CN102539154A - Engine fault diagnosis method and device based on exhaust noise vector quantitative analysis - Google Patents
Engine fault diagnosis method and device based on exhaust noise vector quantitative analysis Download PDFInfo
- Publication number
- CN102539154A CN102539154A CN2011103120232A CN201110312023A CN102539154A CN 102539154 A CN102539154 A CN 102539154A CN 2011103120232 A CN2011103120232 A CN 2011103120232A CN 201110312023 A CN201110312023 A CN 201110312023A CN 102539154 A CN102539154 A CN 102539154A
- Authority
- CN
- China
- Prior art keywords
- code book
- engine
- exhaust noise
- vector sequence
- feature vector
- 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
Images
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention relates to an engine fault diagnosis method and an engine fault diagnosis device based on exhaust noise vector quantitative analysis. The method and the device mainly solve the problems of complex system and high cost because various parameters are required to be combined to complete engine fault diagnosis in the prior art. The diagnosis device comprises a characteristic extractor, a codebook generator, a memory and a recognizer, wherein the characteristic extractor is respectively connected with the codebook generator and the recognizer, the codebook generator is connected with the memory and the memory is connected with the recognizer. The Mel-frequency cepstrum coefficient (MFCC) of exhaust noise signals are used as a characteristic parameter vector sequence of a signal source in a vector quantitative algorithm, the vector quantitative algorithm is used for analyzing the exhaust noise of an engine, the difference of the exhaust noise of the engine under different working conditions is recognized and therefore the faults of the engine are diagnosed. Not only can nondestructive detection be realized, but also the system cost is low and the signals can be acquired more quickly and effectively.
Description
Technical field
The present invention relates to a kind of engine diagnosis field, especially relate to a kind of cost engine diagnosis method and device low, that analyze based on the exhaust noise vector quantization more fast and effectively.
Background technology
At present; The method of engine diagnosis has multiple, and the signal source of most of diagnostic systems is taken from engine start procedure parameter, ignition wave form, intake manifold vacuum waveform, engine speed, cylinder body vibration, electronic control injection procedure parameter and exhaust gas composition etc.These signal sources can only reflect a certain characteristic of engine, need multiparameter to combine to accomplish fault diagnosis, thereby cause system complex, cost higher.As how less detected parameters precisely realizes engine condition monitoring and fault diagnosis, is the matter of utmost importance that high-efficiency vehicle-mounted engine failure diagnosis system design faces.Like publication number is CN201852706U; Name is called a kind of Chinese invention patent application of the automobile failure diagnosis test macro based on the tail gas air draft; Comprise temperature sensor, humidity sensor, barometric pressure sensor, automobile exhaust analyzer, hobby regulations circuit, data collecting card, industrial computer, peripheral ancillary hardware system and software systems; Industrial computer is set up communication through RS-232 serial ports and automobile exhaust analyzer; To obtain CO, HC, CO2, NOx, O2 gas concentration information and the engine speed information in the vehicle exhaust; Humidity, temperature, barometric pressure sensor test environment temperature, humidity, atmospheric pressure information; Send into the signal conditioning circuit input end that the tone links to each other, the signal after the conditioning is sent into industrial computer through data collecting card, software systems realize to the information that industrial computer obtains handle, analyze, show, storage, reasoning diagnostic function.Just there is above-mentioned shortcoming in this patent of invention: these signal sources can only reflect a certain characteristic of engine, and need multiparameter to combine to accomplish fault diagnosis, thereby cause system complex, cost higher.
Summary of the invention
The present invention solves to need to combine multiple parameter to accomplish engine diagnosis in the prior art; Cause system complex, problem that cost is high; Provide a kind of cost engine diagnosis method low, that analyze based on the exhaust noise vector quantization more fast and effectively, and the device that uses this method.
Above-mentioned technical matters of the present invention mainly is able to solve through following technical proposals: a kind of engine diagnosis method of analyzing based on the exhaust noise vector quantization, may further comprise the steps,
A. in training process, in advance the various faults of engine is detected, regard the exhaust noise of engine under each fault as a signal source, from signal source, extract eigenvector, form feature vector sequence through feature extractor; What feature extractor extracted is the Mel frequency cepstral coefficient of exhaust noise, with the feature parameter vector sequence of Mel frequency cepstral coefficient as signal source in the vector quantization process.Be in the training process exhaust noise of engine under each situation all to be carried out characteristic parameter extraction in advance in this step.
B. each feature vector sequence is generated code book through LBG algorithm cluster, each code book all includes the personal characteristics of corresponding exhaust noise, makes the corresponding a kind of malfunction of each code book, then these code books is stored; In training process, in advance the engine various faults is detected and generate code book, and code book is existed formation one code library in the storer, standard as a comparison.These code books perhaps have overlapping at the distribution phase non-overlapping copies of feature space.
C. in test process, engine failure is detected; From the exhaust noise of test, extract feature vector sequence through feature extractor; Each code book that to store then carries out vector quantization to this feature vector sequence successively; Calculate average quantization error separately, each average quantization error is compared, select the corresponding malfunction of the minimum code book of average quantization error as the system diagnostics result.The model training data required based on the method for vector quantization are little, training and recognition time short, the work storage space is little, and system is convenient to hardware and realizes.In the process that feature vector sequence is quantized with code book; At first to select suitable distortion measure; Calculate the distortion that code word replaces vector to cause in the code book, can describe the similarity degree between two or more code books, and the pairing code word of degree of distortion minimum value promptly is the representative vector of this eigenvector; The code book of average quantization error minimum is just the most similar with this feature vector sequence like this, thereby realizes the diagnosis identification to engine failure.The dynamics of engine exhaust air-flow has directly reflected the variation of piping system boundary condition, thereby can reflect the various running statuses of engine.The exhaust noise of engine can be used as the signal source analysis as the direct physical mapping of engine exhaust air-flow, carries out the diagnosis of engine resultant fault.Adopt method of the present invention, use the exhaust noise signal as signal source, realized Non-Destructive Testing, not only cost is low, and signals collecting more fast effectively.
As a kind of preferred version, in training process, the exhaust noise under each fault is repeated to extract feature vector sequence, the code book that generates is revised quantification.Can obtain more accurate code book like this, make that identification is more accurate in the test process.
A kind of EDPAC Engine Diagnostic Package of analyzing based on the exhaust noise vector quantization comprises feature extractor, code book maker, storer and recognizer, and said feature extractor links to each other with code book maker, recognizer respectively; Feature extractor extracts the characteristic parameter of engine exhaust noise; Obtain feature vector sequence, in training process, the feature vector sequence that extracts sends in the code book maker; In test process, the feature vector sequence that extracts sends in the recognizer; Said code book maker links to each other with storer, and the code book maker generates code book with the feature vector sequence that extracts in the training process, and code book is stored in the storer; Said storer links to each other with recognizer; In test process; Each code book that recognizer reads in the storer carries out vector quantization to the feature vector sequence that feature extractor sends successively; And select and the minimum code book of feature vector sequence average quantization error, the engine failure that this code book is corresponding is output as diagnostic result.
The present invention with the MFCC of exhaust noise signal as the signal source feature parameter vector sequence in the Vector Quantization algorithm; The utilization Vector Quantization algorithm is analyzed engine exhaust noise; The difference of engine exhaust noise under the different operating modes of identification, and then diagnosis engine failure have so not only realized Non-Destructive Testing; The simultaneity factor cost is low, and signals collecting more fast effectively.
Description of drawings
Accompanying drawing 1 is a kind of structural representation of the present invention;
Accompanying drawing 2 is a kind of schematic flow sheets in the training process among the present invention;
Accompanying drawing 3 is a kind of schematic flow sheets in the test process among the present invention.
1-feature extractor 2-code book maker 3-storer 4-recognizer.
Embodiment
Pass through embodiment below, and combine accompanying drawing, do further bright specifically technical scheme of the present invention.
Embodiment:
A kind of EDPAC Engine Diagnostic Package of analyzing based on the exhaust noise vector quantization of present embodiment; As shown in Figure 1; Include feature extractor 1, code book maker 2, storer 3 and recognizer 4; This feature extractor links to each other with recognizer with the code book maker respectively, and code book generates the bearing storer and links to each other, and storer links to each other with recognizer.This feature extractor receives engine exhaust noise signal, and from the exhaust noise signal, extracts characteristic parameter, and what feature extractor extracted is the Mel frequency cepstral coefficient of exhaust noise, with it as the signal source feature parameter vector sequence in the Vector Quantization algorithm.In training process, the feature vector sequence that feature extractor extracts sends to the code book maker, by being stored in the storer behind the code book maker generation code book.In test process, the feature vector sequence that feature extractor extracts sends to recognizer.The difference of engine exhaust noise under the recognizer identification different faults state, and then diagnosis engine failure.
The concrete diagnostic method of EDPAC Engine Diagnostic Package is following in detail; As shown in Figure 2, at first to pass through training process, in training process, the exhaust noise under the multiple engine failure state is carried out characteristic parameter extraction; The characteristic parameter that extracts is MFCC; MFCC is the cepstrum coefficient that extracts in Mel scale frequency territory, and its concrete method for distilling is a known technology, and this repeats no more again.From exhaust noise, extract eigenvector; Obtain feature vector sequence; Be provided with N eigenvector; Then feature vector sequence is
; Through the LBG algorithm feature vector sequence is generated code book then;
, the LBG algorithm is a known technology, this repeats no more again.Each exhaust noise all passes through said process, obtains many group codes originally, and code book all includes the personal characteristics of corresponding exhaust noise, and the corresponding a kind of malfunction of each code book interrelates each code book with corresponding malfunction, be stored in the storer then.
As shown in Figure 3, get into test process, the exhaust noise of engine carries out feature extraction to this moment, obtains the feature vector sequence under the test mode, is designated as
X 1,
X 2...,
X M, successively feature vector sequence is carried out vector quantization by each code book in the storer then, and calculate average quantization error separately, establishing Xn is eigenvector,
Y l iBe codebook vectors, the average quantization error formula is:
Wherein,
Y l i,
l=1,2 ...,
L,
i=1,2 ...,
N, be
iIn the individual code book
lIndividual code word, and
d(
X n ,
Y l i ) be vector to be measured
X n , and code vector
Y l iBetween distance, use the Euclidean distance test for distortion measure.Relatively select the minimum code book of average quantization error at last, the malfunction that this code book is corresponding is as the system diagnostics result.
Specific embodiment described herein only is that the present invention's spirit is illustrated.Person of ordinary skill in the field of the present invention can make various modifications or replenishes or adopt similar mode to substitute described specific embodiment, but can't depart from spirit of the present invention or surmount the defined scope of appended claims.
Although this paper has used terms such as extraction apparatus, code book maker, storer, recognizer morely, do not get rid of the possibility of using other term.Using these terms only is in order to describe and explain essence of the present invention more easily; It all is contrary with spirit of the present invention being construed to any additional restriction to them.
Claims (3)
1. engine diagnosis method of analyzing based on the exhaust noise vector quantization is characterized in that: may further comprise the steps,
A. in training process, in advance the various faults of engine is detected, regard the exhaust noise of engine under each fault as a signal source, from signal source, extract eigenvector, form feature vector sequence through feature extractor;
B. each feature vector sequence is generated code book through LBG algorithm cluster, each code book all includes the personal characteristics of corresponding exhaust noise, makes the corresponding a kind of malfunction of each code book, then these code books is stored;
C. in test process, engine failure is detected; From the exhaust noise of test, extract feature vector sequence through feature extractor; Each code book that to store then carries out vector quantization to this feature vector sequence successively; Calculate average quantization error separately, each average quantization error is compared, select the corresponding malfunction of the minimum code book of average quantization error as the system diagnostics result.
2. the engine diagnosis method of analyzing based on the exhaust noise vector quantization according to claim 1 is characterized in that in training process, and the exhaust noise under each fault is repeated to extract feature vector sequence, and the code book that generates is revised quantification.
3. EDPAC Engine Diagnostic Package of analyzing based on the exhaust noise vector quantization; It is characterized in that: comprise feature extractor (1), code book maker (2), storer (3) and recognizer (4); Said feature extractor links to each other with code book maker (2), recognizer (4) respectively, and feature extractor extracts the characteristic parameter of engine exhaust noise, obtains feature vector sequence; In training process; The feature vector sequence that extracts sends in the code book maker (2), and in test process, the feature vector sequence that extracts sends in the recognizer (4); Said code book maker links to each other with storer (3), and the code book maker generates code book with the feature vector sequence that extracts in the training process, and code book is stored in the storer (3); Said storer (3) links to each other with recognizer (4); In test process; Each code book that recognizer reads in the storer carries out vector quantization to the feature vector sequence that feature extractor sends successively; And select and the minimum code book of feature vector sequence average quantization error, the engine failure that this code book is corresponding is output as diagnostic result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011103120232A CN102539154A (en) | 2011-10-16 | 2011-10-16 | Engine fault diagnosis method and device based on exhaust noise vector quantitative analysis |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN2011103120232A CN102539154A (en) | 2011-10-16 | 2011-10-16 | Engine fault diagnosis method and device based on exhaust noise vector quantitative analysis |
Publications (1)
Publication Number | Publication Date |
---|---|
CN102539154A true CN102539154A (en) | 2012-07-04 |
Family
ID=46346594
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN2011103120232A Pending CN102539154A (en) | 2011-10-16 | 2011-10-16 | Engine fault diagnosis method and device based on exhaust noise vector quantitative analysis |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102539154A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103915092A (en) * | 2014-04-01 | 2014-07-09 | 百度在线网络技术(北京)有限公司 | Voice identification method and device |
CN104167207A (en) * | 2014-06-20 | 2014-11-26 | 国家电网公司 | Equipment sound identification method based on transformer substation patrol inspection robot |
CN104952449A (en) * | 2015-01-09 | 2015-09-30 | 珠海高凌技术有限公司 | Method and device for identifying environmental noise sources |
CN112740133A (en) * | 2018-09-24 | 2021-04-30 | Abb瑞士股份有限公司 | System and method for monitoring the technical state of a technical installation |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19855745A1 (en) * | 1998-12-03 | 2000-08-03 | Medav Digitale Signalverarbeit | Fault diagnosis and quality grading system for production test of rotating machines uses sensing of rotation noise and oscillation with training of processor. |
CN101251441A (en) * | 2008-03-20 | 2008-08-27 | 桂林电子科技大学 | Method and system for on-line monitoring car engine state |
CN102156043A (en) * | 2010-12-31 | 2011-08-17 | 北京四方继保自动化股份有限公司 | Online state monitoring and fault diagnosis system of wind generator set |
-
2011
- 2011-10-16 CN CN2011103120232A patent/CN102539154A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19855745A1 (en) * | 1998-12-03 | 2000-08-03 | Medav Digitale Signalverarbeit | Fault diagnosis and quality grading system for production test of rotating machines uses sensing of rotation noise and oscillation with training of processor. |
CN101251441A (en) * | 2008-03-20 | 2008-08-27 | 桂林电子科技大学 | Method and system for on-line monitoring car engine state |
CN102156043A (en) * | 2010-12-31 | 2011-08-17 | 北京四方继保自动化股份有限公司 | Online state monitoring and fault diagnosis system of wind generator set |
Non-Patent Citations (3)
Title |
---|
丁哲等: "发动机故障诊断系统设计", 《噪声与振动控制》 * |
王伟等: "基于MFCC参数和VQ的说话人识别系统", 《仪器仪表学报》 * |
陈国金等: "发动机故障特征量提取方法的研究", 《内燃机学报》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103915092A (en) * | 2014-04-01 | 2014-07-09 | 百度在线网络技术(北京)有限公司 | Voice identification method and device |
CN104167207A (en) * | 2014-06-20 | 2014-11-26 | 国家电网公司 | Equipment sound identification method based on transformer substation patrol inspection robot |
CN104167207B (en) * | 2014-06-20 | 2017-12-12 | 国家电网公司 | A kind of equipment sound identification method based on Intelligent Mobile Robot |
CN104952449A (en) * | 2015-01-09 | 2015-09-30 | 珠海高凌技术有限公司 | Method and device for identifying environmental noise sources |
CN112740133A (en) * | 2018-09-24 | 2021-04-30 | Abb瑞士股份有限公司 | System and method for monitoring the technical state of a technical installation |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US8655571B2 (en) | MFCC and CELP to detect turbine engine faults | |
WO2018005996A1 (en) | System, device, and method for feature generation, selection, and classification for audio detection of anomalous engine operation | |
Devasenapati et al. | Misfire identification in a four-stroke four-cylinder petrol engine using decision tree | |
CN102419974A (en) | Sparse representation features for speech recognition | |
CN111201374A (en) | Calculation of exhaust emissions of a motor vehicle | |
CN102539154A (en) | Engine fault diagnosis method and device based on exhaust noise vector quantitative analysis | |
CN109323754A (en) | A kind of train wheel polygon fault diagnosis detection method | |
CN112735473A (en) | Method and system for identifying unmanned aerial vehicle based on voice | |
CN115824519B (en) | Comprehensive diagnosis method for valve leakage faults based on multi-sensor information fusion | |
CN115324698B (en) | Diesel engine tail gas analysis sensing method based on neural network | |
CN113268833A (en) | Migration fault diagnosis method based on deep joint distribution alignment | |
CN112509599A (en) | Acoustic spectrum fault analysis and diagnosis method based on BP neural network and Mel cepstrum | |
Zhang et al. | Misfire detection of diesel engine based on convolutional neural networks | |
CN116012681A (en) | Method and system for diagnosing motor faults of pipeline robot based on sound vibration signal fusion | |
Yang et al. | A novel multiple feature-based engine knock detection system using sparse bayesian extreme learning machine | |
CN105909979B (en) | Leakage acoustic characteristic extracting method based on Wavelet Transform Fusion blind source separation algorithm | |
Yang et al. | Personalized keyword spotting through multi-task learning | |
JP4886460B2 (en) | Abnormality monitoring device | |
Babu et al. | Misfire detection in a multi-cylinder diesel engine: A machine learning approach | |
KI | Misfire detection in a spark ignition engine using support vector machines | |
CN115758237A (en) | Bearing fault classification method and system based on intelligent inspection robot | |
CN114599870A (en) | System and method for virtual turbocharger speed sensor using neural network | |
CN115310531A (en) | Test result rough detection system and method applied to integrated spacer equipment comprehensive test module | |
Devasenapati et al. | Random forest based misfire detection using kononenko discretiser | |
Alahmer et al. | Monitoring of a spark ignition engine malfunctions using acoustic signal technique |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C12 | Rejection of a patent application after its publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20120704 |