CN103793589A - High-speed train fault handling method - Google Patents
High-speed train fault handling method Download PDFInfo
- Publication number
- CN103793589A CN103793589A CN201210428024.8A CN201210428024A CN103793589A CN 103793589 A CN103793589 A CN 103793589A CN 201210428024 A CN201210428024 A CN 201210428024A CN 103793589 A CN103793589 A CN 103793589A
- Authority
- CN
- China
- Prior art keywords
- fault
- rule
- node
- handling method
- bullet train
- 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
Images
Landscapes
- Test And Diagnosis Of Digital Computers (AREA)
Abstract
The invention relates to a high-speed train fault handling method. The method comprises the steps of 1) querying faults and fault codes in a fault list, which occur during a user set time, to compose a fault set; 2) scanning the fault set to obtain a frequent item set meeting the minimum support level and then obtain a legal fault association rule set; 3) expresses the legal fault association rule set in a rule diagram, and matching the faults in the fault list in the step 1) with the nodes of the rule diagram; 4) traversing the routes to be matched of all the nodes in the matched rule diagram, computing the importance of the faults represented by the nodes on the routes to be matched and performing sequence sorting; 2) according to the importance sequence of the faults to perform treatment. The high-speed train fault handling method can support fault reason analysis under the condition of lack of expert knowledge and meanwhile find out cause and effect in massive faults through rule matching, so that maintenance can be performed preferentially on high-importance source faults, and further the fault maintaining cost of high-speed trains can be greatly reduced.
Description
Technical field
Patent of the present invention relates to fault diagnosis technology, is specifically related to a kind of bullet train failure reason analysis disposal route.
Background technology
The cause-effect relationship of complexity between bullet train network control system fault, to train, maintenance has brought very big difficulty.Therefore, trouble-shooting reason, location root fault, is of great significance saving train maintenance cost tool.
Current failure reason analysis disposal route, conventionally according to expertise, adopts Bayesian network to carry out modeling to the cause-effect relationship between fault, and by solving Bayesian network, obtains possible root fault.Bayesian network is a kind of probability net, and it is the graphical network based on probability inference, and Bayesian formula is the basis of this probability net.Bayesian network is the mathematical model based on probability inference, Bayesian network (Bayesian network) based on probability inference proposes in order to solve ambiguity and imperfection problem, it has advantage very for solving complex apparatus fault uncertain and that relevance causes, in multiple fields, obtains widespread use.
Rule-based reasoning, refers to formal the expertise of association area description out, forms system convention.These Rule Expressions some problems and the corresponding answer of these problems in this field, can utilize them to imitate the related reasoning ability of expert in solving.
The classification of the variable of rule-based middle processing: 1) variable of correlation rule processing can be divided into Boolean type and numeric type.The value of boolean association rule processing is all discrete, kind, and it has shown the relation between these variablees; And numeric type correlation rule can be associated with multidimensional or Multiple-Level Association Rules combine, logarithm value type-word section is processed, cut apart dynamically, or directly original data are processed, certainly in numeric type correlation rule, also can comprise kind of a class variable.2) abstraction hierarchy of rule-based middle data: the abstraction hierarchy of rule-based middle data, can be divided into individual layer correlation rule and Multiple-Level Association Rules.In the correlation rule of individual layer, all variablees all do not consider that real data are to have multiple different levels; And in the correlation rule of multilayer, the multilayer of data has been carried out to sufficient consideration.3) dimension of the data that relate in rule-based: the data in correlation rule, can be divided into one-dimensional with multidimensional.
Adopting Bayesian network to carry out these class methods of modeling to the cause-effect relationship between fault need to be about causal expertise between fault, but in actual applications, the cause-effect relationship between the many faults of bullet train is not known toward contact, does not have expertise to follow.In this case, the existing failure cause disposal route based on Bayesian network is just no longer applicable.How to find between fault still lost causalnexus relation, thereby in the situation that lacking relevant expertise, in a large amount of faults that bullet train is produced, search out root fault, become and have problem to be solved.
Summary of the invention
For deficiency of the prior art, the present invention, by conjunction with association rule mining and rule-based reasoning, analyzes failure cause, greatly saves maintenance cost.
For solving technical matters as above, the invention provides a kind of bullet train failure reason analysis disposal route based on association rule mining and rule-based reasoning, its concrete steps are as follows:
1) fault and the failure code that in inquiry error listing, in user's setting-up time, occur, composition one fault collection;
2) described fault collection is scanned to the Frequent Item Sets that is met minimum support, in described Item Sets, generate fault correlation rule and obtain the legal fault correlation rule set in described fault correlation rule;
3) described legal fault correlation rule set is represented by rule schema, mate the rule schema after being mated according to fault in error listing described in step 1) with the node in this rule schema;
4) in the rule schema after described coupling, traversal finds all nodes all by coupling path; Calculate described by importance the sequence of each node representative fault on coupling path;
5) process from high to low according to fault importance according to described importance ranking.
Importance P (v)=O (the v)+L (v) of described fault, wherein, O (v) is the out-degree of node v in rule schema, L (v) is all by the MAXPATHLEN value of being mated path of node v.
Described fault mode collection is the set of all Frequent Item Sets; Described Frequent Item Sets is the Item Sets that number of times that comprised fault occurs is simultaneously greater than the minimum support that user sets.
Described legal fault correlation rule set comprises some fault correlation rules, and the left minor of described each correlation rule is root fault, and right minor is the fault correlation rule of result fault.
Each node in rule schema after described coupling represents the fault of bullet train, and directed edge sends and points to result fault from root fault, forms a fault correlation rule.
Cause-effect relationship between described fault correlation Rule Expression bullet train fault.
Described frequent item set obtains according to Frequent Itemsets Mining Algorithm Apriori.
Using the fault early of time of occurrence in described frequent item set as root fault, the as a result of fault of fault in time of occurrence evening, forms some fault correlation rules.
Be genuine probability using the degree of confidence of described frequent item set as this fault correlation rule, obtain incidence relation unknown between fault.
Concentrate the root fault of left minor as child node described correlation rule, the result fault of right minor is as father node; For the represented fault of node on same path, according to the fault of first repairing child node representative, then the fault order of repairing father node representative is processed.
Beneficial effect of the present invention
First, the present invention utilizes association rule mining to obtain the incidence relation between fault, so just can, in the situation that expertise lacks, support the processing to failure cause; Meanwhile, the present invention utilizes rule match to find the cause-effect relationship in magnanimity fault, makes maintenance can preferentially repair the root fault that importance is high, thereby greatly reduces the breakdown maintenance cost of bullet train.
Accompanying drawing explanation
Fig. 1 is bullet train fault handling method process flow diagram of the present invention;
Fig. 2 is rule schema in bullet train fault handling method one embodiment of the present invention.
Embodiment:
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme that one's duty is bought in that embodiment is clearly and completely described, and is understandable that, described embodiment is only the present invention's part embodiment, rather than whole embodiment.Based on the embodiment in the present invention, those skilled in the art, not making the every other embodiment obtaining under creative work prerequisite, belong to the scope of protection of the invention.
Inventive principle
The present invention utilizes association rule mining to obtain the incidence relation between fault, and meanwhile, the present invention utilizes rule match to find the cause-effect relationship in magnanimity fault, and first by principle and step, details are as follows:
(1) fault collection bullet train in user's fixed time being occurred carries out association rule mining, the method of association rule mining (can be write referring to Chen Jingmin, data warehouse principle, design and application, water conservancy and hydropower publishing house, 2004,) find the cause-effect relationship of fault collection internal fault, this step specifically comprises:
1.1) error listing within specified time period of user is inquired about, be extracted in the fault and the failure code thereof that in this time period, occur, form fault collection, fault collection is scanned, find the Frequent Item Sets of the given minimum support of all users of meeting; (minimum support can be write referring to Chen Jingmin, data warehouse principle, design and application, water conservancy and hydropower publishing house, 2004; ) Frequent Item Sets refers to that the frequency of occurrences is greater than the item collection of minimum support; At this, Item Sets is the set that has comprised some faults.
1.2) by the Frequent Item Sets obtaining in (1.1), composition fault mode collection (i.e. the set of all Frequent Item Sets), and according to fault mode collection, generate correlation rule, every correlation rule has represented the cause-effect relationship between fault
(2) Association Rules obtaining in (1) is checked, obtain legal fault correlation rule set.A legal fault correlation rule set comprises some fault correlation rules.The left minor of every correlation rule is root fault, and right minor is result fault.
(3) be the rule schema of a directed acyclic by the Rule Expression in legal fault correlation rule set.Each node in figure represents the fault of bullet train; Directed edge sends from a root fault, points to a result fault.
(4) node in fault and rule schema that error listing high speed train user being provided occurs compares one by one, and in the time that the fault of certain node representative in rule schema occurs, this node is mated.What the node in rule schema represented is fault rather than represents a rule.Two nodes add that the directed edge between them represents a fault correlation rule
(5), according to the rule schema after node matching, find the path that all nodes are all mated.The Ingress node of every paths is root fault.That is to say, and leaf node from figure (the Ingress node), find the coupled node being mated, until find the root node (outlet node) in figure, the path of now just having found a node on it all to be mated in rule schema.
(6), for all paths found in (5), calculate the importance of each node representative fault.If P (v) is the importance of node v, P (v)=O (v)+L (v).Wherein, O (v) is the out-degree of node v in (3) described rule schema, in figure, the out-degree of a node refers to the number of the directed edge sending from this node in the drawings, the maximal value of the length in all paths of finding out in (5) that L (v) is node v place.In step 6) in can obtain the importance P (v) of each node in figure; Path maximal value refers to the maximal value of the length in some paths at node v place, is of computing node v importance, i.e. L (v)
(7) sort by the importance of the each Ingress node calculating in (6) order from high to low.When maintenance, preferentially repair the root fault that importance values is high.
Fig. 1 has provided the idiographic flow of implementing.First, user (domain expert) provides the malfunction coding that need to pay close attention to fault, and the value of the correlation parameters such as minimum support is set according to specific needs.(these need to be specifically that in the experience of situation, obtaining appears in the relevant fault that accumulates bullet train maintenance from domain expert simultaneously), at this, minimum support refers to a number of times that collection occurs in given bug list simultaneously.On this basis, using the fault occurring in the parameter of these settings (minimum support) and given bug list as input, adopt Frequent Itemsets Mining Algorithm Apriori algorithm, analyzing and processing goes out all Frequent Item Sets.For simplicity, we claim that the set of these Frequent Item Sets is fault mode collection.The concrete processing procedure of Apriori algorithm can be referring to document Rakesh Agrawal, Ramakrishnan Srikant, " Fast Algorithms for Mining Association Rules ", Proceedings ofthe 20th VLDB Conference, 1994..Suppose, the fault mode collection that now analyzed is for being { { f10, f12}, { f10, f22}, { f10, f23}, { f11, f10}, { f12, f6}, { f12, f10}, { f13, f29}, { f24, f11}, { f25, f11}}.Each frequent item set can degree of confidence be 0.93,0.97,0.65,1,0.61,1,1,1,1.Wherein, degree of confidence refers to the possibility that the simultaneous faults B of fault A appearance also occurs.
Each frequent item set that above-mentioned obtained fault mode is concentrated is analyzed, using the fault early of time of occurrence in each frequent item set as root fault, by the time of occurrence as a result of fault of fault in evening, form some fault correlation rules, it is genuine probability that the degree of confidence of frequent item set can be used as this fault correlation rule, thereby obtains incidence relation unknown between fault.Table 1 has provided the fault correlation rule obtaining.
Table 1
| Fault correlation rule | Fault correlation rule is genuine probability |
| f10→f21 | 0.939189189189189 |
| f10→f22 | 0.965277777777778 |
| f10→f23 | 0.654676258992806 |
| f11→f10 | 1 |
| f12→f6 | 0.614864864864865 |
| f12→f10 | 1 |
| f13→f29 | 1 |
| f24→f11 | 1 |
| f25→f11 | 1 |
Each correlation rule in analytical table 1, using the root fault of left minor in each correlation rule as child node, using the result fault of right minor as father node, thereby is expressed as rule schema as shown in Figure 2 by these correlation rules.Suppose that the fault that now bullet train maintenance personal pays close attention to is f25, f11, f10, f21, f23, f13 and f29, now in Fig. 2 rule schema, searching out the node being mated is accordingly f25, f11, f10, f21, f23, f13 and f29.On this basis, carrying out in the rule schema of coupling, finding by the above path that node formed that these have been mated.At this, find following path, that is: f13 → f29; F25 → f11 → f10 → f21, f22, f23; F12 → f06; F12 → f10 → f21, f22, f23.
According to found path and the represented rule schema of Fig. 2, calculate the importance of each node on each path, can obtain P (f13)=2; P (f29)=0; P (f25)=4; P (f24)=4; P (f11)=3; P (f12)=4; P (f10)=4; P (f6)=0; P (f21)=0; P (f22)=0; P (f23)=0.Like this, for the root fault on each searched out paths, that is, and f12, f13, f25, P (f25) >P (f12) >P (f13).So, when maintenance, will be according to f25, the order of f12 and f13 places under repair.And for the represented fault of the node on same path, according to the fault of first repairing child node representative, then the order of repairing the fault of father node representative is carried out.
Above-described embodiment is only illustrative principle of the present invention and effect thereof, but not for limiting the scope of the invention.Any ripe those skilled in the art in this technology all can, under know-why of the present invention and spirit, make an amendment and change embodiment.Protection scope of the present invention should be as the criterion with described in claims.
Claims (10)
1. a bullet train fault handling method, its step comprises:
1) fault and the failure code that in inquiry error listing, in user's setting-up time, occur, composition one fault collection;
2) described fault collection is scanned to the Frequent Item Sets that is met minimum support, in described Item Sets, generate fault correlation rule and obtain the legal fault correlation rule set in described fault correlation rule;
3) described legal fault correlation rule set is represented by rule schema, mate the rule schema after being mated according to fault in error listing described in step 1) with the node in this rule schema;
4) in the rule schema after described coupling, traversal finds all nodes all by coupling path; Calculate described by importance the sequence of each node representative fault on coupling path;
5) process from high to low according to fault importance according to described importance ranking.
2. bullet train fault handling method as claimed in claim 1, it is characterized in that, importance P (v)=O (the v)+L (v) of described fault, wherein, O (v) is the out-degree of node v in rule schema, and L (v) is all by the MAXPATHLEN value of being mated path of node v.
3. bullet train fault handling method as claimed in claim 1, is characterized in that, described fault mode collection is the set of all Frequent Item Sets; Described Frequent Item Sets is the Item Sets that number of times that comprised fault occurs is simultaneously greater than the minimum support that user sets.
4. bullet train fault handling method as claimed in claim 1, it is characterized in that, described legal fault correlation rule set comprises some fault correlation rules, and the left minor of described each correlation rule is root fault, and right minor is the fault correlation rule of result fault.
5. bullet train fault handling method as claimed in claim 1, is characterized in that, the each node in the rule schema after described coupling represents the fault of bullet train, and directed edge sends and points to result fault from root fault, forms a fault correlation rule.
6. bullet train fault handling method as claimed in claim 1, is characterized in that, the cause-effect relationship between described fault correlation Rule Expression bullet train fault.
7. bullet train fault handling method as claimed in claim 1, is characterized in that, described frequent item set obtains according to Frequent Itemsets Mining Algorithm Apriori.
8. the bullet train fault handling method as described in claim 1 or 7, is characterized in that, using the fault early of time of occurrence in described frequent item set as root fault, the as a result of fault of fault in time of occurrence evening, forms some fault correlation rules.
9. the bullet train fault handling method as described in claim 1 or 7, is characterized in that, is genuine probability using the degree of confidence of described frequent item set as this fault correlation rule, obtains incidence relation unknown between fault.
10. the bullet train fault handling method as described in claim 1 or 4, is characterized in that, concentrates the root fault of left minor as child node described correlation rule, and the result fault of right minor is as father node; For the represented fault of node on same path, according to the fault of first repairing child node representative, then the fault order of repairing father node representative is processed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210428024.8A CN103793589B (en) | 2012-10-31 | 2012-10-31 | High-speed train fault handling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210428024.8A CN103793589B (en) | 2012-10-31 | 2012-10-31 | High-speed train fault handling method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103793589A true CN103793589A (en) | 2014-05-14 |
| CN103793589B CN103793589B (en) | 2017-01-18 |
Family
ID=50669251
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210428024.8A Expired - Fee Related CN103793589B (en) | 2012-10-31 | 2012-10-31 | High-speed train fault handling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103793589B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104460651A (en) * | 2014-10-24 | 2015-03-25 | 北京交控科技有限公司 | ZC double-system downtime fault early-warning method and device based on autonomous learning |
| CN104777827A (en) * | 2015-01-21 | 2015-07-15 | 中国铁路总公司 | Method for diagnosing fault of high-speed railway signal system vehicle-mounted equipment |
| CN105045256A (en) * | 2015-07-08 | 2015-11-11 | 北京泰乐德信息技术有限公司 | Rail traffic real-time fault diagnosis method and system based on data comparative analysis |
| CN105868551A (en) * | 2016-03-28 | 2016-08-17 | 北京交通大学 | Fault association rule construction method |
| CN109358572A (en) * | 2018-09-26 | 2019-02-19 | 深圳壹账通智能科技有限公司 | Method, apparatus, computer equipment and the storage medium of machinery equipment O&M |
| CN110705597A (en) * | 2019-09-04 | 2020-01-17 | 中国科学院计算技术研究所 | Network early event detection method and system based on event cause and effect extraction |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101937447B (en) * | 2010-06-07 | 2012-05-23 | 华为技术有限公司 | Alarm association rule mining method, and rule mining engine and system |
| CN102111296A (en) * | 2011-01-10 | 2011-06-29 | 浪潮通信信息系统有限公司 | Mining method for communication alarm association rule based on maximal frequent item set |
| CN102638100B (en) * | 2012-04-05 | 2014-02-19 | 华北电力大学 | District power network equipment abnormal alarm signal association analysis and diagnosis method |
-
2012
- 2012-10-31 CN CN201210428024.8A patent/CN103793589B/en not_active Expired - Fee Related
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104460651A (en) * | 2014-10-24 | 2015-03-25 | 北京交控科技有限公司 | ZC double-system downtime fault early-warning method and device based on autonomous learning |
| CN104777827A (en) * | 2015-01-21 | 2015-07-15 | 中国铁路总公司 | Method for diagnosing fault of high-speed railway signal system vehicle-mounted equipment |
| CN105045256A (en) * | 2015-07-08 | 2015-11-11 | 北京泰乐德信息技术有限公司 | Rail traffic real-time fault diagnosis method and system based on data comparative analysis |
| CN105045256B (en) * | 2015-07-08 | 2018-11-20 | 北京泰乐德信息技术有限公司 | Rail traffic real-time fault diagnosis method and system based on date comprision |
| CN105868551A (en) * | 2016-03-28 | 2016-08-17 | 北京交通大学 | Fault association rule construction method |
| CN109358572A (en) * | 2018-09-26 | 2019-02-19 | 深圳壹账通智能科技有限公司 | Method, apparatus, computer equipment and the storage medium of machinery equipment O&M |
| CN110705597A (en) * | 2019-09-04 | 2020-01-17 | 中国科学院计算技术研究所 | Network early event detection method and system based on event cause and effect extraction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103793589B (en) | 2017-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ni et al. | A cluster based feature selection method for cross-project software defect prediction | |
| You et al. | Bringing your own view: Graph contrastive learning without prefabricated data augmentations | |
| CN103793589A (en) | High-speed train fault handling method | |
| Karthikeyan et al. | A survey on association rule mining | |
| Qian et al. | Fusing monotonic decision trees | |
| Fleischhacker et al. | Mining RDF data for property axioms | |
| CN107391353A (en) | Complicated software system anomaly detection method based on daily record | |
| Nikitin et al. | Structural evolutionary learning for composite classification models | |
| Liu et al. | Link prediction algorithm for signed social networks based on local and global tightness | |
| He et al. | A heuristic grafting strategy for manufacturing knowledge graph extending and completion based on nature language processing: KnowTree | |
| Khoshgoftaar et al. | Attribute selection using rough sets in software quality classification | |
| Bairagi et al. | Social network comment classification using fuzzy based classifier technique | |
| Khoshraftar et al. | Centrality-based interpretability measures for graph embeddings | |
| Njiwoua et al. | Forwarding the choice of bias LEGAL-F: using feature selection to reduce the complexity of LEGAL | |
| Sun et al. | ERGP: A combined entity resolution approach with genetic programming | |
| Zhang et al. | SARC: Split-and-recombine networks for knowledge-based recommendation | |
| Ellakwa et al. | Integrated ontology for agricultural domain | |
| Luo et al. | An incremental feature selection approach based on information entropy for incomplete data | |
| Sabou et al. | Knowledge engineering in the age of neurosymbolic systems | |
| Zhang et al. | Predicting Wikipedia editor's editing interest based on factor graph model | |
| CN117435246B (en) | Code clone detection method based on Markov chain model | |
| Dong et al. | DGSRC: A Recommendation System Ranking Model Based on Dynamic Graph Neural Networks | |
| Zhang | Mining Complex Feature Correlations from Large Software Product Line Configurations | |
| Yu et al. | Using non-topological node attributes to improve results of link prediction in social networks | |
| Behkamal et al. | Pre-processing ontologies to improve the results of matchers |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170118 Termination date: 20191031 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |