CN1241721A - Fault detecting and diagnosing method based on non-linear spectral analysis - Google Patents
Fault detecting and diagnosing method based on non-linear spectral analysis Download PDFInfo
- Publication number
- CN1241721A CN1241721A CN 99115737 CN99115737A CN1241721A CN 1241721 A CN1241721 A CN 1241721A CN 99115737 CN99115737 CN 99115737 CN 99115737 A CN99115737 A CN 99115737A CN 1241721 A CN1241721 A CN 1241721A
- Authority
- CN
- China
- Prior art keywords
- model
- signal
- identification
- fault
- data
- 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
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The fault detection includes the steps of simplification of the model, design of the exciting signal, data collection and compression, selection of recognition algorithm, pre-treatment of signal, test of the model extending capacity, mode classification and fault diagnosis, and final critical fault state alarm and emergency treatment of fault state. The said method can be used in the state detection and fault diagnosis of both linear and non-linear systems, has high detection and diagnosis precision, and may be used in in-situ application.
Description
The invention belongs to the Control Science and Engineering ambit, further relate to a kind of state-detection and method for diagnosing faults based on non-linear spectral analysis.
At present, about state-detection and fault diagnosis two kinds of fundamental method are arranged, a kind of method is based on signal Processing, and is another kind of based on process analysis procedure analysis.
Based on the state-detection of signal Processing and method for diagnosing faults is fundamental method (see [U.S.] J.S. Mil work, Lin Mingbang etc. translate, " analysis of mechanical disorder and monitoring ", China Machine Press, 1990).Because the variation of parameters such as system output signal amplitude, phase place, frequency, correlativity often and have certain getting in touch between the source of trouble, can obtain the characteristics such as spectrogram of output signal by signal Processing, and can analyze thus the system of drawing of living in state.Method commonly used has Zymography, probability density method, time Sequence Analysis Method etc.The advantage of the method is a simple, intuitive, and is also very effective in many cases.But owing to only utilized output signal, the intrinsic propesties that can not reflect system fully is so this method has its limitation inevitably.Promptly for some systems, variation of output signals is the variation of reflection object feature fully, causes erroneous judgement easily.
Based on the method for process analysis procedure analysis, utilize the input/output signal of object exactly, obtain the variation of object transmission characteristic with identification Method, and judge thus whether system is in malfunction.Parameter model identification or nonparametric model identification all belong to these class methods (seeing Zhou Donghua, Sun Youxian work, " the fault detection and diagnosis technology of control system ", publishing house of Tsing-Hua University, 1994).Directly the method based on the parameter model identification is a kind of popular discrimination method, and advantage is that algorithm is general, realizes easily; Its shortcoming is that the accuracy of parameter estimation depends on the correct of model structure, and the model structure identification is a very difficult thing.
The objective of the invention is to overcome the shortcoming of above-mentioned prior art, a kind of state-detection and method for diagnosing faults based on non-linear spectral analysis proposed, not only go for the state-detection and the fault diagnosis of linear and nonlinear system, and can improve greatly and detect and diagnostic accuracy, can online practical application.
Core technology of the present invention is to utilize the frequency spectrum analysis method of nonlinear dynamic system to carry out state-detection and fault diagnosis.The spectrum analysis of so-called nonlinear dynamic system, utilize the actual measurement inputoutput data of non-linear object to obtain its generalized frequency response function (Generalized Frequency Response Function exactly, be called for short GFRF) model, or be called the non-linear transmission characteristic spectrogram of object.So-called state-detection and method for diagnosing faults based on non-linear spectral analysis carry out feature extraction and pattern classification to the GFRF model that has obtained exactly, with judge to as if be in normal condition, critical fault state, or malfunction.
Technical scheme of the present invention mainly realizes as follows:
1. model simplification promptly according to the characteristics such as symmetry, conjugate symmetry and zero pole distribution that the GFRF model had, is simplified model structure;
2. pumping signal design, promptly according to the identification needs, selected a kind of pumping signal; The present invention adopts has the pseudo-white noise signal of broad frequency band as additional incentive signal input object;
3. choose sampling period and data sample length according to features of the object, and synchronized sampling is carried out in the input and output of practical object, obtain Identification Data; According to the characteristics of non-linear nonparametric model identification, the data that participate in identification are compressed again;
4. choose identification algorithm, promptly according to the finally selected a kind of method of data characteristics, the present invention adopts the whole least-squares algorithm by svd;
5. Signal Pretreatment is promptly carried out conversion and conditioning to inputoutput data, makes these inputoutput datas be more suitable for being used for System Discrimination;
6. model generalization ability test, promptly to identification gained GFRF model, utilize other signal that has neither part nor lot in identification to carry out the model generalization ability test, computation model output is also measured real system output, the check square error is to determine whether model is applicable to state-detection and fault diagnosis;
7. pattern classification and fault diagnostic test promptly adopt the multilayer perceptron BP algorithm of neural network, and the GFRF chromatogram characteristic that has obtained is carried out pattern classification, are in which kind of state with definite object, thereby reach the purpose of state-detection; And utilize classified information to determine whether object is in fault or critical fault state, and the critical fault state is reported to the police, and malfunction is carried out emergency processing.
Fig. 1 is a test system architecture sketch of the present invention.
Fig. 2 is the non-linear spectrogram of a copper plate test film among the embodiment, wherein (i), (ii), (iii) be respectively single order, second order and three rank spectrograms; Last figure is an amplitude-versus-frequency curve, and ordinate is an amplitude; Figure below is the phase-frequency characteristic curve, and ordinate is a phase place; Horizontal ordinate all is frequencies.
Fig. 3 is the assay of embodiment, and wherein last figure is the model prediction of output and the actual curve fitting figure that measures that exports, and figure below is an error curve diagram, and horizontal ordinate is the time, and ordinate is an amplitude.
Test system architecture of the present invention as shown in Figure 1.Wherein 1 is the output signal detecting sensor; The 2nd, the input signal detection sensor; The 3rd, tested non-linear test film; The 4th, drive unit; The 5th, data collector; The 6th, signal source; The 7th, computing machine.
Tested non-linear test film 3 is a collection of reeds, and its vibration has tangible nonlinear characteristic, and along with its nonlinear characteristic of variation of structure and material has significant change, fault also has obvious characteristics.
Get three samples and survey, these three samples are respectively copper plate, brass sheet and the aluminium sheet of long 172mm, wide 20mm.
Embodiments of the invention carry out according to the following steps:
1. model simplification implementation process is: single order is chosen whole points, and second order is chosen 1/4 point, and the point on the center line is chosen on three rank; High-order all omits.
2. pumping signal is designed to pseudo-white noise signal.Utilize computing machine 7 to set pumping signal, make test film 3 vibrations by signal source 6 usefulness drive units 4 again.
3. choosing the sampling period is 0.4 millisecond, and sample length is 1100; Utilize input and output detecting sensor 1 and 2 to obtain input/output signals again, sampling by data sampling device 5 obtains inputoutput data and passes to computing machine 7 again; Because nonparametric model is adopted in identification, and inputoutput data is classified, being divided into is 32 classes; In similar, carry out identification, rather than utilize total data, reach the data compression purpose.
4. the whole least-squares algorithm with svd carries out the GFRF Model Distinguish.
5. adopt the numerical filter method that signal is carried out pre-service, and carry out rapid fourier change (FFT), we carry out identification with 1024 pairs of data handling.
6. carry out actual identification, obtain the spectrogram of test film, as shown in Figure 2.With model the output that other input signal produces is predicted then, carried out actual test simultaneously, with the generalization ability of testing model; In fact we use the 1025th pair to the 1125th pair data and carry out model testing, obtain matched curve shown in Figure 3, wherein solid line is the output of being estimated by model, and dotted line is the output of actual measurement, the two fitting degree is quite desirable, and error of fitting is in ± 5%.
7. adopt neuroid that spectrogram is carried out pattern classification, obtain the classification results shown in the table 1.Be categorized into power and surpass 80%.
Table 1
Annotate: wherein testing time refers to the testing time to test film; The convergence step number refers to each iteration
| Testing time | The convergence step number | Pattern classification error rate (%) | ||||
| ????I(5) | ????II(3) | ????III(3) | ????IV(2) | Subtotal (13) | ||
| ????1 | ????18 | ????0.0 | ????33.3 | ????0.0 | ????0.0 | ????7.7 |
| ????2 | ????18 | ????0.0 | ????33.3 | ????0.0 | ????0.0 | ????7.7 |
| ????3 | ????23 | ????0.0 | ????66.7 | ????33.3 | ????0.0 | ????23.0 |
| ????4 | ????23 | ????0.0 | ????66.7 | ????66.7 | ????0.0 | ????30.7 |
| ????5 | ????50 | ????0.0 | ????33.3 | ????0.0 | ????0.0 | ????7.7 |
| ????6 | ????25 | ????0.0 | ????33.3 | ????33.3 | ????0.0 | ????15.3 |
| ????7 | ????18 | ????0.0 | ????33.3 | ????33.3 | ????0.0 | ????15.3 |
| ????8 | ????66 | ????0.0 | ????66.7 | ????33.3 | ????0.0 | ????23.0 |
| ????9 | ????23 | ????20.0 | ????0.0 | ????66.7 | ????0.0 | ????23.0 |
| ????10 | ????14 | ????0.0 | ????66.7 | ????66.7 | ????0.0 | ????30.7 |
| ????11 | ????16 | ????0.0 | ????33.3 | ????0.0 | ????0.0 | ????7.70 |
| ????12 | ????22 | ????0.0 | ????33.3 | ????33.3 | ????0.0 | ????15.3 |
| ????13 | ????24 | ????0.0 | ????66.7 | ????0.0 | ????0.0 | ????15.3 |
| ????14 | ????34 | ????0.0 | ????66.7 | ????33.3 | ????0.0 | ????23.0 |
| ????15 | ????22 | ????0.0 | ????66.7 | ????0.0 | ????0.0 | ????15.3 |
| ????16 | ????29 | ????20.0 | ????33.3 | ????33.3 | ????0.0 | ????23.0 |
| ????17 | ????17 | ????40.0 | ????33.3 | ????0.0 | ????0.0 | ????15.3 |
| ????18 | ????24 | ????0.0 | ????33.3 | ????0.0 | ????0.0 | ????7.70 |
| ????19 | ????18 | ????0.0 | ????33.3 | ????0.0 | ????0.0 | ????7.70 |
| ????20 | ????18 | ????20.0 | ????33.3 | ????66.7 | ????0.0 | ????30.7 |
Practise used iterations; The pattern classification error rate refers to lose when each test specimen is used for pattern classification
The ratio of the number of times that loses and overall test number of times.
Claims (8)
1. based on the fault detection and diagnosis method of non-linear spectral analysis, it is characterized in that:
(1) model simplification promptly according to the characteristics such as symmetry, conjugate symmetry and zero pole distribution that the GFRF model had, is simplified model structure;
(2) pumping signal design promptly according to the identification needs, is adopted to have the pseudo-white noise signal of broad frequency band as additional incentive signal input object;
(3) choose sampling period and data sample length according to features of the object, and synchronized sampling is carried out in the input and output of practical object, obtain Identification Data, according to the characteristics of non-linear nonparametric model identification, the data that participate in identification are compressed again;
(4) choose identification algorithm, promptly adopt whole least-squares algorithm by svd;
(5) Signal Pretreatment is promptly carried out conversion and conditioning to inputoutput data, makes these inputoutput datas be more suitable for being used for System Discrimination;
(6) model generalization ability test, promptly to identification gained GFRF model, utilize other signal that has neither part nor lot in identification to carry out the model generalization ability test, computation model output is also measured real system output, the check square error is to determine whether model is applicable to state-detection and fault diagnosis;
(7) pattern classification and fault diagnostic test, promptly adopt the multilayer perceptron BP algorithm of neural network, the GFRF chromatogram characteristic that has obtained is carried out pattern classification, to determine which kind of state object is in, and utilize classified information to determine whether object is in fault or critical fault state, the critical fault state is reported to the police, malfunction is carried out emergency processing.
2. method according to claim 1 is characterized in that, the model simplification implementation process is: single order is chosen whole points, and second order is chosen 1/4 point, and the point on the center line is chosen on three rank, and high-order all omits.
3. method according to claim 1 is characterized in that pumping signal is designed to pseudo-white noise signal, utilizes computing machine 7 to set pumping signal then, makes test film 3 vibrations by signal source 6 usefulness drive units 4 again.
4. method according to claim 1, it is characterized in that, choosing the sampling period is 0.4 millisecond, sample length is 1100, utilize input and output detecting sensor 1 and 2 to obtain input/output signal, sampling by data sampling device 5 obtains inputoutput data and passes to computing machine 7 again, and inputoutput data is classified, and being divided into is 32 classes.
5. method according to claim 1 is characterized in that, carries out the GFRF Model Distinguish with the whole least-squares algorithm of svd.
6. method according to claim 1 is characterized in that, adopts the numerical filter method that signal is carried out pre-service, and carries out rapid fourier change (FFT), carries out identification with pretreated data.
7. method according to claim 1 is characterized in that, carries out actual identification, obtains the spectrogram of test film, with model the output that other input signal produces is predicted then, carries out actual test simultaneously, obtains matched curve.
8. method according to claim 1 is characterized in that, adopts neuroid that spectrogram is carried out pattern classification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 99115737 CN1120366C (en) | 1999-03-22 | 1999-03-22 | Fault detecting and diagnosing method based on non-linear spectral analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 99115737 CN1120366C (en) | 1999-03-22 | 1999-03-22 | Fault detecting and diagnosing method based on non-linear spectral analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1241721A true CN1241721A (en) | 2000-01-19 |
| CN1120366C CN1120366C (en) | 2003-09-03 |
Family
ID=5278654
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 99115737 Expired - Fee Related CN1120366C (en) | 1999-03-22 | 1999-03-22 | Fault detecting and diagnosing method based on non-linear spectral analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1120366C (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1301387C (en) * | 2004-06-04 | 2007-02-21 | 广东科龙电器股份有限公司 | Noise source identifying method for air-conditioner based on nervous network |
| CN100335982C (en) * | 2002-08-22 | 2007-09-05 | 气体产品与化学公司 | Equipment quick test based on model control |
| CN100440089C (en) * | 2006-11-23 | 2008-12-03 | 浙江大学 | Industrial process nonlinear fault diagnosis system and method based on fisher |
| CN100461043C (en) * | 2006-12-22 | 2009-02-11 | 浙江大学 | Melt index detection fault diagnosis system and method for industial polypropylene production |
| CN101438249A (en) * | 2006-05-07 | 2009-05-20 | 应用材料股份有限公司 | Ranged fault signatures for fault diagnosis |
| CN102175917A (en) * | 2011-01-19 | 2011-09-07 | 西安交通大学 | Online nonlinear spectrum analysis and fault diagnosis instrument |
| CN1910434B (en) * | 2004-01-14 | 2012-02-15 | Abb公司 | Method and apparatus to diagnose mechanical problems in machinery |
| CN103245524A (en) * | 2013-05-24 | 2013-08-14 | 南京大学 | Acoustic fault diagnosis method based on neural network |
| CN103492850A (en) * | 2011-04-21 | 2014-01-01 | 波音公司 | System and method for simulating high-intensity pyrotechnic shock |
| CN103488204A (en) * | 2013-08-14 | 2014-01-01 | 西北工业大学 | System-identification-based inverse control method for electromagnetic excitation system |
| CN104697796A (en) * | 2015-04-01 | 2015-06-10 | 北京交通大学 | Phased fault detection method for rolling bearing |
| CN105092467A (en) * | 2015-09-09 | 2015-11-25 | 南车工业研究院有限公司 | Rapid monitoring device and method for high-speed railway wheel tread abrasion |
| CN107766879A (en) * | 2017-09-30 | 2018-03-06 | 中国南方电网有限责任公司 | The MLP electric network fault cause diagnosis methods of feature based information extraction |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100393048C (en) * | 2006-01-13 | 2008-06-04 | 武汉大学 | Method for building network fault diagnosis rule base |
-
1999
- 1999-03-22 CN CN 99115737 patent/CN1120366C/en not_active Expired - Fee Related
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100335982C (en) * | 2002-08-22 | 2007-09-05 | 气体产品与化学公司 | Equipment quick test based on model control |
| CN1910434B (en) * | 2004-01-14 | 2012-02-15 | Abb公司 | Method and apparatus to diagnose mechanical problems in machinery |
| CN1301387C (en) * | 2004-06-04 | 2007-02-21 | 广东科龙电器股份有限公司 | Noise source identifying method for air-conditioner based on nervous network |
| CN101438249A (en) * | 2006-05-07 | 2009-05-20 | 应用材料股份有限公司 | Ranged fault signatures for fault diagnosis |
| CN100440089C (en) * | 2006-11-23 | 2008-12-03 | 浙江大学 | Industrial process nonlinear fault diagnosis system and method based on fisher |
| CN100461043C (en) * | 2006-12-22 | 2009-02-11 | 浙江大学 | Melt index detection fault diagnosis system and method for industial polypropylene production |
| CN102175917A (en) * | 2011-01-19 | 2011-09-07 | 西安交通大学 | Online nonlinear spectrum analysis and fault diagnosis instrument |
| CN103492850A (en) * | 2011-04-21 | 2014-01-01 | 波音公司 | System and method for simulating high-intensity pyrotechnic shock |
| CN103492850B (en) * | 2011-04-21 | 2017-06-06 | 波音公司 | System and method for simulating high-intensity pyrotechnic shock |
| CN103245524A (en) * | 2013-05-24 | 2013-08-14 | 南京大学 | Acoustic fault diagnosis method based on neural network |
| CN103488204A (en) * | 2013-08-14 | 2014-01-01 | 西北工业大学 | System-identification-based inverse control method for electromagnetic excitation system |
| CN103488204B (en) * | 2013-08-14 | 2015-10-21 | 西北工业大学 | A kind of inverse control method of the electromagnetic exciting system based on System Discrimination |
| CN104697796A (en) * | 2015-04-01 | 2015-06-10 | 北京交通大学 | Phased fault detection method for rolling bearing |
| CN105092467A (en) * | 2015-09-09 | 2015-11-25 | 南车工业研究院有限公司 | Rapid monitoring device and method for high-speed railway wheel tread abrasion |
| CN105092467B (en) * | 2015-09-09 | 2018-05-18 | 南车工业研究院有限公司 | A kind of quick monitoring device of high-speed railway wheel tread flat and method |
| CN107766879A (en) * | 2017-09-30 | 2018-03-06 | 中国南方电网有限责任公司 | The MLP electric network fault cause diagnosis methods of feature based information extraction |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1120366C (en) | 2003-09-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1120366C (en) | Fault detecting and diagnosing method based on non-linear spectral analysis | |
| CN109460618A (en) | A kind of rolling bearing remaining life on-line prediction method and system | |
| CN100485362C (en) | Analysis method and devicefor localized corroding based on electrochemistry noise | |
| CN111678698B (en) | Rolling bearing fault detection method based on sound and vibration signal fusion | |
| CN110375983A (en) | Failsafe valve real-time diagnosis system and diagnostic method based on time series analysis | |
| CN117517907B (en) | Insulation state monitoring method and system for transformer substation capacitive equipment | |
| CN111898644B (en) | Intelligent identification method for health state of aerospace liquid engine under fault-free sample | |
| CN112036042A (en) | Power equipment abnormality detection method and system based on variational modal decomposition | |
| CN105626502A (en) | Plunger pump health assessment method based on wavelet packet and Laplacian Eigenmap | |
| CN116610998A (en) | Switch cabinet fault diagnosis method and system based on multi-mode data fusion | |
| CN110987431B (en) | Bearing state monitoring and fault diagnosis method based on TQWT (TQWT-assisted SPC) | |
| Gao et al. | A wavelet packet and residual analysis based method for hydraulic pump health diagnosis | |
| CN114739667A (en) | Multi-mode information fusion bearing lubrication state monitoring device and method | |
| CN112801033A (en) | AlexNet network-based construction disturbance and leakage identification method along long oil and gas pipeline | |
| Yao et al. | A low-frequency fault detection method for low-speed planetary gearbox based on acoustic signals | |
| CN112860658A (en) | High-speed train electromagnetic emission prediction method based on data mining | |
| CN203011692U (en) | Portable fault diagnosing device for rolling bearing | |
| CN110160778A (en) | Gearbox fault state identification method based on sequential hypothesis testing | |
| Rui et al. | Research on fault diagnosis and state assessment of vacuum pump based on acoustic emission sensors | |
| CN115655731A (en) | Diesel engine state monitoring method and device and storage medium | |
| Deng et al. | A fractional hilbert transform order optimization algorithm based DE for bearing health monitoring | |
| CN110108489B (en) | Method for acquiring performance degradation trend of rolling bearing | |
| Yu et al. | Research and application of intelligent diagnosis method of mechanical fault based on transformer vibration and noise and BP neural network | |
| Duan et al. | Rolling bearing fault diagnosis method based on multi-information fusion characteristics under complex working conditions | |
| Liu et al. | SAL-CNN: Estimate the remaining useful life of bearings using time-frequency information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C06 | Publication | ||
| PB01 | Publication | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C19 | Lapse of patent right due to non-payment of the annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |