).
In the present embodiment, linear list L={F2, T2, end is T2, then in fig. 2, by token from node F1Move into
Node T2, by T2=0 changes T into2=1, delete linear list L end data T2, now L={F2, forward step ii to)；
Proceed to step ii) after, to be placed with the node T of token2For starting point, do not exist and meet step ii) said two bar
The node of part, therefore be not required to perform step ii) in storing step, be directly entered step iii)；
Enter step iii) after, now linear list L={F2, end is F2, then in fig. 2, by token from node T2Move
Ingress F2, by F2=0 changes F into2=1, delete linear list L end data F2, now chained list is empty, forwards step ii to)；
Proceed to step ii) after, to be placed with the node F of token2For starting point, meet step ii) joint of said two condition
Point is F3, then by F3It is stored in linear list L, then L={F3}；
Enter step iii) after, now linear list L={F3, end is F3, then by token from node F2Move into node
F3, by F3=0 changes F into3=1, delete linear list L end data F3, now chained list is empty, forwards step ii to).
The like, repeat step ii) and iii), until in step iii) in linear list L be sky, searching algorithm terminates.
At the end of searching algorithm, in Fig. 2, the state of each node is: F1=1, F2=1, F3=1, F4=0, F5=0, T1=1, T2=1
, step i)～iii 4. described according to step) terminate after, state be 1 node be FiThe all joints that can arrive
Point.I.e., in the present embodiment, node Fi(i=1) up to node have F1、F2、F3、F4、T1And T2。
5. in the dependency graph representation model shown in Fig. 2, node Fi(i=1) circle representing test item that can arrive
For T1And T2, two test items respective column in matrix D is followed successively by the 1st and the 2nd row, therefore puts d in matrix Di1=1 and di2
=1, i=1.
6. to i from adding 1:i=i+1.If i < m+1, then forward step to 3.；Otherwise, in matrix D, each element determines complete, calculates
Method terminates.
In the present embodiment, now i=1, so far 3.～5. by the 1st time step is finished, the 1st row data in matrix D
Have determined that complete, be followed successively by d11=1 and d12=1；
To i after adding 1: i=2.I < m+1 (m=5) set up, therefore, step need to be forwarded to 3., the 2nd time perform step 3.～
⑤；After 5. 3. step～be finished by the 2nd time, it may be determined that in matrix D, the 2nd row data are followed successively by d21=1 and d22=0；
To i after adding 1: i=3.I < m+1 (m=5) set up, therefore, step need to be forwarded to 3., the 3rd time perform step 3.～
⑤；After 5. 3. step～be finished by the 3rd time, it may be determined that in matrix D, the 3rd row data are followed successively by d31=1 and d32=0；
The like, i being carried out from add operation: when i < m+1 sets up (m=5), forward step to 3., i & lt performs step
3.～5., the determination of the i-th row data in matrix D is completed；3.～the most described circulation otherwise end step.
When, after loop ends, 1.～6. step terminates, so far complete the determination of fault-test correlation matrix D, determine
Correlation matrix D is as follows:
Step 2, sets up Bayesian network node；It is followed successively by the event of system component units corresponding to each row in correlation matrix D
Barrier sets up unique corresponding node, as the root node of Bayesian network；The test item being followed successively by correlation matrix D each row corresponding is built
Vertical unique corresponding node, as the leaf node of Bayesian network.
The Bayesian network node of described foundation, if the fault of system component units corresponding to its representing matrix D the i-th row,
Then for Bayesian network root node numbering F set upi；If the test item that its representing matrix D jth row are corresponding, then for the shellfish set up
This network leaf node code T of leafj。
The Bayesian network node set up the most only comprises 0 and 1 two states, root node FiState is 0 expression fault FiDo not send out
Raw, root node FiState is 1 expression fault FiOccur；Leaf node TjState is 0 expression test item TjMeasured result by or just
Often, leaf node TjState is 1 expression test item TjMeasured result is not passed through or abnormal.
In the present embodiment, correlation matrix D has 5 row data, is corresponding in turn to system component units fault F1、F2、F3、F4With
F5, for F1～F5Set up corresponding node respectively, as the root node of Bayesian network；Correlation matrix D has 2 column data, is corresponding in turn to
Test item T1And T2, for T1And T2Set up corresponding node respectively, as the leaf node of Bayesian network.
As it is shown on figure 3, the root node set up amounts to 5, corresponding to the row data in matrix D, the numbering of root node is successively
For node F1～F5；The leaf node set up amounts to 2, and corresponding to the column data in matrix D, the numbering of leaf node is followed successively by node
T1And T2.Root node and the leaf node set up are two condition: root node Fi(1≤i≤5) state is 0 expression fault FiDo not occur,
Root node FiState is 1 expression fault FiOccur；Leaf node Tj(1≤j≤2) state is 0 expression test item TjMeasured result is passed through
Or normal, leaf node TjState is 1 expression test item TjMeasured result is not passed through or abnormal.
Step 3, according to the two-layer Bayesian network node in element value Connection Step 2 each in matrix D；Described line
Operating procedure is as follows:
The i-th row jth column element d for matrix Dij(1≤i≤m, 1≤j≤n), if dij=1, then by fault FiIn step
2) Bayesian network root node line corresponding in points to test item TjIn step 2) in corresponding Bayesian network leaf node；If
dij=0, do not carry out any line operation；
Traversal i and j, element each in matrix D is carried out step 3 successively) described in line operation.
In the present embodiment, according to described in step 3, concrete operations are as follows:
Matrix D the 1st row the 1st column element d11=1, in figure 3 by node F1Line points to node T1；
Matrix D the 1st row the 2nd column element d12=1, in figure 3 by node F1Line points to node T1；
Matrix D the 2nd row the 1st column element d21=1, in figure 3 by node F2Line points to node T1；
Matrix D the 3rd row the 1st column element d31=1, in figure 3 by node F3Line points to node T1；
Matrix D the 4th row the 2nd column element d42=1, in figure 3 by node F4Line points to node T2；
Matrix D the 5th row the 2nd column element d52=1, in figure 3 by node F5Line points to node T2；
Remaining element of matrix D is 0, does not carry out any line operation.
Step 4, arranges conditional probability table for the Bayesian network root node in step 2.The conditional probability table of each root node
Including 2 conditional probability value, in conditional probability table, each parameter is shown in Table 1.In table 1, as root node FiWhen being 0, conditional probability value
Pr(Fi)=1.0-Pi；As root node FiWhen being 1, conditional probability value Pr (Fi)=Pi.Wherein PiFor matrix D the i-th row corresponding be
The probability of malfunction of system component units.In conditional probability table, FiValue 0 and 1 represents fault F respectivelyiThere are not and occur two kinds of shapes
State.
Table 1 root node conditional probability table
| Root node Fi | Pr(Fi) |
| 0 | 1.0-Pi |
| 1 | Pi |
Conditional probability table refers to the conditional probability table that each node in Bayesian network is corresponding.In conditional probability table
Each row of data include the conditional probability value two that node and the combinations of states of father node, combinations of states are corresponding.Conditional probability table
The implication that middle each row of data is expressed is: when father node state is listed state in table, and node is that listed shape probability of state is equal to table
Conditional probability value listed by.
In the present embodiment, the probability of malfunction of the system component units that matrix D the 1st～5 row is corresponding be followed successively by 0.001,
0.002,0.003,0.004 and 0.005.According to the root node conditional probability method to set up described in step 4, Bayesian network root
In the conditional probability table of node F1, data are: Pr (F1=0)=0.999, Pr (F1=1)=0.001；Root node F2Condition general
In rate table, data are: Pr (F2=0)=0.998, Pr (F2=1)=0.002；Root node F3Conditional probability table in data be: Pr
(F3=0)=0.997, Pr (F3=1)=0.003；Root node F4Conditional probability table in data be: Pr (F4=0)=0.996,
Pr(F4=1)=0.004；Root node F5Conditional probability table in data be: Pr (F5=0)=0.995, Pr (F5=1)=
0.005.Root node F1、F2、F3、F4And F5Conditional probability table as shown in table 3.
Each root node conditional probability table in table 3 Bayesian network
Step 5, for representing that the Bayesian network leaf node of test item arranges conditional probability table；If leaf node Tj(1≤j≤
N) have k father node, then father node combinations of states has 2kThe situation of kind；And when to have any node state in k father node be 1,
Leaf node TjState be 1 probability be set to 1.0；When k father node state that and if only if is 0, leaf node TjState is 1
Probability is set to 0.0.
For the present embodiment, in figure 3, leaf node T1There are 3 father node { F1,F2,F3, when arbitrary father node state is 1
Time, leaf node state be the probability of 1 be 1.0；When 3 father node states are 0, leaf node state be the probability of 1 be 0.0.
In table 4, as j=1, table 4 is leaf node T1Conditional probability table；
In figure 3, leaf node T2There are 3 father node { F1,F4,F5, when arbitrary father node state is 1, leaf node state
Be 1 probability be 1.0；When 3 father node states are 0, leaf node state be the probability of 1 be 0.0.In table 4, j=2 is worked as
Time, table 4 is leaf node T2Conditional probability table.So far, the dependency graph representation model equivalence Bayesian network shown in Fig. 2 is completed
The structure of network.
Table 4 leaf node conditional probability table
Step 6, arranges evidence variable according to the test result of each test item: for testing the leaf node T passed throughjShape is set
State Tj=0, leaf node T unsanctioned for test resultjState T is setj=1,1≤j≤n.
Described evidence variable refers in Bayesian network, variable known to state.Test knot according to each test item
Really, the state of each test item be known state for by or do not pass through, the most in the present invention, evidence variable includes shellfish
All leaf nodes of this network of leaf, i.e. Bayesian network leaf node T1、T2…Tn.Leaf node T1、T2…TnConstitute evidence variables set
Closing T, in set T, each variable-value is designated as t.
The most known test T1Do not pass through, test T2Pass through, then evidence variables collection T={T1, T2, gather T
In each variable-value t={T1=1, T2=0}.
Step 7, it is thus achieved that diagnostic result；The method of described acquisition diagnostic result is as follows:
Calculate Bayesian network posterior probability Pr (Fi=1 | T=t) (1≤i≤m, Fi∈ F), calculate knot according to posterior probability
Fruit is from big to small to fault FiIt is ranked up, according to FiRanking results carries out troubleshooting successively to system.
The existing software of application or algorithm routine all can solve Bayesian network posterior probability problem.In the present embodiment,
Application Bayesian network analysis software SamIam solves posterior probability Pr (F successivelyi=1 | T1=1, T2=0) (1≤i≤m, m=
5), and according to posterior probability result from big to small to faulty FiSequence.Result is as follows:
①Pr(F3=1 | T1=1, T2=0)=0.6072874
②Pr(F2=1 | T1=1, T2=0)=0.5951417
③Pr(F1=1 | T1=1, T2=0)=0.0
Pr(F4=1 | T1=1, T2=0)=0.0
Pr(F5=1 | T1=1, T2=0)=0.0
F is understood according to posterior probability ranking results3It is most likely to occur fault, next to that F2, remaining fault F1、F4And F5All
Can not occur.So in known test result T1=1, T2When=0, the order of system troubleshooting is F3、F2。
So far, the fault diagnosis to Fig. 2 system is completed by above 7 steps.The most in conjunction with the embodiments, from following
Beneficial effect produced by the 3 aspect explanation present invention.
First, the present embodiment diagnostic result possesses reasonability.In the case of test is completely reliable, T2Test is passed through, root
Can determine whether according to the directions of information flow shown in directed line segment in Fig. 3, F1、F4、F5Necessarily will not break down, the posteriority in embodiment
Probability problem solving result is all consistent with this fact；T1Test is not passed through, and can sentence according to the directions of information flow of directed line segment in Fig. 3
Disconnected, F1、F2、F3All it is likely to occur fault, and by T2Test is by having got rid of F1The probability of fault, therefore at evidence variable
{T1=1, T2Under the support of=0}, only F2And F3It may happen that fault, the posterior probability problem solving result in embodiment shows
The probability that both each break down, more than 0.5, is true to life.
Secondly, the probability that each fault of the method effective integration that the present invention proposes occurs, diagnostic result is more quantitative, relative to
It is directly based upon the method for correlation matrix D at quantitative aspect more superiority.Application correlation matrix D carries out event to system directly below
Barrier is analyzed.Know evidence variable { T1=1, T2=0}, matches with the 2nd row and the 3rd row data in matrix D, and this shows at this
Under evidence variable is supported, matrix D the 2nd row and fault F corresponding to the 3rd row2And F3All it is likely to occur, but cannot deduce further
Both more likely break down whom, thus increase the probability that troubleshooting mistiming is torn open, and delay system recovers the time of normal work.
And the diagnostic method that the present invention proposes, at known evidence variable { T1=1, T2In the case of=0}, it is possible to each fault of effective integration
The probability occurred, by solving of posterior probability problem, obtains F3Compare F2It is more likely to occur this result of fault, and then can give
Go out troubleshooting order F3And F2, reduce the probability by mistake torn open, improve the time that system recovery normally works.
Finally, the method for diagnosing faults based on fault-test correlation matrix that the present invention proposes, the most directly depend on
Connect root node and the leaf node of two-layer Bayesian network according to element value each in correlation matrix D, utilize the most in step 4
Conditional probability table has carried out quantitative expression to the bonding strength of root node in step 3 Yu leaf node, overcomes existing for event
In the Bayesian network of barrier diagnosis, between failure classes node and test class node, incidence relation is difficult to the difficulty directly given, and can promote
The method for diagnosing faults based on Bayesian network extensive application in engineering, can assist the intellectuality realizing fault diagnosis, carry
High diagnosis efficiency, has reference to exploitation system fault diagnosis instrument.
Claims (4)
1. a method for diagnosing faults based on fault-test correlation matrix, it is characterised in that: described method includes following step
Rapid:
1) system failure-test correlation matrix D is determined by the dependency graph representation model of system；Wherein, in matrix D, every a line
The fault of the only element in correspondence system, the unique test item in every string correspondence system；That is, for there being m component units,
The system of n test item, correlation matrix D has m row n column data；
2) Bayesian network node is set up；The fault being followed successively by system component units corresponding to each row in correlation matrix D is set up only
One corresponding node, as the root node of Bayesian network；The test item foundation being followed successively by correlation matrix D each row corresponding is the most right
Answer node, as the leaf node of Bayesian network；
3) according to element value Connection Step 2 each in matrix D) in two-layer Bayesian network node；The operating procedure of line is such as
Under:
The i-th row jth column element d for matrix Dij(1≤i≤m, 1≤j≤n), if dij=1, then by fault FiIn step 2) in
Corresponding Bayesian network root node line points to test item TjIn step 2) in corresponding Bayesian network leaf node；If dij=
0, do not carry out any line operation；
Traversal i and j, carries out described line operation successively to element each in matrix D；
4) for representing that the Bayesian network root node of fault arranges conditional probability table；
5) for representing that the Bayesian network leaf node of test item arranges conditional probability table；If leaf node Tj(1≤j≤n) has k father
Node, then father node combinations of states has 2kThe situation of kind；And when to have any node state in k father node be 1, leaf node TjShape
State be 1 probability be set to 1.0；When k father node state that and if only if is 0, leaf node TjState be 1 probability be set to
0.0；
6) evidence variable is set according to the test result of each test item: for testing the leaf node T passed throughjState T is setj=0,
Leaf node T unsanctioned for test resultjState T is setj=1,1≤j≤n；
7) diagnostic result is obtained；The method of described acquisition diagnostic result is as follows:
Calculate Bayesian network posterior probability Pr (Fi=1 | T=t) (1≤i≤m, Fi∈ F), according to posterior probability result of calculation from
Big to little to fault FiIt is ranked up, according to FiRanking results carries out troubleshooting successively to system.
Method for diagnosing faults based on fault-test correlation matrix the most according to claim 1, it is characterised in that: described
Step 1) determine comprising the concrete steps that of the system failure-test correlation matrix D by the dependency graph representation model of system:
1.1] for having m unit, the system of n test item, set up two-dimensional correlation matrix D, matrix D size for m row n
Row, the fault of the unique component units in each row correspondence system dependency graph representation model, in each row correspondence system correlation models
Unique test item；Initializing each element in D is 0；
1.2] i=1 is made, wherein, i representing matrix D the i-th row data, FiThe system component units that representing matrix D the i-th row data are corresponding
Fault；
1.3] in dependency graph representation model, all square frames and circle are regarded as node, all node states are set and are 0, i.e.
Put Fi=0 and Tj=0；
1.4] in dependency graph representation model, at numbered FiNode in place a token, put numbered FiNode state
Be 1, i.e. Fi=1, and search for, by searching algorithm, whole nodes that this token can arrive along directed line；
1.5] in dependency graph representation model, if node FiThe node that can arrive is the circle representing test item, then determine
The test item represented by this circle respective column j in matrix D, and in matrix D, put dij=1；
1.6] to i from adding 1:i=i+1；If i < m+1, then forward step 1.3 to]；Otherwise, in matrix D, each element determines complete, calculates
Method terminates.
Method for diagnosing faults based on fault-test correlation matrix the most according to claim 2, it is characterised in that: step
1.4] searching algorithm specifically includes following steps:
1.4.1] an empty linear list L is set up；
1.4.2] with token place node as starting point, point to along directed line segment and determine whether the adjacent node of starting point meets two bars
Part: the adjacent node state of starting point is 0, the adjacent node of starting point is not present in linear list L；For meeting the starting point of two conditions
Adjacent joint be stored in linear list L；
1.4.3] if linear list L is empty, reject the token in dependency graph representation model, terminate searching algorithm；Otherwise, relevant
In property figure representation model, token is moved in the node that linear list L end data represents, the joint that linear list L end data is represented
The state of point changes 1 into from 0, deletes linear list L end data, forwards step 1.4.2 to].
Method for diagnosing faults based on fault-test correlation matrix the most according to claim 3, it is characterised in that: described
Step 4) if fault FiProbability of happening be Pi, then it represents that fault FiNode condition probability tables parameter be provided that
Root node Fi Pr(Fi)
0 1.0-Pi
1 Pi。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610629637.6A CN106250631B (en) | 2016-08-03 | 2016-08-03 | It is a kind of based on failure-test correlation matrix method for diagnosing faults |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610629637.6A CN106250631B (en) | 2016-08-03 | 2016-08-03 | It is a kind of based on failure-test correlation matrix method for diagnosing faults |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106250631A true CN106250631A (en) | 2016-12-21 |
| CN106250631B CN106250631B (en) | 2019-03-08 |
Family
ID=57606667
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610629637.6A Expired - Fee Related CN106250631B (en) | 2016-08-03 | 2016-08-03 | It is a kind of based on failure-test correlation matrix method for diagnosing faults |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106250631B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108375742A (en) * | 2018-02-05 | 2018-08-07 | 哈尔滨工业大学 | A kind of flyback sourse and test method with testable parameter and fault diagnosis functions |
| CN109298706A (en) * | 2018-11-02 | 2019-02-01 | 中国航空工业集团公司西安飞机设计研究所 | A kind of flight control system method for diagnosing faults based on Bayesian network |
| CN109872004A (en) * | 2019-03-08 | 2019-06-11 | 北京工商大学 | A kind of equipment fault prediction based on fuzzy Bayesian network and health evaluating method |
| CN110479636A (en) * | 2019-07-19 | 2019-11-22 | 深圳市微蓝智能科技有限公司 | Method and device based on neural network automatic sorting tobacco leaf |
| CN110989558A (en) * | 2019-12-19 | 2020-04-10 | 浙江中控技术股份有限公司 | Fault diagnosis item processing method and system |
| CN111027721A (en) * | 2019-11-22 | 2020-04-17 | 中国铁道科学研究院集团有限公司通信信号研究所 | System fault positioning method |
| CN111061293A (en) * | 2019-11-12 | 2020-04-24 | 中国运载火箭技术研究院 | Multi-parameter coupled aircraft fault positioning method, aircraft and storage medium |
| CN113962394A (en) * | 2021-09-10 | 2022-01-21 | 重庆邮电大学 | Automatic generation method of correlation model based fault diagnosis inference machine |
| CN114978876A (en) * | 2022-04-21 | 2022-08-30 | 北京航空航天大学 | Function test-oriented Bayesian fault diagnosis method for civil aircraft communication system |
| CN117061332A (en) * | 2023-10-11 | 2023-11-14 | 中国人民解放军国防科技大学 | Fault diagnosis method and system based on probability directed graph deep learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050160324A1 (en) * | 2003-12-24 | 2005-07-21 | The Boeing Company, A Delaware Corporation | Automatic generation of baysian diagnostics from fault trees |
| CN101814114A (en) * | 2010-01-07 | 2010-08-25 | 北京航空航天大学 | Method for performing fault diagnosis by using model conversion |
| CN102184291A (en) * | 2011-05-06 | 2011-09-14 | 北京航空航天大学 | System level fault diagnosis method for full-test non-feedback system |
| CN106342306B (en) * | 2011-06-24 | 2013-01-16 | 中国人民解放军国防科学技术大学 | Product test index selection method in undetected situation |
-
2016
- 2016-08-03 CN CN201610629637.6A patent/CN106250631B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20050160324A1 (en) * | 2003-12-24 | 2005-07-21 | The Boeing Company, A Delaware Corporation | Automatic generation of baysian diagnostics from fault trees |
| CN101814114A (en) * | 2010-01-07 | 2010-08-25 | 北京航空航天大学 | Method for performing fault diagnosis by using model conversion |
| CN102184291A (en) * | 2011-05-06 | 2011-09-14 | 北京航空航天大学 | System level fault diagnosis method for full-test non-feedback system |
| CN106342306B (en) * | 2011-06-24 | 2013-01-16 | 中国人民解放军国防科学技术大学 | Product test index selection method in undetected situation |
Non-Patent Citations (1)
| Title |
|---|
| 贾兴亮: "基于贝叶斯网络的雷达可靠性评估", 《四川兵工学报》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108375742A (en) * | 2018-02-05 | 2018-08-07 | 哈尔滨工业大学 | A kind of flyback sourse and test method with testable parameter and fault diagnosis functions |
| CN109298706A (en) * | 2018-11-02 | 2019-02-01 | 中国航空工业集团公司西安飞机设计研究所 | A kind of flight control system method for diagnosing faults based on Bayesian network |
| CN109872004A (en) * | 2019-03-08 | 2019-06-11 | 北京工商大学 | A kind of equipment fault prediction based on fuzzy Bayesian network and health evaluating method |
| CN109872004B (en) * | 2019-03-08 | 2021-06-04 | 北京工商大学 | Equipment fault prediction and health assessment method based on fuzzy Bayesian network |
| CN110479636A (en) * | 2019-07-19 | 2019-11-22 | 深圳市微蓝智能科技有限公司 | Method and device based on neural network automatic sorting tobacco leaf |
| CN111061293A (en) * | 2019-11-12 | 2020-04-24 | 中国运载火箭技术研究院 | Multi-parameter coupled aircraft fault positioning method, aircraft and storage medium |
| CN111061293B (en) * | 2019-11-12 | 2023-11-17 | 中国运载火箭技术研究院 | Multi-parameter coupled aircraft fault positioning method, aircraft and storage medium |
| CN111027721A (en) * | 2019-11-22 | 2020-04-17 | 中国铁道科学研究院集团有限公司通信信号研究所 | System fault positioning method |
| CN111027721B (en) * | 2019-11-22 | 2023-07-18 | 中国铁道科学研究院集团有限公司通信信号研究所 | System fault positioning method |
| CN110989558B (en) * | 2019-12-19 | 2021-03-19 | 浙江中控技术股份有限公司 | Fault diagnosis item processing method and system |
| CN110989558A (en) * | 2019-12-19 | 2020-04-10 | 浙江中控技术股份有限公司 | Fault diagnosis item processing method and system |
| CN113962394A (en) * | 2021-09-10 | 2022-01-21 | 重庆邮电大学 | Automatic generation method of correlation model based fault diagnosis inference machine |
| CN114978876A (en) * | 2022-04-21 | 2022-08-30 | 北京航空航天大学 | Function test-oriented Bayesian fault diagnosis method for civil aircraft communication system |
| CN117061332A (en) * | 2023-10-11 | 2023-11-14 | 中国人民解放军国防科技大学 | Fault diagnosis method and system based on probability directed graph deep learning |
| CN117061332B (en) * | 2023-10-11 | 2023-12-29 | 中国人民解放军国防科技大学 | Fault diagnosis method and system based on probability directed graph deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106250631B (en) | 2019-03-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106250631A (en) | A kind of method for diagnosing faults based on fault test correlation matrix | |
| Contant et al. | Diagnosability of discrete event systems with modular structure | |
| CN103235881B (en) | A kind of nuclear reactor fault monitoring system based on minimal cut set | |
| CN107272646A (en) | Forcing press fault diagnosis system based on expert system | |
| CN104123407B (en) | A kind of testability model auto-creating method based on circuit simulation | |
| CN105629156A (en) | Genetic programming-based analog circuit fault test optimal sequential search method | |
| CN106650942A (en) | Fault diagnosis method based on electronic equipment testability model | |
| CN102662831B (en) | Method of diagnosis strategy optimization for fault tolerant system | |
| CN104361169A (en) | Method for monitoring reliability of modeling based on decomposition method | |
| CN113361136B (en) | Diagnostic verification method for automatic manufacturing system based on Petri network | |
| CN107844641A (en) | A kind of Reliability Modeling being combined based on failure mechanism tree and fault tree | |
| CN106291339A (en) | A kind of circuit breaker failure diagnostic expert system based on artificial neural network | |
| CN110209118A (en) | With the robustness control method based on Petri network for using multiple types multi-quantity resource and flexible path feature | |
| Fotuhi-Firuzabad et al. | A novel approach to determine minimal tie-sets of complex network | |
| Su et al. | Diagnosability of Discrete-Event Systems with Uncertain Observations. | |
| CN107490758B (en) | Modularization BP neural network circuit failure diagnosis method based on fault propagation | |
| CN102736616A (en) | Dulmage-Mendelsohn (DM)-decomposition-based measuring point optimal configuration method for closed loop system | |
| CN111190759A (en) | Hybrid diagnosis strategy generation method based on function and fault mode | |
| Kolcio et al. | Model-based approach to rover health assessment for increased productivity | |
| CN104515950B (en) | A kind of build-in self-test method of integrated circuit and application | |
| CN109581194A (en) | A kind of electronic system malfunction Test Strategy dynamic creation method | |
| Bobalo et al. | Algorithms of automated formulation of the operability condition of complex technical systems | |
| US7739213B1 (en) | Method for developing complex probabilistic models | |
| Li et al. | Polynomial time verification of modular diagnosability of discrete event systems | |
| CN113052431A (en) | Effectiveness evaluation method of aerospace test and launch system based on ADCP model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190308 Termination date: 20190803 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |