CN114692495A - Efficient complex system reliability evaluation method based on reliability block diagram - Google Patents

Efficient complex system reliability evaluation method based on reliability block diagram Download PDF

Info

Publication number
CN114692495A
CN114692495A CN202210290231.5A CN202210290231A CN114692495A CN 114692495 A CN114692495 A CN 114692495A CN 202210290231 A CN202210290231 A CN 202210290231A CN 114692495 A CN114692495 A CN 114692495A
Authority
CN
China
Prior art keywords
node
turning
model
stack
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.)
Pending
Application number
CN202210290231.5A
Other languages
Chinese (zh)
Inventor
李志峰
任羿
杨德真
王自力
冯强
孙博
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beihang University
Original Assignee
Beihang University
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 Beihang University filed Critical Beihang University
Priority to CN202210290231.5A priority Critical patent/CN114692495A/en
Publication of CN114692495A publication Critical patent/CN114692495A/en
Pending legal-status Critical Current

Links

Images

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
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Computational Mathematics (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Algebra (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • Computer Hardware Design (AREA)
  • Geometry (AREA)
  • Computational Linguistics (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The invention provides a method for efficiently solving a reliability block model based on a structure identification method and a Bayesian network. Aims and solves the problems that: and the modeling capability and the calculation efficiency of the reliability block diagram model of the large-scale complex system are improved. The method comprises the steps of firstly identifying basic structures in a reliability block diagram model, wherein the structures comprise series connection, parallel connection, voting and connection, taking the basic structures as modules, and repeating the identification process to finally form the reliability block diagram model with a hierarchical structure; secondly, traversing the hierarchical reliability block diagram model from top to bottom until reaching the bottommost module, converting the modules into corresponding Bayesian networks according to the structure types, backtracking upwards, and converting layer by layer to form a hierarchical Bayesian network model; and finally, obtaining the probability distribution of all nodes in the Bayesian network by using a Bayesian inference algorithm, wherein the probability distribution of leaf nodes of the Bayesian network is the failure probability distribution of the system, and the probability distribution of middle nodes is the failure probability distribution of different modules.

Description

Efficient complex system reliability evaluation method based on reliability block diagram
Technical Field
The invention provides a high-efficiency resolving method for a reliability block diagram of a large-scale complex system. It is suitable for reliability evaluation of complex systems. The invention belongs to the technical field of system reliability evaluation.
Background
The reliability block diagram model is a model commonly used in system reliability modeling, and is widely applied to the fields of aviation, aerospace, ships and the like. The structure of the reliability block model comprises series connection, parallel connection, voting and parallel connection. In addition, the GJB813-1900 has expanded the original reliability block model and proposed a model representing the task reliability of the multifunctional system. The reliability block diagram model represents the logical relationship between the system and the components, so that the reliability of the system can be calculated according to the model structure after the reliability of the components is obtained.
Currently, common reliability block model calculation methods include a total probability-based method, a way set-based method, a simulation-based method, a fault tree-based method, a binary decision diagram-based method, and a bayesian network-based method. The full probability based method is difficult to apply to large scale systems and difficult to program; the way set based approach fails to solve and join structure and multi-functional models; simulation-based methods generally fail to obtain accurate results and require long time to solve; the method based on the fault tree is a method for converting a reliability block diagram into a fault tree model and then utilizing the fault tree model to solve the fault tree, while the current most efficient method for solving the fault tree is a method based on a binary decision diagram, but the method based on the binary decision diagram cannot process and connect structures; the method based on the Bayesian network generally converts a reliability block diagram model into a three-layer Bayesian network, wherein the bottommost node is all nodes forming a path in the reliability block diagram, the middle node represents a path of the reliability block diagram, and the top node represents a system; the method is difficult to solve the reliability evaluation of a large-scale system, the difference between a converted model and an original reliability block model is large, and the existing literature has no Bayesian network-based solution and union and multifunctional model method.
The invention provides an efficient solving method of a reliability block diagram model of a large-scale complex system based on a structure recognition algorithm and a Bayesian network. The method can solve the reliability block model of the structure including series-parallel connection, voting, connection and the like, can also solve the multifunctional model, has greatly improved calculation efficiency compared with the original Bayes method, and can be used for evaluating the reliability of a large-scale complex system.
Disclosure of Invention
The invention provides a method for efficiently solving a reliability block model based on a structure identification method and a Bayesian network. Aims and solves the problems that: and the modeling capability and the calculation efficiency of the reliability block diagram model of the large-scale complex system are improved. The method comprises the steps of firstly identifying basic structures in a reliability block diagram model, wherein the structures comprise series connection, parallel connection, voting and connection, taking the basic structures as modules, and repeating the identification process to finally form the reliability block diagram model with a hierarchical structure; secondly, traversing the hierarchical reliability block model from top to bottom until reaching the bottommost module, converting the modules into corresponding Bayesian networks according to the structure types, backtracking upwards, and converting layer by layer to form a hierarchical Bayesian network model; and finally, obtaining the probability distribution of all nodes in the Bayesian network by using a Bayesian inference algorithm, wherein the probability distribution of leaf nodes of the Bayesian network is the failure probability distribution of the system, and the probability distribution of middle nodes is the failure probability distribution of different modules.
The invention relates to a method for efficiently solving a reliability block diagram model based on a structure identification method and a Bayesian network, which comprises the following three parts:
a first part: and identifying the reliability block diagram model structure.
The structure recognition of the reliability block model is the basis for developing the reliability block model evaluation of the large-scale complex system, and the basic composition modules and the hierarchical reliability block model of the model are obtained through the structure recognition. The structure identification process is as follows:
step 1: initializing a current position node P, assigning the current position node P as the end point of the reliability block diagram model, initializing an empty list, and turning to the step 2 if the node P is not traced back; otherwise, the process is exited, and the model after structure recognition and layering is returned;
step 2: accessing the node P, judging whether the node P has a precursor node, if not, turning to the step 3, and if so, turning to the step 4:
and step 3: judging whether the list is empty or not, if so, backtracking to the step before the step 2 is carried out; otherwise, turning to the step 5;
and 4, step 4: if the number of the precursor nodes of the node P is equal to 1, turning to the step 6; otherwise, go to step 7
And 5: and traversing the elements in the list, wherein the elements are of a stack structure, and model nodes pressed in the traversal reliability block diagram model are stored in the stack. For each stack, extracting a stack top element, if the stack top element is a successor node of the current node P and a predecessor node of the stack top element is P, removing the whole stack from the list, identifying the nodes in the stack as a serial structure, and backtracking to the step before the step 2; if no such stack can be found, then a trace is made back to the step before proceeding to step 2.
Step 6: if the node with the same predecessor as the node P exists, turning to the step 8, otherwise, turning to the step 9;
and 7: judging whether all precursor nodes of the current node have common precursor nodes, if not, traversing the precursor nodes of P, and turning to the step 12; if yes, go to step 13;
and 8: if the list is empty, backtracking; if not, traversing the stack structure in the list, taking out the stack top element, judging whether the stack top element and the current node P are in a precursor successor relationship, if so, pressing the node P into the stack, removing the stack from the list, then processing the elements in the stack, after finishing, assigning P as a precursor node, and turning to the step 2;
and step 9: if the stack top element S and the node P exist in the list as a stack of a precursor successor relationship, turning to the step 10, otherwise, turning to the step 11;
step 10: pressing the node P into a stack, assigning P as a precursor node of the node P, and then turning to the step 2;
step 11: building a stack, pressing the node P into the stack, adding the newly built stack into the list, assigning P as a precursor node of P, and then turning to the step 2;
step 12: whether the precursor nodes of P are traversed or not is judged, if not, one precursor node is selected from the non-traversed precursor nodes of P and assigned to P, and then the step 2 is carried out; if yes, directly turning to the step 2;
step 13: processing the node set with the same predecessor successor node into a new module, wherein the processing flow is as follows: and newly building a virtual node V, taking the precursor nodes of P as child nodes of V, and setting the type of the node V according to the structural type formed by the node P and the precursor nodes thereof. The node type of P is 'AND JOIN', 'voting' and 'PARALLEL', if the node type of P is 'AND JOIN' or 'voting', the node type is set according to the type, otherwise, the node type is set to be PARALLEL, and the predecessor successor relationship between the virtual node V and the node P and the successive relationship between the virtual node and the original predecessor node of P are reestablished. After finishing, turning to the step 2;
a second part: and (4) converting the hierarchical reliability block diagram model into a Bayesian network model.
The Bayesian network is a calculation tool for quantitative analysis of the reliability block model, so that after the hierarchical reliability block model is constructed, the reliability block model needs to be converted into the Bayesian network model, and the process is to convert the basic structure forming the reliability block model into the corresponding Bayesian network from bottom to top, and the method comprises the following five steps:
step 1: traversing the hierarchical reliability block diagram model from top to bottom, judging whether the current unit is converted, if so, backtracking, otherwise, turning to the step 2;
step 2: judging whether the current unit has a lower layer unit, if so, turning to the step 3, and if not, turning to the step 4;
and step 3: turning to a lower model of the unit, traversing the unit of the lower model to serve as a current unit, turning to the step 2, backtracking after the conversion of all nodes in the lower model is completed, and turning to the step 5;
and 4, step 4: the model unit is converted into a node in the Bayesian network model, and the state of the Bayesian network node and the corresponding probability distribution are obtained according to the state and the reliability R of the unit. When the state "0" represents that the unit is failed, and the state "1" represents that the unit works normally, the probability distribution of the Bayes nodes is as follows:
p (unit state 0) ═ l-R
p (unit state 1) ═ R
After the conversion is finished, turning to the step 3;
and 5: generating a Bayesian network model corresponding to the reliability block diagram unit with the lower model according to the unit type and the following conversion rule, marking the reliability block diagram unit as converted after the conversion is finished, and turning to the step 3;
a) series connection structure
Each node in the series structure acts as a root node of the bayesian network and the entire series structure is represented as a leaf node of the bayesian network. An example of a Bayesian network model corresponding to a two-cell series configuration is shown in FIG. 1.
b) Parallel structure
Each node in the parallel structure is used as a root node of the Bayesian network, and the whole parallel structure is a leaf node of the Bayesian network. An example of a bayesian network model corresponding to a two-unit parallel configuration is shown in fig. 2.
c) A voting structure;
each node in the voting structure is used as a root node of the Bayesian network, and the whole voting structure is a leaf node of the Bayesian network. An example of a Bayesian network model corresponding to the 2-out-of-3 voting structure is shown in FIG. 3.
d) And a coupling structure
Each unit in the union structure is used as a root node of the Bayesian network, and a virtual root node is added to represent the weight of each unit in the union structure, wherein the probability distribution of the virtual root node is the weight value of the unit in the union structure. An example of a Bayesian network model for which the two-element and join structure corresponds is shown in FIG. 4.
And a third part: and (4) evaluating the reliability of the system based on a Bayesian network inference process.
The inference process of the Bayesian network model is carried out to obtain the probability distribution of each node in the Bayesian network model, and the probability value representing the normal state of the system and the unit in the node in the Bayesian network model is obtained, wherein the probability value representing the normal state of the system is the reliability of the complex system.
Drawings
FIG. 1 Bayesian network model corresponding to the series model
FIG. 2 Bayesian network model corresponding to parallel model
Bayesian network model corresponding to voting model in FIG. 3
FIG. 4 is a Bayesian network model corresponding to a connected model
FIG. 5 is a flow chart of a method for efficiently calculating reliability of a complex system
FIG. 6 is an example of a reliability block diagram for a complex system
FIG. 7 is a block diagram illustrating a complex system reliability block diagram identification process and a hierarchical process schematic
FIG. 8 is a partial process schematic and final result of conversion of a hierarchical reliability block model into a Bayesian network
Detailed Description
Description of the embodiments: a method for efficiently solving a reliability block model based on a structure identification method and a bayesian network is provided, a complete flow of the method is shown in fig. 5, and a specific implementation manner is described as follows:
a first part: and identifying the reliability block diagram model structure.
The reliability block diagram model after the complex system modeling is shown as 6, wherein the reliability of the units is 0.9. Wherein the node X3 is represented as a union model, the weight is shown as a numerical value on a connecting line, the node X4 is represented as a model of 2 in 3, and the construction process comprises the following steps:
step 1: the current location node P is initialized to X5, and an empty list is initialized { }.
Step 2: p is X5, the precursor node is X4, go to step 4;
and 4, step 4: if the precursor node of P is only X4, turning to step 6;
step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: if the current list is empty, turning to step 11;
step 11: newly building a Stack1<>If X5 is added to the Stack, then there is Stack1<X5>If the Stack is added to list, there is list { Stack1<X5>Step 2, changing P to X4;
step 2: p is X4, and its predecessor nodes are: G. h, I, go to step 4;
and 4, step 4: the number of the precursor nodes of P is not 1, and the step is 7;
and 7: the predecessor nodes of unit G, H, I are all F, go to step 13;
step 13: newly building a virtual module V1Unit G, H, I is regarded as V1And marking the child nodes as voting models, and marking V1Setting the predecessor node and successor node of (1) as F and X4 respectively, and turning to step 2;
step 2: p is still X4 at this time, its precursor is V1Turning to the step 4;
and 4, step 4: the number of the precursor nodes of P is 1, and turning to the step 6;
and 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: if the list is not empty, traversing the stack in the list, taking X5 and X4 as predecessor successor relations, and turning to step 10;
step 10: adding X4 to the Stack, there is Stack1<X5,X4>Let P equal V1(ii) a Turning to the step 2;
step 2: access node P, V1The precursor node of (1) is F, turning to step 4;
and 4, step 4: v1The number of the precursor nodes is 1, and step 6 is carried out;
step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: the subsequent relation that the stack top element X4 and the node V1 are predecessors exists, and the step 10 is carried out;
step 10: pushing the node V1 into a Stack1, and turning to the step 2 when P is equal to F;
step 2: the precursor node of P is X3, and step 4 is switched;
and 4, step 4: the number of the precursor nodes of P is 1, and step 6 is carried out;
step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: nodes V1 and F have predecessor successor relationship, go to step 10;
step 10: pressing P into Stack1, if P is Stack1< X5, X4, V1, F >, making P equal to X3, and turning to step 2;
step 2: current P is X3 with its predecessor node C, D, E; turning to the step 4;
and 4, step 4: p, turning to step 7 if the number of the predecessors is not 1;
and 7: all predecessor nodes of the current node do not have a common predecessor node, and then the step 12 is carried out;
step 12: turning to step 2 when P is equal to C;
step 2: when the current P is a unit C, the precursor node of the current P is X2, and the step 4 is switched to:
and 4, step 4: the number of the precursor nodes of P is 1, and step 6 is carried out;
and 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: the Stack in the list is Stack1, no Stack top element having a successor relationship with the P predecessor exists, and the step 11 is carried out;
step 11: newly building a Stack2, pressing a node P into the Stack to obtain Stack2< C >, updating list { Stack1< X5, X4, V1, F >, Stack2< C > }, making P equal to X2, and going to step 2;
step 2: accessing a node P, wherein the precursor node of the node P is A, B, and turning to the step 4;
and 4, step 4: p, the number of the precursor nodes is 2, and the step 7 is switched;
and 7: A. b has the same predecessor node X1, go to step 13;
step 13: newly building a virtual node V2, setting the attribute of V2 in parallel, adding A, B to the child node of V2, updating the predecessor and successor nodes of V2 to be X1 and X2 respectively, and turning to step 2;
step 2: p, the precursor node is V2, go to step 4;
and 4, step 4: the number of the precursor nodes of P is 1, and turning to step 6;
step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: c and P in Stack2< C > have a predecessor successor relationship, go to step 10;
step 10: adding P into Stack2, with Stack2< C, X2>, changing P to V2, and turning to step 2;
step 2: accessing the node V2, wherein the precursor node of the V2 is X1, and turning to the step 4;
and 4, step 4: the number of the precursor nodes of V2 is 1, and step 6 is switched;
step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: the Stack2 with the Stack top element and the node P as the precursor successor relationship exists, and the step 10 is carried out;
step 10: adding P into the station Stack2, if P is Stack2< C, X2, V2>, making P equal to X1, and turning to step 2;
step 2: accessing the node X1, wherein the predecessor node is X0, and turning to step 4;
and 4, step 4: the number of the precursor nodes of P is 1, and turning to the step 6;
step 6: there is node D, E with the same predecessor as P, go to step 8;
and 8: if the list is not empty, the top element of Stack2 has a predecessor successor relationship with P, Stack2 is updated, if there is Stack2< C, X2, V2, X1>, and it is removed from the list, and at the same time, a new virtual unit V3 is generated, C, X2, V2, X1 are used as its lower node, the type is set to be a serial structure, if there is a list { Stack1}, let P be X0, and go to step 2;
step 2: x0 has no precursor node, and go to step 3;
and step 3: if the list is not empty, turning to the step 5;
and 5: if the stack meeting the condition cannot be found, backtracking to the child node of the access node X3, and turning to step 12;
step 12: accessing the node D, and turning to the step 2;
step 2: p has a predecessor node X0, go to step 4;
and 4, step 4: the number of the precursor nodes of P is 1, and turning to the step 6;
and 6: there is a node E, V3 with the same predecessor as P; turning to step 8;
and 8: if the list is not empty, the condition that the stack top element and the node P have a precursor successor relationship does not exist, and the step 2 is switched to when the P is X0;
and 2, step: p has no front-driving node, and step 3 is switched;
and step 3: if the list is not empty, backtracking to step 12, and continuing to access the node E; e is the same as the processing of D, and is not repeated again, the node X3 is directly traced back, and step 2 is performed when P is X3;
step 2: the precursor node of X3 is V3, D, E, go to step 4;
and 4, step 4: the number of the node P predecessors is 3, and the step 7 is carried out;
and 7: p has the same precursor node X0, go to step 13;
step 13: newly building a virtual node V4, taking V3 and V D, E as child nodes of V4, setting the type of V4 as a union, updating predecessor and successor nodes of V4 as X0 and X3, and turning to step 2;
step 2: p is X3, the precursor node of P is V4, go to step 4;
and 4, step 4: the number of the precursor nodes of P is 1, and turning to the step 6;
step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: f and P in Stack1 are precursor successor relations, and go to step 10;
step 10: pressing P into a Stack1 to obtain Stack1< X5, X4, V1, F, X3>, enabling P to be V4, and turning to the step 2;
and 2, step: accessing the node V4, wherein the precursor node of the V4 is X0, and turning to the step 4;
and 4, step 4: the number of the precursor nodes of the node V4 is 1, and then the step 6 is switched;
step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: the top Stack element X3 and V4 with the Stack1 having a predecessor successor relationship, go to step 10;
step 10: pressing V4 into a Stack1 to obtain Stack1< X5, X4, V1, F, X3 and V4>, enabling P to be X0, and turning to step 2;
step 2: x0 has no precursor node, and go to step 3;
and step 3: if the list is not empty, turning to the step 5;
and 5: stack1 is removed and used as a series structure to generate a virtual cell V5, X5, X4, V1, F, X3, V4 as its lower child node, the predecessor node of V5 is X0, and the successor node is empty. Turning to step 2 when P is V5;
step 2: the precursor node of V5 is X0, go to step 4;
and 4, step 4: the number of the precursor nodes of V5 is 1, and step 6 is switched to
Step 6: if no node with the same predecessor as P exists, turning to step 9;
and step 9: if the list is empty, turning to the step 11;
step 11: newly building a Stack3< >, adding P, wherein the number of P is Stack3< V5>, updating list, wherein the list is { Stack3< V5 }, making P be X0, and turning to the step 2;
step 2: the node X0 has no front-driving node, and the step 3 is switched;
and step 3: if the list is not empty, turning to the step 5;
and 5: the entire structure recognition process can be terminated by removing Stack3, converting element V5 in the Stack to a serial structure, creating a virtual cell V6, having V5 and X0 as its child nodes, and updating the predecessor successor nodes of V6 to null.
The reliability block model shown in fig. 6 is shown in fig. 7 according to the above steps of the recognition process and the finally formed hierarchical reliability block model.
A second part: and (4) converting the hierarchical reliability block diagram model into a Bayesian network model.
The hierarchical reliability block model is shown in fig. 7.
Step 1: traversing a hierarchical reliability block diagram model from top to bottom, firstly accessing a node V6, and turning to the step 2 if the node V6 is not converted;
step 2: the lower level nodes of V6 include: v4, F, V1, X0, X3, X4 and X5, and turning to step 3;
and step 3: entering a lower layer model, taking V4 as a current access node, and turning to step 2;
step 2: the lower level nodes of V4 include: v3, D, E, go to step 3;
and step 3: entering a lower layer model of V4, taking V3 as a currently accessed node, and turning to step 2;
step 2: the lower level nodes of V3 include: v2, C, X1 and X2, turning to step 3;
and step 3: entering a lower layer model of V3, taking V2 as a currently accessed node, and turning to step 2;
step 2: the lower level nodes of V2 include: A. b, turning to the step 3;
and step 3: entering a lower layer model, then taking A as an access node, and turning to the step 2;
step 2: step A, turning to step 4 without a lower node;
and 4, step 4: generating a bayesian network node corresponding to the node a, knowing that the reliability of the unit a is 0.9, knowing that p (the state of the unit a is 0) is 0.1 and p (the state of the unit a is 1) is 0.9, after the conversion is completed, backtracking, and turning to step 3;
and step 3: continuing to access the unit B, and turning to the step 2;
step 2: the unit B has no lower node and the step 4 is switched;
and 4, step 4: the conversion unit B is a node of the bayesian network model, and if the reliability of the unit B is known to be 0.9, it is known that p (the state of the unit B is 0) is 0.1 and p (the state of the unit B is 1) is 0.9, after the conversion is completed, backtracking is performed, and the step 3 is returned;
and step 3: completing traversing of the sub-model of the cell V2, and turning to the step 5;
and 5: the type of the unit V2 is a parallel structure, conversion is carried out according to the rule of the parallel structure, after the conversion is finished, the V2 is marked as converted, and the step 3 is carried out;
and step 3: continuing to access the subunit of V3, accessing the unit C, and turning to the step 2;
step 2: the unit C has no lower layer unit, and the step 4 is switched;
and 4, step 4: the conversion unit C is a Bayesian network node and converts to the step 3;
and step 3: continuously traversing other sub-units of the V3, wherein the unit X1 and the unit X2 are auxiliary modeling units which have no influence on the evaluation result of the model, so that the conversion of the nodes is directly ignored, and the upper-layer model is returned to the step 5 after the sub-units of the V3 are completely traversed;
and 5: the type of the unit V3 is a series structure, a corresponding Bayesian network model is generated according to a conversion rule of the series structure, and then the step 3 is switched;
the bayesian network model corresponding to the hierarchical reliability block diagram model is obtained according to the above method, as shown in fig. 8 (6).
And a third part: and (4) evaluating the reliability of the system based on a Bayesian network inference process.
The inference process of the Bayesian network model is carried out to obtain the probability distribution of each node in the Bayesian network model, the probability value of the normal state of the representation system and the unit in the node in the Bayesian network model is obtained, the probability distribution of the node W representing the weight in the model is shown in the table 1, and the probability distribution of each node is shown in the table 2.
TABLE 1 weight distribution of nodes W
Node name V3 weight D weight E weight
W 0.2 0.3 0.5
TABLE 2 State probability distribution of nodes in the model
Figure BDA0003559676600000101
Figure BDA0003559676600000111
When the state probability distribution of V6 is the probability distribution of the system, it is found that the reliability of the system is 0.786

Claims (3)

1. A method for efficiently solving a reliability block diagram model based on a structure identification method and a Bayesian network is characterized by comprising the following two components:
(1) identifying a reliability block diagram model structure;
(2) and converting the hierarchical reliability block diagram model into a Bayesian network model.
2. The method for efficiently solving the reliability block model based on the structure recognition method and the Bayesian network as recited in claim 1, wherein in the component (1), the structure recognition process is as follows:
step 1: initializing a current position node P, assigning the current position node P as the end point of the reliability block diagram model, initializing an empty list, and turning to the step 2 if the node P is not traced back; otherwise, the process is exited, and the model after structure recognition and layering is returned;
step 2: accessing the node P, judging whether the node P has a precursor node, if not, turning to the step 3, and if so, turning to the step 4:
and step 3: judging whether the list is empty or not, if so, backtracking to the step before the step 2 is carried out; otherwise, turning to the step 5;
and 4, step 4: if the number of the precursor nodes of the node P is equal to 1, turning to the step 6; otherwise, go to step 7
And 5: traversing elements in the list, wherein the elements are of a stack structure, and model nodes pressed in the traversal reliability block diagram model are stored in the stack; for each stack, extracting a stack top element, if the stack top element is a successor node of the current node P and a predecessor node of the stack top element is P, removing the whole stack from the list, identifying the nodes in the stack as a serial structure, and backtracking to the step before the step 2; if the stack cannot be found, backtracking to the step before the step 2 is carried out;
step 6: if the node with the same predecessor as the node P exists, turning to the step 8, otherwise, turning to the step 9;
and 7: judging whether all precursor nodes of the current node have common precursor nodes, if not, traversing the precursor nodes of P, and turning to the step 12; if yes, go to step 13;
and 8: if the list is empty, backtracking; if not, traversing the stack structure in the list, taking out the stack top element, judging whether the stack top element and the current node P are in a precursor successor relationship, if so, pressing the node P into the stack, removing the stack from the list, then processing the elements in the stack, after finishing, assigning P as a precursor node, and turning to the step 2;
and step 9: if the stack top element S and the node P are stacks in a precursor successor relationship in the list, turning to the step 10, otherwise, turning to the step 11;
step 10: pressing the node P into a stack, assigning P as a precursor node of the node P, and then turning to the step 2;
step 11: building a stack, pressing the node P into the stack, adding the newly built stack into the list, assigning P as a precursor node of P, and then turning to the step 2;
step 12: whether the precursor nodes of P are traversed or not is judged, if not, one precursor node is selected from the non-traversed precursor nodes of P and assigned to P, and then the step 2 is carried out; if yes, directly turning to the step 2;
step 13: processing the node set with the same predecessor successor node into a new module, wherein the processing flow is as follows: newly building a virtual node V, taking the precursor nodes of P as child nodes of V, and setting the type of the node V according to the structural type formed by the node P and the precursor nodes thereof; the node type of P is 'AND JOINT', 'voting' and 'PARALLEL', if the node type of P is 'AND JOINT', 'voting' or 'PARALLEL', the node is set according to the type, if not, the node type of P is set to be PARALLEL, the predecessor successor relationship between the virtual node V and the node P and the successive relationship between the virtual node and the original predecessor node of P are reestablished, and the step 2 is completed.
3. The method for efficiently solving the reliability block model based on the structure recognition method and the Bayesian network as recited in claim 1, wherein the transformation process for transforming the basic structure forming the reliability block model into the Bayesian network model from top to bottom in the component (2) comprises the following five steps:
step 1: traversing the hierarchical reliability block diagram model from top to bottom, judging whether the current unit is converted, if so, backtracking, otherwise, turning to the step 2;
step 2: judging whether the current unit has a lower layer unit, if so, turning to the step 3, and if not, turning to the step 4;
and step 3: turning to a lower model of the unit, traversing the unit of the lower model to serve as a current unit, turning to the step 2, backtracking after the conversion of all nodes in the lower model is completed, and turning to the step 5;
and 4, step 4: converting the model unit into a node in the Bayesian network model, and obtaining the state of the Bayesian network node and the corresponding probability distribution thereof according to the state and the reliability R of the unit; the state "0" represents that the unit is failed, and the state "1" represents that the unit is normally operated, then the probability distribution of the Bayesian nodes is as follows:
p (unit state 0) ═ 1-R
p (unit state 1) ═ R
After the conversion is finished, turning to the step 3;
and 5: generating a Bayesian network model corresponding to the reliability block diagram unit with the lower layer model according to the unit type and the following conversion rules, marking the reliability block diagram unit as converted after the conversion is finished, and turning to the step 3;
a) series connection structure
Each node in the series structure is used as a root node of the Bayesian network, and the whole series structure is a leaf node of the Bayesian network;
b) parallel structure
Each node in the parallel structure is used as a root node of the Bayesian network, and the whole parallel structure is a leaf node of the Bayesian network;
c) a voting structure;
each node in the voting structure is used as a root node of the Bayesian network, and the whole voting structure is a leaf node of the Bayesian network;
d) and a coupling structure
Each unit in the union structure is used as a root node of the Bayesian network, and a virtual root node is added to represent the weight of each unit in the union structure, wherein the probability distribution of the virtual root node is the weight value of the unit in the union structure.
CN202210290231.5A 2022-03-23 2022-03-23 Efficient complex system reliability evaluation method based on reliability block diagram Pending CN114692495A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210290231.5A CN114692495A (en) 2022-03-23 2022-03-23 Efficient complex system reliability evaluation method based on reliability block diagram

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210290231.5A CN114692495A (en) 2022-03-23 2022-03-23 Efficient complex system reliability evaluation method based on reliability block diagram

Publications (1)

Publication Number Publication Date
CN114692495A true CN114692495A (en) 2022-07-01

Family

ID=82138426

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210290231.5A Pending CN114692495A (en) 2022-03-23 2022-03-23 Efficient complex system reliability evaluation method based on reliability block diagram

Country Status (1)

Country Link
CN (1) CN114692495A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115408844A (en) * 2022-08-25 2022-11-29 军事科学院系统工程研究院网络信息研究所 Computable high-reliability architecture design method and system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115408844A (en) * 2022-08-25 2022-11-29 军事科学院系统工程研究院网络信息研究所 Computable high-reliability architecture design method and system

Similar Documents

Publication Publication Date Title
CN111462282A (en) Scene graph generation method
Zheng et al. Fault diagnosis system of bridge crane equipment based on fault tree and Bayesian network
CN112685504B (en) Production process-oriented distributed migration chart learning method
CN110765582B (en) Self-organization center K-means microgrid scene division method based on Markov chain
CN115563674B (en) Initial planar arrangement generating method and device, electronic equipment and storage medium
CN104506340A (en) Creation method of decision tree in industrial Ethernet fault diagnosis method
CN115544264B (en) Knowledge-driven intelligent construction method and system for digital twin scene of bridge construction
CN112182306B (en) Uncertain graph-based community discovery method
CN111597943B (en) Table structure identification method based on graph neural network
CN110532291B (en) Method and system for converting deep learning frame model based on minimum execution cost
CN114692495A (en) Efficient complex system reliability evaluation method based on reliability block diagram
CN112217674A (en) Alarm root cause identification method based on causal network mining and graph attention network
CN113343427A (en) Structural topology configuration prediction method based on convolutional neural network
CN114692403A (en) Binary decision diagram-based efficient complex system reliability evaluation method
CN112149949A (en) Satellite communication system efficiency evaluation method based on hesitation fuzzy and analytic hierarchy process
Astanaliev Important principles of innovative reforms in the process of electronic document management in railway automation and telemechanics
CN117010373A (en) Recommendation method for category and group to which asset management data of power equipment belong
CN110517350B (en) Preprocessing method and processor of Monte Carlo particle transport calculation program
CN108198084A (en) A kind of complex network is overlapped community discovery method
CN116432077A (en) Intelligent station secondary system fault positioning method based on optimized BP neural network
CN111275079A (en) Crowdsourcing label speculation method and system based on graph neural network
CN113051408A (en) Sparse knowledge graph reasoning method based on information enhancement
Ping et al. An effective hybrid method for analysis the large-scale reliability block diagram model
CN110889614A (en) Power grid system important user power supply risk analysis method based on SCADA big data
CN109816228A (en) A kind of enterprise B IM technical application ability quantitative evaluation system and method

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