CN104680232A - RVM (Relevance Vector Machine)-based engine failure detecting method - Google Patents
RVM (Relevance Vector Machine)-based engine failure detecting method Download PDFInfo
- Publication number
- CN104680232A CN104680232A CN201410586671.0A CN201410586671A CN104680232A CN 104680232 A CN104680232 A CN 104680232A CN 201410586671 A CN201410586671 A CN 201410586671A CN 104680232 A CN104680232 A CN 104680232A
- Authority
- CN
- China
- Prior art keywords
- engine
- rvm
- fish
- individuality
- behavior
- 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
Links
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to an RVM (Relevance Vector Machine)-based engine failure detecting method and belongs to the field of engine failure detection. The method comprises the following steps of step 1, pre-processing, i.e., normalizing the running speed of an engine and amplitude corresponding to the running speed, and the prior data of a frequency sample, and building corresponding relation between the running speed and an engine state, between the amplitude corresponding to the running speed and engine states, and between the frequency corresponding to the running speed and the engine state; step 2, performing machining training, i.e., selecting a proper kernel function, performing fish-swarm algorithm optimization training on a hyper-parameter of the kernel function, and establishing a proper RVM model; step 3, performing failure diagnosis, i.e., adopting a one-to-one RVM classifier to perform failure diagnosis on a to-be-detected sample and outputting a result. According to the detecting method provided by the invention, a vehicle engine parameter analysis method is used, the parameter optimized RVM model is established through relevance vector training normalizing engine state sample data, an engine failure detecting experiment is performed by the model, the method is good in robustness, and the generalization ability is reinforced.
Description
Technical field
The invention belongs to automobile technical field, be specifically related to a kind of engine failure detection method.
Background technology
Engine is the heart of motor racing, determines the quality of the performance of automobile.Modern Engine has become set electron technology, computer technology, infotech in the intelligent control system of one, and integration degree is more and more higher, structure also becomes increasingly complex; But the intelligent fault diagnosis but making to start of engine and maintenance become the bottleneck of restriction automobile industry development.
Automobile engine system is optimized control by the ratio, discharging waste gas etc. of electronic control means to engine ignition, oil spout, air and fuel oil, makes engine operation in optimum condition.Automobile engine system mainly comprises electric control fuel oil jet system, electronic control ignition system, warning prompt system etc.
Summary of the invention
In order to carry out Real-Time Monitoring to the duty of engine, avoiding engine failure to bring security incident, the invention provides a kind of engine failure detection method based on RVM, judge whether engine has fault according to collection engine operating parameter.
Technical scheme of the present invention is: a kind of engine diagnosis algorithm, comprise the steps: step one: pre-service: engine operational speed and the amplitude of correspondence thereof, the priori data of frequency samples are normalized, and set up the amplitude of travelling speed and correspondence thereof, the corresponding relation between frequency and engine condition; Step 2: machine is trained: select suitable kernel function and carry out fish-swarm algorithm optimization training to its hyper parameter, setting up suitable RVM model; Step 3: fault diagnosis: adopt " one to one " RVM sorter to carry out sample to be tested fault diagnosis and Output rusults.In described step 2, kernel function is gaussian radial basis function kernel function.The implementation procedure of described fish-swarm algorithm is: a. initialization shoal of fish population number, the crossing-over rate in each stage and the ginseng such as aberration rate, maximum iteration time; B. use mutual information calculates the mutual information between each variable; C. adopt MWST algorithm to generate initial non-directed graph, and specify any one node to be that root node generates initial population; D. calculate the BIC scoring of all initial population, find out the maximum score value of scoring and individuality; E. the result of d is put into the bulletin board after initialization; While(iterations < maximum iteration time) if the fish individuality of this numbering of for k=1:fishnumf. meets condition of bunching, then perform behavior of bunching; Otherwise execution foraging behavior; If g. the fish individuality of this numbering meets the condition that knocks into the back, then perform behavior of knocking into the back; Otherwise execution foraging behavior; H. the fish individuality executing bunch behavior and behavior of knocking into the back is compared, obtain the mode that this fish individuality " predation " is optimum, and record; I. the optimum individual in the fishinum after " predation " fish individuality is used once to upgrade bulletin board; J. judge whether to reach maximum iteration time.If reach maximum iteration time, then algorithm terminates; Otherwise proceed; K. export the top score value after calculating, obtain optimum optimized network.
The present invention has following good effect: utilize motor car engine Parameter analysis method, normalized engine condition sample data is trained by associated vector, establish the RVM model of parameter optimization, utilize this model to carry out engine failure test experience, demonstrate the validity of the method.Experimental result shows, adopt the Engine Failure Diagnostic Technology of " fish-swarm algorithm-RVM of optimization " more accurate than traditional diagnostic result based on the method such as neural network, GA-BP, GA-SVM, robustness is good, generalization ability is strengthened further, has good replicability and certain practical value.
Accompanying drawing explanation
Fig. 1 is specific embodiment of the invention engine intelligent diagnostic system structural drawing.
Embodiment
Contrast accompanying drawing below, by the description to embodiment, the specific embodiment of the present invention is as the effect of the mutual alignment between the shape of involved each component, structure, each several part and annexation, each several part and principle of work, manufacturing process and operation using method etc., be described in further detail, have more complete, accurate and deep understanding to help those skilled in the art to inventive concept of the present invention, technical scheme.
During engine non-normal working, can produce vibration, at different speeds, the vibration amplitude that engine operation causes and frequency are also not quite similar.The duty that therefore can judge residing for engine by the vibration amplitude of the operating rate of motor car engine and correspondence and frequency, completes fault diagnosis.Theoretical according to this, the method for machine learning can be utilized to carry out learning training to priori data sample, the machine mould trained is used for engine diagnosis.
RVM algorithm by " kernel function " nonlinear problem of lower dimensional space is mapped to higher dimensional space and changes into linear problem, and the ingenious part of kernel function is that mapping process does not increase a lot of calculated amount.Compared with the SVM algorithm relatively more early proposed, the selection of RVM algorithm Kernel Function is not by the restriction of Mercer condition, due to its separate openness and probability, be the algorithm that in machine learning method, Generalization Capability is the most outstanding theoretically, especially effective to the machine learning of small sample.
The method obtains the outstanding sparse solution of Generalization Capability, is be based upon on the basis of suitable Selection of kernel function and optimum configurations.In order to obtain best setting, the optimization of RVM hyper parameter be have employed to the fish-swarm algorithm of improvement herein, and " fish-swarm algorithm-RVM model of improvement " is applied among RVM engine diagnosis.
Fault diagnosis system composition based on " fish-swarm algorithm-RVM model of improvement " can be divided into 3 main process:
1) pre-service: engine operational speed and the amplitude of correspondence thereof, the priori data of frequency samples are normalized, and set up the amplitude of travelling speed and correspondence thereof, the corresponding relation between frequency and engine condition;
2) machine training: select suitable kernel function and fish-swarm algorithm optimization training is carried out to its hyper parameter, setting up suitable RVM model;
3) fault diagnosis: adopt " one to one " RVM sorter to carry out sample to be tested fault diagnosis and Output rusults.
Conventional RVM kernel function has 4 kinds:
Linear kernel function:
K( x，z) = x·z ( 1)
Polynomial kernel function:
K( x，z) = [ s( x·z) + c]
q( 2)
Gaussian radial basis function (RBF) kernel function:
K( x，z ) = exp(-λ‖x － z‖
2) ( 3)
Sigmoid kernel function:
K (x，z) = tanh［s( x·z) + c］ ( 4)
Select suitable kernel function to be the key that the method can successfully use, trained by testing authentication, more respective Generalization Capability, select RBF kernel function as the RVM model of fault diagnosis herein.
To the frequency signal gathered according to low frequency, engine operation frequency range, high band three frequency range automatic classifications, by vibration amplitude according to statistics amplitude section when little amplitude, engine work, the classification of extraordinary amplitude; To comprehensively analyzing for information about, judge whether to report to the police and the related data of tissue registration's collection.
Engine intelligent fault diagnosis system structure is as shown in Fig. 1.
(1) implementation procedure of the fish-swarm algorithm improved
1) to bunch behavior: the state of establishing Artificial Fish current is X
1, the number of partners in its visible range is n
1, the center individuality in its horizon range is X
c.At n
1>=1 and crowding m < delta now when, if X
1the X of the center of partner in neighborhood
clegal individuality, then an X
1
With crossing-over rate P
1with X
cintersect; If X
1the X of the center of partner in neighborhood
cnot legal individuality, then an X
csearch its nearest individual X of distance in global scope
s, X
1with crossing-over rate P
1with X
sintersect. at n
1when < 1 or m>=delta, individual X
1all perform foraging behavior.
Crossing-over rate P
1generally be set as 0.5-0.7, because individual X now
cor X
sbe the center of horizon range, can think that this individuality is for locally more excellent, so larger P
1the characteristic of optimum individual can be retained better. wherein m ∈ (0,1) is the random random number produced.
2) to knock into the back behavior: the state of establishing Artificial Fish current is X
1, the number of partners in its visible range is n
2, the optimum individual in its horizon range is X
b.At n
2>=1 and m < delta when, X
1with crossing-over rate P
2with X
bintersect; At n
2when < 1 or m>=delta, individual X
1all perform foraging behavior.
P
2generally be set to 0.1-0.4, individual X now
bthe optimum individual in horizon range, but in order to keep the diversity of population, P
2should not establish too large.
3) foraging behavior: the state of establishing Artificial Fish current is X
1, the mutation probability P produced with self-adaptation
m1if carry out self variation. variation after individuality legal and more excellent than current individuality, then using this individuality as perform look for food after new individuality and foraging behavior terminate; If still more excellent individuality can not be searched out, then by X after performing try_number variation
1with the mutation probability P artificially specified
m2legal individuality after variation is as performing the new individuality of foraging behavior and foraging behavior terminates.
P
m2generally be set to 0 ~ 0.5, if the mode that cloud self-adaptation produces aberration rate lost efficacy, then individuality must be absorbed in local optimum and jump not out, therefore wanted human intervention, but can not destroy the more excellent fish individuality obtained.
P
m1be adopt cloud adaptive approach to produce, during generation, parameter agree as follows: f is the fitness of single individuality, and Yan f is the average fitness of population, f
maxfor the maximum adaptation degree of population.
１）Ｅ
ｘ＝ｆ，
E
n=(f
max-f)/c
1∥ c
1for control coefrficient,
H
e=E
n/ c
2∥ c
2for control coefrficient.
２）Ｅ
ｍ＝ＲＡＮＤＮ （Ｅ
ｎ，Ｈ
ｅ）。
３）ＩＦf≥ｆ ，
Ｐ
ｍ１＝ｋ
１×exp［－（ｆ－Ｅ
ｘ）２／（２Ｅ
２ ｍ）］，
ＥＬＳＥ
Ｐ
ｍ１＝ｋ
２。
RANDN(E
n, H
e) represent with E
nfor expecting, with H
ethe random number obtained for the normal distribution of standard deviation, k
1, k
2∈ (0,1) is constant.E
naffect the steep of cloud, E
nthe horizontal width that larger then cloud covers is larger, and larger cover width can make more more excellent individuality get less aberration rate; H
edetermine the dispersion degree of water dust, H
ethe excessive steady tendency may losing cloud model, H
etoo smallly may lose randomness ［ 11 ］, therefore c
1∈ (2.7,3.2), c
2∈ (5,15).
(2) specific implementation of the fish-swarm algorithm improved
1) crossing-over rate in initialization shoal of fish population number, each stage and the parameter such as aberration rate, maximum iteration time.
2) use mutual information calculates the mutual information between each variable.
3) adopt MWST algorithm to generate initial non-directed graph, and specify any one node to be that root node generates initial population.
4) calculate the BIC scoring of all initial population, find out the maximum score value of scoring and individuality.
5) by 4) result put into the bulletin board after initialization.
While(iterations < maximum iteration time)
for ｋ＝１：fishnum
6) if the fish individuality of this numbering meets condition of bunching, then behavior of bunching is performed; Otherwise execution foraging behavior.
7) if the fish individuality of this numbering meets the condition that knocks into the back, then behavior of knocking into the back is performed; Otherwise execution foraging behavior.
8) the fish individuality executing bunch behavior and behavior of knocking into the back is compared, obtain the mode that this fish individuality " predation " is optimum, and record.
9) optimum individual in fishinum fish individuality after using once " predation " upgrades bulletin board.
10) judge whether to reach maximum iteration time.If reach maximum iteration time, then algorithm terminates; Otherwise proceed.
11) export the top score value after calculating, obtain optimum optimized network.
Above by reference to the accompanying drawings to invention has been exemplary description; obvious specific implementation of the present invention is not subject to the restrictions described above; as long as have employed the improvement of the various unsubstantialities that method of the present invention is conceived and technical scheme is carried out; or design of the present invention and technical scheme directly applied to other occasion, all within protection scope of the present invention without to improve.
Claims (3)
1. based on a Fault Diagnosis of Engine of RVM, it is characterized in that: comprise the steps:
Step one: pre-service: engine operational speed and the amplitude of correspondence thereof, the priori data of frequency samples are normalized, and set up the amplitude of travelling speed and correspondence thereof, the corresponding relation between frequency and engine condition;
Step 2: machine is trained: select suitable kernel function and carry out fish-swarm algorithm optimization training to its hyper parameter, setting up suitable RVM model;
Step 3: fault diagnosis: adopt " one to one " RVM sorter to carry out sample to be tested fault diagnosis and Output rusults.
2. the Fault Diagnosis of Engine based on RVM according to claim 1, is characterized in that: in described step 2, kernel function is gaussian radial basis function kernel function.
3. the Fault Diagnosis of Engine based on RVM according to claim 1, is characterized in that: the implementation procedure of described fish-swarm algorithm is:
A. the crossing-over rate in initialization shoal of fish population number, each stage and aberration rate, maximum iteration time;
B. use mutual information calculates the mutual information between each variable;
C. adopt MWST algorithm to generate initial non-directed graph, and specify any one node to be that root node generates initial population;
D. calculate the BIC scoring of all initial population, find out the maximum score value of scoring and individuality;
E. the result of d is put into the bulletin board after initialization;
While(iterations < maximum iteration time)
for ｋ＝１：fishnum
If f. the fish individuality of this numbering meets condition of bunching, then perform behavior of bunching; Otherwise execution foraging behavior;
If g. the fish individuality of this numbering meets the condition that knocks into the back, then perform behavior of knocking into the back; Otherwise execution foraging behavior;
H. the fish individuality executing bunch behavior and behavior of knocking into the back is compared, obtain the mode that this fish individuality " predation " is optimum, and record;
I. the optimum individual in the fishinum after " predation " fish individuality is used once to upgrade bulletin board;
J. judge whether to reach maximum iteration time, if reach maximum iteration time, then algorithm terminates; Otherwise proceed;
K. export the top score value after calculating, obtain optimized parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410586671.0A CN104680232A (en) | 2014-10-28 | 2014-10-28 | RVM (Relevance Vector Machine)-based engine failure detecting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410586671.0A CN104680232A (en) | 2014-10-28 | 2014-10-28 | RVM (Relevance Vector Machine)-based engine failure detecting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104680232A true CN104680232A (en) | 2015-06-03 |
Family
ID=53315243
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410586671.0A Pending CN104680232A (en) | 2014-10-28 | 2014-10-28 | RVM (Relevance Vector Machine)-based engine failure detecting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104680232A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106571016A (en) * | 2016-11-03 | 2017-04-19 | 西安交通大学 | Early-stage fault discrimination method of machinery based on alarm frequency jump triggering mechanism |
| CN109948237A (en) * | 2019-03-15 | 2019-06-28 | 中国汽车技术研究中心有限公司 | A method of for predicting bicycle discharge amount |
| CN111523603B (en) * | 2020-04-27 | 2024-02-02 | 江苏科技大学 | Ship power equipment fault identification method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1997049977A1 (en) * | 1996-06-24 | 1997-12-31 | Arcelik A.S. | Model-based fault detection system for electric motors |
| CN102393733A (en) * | 2011-10-28 | 2012-03-28 | 北京清佰华通科技有限公司 | Failure diagnosis method, fault diagnosis instrument and system thereof |
-
2014
- 2014-10-28 CN CN201410586671.0A patent/CN104680232A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1997049977A1 (en) * | 1996-06-24 | 1997-12-31 | Arcelik A.S. | Model-based fault detection system for electric motors |
| CN102393733A (en) * | 2011-10-28 | 2012-03-28 | 北京清佰华通科技有限公司 | Failure diagnosis method, fault diagnosis instrument and system thereof |
Non-Patent Citations (2)
| Title |
|---|
| 毕晓君等: "基于PSO-RVM算法的发动机故障诊断", 《哈尔滨工程大学学报》 * |
| 郭童等: "基于混合遗传鱼群算法的贝叶斯网络结构学习", 《浙江大学学报（工学版）》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106571016A (en) * | 2016-11-03 | 2017-04-19 | 西安交通大学 | Early-stage fault discrimination method of machinery based on alarm frequency jump triggering mechanism |
| CN109948237A (en) * | 2019-03-15 | 2019-06-28 | 中国汽车技术研究中心有限公司 | A method of for predicting bicycle discharge amount |
| CN111523603B (en) * | 2020-04-27 | 2024-02-02 | 江苏科技大学 | Ship power equipment fault identification method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huang et al. | Rolling bearing fault diagnosis and performance degradation assessment under variable operation conditions based on nuisance attribute projection | |
| CN106817248B (en) | APT attack detection method | |
| Hoang et al. | An efficient hidden Markov model training scheme for anomaly intrusion detection of server applications based on system calls | |
| CN111458142A (en) | Sliding bearing fault diagnosis method based on generation of countermeasure network and convolutional neural network | |
| CN101494567B (en) | Detection method for distributed abnegation service attack based on load prediction | |
| CN112087442B (en) | Time sequence related network intrusion detection method based on attention mechanism | |
| Zhang et al. | Dual generative adversarial networks based unknown encryption ransomware attack detection | |
| Mbow et al. | An intrusion detection system for imbalanced dataset based on deep learning | |
| CN104680232A (en) | RVM (Relevance Vector Machine)-based engine failure detecting method | |
| CN109813542A (en) | The method for diagnosing faults of air-treatment unit based on production confrontation network | |
| CN106339322A (en) | Method for software behavior prediction based on HMM-ACO | |
| CN108896330B (en) | Fault diagnosis method for hydroelectric generating sets | |
| CN110532613A (en) | Ship power system operation mode recognition method and device | |
| CN104680234A (en) | Engine fault diagnosis method | |
| CN106339628A (en) | Hardware anti-virus device based on microarchitecture level | |
| Kang et al. | A transfer learning based abnormal can bus message detection system | |
| CN108376266A (en) | One-class support vector machines Optimization Method of Kernel Parameter based on sample edge point internal point | |
| CN117155701A (en) | Network flow intrusion detection method | |
| CN116916317A (en) | Invasion detection method based on white shark and random forest | |
| Bui et al. | A clustering-based shrink autoencoder for detecting anomalies in intrusion detection systems | |
| Guo | An intrusion detection system for the internet of things using machine learning models | |
| CN116647844A (en) | Vehicle-mounted network intrusion detection method based on stacking integration algorithm | |
| CN108734009A (en) | The optimal inspection vector generation method merged based on genetic algorithm and MERS | |
| CN114357864A (en) | Phase modulator state evaluation method and evaluation system based on fuzzy reasoning | |
| CN114707566A (en) | Intelligent networking automobile abnormity intelligent detection method and device and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20150603 |