CN105676842A - High-speed railway train control vehicle-mounted equipment fault diagnosis method - Google Patents

Abstract

The invention relates to a high-speed railway train control vehicle-mounted equipment fault diagnosis method. The high-speed railway train control vehicle-mounted equipment fault diagnosis method is characterized in that the method comprises the main steps of: step 1, analyzing high-speed railway train control vehicle-mounted equipment fault data and extracting features; step 2, extracting a minimal decision information table through reducing fault decision tables; and step 3, establishing a Bayesian fault diagnosis network, carrying out fault diagnosis and providing emergency decision support.

Description

A kind of high ferro train control on board equipment method for diagnosing faults

Technical field

The invention belongs to technical field of rail transportation operation control, relate to the method for diagnosing faults of a kind of high ferro train control on board equipment.

Background technology

CTCS-3 level train control on board equipment is China Express Railway important technology equipment, is to ensure that bullet train runs one of safe, reliable, efficient key technology and equipment. Mobile unit is received from the RBC on ground, transponder, track circuit data by GSM-R, BTM, TCR. The informixs such as train travel license, track data, train interface, driver operation are processed, according to target range continuous velocity control model, generates least favorable speed control curve. Take sound and light alarm, excision pull strength, three grades of service braking (weak, medium, strong) and brake hard measure, monitor train operation, it is ensured that train operating safety.

In the prior art, the troubles diagnosis and fixing to high ferro mobile unit, the Main Diagnosis measure of onsite application has:

(1) desk checking, this is a kind of traditional diagnosis and maintenance mode, because its labor intensity is big, the low inferior reason of diagnosis efficiency, increasingly it is not suitable with modern demand for development.

(2) data monitoring system, this system is by the operational factor of specific data acquisition unit Real-time Collection mobile unit, and is pooled to data monitoring center, and plant maintenance personnel comprehensively analyze the data collected to carry out fault diagnosis and maintenance. Data monitoring system achieves the Real-time Collection of mobile unit and possesses certain trouble diagnosibility, is the important means of current telecommunication and signaling branch regular maintenance and accident analysis.

Summary of the invention

The invention aims to solve the problems referred to above, it is provided that a kind of high ferro train control on board equipment method for diagnosing faults, solve problems of the prior art.

(1) traditional desk checking method, labor intensity is big, and the low inferior reason of diagnosis efficiency is increasingly not suitable with the demand for development of high ferro train control on board equipment.

(2) data monitoring system still suffers from misdiagnosis rate height, the coarse problem of fault location in actual applications, and its data analysis needs a large amount of artificial participation simultaneously, and this brings heavy burden to attendant undoubtedly.

A kind of method for diagnosing faults of high ferro train control on board equipment, described method comprises following several key step:

The first step, high ferro train control on board equipment failure data analyzing and feature extraction;

Second step, by Decision Table for Fault yojan, extract minimum decision information table;

3rd step, set up Bayes's fault diagnosis network, carry out fault diagnosis and Emergency decision support is provided.

In the described first step, set up mobile unit domain features dictionary, described feature dictionary includes, transponder, message, mistake, BTM, DMI, main frame, communication disruption, output, stop, wireless time-out, brake hard, C3 turns C2, A system, B system, A system, restart, B system, RBC, set up, connect, ATP, disconnecting, ground installation, ATPCU, fail-safe software, service braking, DMI blank screen, JRU fault, SDU fault, speed sensor fault, change is restart, braking test is unsuccessful, MVB, C2CU, fault, C2, degradation, CTCS-2, CTCS-3, grade is changed, failure, advance rashly, hardware, core code, braking bypass, A/B code, inconsistent, initialize, safety, software, transfer district, C3.

With feature dictionary for foundation, mobile unit failure logging is carried out fault signature extraction, adopt the property value after quantifying to form a two-dimensional table, one object of each line description, one attribute of every string description object, including conditional attribute and decision attribute, builds Decision Table for Fault.

The shortcoming that described second step exists decision information redundancy, decision attribute conflict either directly through the Decision Table for Fault that participle obtains, row control vehicle-mounted system failure information table is shown as a knowledge system U, and wherein C represents conditional attribute, and D represents decision attribute, R is expressed as the equivalence relation on U, POS_C(D) represent positive territory, then attribute dependability and Attribute Significance formula are respectively as shown in formula one and formula two:

V (C, D)=card (POS_C(D)/card (U) formula one

SIG (a, R, D)=V (R ∪ a}, D)-V (R, D) formula two

After increasing an attribute a ∈ C-R in R, the increment of mutual information is:

Δ I=I (U ∪ { a}, D)-I (R, D)=(H (D)-H (D | R ∪ { a}))-(H (D)-H (D | R))=H (D | R)-H (D | R ∪ a})

Wherein, the mutual information that I (R, D) is R and D, when H (D | R) is for known R, the conditional entropy of D; Therefore, the importance degree of any attribute a ∈ C-R can be defined as:

SIG (a, R, D)=H (D | R)-H (D | R ∪ a}) formula three

Due to 0≤V (R, D)≤1 and 0≤H (D | R)≤1bn, can to H (D | R) and V (R, D) change, then H (D | R)=1-V (R, D)+1bn, is weighted the structure of meansigma methods to row control vehicle-mounted fault attribute importance degree algorithm, then the Attribute Significance of weighted mean is expressed as:

SIG (a, R, D)=k (V (R ∪ a}, D)-V (R, D))+(1-k) (H (D | R)-H (D | R ∪ a})) formula four

SIG (a, R, D)=V (R ∪ a}, D)+kV (R, D) formula five

Calculate by experiment, during k=0.96, closest to practical situation;

By above-mentioned calculating, using attribute dependability as main decision criteria, adopt the method for weighted mean to calculate each Attribute Significance of row control vehicle-mounted system failure decision table, it is carried out attribute reduction.

With the decision table after yojan for object, show according to its fault, fault type and failure cause set up Bayesian network model, ground floor is fault presentation layer, presentation layer builds based on mobile unit domain features dictionary, realize the expression of fault signature, complete mobile unit and can gather the extraction work of fault signature in malfunction; The second layer is fault type layer, and failure modes layer realizes the fault signature mapping to fault type; Third layer is failure cause layer, concrete failure mode type and possible failure cause is mapped, utilizes Bayesian Network Learning and reasoning, fault is directly targeted to parts or code lost efficacy; 4th layer is decision-making level, concrete failure cause is combined with extraneous factor, finds out maximally effective Emergency decision method.

Concrete fault diagnosis algorithm is as follows:

(1) structure fault presentation layer, corresponding with the Feature Words in mobile unit fault signature dictionary by the concrete phenomenon of the failure of mobile unit;

(2) according to mobile unit failure modes table, set up fault type node, and with Decision Table for Fault for foundation, build the corresponding relation of fault type and fault signature;

(3) by the Bayesian Structure Learning method (K2 and MCMC algorithm combines) improved and domain knowledge combined structure Bayesian network;

(4) utilize fault data sample, by Bayesian network is carried out parameter learning, obtain diagnostic Bayesian network model;

(5) utilize structure and the CPT of Bayesian network, according to Bayesian Network Inference, row control vehicle-mounted system is carried out fault diagnosis;

(6) diagnostic result is combined with emergency condition, adds environmental element, communications status, time of origin, power supply state and rescue outfit node, set up the row control vehicle-mounted fault diagnosis system with Emergency decision function, utilize it to carry out Decision Inference.

Beneficial effect

The present invention is directed to the fault diagnosis of high ferro mobile unit, propose a kind of intelligentized method for diagnosing faults, utilize the priori fault knowledge of mobile unit, fault attribute yojan is carried out by dimensionality reduction, and construct a kind of three layers Bayesian diagnostic network, solve poor fault tolerance in existing monolayer Bayesian network method for diagnosing faults, problem that efficiency is on the low side.

1, contain the structure of the high ferro train control on board equipment feature dictionary of domain features, owing to the failure logging of high ferro train control on board equipment is the form of text, set up the dictionary that high ferro mobile unit is exclusive, trace table is carried out feature extraction;

2, the Attribute Significance computational methods of weighted mean, solve the problem that inconsistent decision table does not produce the relation that is completely dependent on, and individually use the attribute reduction method based on attribute dependability to cause the problem that yojan resultant error is bigger, row control vehicle-mounted fault message is carried out attribute reduction, reduces the dimension of Decision Table for Fault;

3, there is the train control on board equipment Bayesian diagnostic network of decision support, comprise high ferro train control on board equipment Bayes's fault diagnosis network structure of phenomenon of the failure layer, fault type layer and failure cause layer and Emergency decision layer, provide decision support in conjunction with extraneous factor for emergency management and rescue.

Accompanying drawing explanation

Fig. 1 is row control vehicle-mounted system fault diagnosis flow chart figure;

Fig. 2 is multilamellar Bayesian network trouble-shooting chart.

Detailed description of the invention

Embodiment 1

The present invention utilizes the priori fault knowledge of mobile unit, and extract row control vehicle-mounted system failure feature, set up Decision Table for Fault, use Algorithm for Reduction to carry out attribute reduction, reduce the attribute number in decision table, and then reduce the complexity of BN modeling and reasoning. Set up Bayesian network model, utilize Bayesian network model to carry out diagnostic reasoning. The present invention comprises following several committed step:

The first, high ferro train control on board equipment failure data analyzing and feature extraction

Set up mobile unit domain features dictionary. Feature dictionary is as follows:

{ transponder, message, mistake, BTM, DMI, main frame, communication disruption, output, stop, wireless time-out, brake hard, C3 turns C2, A system, B system, A system, restart, B system, RBC, set up, connect, ATP, disconnecting, ground installation, ATPCU, fail-safe software, service braking, DMI blank screen, JRU fault, SDU fault, speed sensor fault, change is restart, braking test is unsuccessful, MVB, C2CU, fault, C2, degradation, CTCS-2, CTCS-3, grade is changed, failure, advance rashly, hardware, core code, braking bypass, A/B code, inconsistent, initialize, safety, software, transfer district, C3 etc. }

With feature dictionary for foundation, mobile unit failure logging is carried out fault signature extraction. Adopting the property value after quantifying to form a two-dimensional table, one object of each line description, an attribute of every string description object, including conditional attribute and decision attribute. Build Decision Table for Fault.

For high ferro mobile unit, fault case feature is carried out decision table structure. Wherein decision attribute represents fault type, and a1～a60 represents conditional attribute, and 1 represents generation, and 0 expression does not occur, and wherein, train control on board equipment fault message can show with decision table.

Table 1 onboard system Decision Table for Fault

Second: by Decision Table for Fault yojan, extracting minimum decision information table.

There is the shortcoming such as decision information redundancy, decision attribute conflict in the Decision Table for Fault obtained either directly through participle. The present invention failure modes information according to priori, carries out dimensionality reduction by Decision Table for Fault, completes fault attribute yojan.

Row control vehicle-mounted system failure information table is shown as a knowledge system U, and wherein C represents conditional attribute, and D represents decision attribute, and R is expressed as the equivalence relation on U, POS_C1. and 2. (D) represent positive territory, then attribute dependability and Attribute Significance formula are respectively as shown in:

V (C, D)=card (POS_C(D)/card(U)①

SIG (a, R, D)=V (R ∪ a}, D)-V (R, D) is 2.

After increasing an attribute a ∈ C-R in R, the increment of mutual information is:

Δ I=I (U ∪ { a}, D)-I (R, D)=(H (D)-H (D | R ∪ { a}))-(H (D)-H (D | R))=H (D | R)-H (D | R ∪ a})

Wherein, the mutual information that I (R, D) is R and D, when H (D | R) is for known R, the conditional entropy of D. Therefore, the importance degree of any attribute a ∈ C-R can be defined as:

SIG (a, R, D)=H (D | R)-H (D | R ∪ a}) 3.

Due to 0≤V (R, D)≤1 and 0≤H (D | R)≤1bn, can to H (D | R) and V (R, D) change, then H (D | R)=1-V (R, D)+1bn, is weighted the structure of meansigma methods to row control vehicle-mounted fault attribute importance degree algorithm, then the Attribute Significance of weighted mean is expressed as:

SIG (a, R, D)=k (V (R ∪ a}, D)-V (R, D))+(1-k) (H (D | R)-H (D | R ∪ a})) 4.

SIG (a, R, D)=V (R ∪ a}, D)+kV (R, D) is 5.

Calculate by experiment, during k=0.96, closest to practical situation.

By said method, using attribute dependability as main decision criteria, adopt the method for weighted mean to calculate each Attribute Significance of row control vehicle-mounted system failure decision table, it is carried out attribute reduction.

3rd: set up Bayes's fault diagnosis network, carry out fault diagnosis and Emergency decision support is provided.

With the decision table after yojan for object, showing according to its fault, fault type and failure cause set up Bayesian network model. Ground floor is fault presentation layer: presentation layer builds based on mobile unit domain features dictionary, it is achieved the expression of fault signature, completes mobile unit and can gather the extraction work of fault signature in malfunction; The second layer is fault type layer, and failure modes layer realizes the fault signature mapping to fault type; Third layer is failure cause layer, concrete failure mode type and possible failure cause is mapped, utilizes Bayesian Network Learning and reasoning, fault is directly targeted to parts or code lost efficacy; 4th layer is decision-making level, concrete failure cause is combined with extraneous factor, finds out maximally effective Emergency decision method. Specific algorithm is as follows:

1. structure fault presentation layer, corresponding with the Feature Words in mobile unit fault signature dictionary by the concrete phenomenon of the failure of mobile unit.

2. according to mobile unit failure modes table, set up fault type node, and with Decision Table for Fault for foundation, build the corresponding relation of fault type and fault signature.

3. by the Bayesian Structure Learning method (K2 and MCMC algorithm combines) improved and domain knowledge combined structure Bayesian network.

4. utilize fault data sample, by Bayesian network is carried out parameter learning, obtain diagnostic Bayesian network model.

5. utilize structure and the CPT of Bayesian network, according to Bayesian Network Inference, row control vehicle-mounted system is carried out fault diagnosis.

6. diagnostic result is combined with emergency condition, adds environmental element, communications status, time of origin, power supply state and rescue outfit node, set up the row control vehicle-mounted fault diagnosis system with Emergency decision function, utilize it to carry out Decision Inference.

The key problem in technology point of the present invention is in that:

1, contain the structure of the high ferro train control on board equipment feature dictionary of domain features, owing to the failure logging of high ferro train control on board equipment is the form of text, set up the dictionary that high ferro mobile unit is exclusive, trace table is carried out feature extraction;

2, the Attribute Significance computational methods of weighted mean, solve the problem that inconsistent decision table does not produce the relation that is completely dependent on, and individually use the attribute reduction method based on attribute dependability to cause the problem that yojan resultant error is bigger, row control vehicle-mounted fault message is carried out attribute reduction, reduces the dimension of Decision Table for Fault;

3, there is the train control on board equipment Bayesian diagnostic network of decision support, comprise high ferro train control on board equipment Bayes's fault diagnosis network structure of phenomenon of the failure layer, fault type layer and failure cause layer and Emergency decision layer, provide decision support in conjunction with extraneous factor for emergency management and rescue.

Last it is noted that obvious, above-described embodiment is only for clearly demonstrating example of the present invention, and is not the restriction to embodiment. For those of ordinary skill in the field, can also make other changes in different forms on the basis of the above description. Here without also cannot all of embodiment be given exhaustive. And the apparent change thus extended out or variation are still among protection scope of the present invention.

Claims (6)

1. the method for diagnosing faults of a high ferro train control on board equipment, it is characterised in that described method comprises following several key step:

The first step, set up high ferro train control on board equipment failure data analyzing and feature extraction;

Second step, by Decision Table for Fault yojan, extract minimum decision information table;

3rd step, set up Bayes's fault diagnosis network, carry out fault diagnosis and Emergency decision support is provided.

2. method for diagnosing faults according to claim 1, it is characterized in that: in the described first step, described failure data analyzing and feature extraction refer to, set up mobile unit domain features dictionary, described feature dictionary includes, transponder, message, mistake, BTM, DMI, main frame, communication disruption, output, stop, wireless time-out, brake hard, C3 turns C2, A system, B system, A system, restart, B system, RBC, set up, connect, ATP, disconnecting, ground installation, ATPCU, fail-safe software, service braking, DMI blank screen, JRU fault, SDU fault, speed sensor fault, change is restart, braking test is unsuccessful, MVB, C2CU, fault, C2, degradation, CTCS-2, CTCS-3, grade is changed, failure, advance rashly, hardware, core code, braking bypass, A/B code, inconsistent, initialize, safety, software, transfer district and C3.

3. method for diagnosing faults according to claim 2, it is characterized in that: with feature dictionary for foundation, mobile unit failure logging is carried out fault signature extraction, the property value after quantifying is adopted to form a two-dimensional table, one object of each line description, one attribute of every string description object, including conditional attribute and decision attribute, builds Decision Table for Fault.

4. method for diagnosing faults according to claim 1, it is characterized in that: the shortcoming that described second step exists decision information redundancy, decision attribute conflict either directly through the Decision Table for Fault that participle obtains, row control vehicle-mounted system failure information table is shown as a knowledge system U, wherein C represents conditional attribute, D represents decision attribute, R is expressed as the equivalence relation on U, POS_C(D) represent positive territory, then attribute dependability and Attribute Significance formula are respectively as shown in formula one and formula two:

V (C, D)=card (POS_C(D)/card (U) formula one

SIG (a, R, D)=V (R ∪ a}, D)-V (R, D) formula two

After increasing an attribute a ∈ C-R in R, the increment of mutual information is:

Δ I=I (U ∪ { a}, D)-I (R, D)=(H (D)-H (D | R ∪ { a}))-(H (D)-H (D | R))=H (D | R)-H (D | R ∪ a})

Wherein, the mutual information that I (R, D) is R and D, when H (D | R) is for known R, the conditional entropy of D; Therefore, the importance degree of any attribute a ∈ C-R can be defined as:

SIG (a, R, D)=H (D | R)-H (D | R ∪ a}) formula three

Due to 0≤V (R, D)≤1 and 0≤H (D | R)≤1bn, can to H (D | R) and V (R, D) change, then H (D | R)=1-V (R, D)+1bn, is weighted the structure of meansigma methods to row control vehicle-mounted fault attribute importance degree algorithm, then the Attribute Significance of weighted mean is expressed as:

SIG (a, R, D)=k (V (R ∪ a}, D)-V (R, D))+(1-k) (H (D | R)-H (D | R ∪ a})) formula four

SIG (a, R, D)=V (R ∪ a}, D)+kV (R, D) formula five

Calculate by experiment, during k=0.96, closest to practical situation;

By above-mentioned calculating, using attribute dependability as main decision criteria, adopt the method for weighted mean to calculate each Attribute Significance of row control vehicle-mounted system failure decision table, it is carried out attribute reduction.

5. method for diagnosing faults according to claim 4, it is characterized in that: with the decision table after yojan for object, show according to its fault, fault type and failure cause set up Bayesian network model, ground floor is fault presentation layer, presentation layer builds based on mobile unit domain features dictionary, it is achieved the expression of fault signature, completes mobile unit and can gather the extraction work of fault signature in malfunction; The second layer is fault type layer, and failure modes layer realizes the fault signature mapping to fault type; Third layer is failure cause layer, concrete failure mode type and possible failure cause is mapped, utilizes Bayesian Network Learning and reasoning, fault is directly targeted to parts or code lost efficacy; 4th layer is decision-making level, concrete failure cause is combined with extraneous factor, finds out maximally effective Emergency decision method.

6. method for diagnosing faults according to claim 5, it is characterised in that: concrete fault diagnosis algorithm is as follows:

(1) structure fault presentation layer, corresponding with the Feature Words in mobile unit fault signature dictionary by the concrete phenomenon of the failure of mobile unit;

(2) according to mobile unit failure modes table, set up fault type node, and with Decision Table for Fault for foundation, build the corresponding relation of fault type and fault signature;

(3) by the Bayesian Structure Learning method (K2 and MCMC algorithm combines) improved and domain knowledge combined structure Bayesian network;

(4) utilize fault data sample, by Bayesian network is carried out parameter learning, obtain diagnostic Bayesian network model;

(5) utilize structure and the CPT of Bayesian network, according to Bayesian Network Inference, row control vehicle-mounted system is carried out fault diagnosis;

(6) diagnostic result is combined with emergency condition, adds environmental element, communications status, time of origin, power supply state and rescue outfit node, set up the row control vehicle-mounted fault diagnosis system with Emergency decision function, utilize it to carry out Decision Inference.