CN101634605A - Intelligent gearbox fault diagnosis method based on mixed inference and neural network - Google Patents
Intelligent gearbox fault diagnosis method based on mixed inference and neural network Download PDFInfo
- Publication number
- CN101634605A CN101634605A CN200910081750A CN200910081750A CN101634605A CN 101634605 A CN101634605 A CN 101634605A CN 200910081750 A CN200910081750 A CN 200910081750A CN 200910081750 A CN200910081750 A CN 200910081750A CN 101634605 A CN101634605 A CN 101634605A
- Authority
- CN
- China
- Prior art keywords
- neural network
- intelligent
- fault diagnosis
- analysis
- fault
- 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
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The invention relates to an intelligent gearbox fault diagnosis method based on mixed inference and neural network, mainly comprising the following steps: enriching a knowledge base of an intelligent diagnosis system; storing expert knowledge into a computer; combining the neural network to recognize patterns; analyzing a main abnormal phenomenon of a gearbox; systematically researching generation mechanisms of vibration and noise of a gear; analyzing a mathematical model of gear vibration; analyzing the vibration mechanisms of various fault types by fully using on-site fault data; enriching time domain and frequency domain characteristics of various faults; and ensuring a characteristic extraction method of gear fault signals. A data analysis method mainly comprises the steps: band-pass filter frequency analysis, double-meaning analysis, envelope demodulation analysis and demodulation analysis based on EMD. The intelligent gearbox fault diagnosis method based on mixed inference and neural network solves the problems of the traditional serial information processing in the fields of pattern recognition, artificial intelligence, and the like so that a pattern recognition network is completely realized by the computer; and the intelligent gearbox fault diagnosis method based on mixed inference and neural network is convenient to realize the automation and the intelligence of the fault diagnosis and can provide a basis for a diagnosis decision in time.
Description
Technical field
The present invention relates to a kind of intelligent gearbox fault diagnosis method, can be further developed into the more perfect intelligent diagnosis system of function.Based on the rule-based reasoning (BRB) of fault with based on the reasoning by cases (CBR) of a large amount of real case, utilize artificial neural network (ANN) again with powerful pattern classification function, form the three-dimensional diagnostic method of the multi-method of physical fault.Belong to the Fault Diagnosis of Gear Case field.
Background technology
Artificial intelligence has formed intelligent diagnostics with combining of diagnosis theory, the simulation human brain thinking reasoning of drought period development, knowledge-based expert system is with the form access arrangement diagnostic field of serial operation, formed the diagnostic reasoning expert system based on knowledge.Intelligent trouble diagnosis technology based on knowledge is one of the most noticeable developing direction in the Device Diagnostic field, also is maximum, the most widely used class intelligent diagnostics technology of research.It has roughly experienced two developing stage: based on the first generation fault diagnosis expert system of shallow knowledge (human expert's experimental knowledge) with based on the second generation fault diagnosis expert system of knowing very well knowledge (the model knowledge of diagnosis object).A large amount of cases of diagnosis object can constitute new inference rule, add nearest neighbor method and search in existing a large amount of cases in system, find immediate fault mode to infer new failure modes.The expert system of the mixed structure that occurs is that above-mentioned two kinds of methods are used in combination in the recent period, and is complementary not enough, brings out the best in each other.The applied research of neural network in the equipment fault diagnosis field mainly concentrates on aspect three: a. uses neural network from the pattern-recognition angle and carries out fault diagnosis as sorter; B. use neural network from the prediction angle and carry out failure prediction as dynamic prediction model; C. from the diagnostic expert system of knowledge processing angle foundation based on neural network.Expert system is based on the inference system of symbol, and there is the shortcoming of knowledge acquisition difficulty in it, but possesses explanation function.The input of neural network is that the sign of diagnosed object is an eigenwert, export the possibility that then expression is broken down, but it does not possess explanation function.Neural network is combined utilization with the mixed inference expert system, and performance advantage separately is a kind of new way of carrying out fault diagnosis.
To the history in existing more than 40 year of study of neural networks, the one section winding raod journey of having passed by.As far back as the forties, some scientists just study neural network from different angles such as Neuscience, mathematics, physics, psychology, biology, cognitive science and bionics, have obtained certain achievement.By 1969, the founder Minsky of artificial intelligence etc. are to being that the function and the limitation thereof of the network system of representative done deep research from mathematics with the perceptron, and published " Perceptron " book in 1969, his conclusion is pessimistic, negated the effect of artificial neural network, the research of this respect was disappeared basically in the latter stage sixties.Certainly more chief reason is the period of great prosperity of traditional just at that time digital machine, thereby has hindered the development of neural network.To the eighties, basic variation has taken place in situation, has run into great difficulty because traditional serial information is handled in fields such as pattern-recognition and artificial intelligence, impels people to remove to study the artificial neural network that is treated to feature with parallel information with bigger interest.
Neural network has the unexistent advantage of a series of traditional linear systems, as non-linear, adaptivity, concurrency and fault-tolerance etc., owing to had these new features, the application of neural network expands to control, the selection of communication network dynamic route and the communication controler etc. of the differentiation of automatic retrieval, postcode and identification, literal identification, data compression, message volume; At military aspect multiple target tracking, fighter flight control, auditory localization and combat decision-making etc. are arranged; Neural network also has application in other field, no longer sets forth.
Therefore, in the Fault Diagnosis of Gear Case field, demand developing a kind of artificial neural network urgently as a kind of adaptive mode identification technology, this technology does not need to provide in advance priori and the discriminant function about pattern, and it forms desired decision region automatically by the study mechanism of self.The characteristic of network by its topological structure, node characteristic, study or training rule determine that it can make full use of status information, to training one by one from the information of different conditions obtaining certain mapping relations, and network can be learnt continuously.Work as environment change, this mapping relations can self-adaptation, in the hope of further approaching object.For making neural network reach the purpose of pattern-recognition, use, is found out for fault and is reflected the input vector of the most responsive characteristic signal as neural network by feature selecting from the vibration signal of machine different conditions, set up the fault mode training sample set, network is trained.When network training finishes, for the status information of each new input, network will provide classification results rapidly.
Summary of the invention
The objective of the invention is to, by a kind of intelligent gearbox fault diagnosis method based on mixed inference and neural network is provided, make the pattern-recognition complete network by computer realization, be convenient to realize the robotization and the intellectuality of fault diagnosis, raise the efficiency, can in time provide foundation for the diagnosis decision-making.
The present invention adopts following technological means to realize:
Based on the intelligent gearbox fault diagnosis method of mixed inference and neural network,, add the diagnostic rule that demodulation analysis forms by enriching and improving existing intelligent diagnostics knowledge base; And the feature that various data analysing methods extract carried out the identification of fault mode as the input of neural network, with existing case and these two kinds of inference methods of rule again in conjunction with Application of Neural Network in the intelligent fault diagnosis system of high-speed rod-rolling mill; This method is replenished inference mechanism by enriching knowledge base, sets up the emulation neural network module, sets up a Practical Intelligent diagnostic system; Mainly may further comprise the steps:
Enriching of intelligent diagnosis system knowledge base; Expertise is deposited in computing machine, and carry out pattern-recognition in conjunction with neural network;
Analyze the main abnormal phenomenon of gear case, systematically study the vibration of gear, the mechanism that noise produces;
Analyze the mathematical model of gear vibration, make full use of the field failure data, analyze the vibration mechanism of various fault types; Enrich the time domain and the frequency domain character of various faults, and the feature extracting method of definite gear distress signal;
Described data analysing method mainly comprises: bandpass filtering spectrum analysis, having a double meaning analysis, envelope demodulation analysis, based on the demodulation analysis of EMD.
Aforesaid demodulation analysis includes the feature of envelope time domain parameter, demodulation spectrogram.
Aforesaid emulation neural network module, adopt the network structure of propagating neural network algorithm backward to realize that this network structure implementation method comprises:
(1) sets up network: determine the number of input layer, hidden layer and output layer, determine sample space total amount and study maximum cycle;
(2) training sample database: store a sample earlier, and then read other samples;
(3) measured data: read the characteristic in the gear case experiment, as the input data;
(4) data normalization: input feature value is carried out normalized;
(5) training network: train the weights that finish to the knowledge base storage;
(6) knowledge base: access right value matrix;
(7) diagnostic module: the measured data and the weight matrix that read after the normalization come the actual condition of gear case is diagnosed.
Aforesaid intelligent gearbox fault diagnosis method based on mixed inference and neural network is characterized in that: be optimized with combining of intelligent diagnosis system based on neural network based on the fault diagnosis expert system of knowledge.
Aforesaid optimization may further comprise the steps: data are earlier through the arest neighbors retrieval of case library, obtain exporting diagnosis report behind the similar solution; Not obtaining similar solution selects to carry out reasoning by predetermined rule in the knowledge base again; If there is not corresponding rule, select the neural network reasoning again.
Aforesaid optimization also comprises with the data of known conclusion to be verified.
Aforesaid checking comprises:
(1) simulating, verifying;
(2) experimental verification.
The different demodulation effects of aforesaid simulating, verifying by empirical mode decomposition, related function and Hilbert envelope illustrates that related function combines the envelope information of extraction fault-signal with Hilbert envelope demodulation method based on empirical mode decomposition.
The present invention is based on the intelligent gearbox fault diagnosis method of mixed inference and neural network, compared with prior art, have following remarkable advantages and beneficial effect:
The present invention has overcome traditional serial information and has handled in the existing problem in fields such as pattern-recognition and artificial intelligence, make full use of status information, to training one by one to obtain certain mapping relations from the information of different conditions, this mapping relations can self-adaptation, in the hope of further approaching object.For making neural network reach the purpose of pattern-recognition, use, is found out for fault and is reflected the input vector of the most responsive characteristic signal as neural network by feature selecting from the vibration signal of machine different conditions, set up the fault mode training sample set, network is trained.When network training finishes, for the status information of each new input, network will provide classification results rapidly.Make this pattern-recognition complete network by computer realization, be convenient to realize the robotization and the intellectuality of fault diagnosis, improved efficient, can in time provide foundation for the diagnosis decision-making.
Description of drawings
Fig. 1 knowledge base is replenished approach;
The position of Fig. 2 neural network in system and the relation of other modules;
Fig. 3 neural network design diagram;
Fig. 4 is based on the method for diagnosing faults process flow diagram of empirical mode decomposition and artificial neural network;
Three layers of back-propagating algorithm of Fig. 5 network model;
The high line finishing mill of Fig. 6 gearbox drive chain figure;
The time domain waveform of Fig. 7 simulate signal and frequency spectrum;
Time domain waveform after Fig. 8 correlation analysis and Hilbert frequency spectrum;
Fig. 9 stand-alone mode component;
The envelope demodulation spectrum of first three rank stand-alone mode component of Figure 10;
Figure 11 is through the stand-alone mode component of correlation analysis;
Figure 12,14 envelope demodulation spectrums based on correlation analysis and empirical mode decomposition;
The time domain of Figure 13 experimental data and frequency spectrum thereof.
Embodiment
Content in conjunction with the inventive method provides embodiment:
1, knowledge base replenishes
Knowledge base is basis and the prerequisite that inference machine carries out diagnostic reasoning.It deposits the knowledge that the domain expert has in computing machine with the certain expression form, and knowledge is carried out facility and effectively management.
At first, analyze the main abnormal phenomenon of gear case, systematically study the vibration of gear, the mechanism that noise produces.Analyze the mathematical model of gear vibration, the principal element that influences the gear train vibration is discussed.Analyze the vibration mechanism of various fault types on this basis.Enrich the time domain and the frequency domain character of various faults, and discuss the feature extracting method of gear distress signal in detail.Choose sensitive parameter on this basis meticulously, gather expertise.Traditional method is the theoretical formation rule of time frequency analysis, using the several data analytical approach in the present invention analyzes the fault and the hidden danger fault in gear case each period, data analysing method mainly comprises the bandpass filtering spectrum analysis, envelope demodulation is analyzed, correlation analysis, based on demodulation analysis of empirical mode decomposition etc., the criterion of demodulation analysis has the feature of envelope time domain parameter, demodulation spectrogram etc.On existing research basis, enrich knowledge base from point of theory.
Secondly, case library is as the basis of knowledge base, also for the extracting rule service.By the existing case of sorting-out in statistics, research can obtain the distinctive fault rule of on-the-spot case.The main diagnosis that provides according to the various places monitoring system case library that upgrades in time.In addition, original case also needs certain modification and storage.Constantly enriching on the basis of on-the-spot case, the angle of extracting rule is enriched knowledge base from existing case.
See also shown in Figure 1ly, be the additional approach of knowledge base.Enriching of knowledge base is main by this two aspect, and the process need of substantial knowledge base is verified repeatedly and adhered to that for a long time fault can not only take place according to intrinsic pattern, but also can extract regularity from seem the various faults that mix.
Exploitation and the application of neural network in the gear case intelligent diagnosis system
Obtain in experiment to use the intelligent diagnostics that neural network is carried out gearbox fault on the basis of data.
See also shown in Figure 2ly, be the position of neural network in system and the relation of other modules.Study by to the input sample is stored in the form of knowledge with weights and threshold value in the network.Such as, choose of the input of some sensitive factors by the analysis of amplitude parameter as neural network, kurtosis index, nargin index and pulse index impact pulse class failure ratio are responsive, and particularly when fault took place in early days, they were significantly increased; But after rising to a certain degree, the development gradually with fault can descend on the contrary, shows that they have higher susceptibility to initial failure, but stability is bad.Can make full use of this point, in the neural network reasoning, pay close attention to the trend of sensitive factor and just can make correct diagnosis.The method that all right application experience pattern is decomposed, the energy feature that comprises most of dominance failure message that extracts from a series of empirical mode components is as the input of neural network.Have functions such as knowledge is obtained automatically, parallel processing, adaptive learning, association's reasoning and fault-tolerance preferably based on the Fault Diagnosis of Gear Case expert system of neural network.
See also shown in Figure 3ly, be this Module Design synoptic diagram.Neural network module is one of nucleus module of this Fault Diagnosis of Gear Case expert system, and it adopts back-propagating algorithm network structure to realize.This module also can independently realize fault diagnosis as the important component part of expert system.
The implementation method of this module is as follows:
(1) sets up network: determine the number of input layer, hidden layer and output layer, determine sample space total amount and study maximum cycle;
(2) training sample database: store a sample earlier, and then read other samples;
(3) measured data: read the characteristic in the gear case experiment, as the input data;
(4) data normalization: input feature value is carried out normalized;
(5) training network: train the weights that finish to the knowledge base storage;
(6) knowledge base: access right value matrix;
(7) diagnostic module: the measured data and the weight matrix that read after the normalization come the actual condition of gear case is diagnosed.
Set up neural network model:
See also shown in Figure 4ly, be the diagnostic method process flow diagram.By the demodulation method of research based on empirical mode decomposition, obtain fault characteristic information, can be used as the input vector of neural network.At first, former starting acceleration vibration signal is broken down into the natural mode function (IMFs) of limited quantity.When fault takes place, the energy of vibration signal will change between different frequency bands.Therefore, discern the various fault modes of gear case, the energy feature that comprises most of dominance failure message that from a series of IMFs, extracts, these features can save as the input vector of artificial neural network.
Different with the logic high level model of traditional expert system based on the fault diagnosis expert system of neural network, it is a kind of low layer numerical model, and information processing is to be undertaken by the interaction between the simple process unit that is referred to as node in a large number.Because its distributed information hold mode, for expertise obtain and expression and reasoning provide brand-new mode.By study to the experience sample, the form of expertise with weights and threshold value is stored in the network, and utilize the information retentivity of network to finish the out of true diagnostic reasoning, simulated preferably the expert by rule of thumb, the reasoning process of intuition rather than complicated calculating.
See also the back-propagating algorithm neural network model schematic diagram that Figure 5 shows that three layers, in actual applications, generally getting hidden layer is that one deck constitutes one three layers BP network, to reduce calculated amount and to avoid the complicated network structureization.With three layers BP networks is its principle of example explanation.Except that input, output layer node, there is one deck to conceal node, neuronic characteristic function is chosen to be continuously differentiable Sigmoid function.
Neural network possessed extracts the required characteristic parameter of diagnosis from the non-linear mapping capability that is input to output and signal processing technology from the signal of gathering function is of equal value.Neural network can be used for signal Processing.But in most of the cases neural network is as the state recognition instrument, and as the follow-up phase of signal Processing, signal Processing then is the pre-process of neural network in diagnostic procedure.
3, intelligent diagnostics is globality optimization
Based on the fault diagnosis expert system of knowledge with based on the combining of the intelligent diagnosis system of neural network, the subsystem in the total system all has the function of independently carrying out intelligent diagnostics.Data are earlier through the arest neighbors retrieval of case library, obtain exporting diagnosis report behind the similar solution; Not obtaining similar solution selects to carry out reasoning by predetermined rule in the knowledge base again; If there is not corresponding rule, select the neural network reasoning again.This flow process is seen top shown in Figure 2.Can verify that the validity of contrast rule-based reasoning and neural network reasoning is by substantial respectively again rule or the neural network improved of contrast with the data of known conclusion.By the progressively optimization and the improvement of each subsystem, also comprise local optimum and be connected optimization, finally make whole intelligence system function be tending towards powerful and perfect, improve the accurate rate of intelligent diagnostics.
Hybrid intelligent diagnostic system based on reasoning by cases, rule-based reasoning and back-propagating neural network.At first be the case library of bottom, the fault diagnosis expert system that is based on knowledge again with based on the combining of the intelligent diagnosis system of neural network, the subsystem in the total system all has the function of independently carrying out intelligent diagnostics.Data are earlier through the arest neighbors retrieval of case library, obtain exporting diagnosis report behind the similar solution; Not obtaining similar solution selects to carry out reasoning by predetermined rule in the knowledge base again; Perhaps select the neural network reasoning.
At first, prove to use to have high sensitivity based on signal processing methods such as empirical mode decomposition envelope demodulations, the analysis of failure data are carried out fault signature and are extracted better.
(1) simulating, verifying
See also and Figure 6 shows that high line finishing mill gearbox drive chain figure;
Following emulation is by the different demodulation effects of empirical mode decomposition, related function and Hilbert envelope, prove absolutely that related function combines with Hilbert envelope demodulation method based on empirical mode decomposition, can farthest extract the envelope information of fault-signal.
Simulate signal is as follows, 300 hertz of carrier frequencies, and 50 hertz of amplitude modulation(PAM) frequencies, E is a noise figure, gets E=1.5 here.The simulation process process is undertaken by three schemes, and every kind of scheme all is divided into noise and contrasts than big and the less two kinds of situations of noise.
y＝(1+cos50πt)cos(300πt)+cos(100πt)+En(t)
Scheme one: correlation analysis+Hilbert envelope
Fig. 7 is the original waveform and the frequency spectrum of simulate signal, and this moment is because its spectrum component more complicated of interference of noise.Through can seeing that noise obviously reduces after the correlation analysis, in the frequency spectrum modulation intelligence can be more clearly reflect.Fig. 8 (b) is the result that the signal after the relevant noise reduction is carried out envelope demodulation.Because the existence of addition composition is arranged, beyond all recognition other frequency contents have appearred in the demodulation spectra, rather than 25 hertz of our needed modulation intelligences.Therefore need carry out empirical mode decomposition to signal, obtain the stand-alone mode component that wherein comprises modulation intelligence, carry out demodulation analysis again.
Scheme two: empirical mode decomposition+Hilbert envelope
Cause frequency interferences owing to comprise the additivity factor in the signal, therefore earlier simulate signal is carried out empirical mode decomposition, the stand-alone mode component of isolating wherein carries out demodulation process again.
Fig. 9 is the result of simulate signal through empirical mode decomposition, comprises 9 stand-alone mode components altogether, and what provide among the figure is preceding 5 rank.Fig. 5-the 4th, the envelope demodulation spectrum of first three rank stand-alone mode component.Wherein the second rank demodulation spectra is put on and can be reflected certain failure message.But because interference of noise, this part message reflection be not fairly obvious.Therefore before carrying out empirical mode decomposition and Hilbert envelope demodulation, be necessary earlier analyzed signal to be carried out noise reduction.
Scheme three: correlation analysis+empirical mode decomposition+Hilbert
Figure 10 and Figure 11 are respectively empirical mode decomposition stand-alone mode component and corresponding first three the rank Hilbert demodulation spectra after correlation analysis.Gem-pure 25 hertz of modulating frequencies that reflected of the demodulation spectra of the first rank stand-alone mode component wherein.Two kinds of schemes are compared, and scheme three has significantly improved the validity and the sensitivity of Hilbert demodulation method.
Hilbert envelope demodulation method based on correlation analysis and empirical mode decomposition is the effective ways that carry out Gear Fault Diagnosis.Simple Hilbert envelope demodulation method is because the unicity of technological means and the limitation of self, often is difficult to extracting promptly and accurately for the fault under the complex working condition.Correlation analysis and empirical mode decomposition (EMD) then can overcome these shortcomings.Correlation analysis can effectively be removed noise under the prerequisite that does not change frequency content, show fault signature; Empirical mode decomposition then can be carried out auto adapted filtering to signal, decomposites a plurality of stand-alone mode components and then handles, and has avoided the additivity composition to produce situation about disturbing, and makes diagnostic result more accurate.
(2) experimental verification
Adopt bearing test data analysis that platform is surveyed, the result is shown in Figure 13,14.
Figure 13 (a) and 13 (b) be respectively experiment table adopt bearing vibration signal time domain waveform and frequency spectrum, Figure 14 is for through based on the result after the Hilbert demodulation of empirical mode decomposition.As can be seen from the figure, the frequency spectrum of first three rank stand-alone mode component has all demodulated 76.4 hertz of bearing outer ring faults and frequency multiplication composition thereof.
Secondly, new data analysing method can be enrolled expert system on the one hand, as diagnostic rule.Can be used as case library carries out the fault mode nearest neighbor search on the one hand.In the time of data analysis, retrieve the intelligence system that draws diagnosis report automatically according to expertise and actual diagnosis case, but data analysis only limits to time-domain analysis and frequency-domain analysis.
At last, the training of network is carried out in all kinds of fault datas that normal data and various analysis the are obtained good input as the back-propagating neural network of classifying.Go to diagnose new data with the network that trains then, carry out pattern classification.Enrich in intelligent diagnosis system as standalone module.Use matlab and write the back-propagating network, Network Design is similar, and concrete network training and emulation are no longer set forth.This step needs lot of data to carry out network training.Fact proved that this pattern-recognition complete network is convenient to realize the robotization and the intellectuality of fault diagnosis by computer realization, has improved efficient, can in time provide foundation for the diagnosis decision-making.
Claims (8)
1, based on the intelligent gearbox fault diagnosis method of mixed inference and neural network,, adds the diagnostic rule that demodulation analysis forms by enriching and improving existing intelligent diagnostics knowledge base; And the feature that various data analysing methods extract carried out the identification of fault mode as the input of neural network, with existing case and these two kinds of inference methods of rule again in conjunction with Application of Neural Network in the intelligent fault diagnosis system of high-speed rod-rolling mill; It is characterized in that: this method is replenished inference mechanism by enriching knowledge base, sets up the emulation neural network module, sets up a Practical Intelligent diagnostic system; Mainly may further comprise the steps:
Enriching of intelligent diagnosis system knowledge base;
Deposit expertise in computing machine, and carry out pattern-recognition in conjunction with neural network;
Analyze the main abnormal phenomenon of gear case, systematically study the vibration of gear, the mechanism that noise produces;
Analyze the mathematical model of gear vibration, make full use of the field failure data, analyze the vibration mechanism of various fault types;
Enrich the time domain and the frequency domain character of various faults, and the feature extracting method of definite gear distress signal;
Described data analysing method mainly comprises: bandpass filtering spectrum analysis, having a double meaning analysis, envelope demodulation analysis, based on the demodulation analysis of EMD.
2, the intelligent gearbox fault diagnosis method based on mixed inference and neural network according to claim 1 is characterized in that: described demodulation analysis includes the feature of envelope time domain parameter, demodulation spectrogram.
3, the intelligent gearbox fault diagnosis method based on mixed inference and neural network according to claim 1, it is characterized in that: described emulation neural network module, adopt the network structure of propagating neural network algorithm backward to realize that this network structure implementation method comprises:
(1) sets up network: determine the number of input layer, hidden layer and output layer, determine sample space total amount and study maximum cycle;
(2) training sample database: store a sample earlier, and then read other samples;
(3) measured data: read the characteristic in the gear case experiment, as the input data;
(4) data normalization: input feature value is carried out normalized;
(5) training network: train the weights that finish to the knowledge base storage;
(6) knowledge base: access right value matrix;
(7) diagnostic module: the measured data and the weight matrix that read after the normalization come the actual condition of gear case is diagnosed.
4, the intelligent gearbox fault diagnosis method based on mixed inference and neural network according to claim 1 is characterized in that: be optimized with combining of intelligent diagnosis system based on neural network based on the fault diagnosis expert system of knowledge.
5, the intelligent gearbox fault diagnosis method based on mixed inference and neural network according to claim 1, it is characterized in that: described optimization may further comprise the steps: data are earlier through the arest neighbors retrieval of case library, obtain exporting diagnosis report behind the similar solution; Not obtaining similar solution selects to carry out reasoning by predetermined rule in the knowledge base again; If there is not corresponding rule, select the neural network reasoning again.
6, according to claim 1 or 4 described intelligent gearbox fault diagnosis methods based on mixed inference and neural network, it is characterized in that: described optimization also comprises with the data of known conclusion to be verified.
7, according to claim 1 or 6 described intelligent gearbox fault diagnosis methods based on mixed inference and neural network, it is characterized in that: described checking comprises:
(1) simulating, verifying;
(2) experimental verification.
8, the intelligent gearbox fault diagnosis method based on mixed inference and neural network according to claim 7, it is characterized in that: described simulating, verifying is by the different demodulation effects of empirical mode decomposition, related function and Hilbert envelope, illustrate that related function combines with Hilbert envelope demodulation method based on empirical mode decomposition, extracts the envelope information of fault-signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100817505A CN101634605B (en) | 2009-04-10 | 2009-04-10 | Intelligent gearbox fault diagnosis method based on mixed inference and neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100817505A CN101634605B (en) | 2009-04-10 | 2009-04-10 | Intelligent gearbox fault diagnosis method based on mixed inference and neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101634605A true CN101634605A (en) | 2010-01-27 |
| CN101634605B CN101634605B (en) | 2011-03-30 |
Family
ID=41593864
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009100817505A Expired - Fee Related CN101634605B (en) | 2009-04-10 | 2009-04-10 | Intelligent gearbox fault diagnosis method based on mixed inference and neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101634605B (en) |
Cited By (41)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101846992A (en) * | 2010-05-07 | 2010-09-29 | 上海理工大学 | Fault tree construction method based on fault case of numerical control machine |
| CN101870075A (en) * | 2010-07-02 | 2010-10-27 | 西南交通大学 | Method for predicting service life of screw pair of numerical control machine on basis of performance degradation model |
| CN102243497A (en) * | 2011-07-25 | 2011-11-16 | 江苏吉美思物联网产业股份有限公司 | Networking technology-based remote intelligent analysis service system used for engineering machinery |
| CN102507186A (en) * | 2011-11-01 | 2012-06-20 | 西安交通大学 | Characteristic parameter-based method for condition monitoring and fault identification of planetary gearbox |
| CN102620928A (en) * | 2012-03-02 | 2012-08-01 | 燕山大学 | Wind-power gear box fault diagnosis method based on wavelet medium-soft threshold and electronic-magnetic diaphragm (EMD) |
| CN102768115A (en) * | 2012-06-27 | 2012-11-07 | 华北电力大学 | Method for dynamically monitoring health status of wind turbine gearbox in real time |
| CN102840976A (en) * | 2012-08-24 | 2012-12-26 | 武汉钢铁(集团)公司 | Detection method for rolling machine main drive system |
| CN102930285A (en) * | 2012-09-18 | 2013-02-13 | 四川大学 | Early failure identification method based on SILLE (Supervised Increment Locally Linear Embedding) dimensionality reduction |
| CN102937522A (en) * | 2012-08-30 | 2013-02-20 | 桂林电子科技大学 | Composite fault diagnosis method and system of gear case |
| CN103267652A (en) * | 2013-05-24 | 2013-08-28 | 北京工业大学 | Intelligent online diagnosis method for early failures of equipment |
| CN103336198A (en) * | 2013-06-21 | 2013-10-02 | 中国人民解放军国防科学技术大学 | Electrical system fault diagnosing device |
| CN103344430A (en) * | 2013-07-09 | 2013-10-09 | 上海电机学院 | Fault diagnosis method of gearbox |
| CN103488887A (en) * | 2013-09-16 | 2014-01-01 | 北京理工大学 | Product reliability assessment method based on hybrid expert network |
| CN103761569A (en) * | 2013-12-20 | 2014-04-30 | 上海电机学院 | Fault diagnosis method and device for wind driven generator |
| CN104048825A (en) * | 2014-05-28 | 2014-09-17 | 北京航空航天大学 | Multi-sensor fusion type fault locating method for gearbox of wind turbine generator |
| CN104363129A (en) * | 2014-12-10 | 2015-02-18 | 国家电网公司 | Network event correlation analysis and dynamic early-warning method |
| CN104458250A (en) * | 2014-12-02 | 2015-03-25 | 上海电机学院 | Intelligent gearbox fault diagnosis method |
| CN104700148A (en) * | 2015-04-02 | 2015-06-10 | 刘岩 | Motor fault diagnosis method based on behavior |
| CN104699994A (en) * | 2015-04-02 | 2015-06-10 | 刘岩 | FBF (fuzzy basis function) neural network based motor air gap eccentricity fault diagnosis method |
| CN105204492A (en) * | 2015-09-08 | 2015-12-30 | 珠海格力电器股份有限公司 | Component fault diagnosis method, controller and air conditioner |
| CN105467975A (en) * | 2015-12-29 | 2016-04-06 | 山东鲁能软件技术有限公司 | Equipment fault diagnosis method |
| WO2016053748A1 (en) * | 2014-09-29 | 2016-04-07 | Sikorsky Aircraft Corporation | Vibration signatures for prognostics and health monitoring of machinery |
| CN106092564A (en) * | 2016-06-06 | 2016-11-09 | 电子科技大学 | The gear failure diagnosing method demodulated based on ESMD and energy operator |
| CN106124982A (en) * | 2016-06-14 | 2016-11-16 | 都城绿色能源有限公司 | Automatic expert's resultant fault diagnostic system of a kind of Wind turbines and diagnostic method |
| CN106610584A (en) * | 2015-10-27 | 2017-05-03 | 沈阳工业大学 | Remanufacturing process quality control method based on neural network and expert system |
| CN107101829A (en) * | 2017-04-11 | 2017-08-29 | 西北工业大学 | A kind of intelligent diagnosing method of aero-engine structure class failure |
| CN107193275A (en) * | 2017-07-01 | 2017-09-22 | 浙江三青环保科技有限公司 | A kind of equipment operation intelligent diagnosis system and diagnostic method |
| CN107390097A (en) * | 2017-07-17 | 2017-11-24 | 天津科技大学 | A kind of acoustoelectric combined shelf depreciation simulation detection system of GIS and its detection method |
| CN107862375A (en) * | 2017-10-30 | 2018-03-30 | 北京计算机技术及应用研究所 | A kind of two stage equipment fault diagnosis method |
| CN108363952A (en) * | 2017-01-27 | 2018-08-03 | 大隈株式会社 | Diagnostic device |
| CN108375971A (en) * | 2018-03-18 | 2018-08-07 | 哈尔滨工程大学 | Integrated Electronic System health control module and health control method for moonlet |
| CN108445868A (en) * | 2018-03-26 | 2018-08-24 | 安徽省爱夫卡电子科技有限公司 | A kind of automobile intelligent fault diagnosis system and method based on modern signal processing technology |
| CN108470084A (en) * | 2018-02-09 | 2018-08-31 | 浙江大学 | A kind of turbine shafting vibrating failure diagnosis method based on neural network and reasoning by cases |
| CN108613802A (en) * | 2018-05-10 | 2018-10-02 | 重庆大学 | A kind of mechanical failure diagnostic method based on depth mixed network structure |
| CN108647485A (en) * | 2018-05-18 | 2018-10-12 | 上海交通大学 | Catalyst carbon deposition measurement method, system, medium and equipment in fluid catalytic cracking |
| CN109061495A (en) * | 2018-08-07 | 2018-12-21 | 中国电建集团福建省电力勘测设计院有限公司 | A kind of hybrid energy-storing battery failure diagnostic method |
| CN109923558A (en) * | 2016-11-04 | 2019-06-21 | 谷歌有限责任公司 | Mixture of expert neural network |
| CN109932179A (en) * | 2019-04-09 | 2019-06-25 | 东南大学 | A kind of rolling bearing fault testing method based on the reconstruct of DS Adaptive spectra |
| CN111611279A (en) * | 2020-04-24 | 2020-09-01 | 中国电子科技集团公司第二十九研究所 | Microwave assembly fault diagnosis system and method based on test index similarity |
| CN112879278A (en) * | 2021-01-11 | 2021-06-01 | 苏州欣皓信息技术有限公司 | Pump station unit fault diagnosis method based on noise signal A weighting analysis |
| CN116538092A (en) * | 2023-07-06 | 2023-08-04 | 中国科学院理化技术研究所 | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE112018007597B4 (en) | 2018-06-18 | 2022-06-09 | Mitsubishi Electric Corporation | Diagnostic device, diagnostic method and program |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2681138Y (en) * | 2004-03-11 | 2005-02-23 | 武汉科技大学 | Fault diagnosis equipment for high power gear box |
| CN100507490C (en) * | 2006-12-29 | 2009-07-01 | 北京工业大学 | Large-sized low-speed heavy-loaded gear box fault diagnosis method |
-
2009
- 2009-04-10 CN CN2009100817505A patent/CN101634605B/en not_active Expired - Fee Related
Cited By (58)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101846992B (en) * | 2010-05-07 | 2011-12-07 | 上海理工大学 | Fault tree construction method based on fault case of numerical control machine |
| CN101846992A (en) * | 2010-05-07 | 2010-09-29 | 上海理工大学 | Fault tree construction method based on fault case of numerical control machine |
| CN101870075A (en) * | 2010-07-02 | 2010-10-27 | 西南交通大学 | Method for predicting service life of screw pair of numerical control machine on basis of performance degradation model |
| CN102243497B (en) * | 2011-07-25 | 2013-10-02 | 江苏吉美思物联网产业股份有限公司 | Networking technology-based remote intelligent analysis service system used for engineering machinery |
| CN102243497A (en) * | 2011-07-25 | 2011-11-16 | 江苏吉美思物联网产业股份有限公司 | Networking technology-based remote intelligent analysis service system used for engineering machinery |
| CN102507186A (en) * | 2011-11-01 | 2012-06-20 | 西安交通大学 | Characteristic parameter-based method for condition monitoring and fault identification of planetary gearbox |
| CN102620928A (en) * | 2012-03-02 | 2012-08-01 | 燕山大学 | Wind-power gear box fault diagnosis method based on wavelet medium-soft threshold and electronic-magnetic diaphragm (EMD) |
| CN102768115A (en) * | 2012-06-27 | 2012-11-07 | 华北电力大学 | Method for dynamically monitoring health status of wind turbine gearbox in real time |
| CN102768115B (en) * | 2012-06-27 | 2016-04-20 | 华北电力大学 | A kind of gearbox of wind turbine health status real-time dynamic monitoring method |
| CN102840976A (en) * | 2012-08-24 | 2012-12-26 | 武汉钢铁(集团)公司 | Detection method for rolling machine main drive system |
| CN102840976B (en) * | 2012-08-24 | 2015-10-28 | 武汉钢铁(集团)公司 | A kind of detection method of mill main drive system |
| CN102937522A (en) * | 2012-08-30 | 2013-02-20 | 桂林电子科技大学 | Composite fault diagnosis method and system of gear case |
| CN102930285A (en) * | 2012-09-18 | 2013-02-13 | 四川大学 | Early failure identification method based on SILLE (Supervised Increment Locally Linear Embedding) dimensionality reduction |
| CN103267652A (en) * | 2013-05-24 | 2013-08-28 | 北京工业大学 | Intelligent online diagnosis method for early failures of equipment |
| CN103267652B (en) * | 2013-05-24 | 2015-05-20 | 北京工业大学 | Intelligent online diagnosis method for early failures of equipment |
| CN103336198A (en) * | 2013-06-21 | 2013-10-02 | 中国人民解放军国防科学技术大学 | Electrical system fault diagnosing device |
| CN103336198B (en) * | 2013-06-21 | 2015-06-17 | 中国人民解放军国防科学技术大学 | Electrical system fault diagnosing device |
| CN103344430A (en) * | 2013-07-09 | 2013-10-09 | 上海电机学院 | Fault diagnosis method of gearbox |
| CN103488887B (en) * | 2013-09-16 | 2016-06-22 | 北京理工大学 | A kind of Reliability Assessment method based on Mixture of expert network |
| CN103488887A (en) * | 2013-09-16 | 2014-01-01 | 北京理工大学 | Product reliability assessment method based on hybrid expert network |
| CN103761569A (en) * | 2013-12-20 | 2014-04-30 | 上海电机学院 | Fault diagnosis method and device for wind driven generator |
| CN104048825A (en) * | 2014-05-28 | 2014-09-17 | 北京航空航天大学 | Multi-sensor fusion type fault locating method for gearbox of wind turbine generator |
| CN104048825B (en) * | 2014-05-28 | 2016-05-25 | 北京航空航天大学 | A kind of gearbox of wind turbine Fault Locating Method of Multi-sensor Fusion |
| WO2016053748A1 (en) * | 2014-09-29 | 2016-04-07 | Sikorsky Aircraft Corporation | Vibration signatures for prognostics and health monitoring of machinery |
| CN104458250A (en) * | 2014-12-02 | 2015-03-25 | 上海电机学院 | Intelligent gearbox fault diagnosis method |
| CN104363129A (en) * | 2014-12-10 | 2015-02-18 | 国家电网公司 | Network event correlation analysis and dynamic early-warning method |
| CN104363129B (en) * | 2014-12-10 | 2017-08-04 | 国家电网公司 | A kind of network event association analysis and dynamic early-warning method |
| CN104700148A (en) * | 2015-04-02 | 2015-06-10 | 刘岩 | Motor fault diagnosis method based on behavior |
| CN104699994A (en) * | 2015-04-02 | 2015-06-10 | 刘岩 | FBF (fuzzy basis function) neural network based motor air gap eccentricity fault diagnosis method |
| CN105204492A (en) * | 2015-09-08 | 2015-12-30 | 珠海格力电器股份有限公司 | Component fault diagnosis method, controller and air conditioner |
| CN106610584A (en) * | 2015-10-27 | 2017-05-03 | 沈阳工业大学 | Remanufacturing process quality control method based on neural network and expert system |
| CN105467975A (en) * | 2015-12-29 | 2016-04-06 | 山东鲁能软件技术有限公司 | Equipment fault diagnosis method |
| CN105467975B (en) * | 2015-12-29 | 2018-08-03 | 山东鲁能软件技术有限公司 | A kind of equipment fault diagnosis method |
| CN106092564A (en) * | 2016-06-06 | 2016-11-09 | 电子科技大学 | The gear failure diagnosing method demodulated based on ESMD and energy operator |
| CN106124982A (en) * | 2016-06-14 | 2016-11-16 | 都城绿色能源有限公司 | Automatic expert's resultant fault diagnostic system of a kind of Wind turbines and diagnostic method |
| US12067476B2 (en) | 2016-11-04 | 2024-08-20 | Google Llc | Mixture of experts neural networks |
| US11790214B2 (en) | 2016-11-04 | 2023-10-17 | Google Llc | Mixture of experts neural networks |
| CN109923558A (en) * | 2016-11-04 | 2019-06-21 | 谷歌有限责任公司 | Mixture of expert neural network |
| CN108363952A (en) * | 2017-01-27 | 2018-08-03 | 大隈株式会社 | Diagnostic device |
| CN107101829B (en) * | 2017-04-11 | 2019-03-29 | 西北工业大学 | A kind of intelligent diagnosing method of aero-engine structure class failure |
| CN107101829A (en) * | 2017-04-11 | 2017-08-29 | 西北工业大学 | A kind of intelligent diagnosing method of aero-engine structure class failure |
| CN107193275A (en) * | 2017-07-01 | 2017-09-22 | 浙江三青环保科技有限公司 | A kind of equipment operation intelligent diagnosis system and diagnostic method |
| CN107390097A (en) * | 2017-07-17 | 2017-11-24 | 天津科技大学 | A kind of acoustoelectric combined shelf depreciation simulation detection system of GIS and its detection method |
| CN107862375A (en) * | 2017-10-30 | 2018-03-30 | 北京计算机技术及应用研究所 | A kind of two stage equipment fault diagnosis method |
| CN108470084A (en) * | 2018-02-09 | 2018-08-31 | 浙江大学 | A kind of turbine shafting vibrating failure diagnosis method based on neural network and reasoning by cases |
| CN108375971A (en) * | 2018-03-18 | 2018-08-07 | 哈尔滨工程大学 | Integrated Electronic System health control module and health control method for moonlet |
| CN108445868A (en) * | 2018-03-26 | 2018-08-24 | 安徽省爱夫卡电子科技有限公司 | A kind of automobile intelligent fault diagnosis system and method based on modern signal processing technology |
| CN108613802A (en) * | 2018-05-10 | 2018-10-02 | 重庆大学 | A kind of mechanical failure diagnostic method based on depth mixed network structure |
| CN108613802B (en) * | 2018-05-10 | 2019-06-07 | 重庆大学 | A kind of mechanical failure diagnostic method based on depth mixed network structure |
| CN108647485A (en) * | 2018-05-18 | 2018-10-12 | 上海交通大学 | Catalyst carbon deposition measurement method, system, medium and equipment in fluid catalytic cracking |
| CN108647485B (en) * | 2018-05-18 | 2021-04-02 | 上海交通大学 | Method, system, medium and equipment for measuring catalyst carbon deposition in fluidized catalytic cracking |
| CN109061495A (en) * | 2018-08-07 | 2018-12-21 | 中国电建集团福建省电力勘测设计院有限公司 | A kind of hybrid energy-storing battery failure diagnostic method |
| CN109932179A (en) * | 2019-04-09 | 2019-06-25 | 东南大学 | A kind of rolling bearing fault testing method based on the reconstruct of DS Adaptive spectra |
| CN111611279A (en) * | 2020-04-24 | 2020-09-01 | 中国电子科技集团公司第二十九研究所 | Microwave assembly fault diagnosis system and method based on test index similarity |
| CN111611279B (en) * | 2020-04-24 | 2023-09-12 | 中国电子科技集团公司第二十九研究所 | Microwave component fault diagnosis system and method based on test index similarity |
| CN112879278A (en) * | 2021-01-11 | 2021-06-01 | 苏州欣皓信息技术有限公司 | Pump station unit fault diagnosis method based on noise signal A weighting analysis |
| CN116538092A (en) * | 2023-07-06 | 2023-08-04 | 中国科学院理化技术研究所 | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium |
| CN116538092B (en) * | 2023-07-06 | 2023-11-14 | 中国科学院理化技术研究所 | Compressor on-line monitoring and diagnosing method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101634605B (en) | 2011-03-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101634605B (en) | Intelligent gearbox fault diagnosis method based on mixed inference and neural network | |
| Yang et al. | Fault diagnosis of gearbox based on RBF-PF and particle swarm optimization wavelet neural network | |
| de Oliveira et al. | Deep learning for power quality | |
| CN106769052B (en) | A kind of mechanical system rolling bearing intelligent failure diagnosis method based on clustering | |
| CN102112933B (en) | Error detection method and system | |
| CN109543210A (en) | A kind of Wind turbines failure prediction system based on machine learning algorithm platform | |
| CN109165604A (en) | The recognition methods of non-intrusion type load and its test macro based on coorinated training | |
| CN108256556A (en) | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network | |
| CN110455537A (en) | A kind of Method for Bearing Fault Diagnosis and system | |
| CN117609836A (en) | Electromagnetic sensitivity prediction and health management method for integrated module | |
| CN109214356A (en) | A kind of fan transmission system intelligent fault diagnosis method based on DCNN model | |
| CN105319071A (en) | Diesel engine fuel oil system fault diagnosis method based on least square support vector machine | |
| CN109934335A (en) | High-speed railway track switch method for diagnosing faults based on interacting depth study | |
| CN111898644B (en) | Intelligent identification method for health state of aerospace liquid engine under fault-free sample | |
| CN110059845A (en) | Metering device clocking error trend forecasting method based on timing evolved genes model | |
| Wang et al. | A remaining useful life prediction model based on hybrid long-short sequences for engines | |
| CN116910574A (en) | Mechanical equipment diagnosis method and device based on meta-learning and time convolution network | |
| Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
| Rui et al. | Signal processing collaborated with deep learning: An interpretable FIRNet for industrial intelligent diagnosis | |
| CN103603794A (en) | Method and device for adaptive fault diagnosis of gas storage injection-production compressor unit | |
| CN106909136A (en) | A kind of method for diagnosing faults based on index regularization kernel linear discriminant analysis | |
| CN110222610A (en) | A kind of method and device of Modulation recognition | |
| Wang | Research on the fault diagnosis of mechanical equipment vibration system based on expert system | |
| CN110780660A (en) | Tobacco production industry control system fault diagnosis method based on production state | |
| Mengjiao et al. | Wind turbine bearing fault diagnosis method based on multi-domain feature extraction |
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 | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110330 Termination date: 20120410 |