CN105865794B - Engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis - Google Patents
Engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis Download PDFInfo
- Publication number
- CN105865794B CN105865794B CN201610317426.9A CN201610317426A CN105865794B CN 105865794 B CN105865794 B CN 105865794B CN 201610317426 A CN201610317426 A CN 201610317426A CN 105865794 B CN105865794 B CN 105865794B
- Authority
- CN
- China
- Prior art keywords
- principal component
- engine
- fourier transform
- signal
- discrete fourier
- 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.)
- Expired - Fee Related
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M15/00—Testing of engines
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Combined Controls Of Internal Combustion Engines (AREA)
- Testing Of Engines (AREA)
Abstract
The present invention provides a kind of engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis：It is primarily based on short time discrete Fourier transform and accurately obtains the transient speed waveform of each cylinder working state of reflection in working cycles, and be operated the characteristic parameter extraction of waveform in circulation on this basis, extraction reflects the high dimensional feature vector of each cylinder working state.High dimensional feature vector of the present invention to extract utilizes the T of principal component analysis for analysis object2Misfire fault is monitored with SPE monitoring indexes, and Cylinder is accurately positioned using the contribution plot method of principal component analysis.
Description
Technical field
The invention belongs to engine diagnosis field, and in particular to one kind is based on short time discrete Fourier transform and principal component point
The engine fire fault diagnosis method of analysis.
Background technology
Misfire fault is a kind of most common failure of multicylinder engine, and it not only reduces the dynamic property of engine, economy,
And aggravation exhaust emission.Therefore engine fire failure is carried out effectively monitoring and diagnosing very necessary.Engine is stable
When being worked under operating mode, its bent axle mean speed is constant, and when misfire fault occurs, the working condition of Cylinder becomes
Change, this can cause the transient speed of bent axle also to occur significantly to fluctuate.Therefore can be to event of catching fire using the Instantaneous Speed Fluctuations of bent axle
Barrier carries out efficient diagnosis, and it is that the key of effective Misfire Fault Diagnosis is carried out to engine accurately and timely to obtain instantaneous speed of crankshaft
And basis.Presently the most common Transient Rotate Speed Measuring means are to use hall sensor, intrinsic directly using internal combustion engine
Coding disk (such as flywheel ring gear) is measured by counting method.This measuring method influence measurement accuracy factor it is more, it is necessary to
Sensor raw signals are handled.The method of processing mainly includes two kinds of hardware counting method and software counting method.The former surveys
Measure with high costs, the latter mainly calculates transient speed by the method for software interpolation, and the method is had a great influence by noise signal, essence
Degree is poor, (is used to indicate cylinder compression top dead centre) when gear ring has hypodontia, Severe distortion occurs in transient speed result of calculation.
The content of the invention
It is an object of the invention to provide a kind of based on short time discrete Fourier transform and the event of the engine fire of principal component analysis
Hinder diagnostic method.
To achieve these goals, present invention employs following technical scheme：
1) engine rotational speed signal is divided into the continuous signal that length is L at equal intervals, then to every segment length of division
Short time discrete Fourier transform is carried out for L continuous signal, obtains the time-frequency distributions of tach signal；
2) according to the time-frequency distributions of tach signal, wink is calculated using the time-frequency distributions crestal line extracting method based on cost function
When frequency；The transient speed of engine is calculated according to instantaneous frequency；
3) transient speed waveform is divided according to difference cylinder acting section, rotating speed is calculated respectively in respective section
Fluctuation parameters value, peak value and the kurtosis of waveform, wherein fluctuation parameters value are defined as：
B=| Mright-Mleft|
Wherein MrightAnd MleftThe single cylinder each amplitude of section speed waveform right endpoint and the amplitude of left end point are represented respectively；
So, the characteristic on following 3N dimensional feature vectors F is obtained by transient speed waveform extracting：
F=[B P K], B=[Bi], P=[Pi], K=[Ki]；I=1,2 ... N
Wherein P is peak value, and K is kurtosis, and N is number of cylinders, and i is arranged according to cylinder operation sequential；
4) characteristic obtained according to step 3), misfire fault monitoring and diagnosis are carried out using principal component analysis.
The acquisition methods of the engine rotational speed signal comprise the following steps：The original tach signal of engine of collection is led to
Cross LPF and remove recurrent pulses type interference signal, then remove as the exception caused by engine compresses top dead centre hypodontia
Signal.
The original tach signal of engine is to be installed on the voltage that engine flywheel Pan Chu speed probe is exported
Signal.
The LPF use exponent number for 50~150 low pass filter, the cut-off frequency of the wave filter for 3000~
5000Hz。
The minimizing technology of the abnormal signal specifically includes following steps：Peak value is carried out to the tach signal after LPF
Search, when the data points between adjacent peak are less than given threshold, remove all data points between the adjacent peak.
The value of the L is 4000~6000.
The short time discrete Fourier transform is carried out according to the following formula：
Wherein x (t) is the continuous signal that length is L, and h (t) is a window function along time shaft sliding time, and τ is
The time-shifting factor, t represent signal time, and f represents signal frequency.
The computational methods of the instantaneous frequency specifically include following steps：
2.1) cost function being shown below is built：
CFk=-| S (k, a (k)) |2+|a(k)-a(k-1)|2, k=2,3 ..., n
Wherein k is the index value on time shaft, and a (k) is the frequency values at k moment, and n is the maximum of index value, and S is corresponding
Moment Short-time Fourier is composed；
2.2) by minimizing cost function, searching spectral peak modulus value is larger on a timeline and changes smooth part most
Big value is used as instantaneous frequency values.
The transient speed is calculated according to the following formula：
Sn=60Cffs/zNft
Wherein SnFor transient speed, CfFor instantaneous frequency, z is the Flywheel disc gear ring number of teeth, fsFor sample rate, NftFor Fourier
Transform length.
The step 4) specifically includes following steps：
4.1) it is X by characteristic definition datap∈Rm×d, R is characterized data set, and m is the number of sample, and d is monitored parameterses
Number；By XpIt is decomposed into vectorial tiAnd piApposition and remaining matrix EpSum：
Wherein tiIt is the score vector for including information between sample, piIt is the load vector of information between comprising variable；
4.2) according to characteristic decomposition result, the monitoring index T of each sample is calculated2Or SPE value and monitoring refer to
Upper control limit corresponding to mark；
4.3) as monitoring index T2Or SPE value is when exceeding upper control limit, to the characteristic of each cylinder of corresponding sample
According to monitoring index T2Or SPE contribution margin is calculated, and according to contribution margin size, the cylinder that misfire fault occurs is carried out
Positioning.When misfire fault occurs for single cylinder, the characteristic according to corresponding to the cylinder that misfire fault occurs be calculated to prison
The contribution margin for surveying index is maximum.
Beneficial effects of the present invention are embodied in：
The inventive method is primarily based on short time discrete Fourier transform and accurately obtains each cylinder working state of reflection in working cycles
Transient speed waveform, and the characteristic parameter extraction of waveform in circulation is operated on this basis, extraction reflects each cylinder working shape
The high dimensional feature vector of state.High dimensional feature vector of the present invention to extract is referred to for analysis object using the monitoring of principal component analysis
Mark is monitored to misfire fault, and Cylinder is accurately positioned using the contribution plot method of principal component analysis.
The advantages of the inventive method, is as follows：
(1) the transient speed extracting method of short time discrete Fourier transform (STFT) is based on, obtains I. C. Engine Transient Speed precision
High, method strong antijamming capability.
(2) misfire fault is monitored using principal component analysis (PCA), monitors characteristic value and monitoring index operand
It is small, it is high to misfire fault verification and measurement ratio, it can be accurately positioned Cylinder using contribution plot method.
(3) the inventive method is directly modeled using normal work data, therefore need not obtain fault data in advance,
To the Misfire Fault Diagnosis of engine, there is stronger engineering practicability.
Brief description of the drawings
Fig. 1 is speed probe original waveform.
Fig. 2 is speed probe waveform after filtering.
Fig. 3 is the time-frequency distributions of speed probe.
Fig. 4 is the transient speed ratio in the full employment circulation of different faults.
Fig. 5 is 1 cylinder misfire fault monitoring figure.
Fig. 6 is 1 cylinder misfire fault T2Monitoring index contribution plot.
Embodiment
The present invention is elaborated with reference to the accompanying drawings and examples.
The present invention is using the hall-effect speed sensor raw output voltage signal of engine pedestal as Misfire Fault Diagnosis pair
As because engine crankshaft rotating speed signal is a kind of typical non-stationary signal, and the instantaneous frequency of non-stationary signal has because of it
The physical concept of body is also widely used in terms of I. C. Engine Transient Speed calculating.By this inspiration, the present invention uses first
Transient speed extracting method based on short time discrete Fourier transform obtains accurate instantaneous speed of crankshaft, on this basis, to waveform
Rotating speed characteristic parameter is extracted, using the multivariable monitoring based on principal component analysis and diagnostic method to engine fire failure
Diagnosed.Because fluctuation of speed information of the bent axle in a working cycles can reflect the working condition of each cylinder, thus it is right
The high dimensional feature vector of each cylinder carries out depth excavation, internal combustion engine malfunction can accurately be not only identified, can more utilized
The corresponding relation of characteristic vector and each cylinder carries out being accurately positioned for Cylinder.
The present invention is tested using model 4G15S gasoline engine machine frame, and checking is based on short time discrete Fourier transform
With the validity of the internal combustion engine Misfire Fault Diagnosis method of principal component analysis.The engine flywheel disk gear ring number of teeth is 60.Wherein
Two teeth are removed for indicating cylinder compression top dead center position.(Flywheel disc connects engine crankshaft, and collection is just at Flywheel disc
It is engine rotational speed signal) hall-effect speed sensor is installed, setting sample rate is 200K, and tach signal is adopted
Collection, the original speed waveform of collection as shown in figure 1, sensor output value place position less than normal is cylinder compression top dead center position, this
Place causes signal waveform amplitude substantially less than normal due to gear ring hypodontia.Due to by Flywheel disc gear ring profile error, error of division etc.
The influence of factor, recurrent pulses type interference signal is produced, causes speed waveform to distort.The present invention use exponent number for 100 FIR
Low pass filter filters out impulse disturbances, and filter window function is arranged to Hanning window, and cut-off frequency is arranged to 4000Hz.Remove distortion
Speed waveform is as shown in Figure 2 after interference.In order to remove cylinder compression top dead centre low-voltage output to sensor overall output voltage
Influence, to filtered speed waveform carry out peak value searching, differentiate and remove the low-voltage caused by compression top center hypodontia
Data point.Detailed process is：Counted by calculating the data between peak value, when the points between adjacent peak are less than given threshold
When (threshold value basis of design is：Points between the tach signal adjacent peak of non-edentulous site are w, then threshold value is arranged to w/2),
By data points between deleting adjacent peak less than all tach signal data points between the adjacent peak of threshold value, so as to go
Except low-voltage data point caused by hypodontia.Then the tach signal of low-voltage data point will be removed, head and the tail are attached successively, real
Existing tach signal restructuring.
In one working cycle process of engine, its output torque is influenceed by current acting cylinder successively, for one
Individual N cylinder IC engines, cylinder i (i=1,2 ..., N) influence output torque in bent axle change angular region.TDCiBe cylinder i (i=1,
2 ..., N) bent axle start angle.Due to the transient speed that cylinder i output torque can be had influence in above-mentioned bent axle change angle,
Therefore the transient speed that can be changed by cylinder i (i=1,2 ..., N) in bent axle in angular region, to cylinder i (i=1,2 ..., N)
Working condition make assessment.Because tach signal is typical non-stationary signal, thus it is (short using time frequency analysis algorithm STFT
When Fourier transform), time frequency analysis is carried out to the speed waveform data point of restructuring, its detailed process is as follows：First will be analyzed
Signal is divided into length L=5000 continuous signal at equal intervals, then carries out following Short-time Fourier change to every segment signal x (t)
Change：
Wherein h (t) is a window function along time shaft sliding time, and it is arranged to width l=1000 Hanning window,
τ is the time-shifting factor, and it is arranged to 900, t and represents signal time, and f represents signal frequency, and j represents complex unit.Obtain
Tach signal time-frequency distributions are as shown in Figure 3.As can be seen from Figure 3, the time-frequency energy of tach signal is concentrated mainly on 1900Hz-
In 2100Hz sections.
Due to disturbing and the influence of other error components, time-frequency distributions of the signal in a window there is multiple peak values,
Larger error can be produced by directly calculating instantaneous frequency using the method for spectral peak maximum.And because the fluctuation of speed is in steady working condition
Under the conditions of fluctuate it is smaller, it is believed that the signal transient frequency in adjacent windows is continuous, therefore the present invention is using being based on cost
The time-frequency distributions crestal line extracting method of function, accurately calculates instantaneous frequency.Cost function structure is as follows：
CFk=-| S (k, a (k)) |2+|a(k)-a(k-1)|2, k=2,3 ..., n (2)
S is corresponding moment Short-time Fourier spectrum in formula, and it is obtained by formula 1, and k is the index value on time shaft, corresponding public
The time t of formula 1, a (k) are the frequency values at k moment, and n is the maximum of index value, and it is determined by signal length.The cost function
It is made up of two parts, the modulus value of the former spectral peak, the latter represents the rate of change of spectrum peak, by minimizing cost function,
Searching spectral peak modulus value is larger on time shaft and changes smooth local maximum as instantaneous frequency values, now corresponding instantaneous
Rotating speed is：
Sn=60Cffs/zNft (3)
S in formulanFor transient speed, unit is rpm (rev/min), CfFor instantaneous frequency, z is the Flywheel disc gear ring number of teeth.fsFor
Signal sampling rate (value is 200K), NftFor Fourier transformation length.
When engine work and cylinder 1,2,4 misfire fault occur respectively, signals of rotational speed sensor is carried out
As a result as shown in Figure 4 collection, transient speed is extracted using above method,.Figure 4, it can be seen that during engine work,
With reciprocating motion of the piston in cylinder, transient speed is also in regular fluctuation, and transient speed waveform is in cylinder acting area
Between there is corresponding crest, and crest amplitude approximately equal.When there is single cylinder misfire fault in engine, its each cylinder working state
Uniformity significant change, Cylinder correspond to crest disappearance or amplitude and substantially reduced.At work in sequence, Cylinder adjacent cylinder
Working condition is also affected, and its crest amplitude is significantly lower than other cylinder spectral peaks, therefore by comparing the work of adjacent cylinder
State, can not only effective detection be carried out to single cylinder misfire fault, and can provide useful information for the positioning of Cylinder.
In order to reflect speed waveform feature corresponding to difference cylinder, in a working cycles, transient speed waveform is pressed
Divided according to difference cylinder acting section, according to the variation tendency of transient speed, when transient speed is risen to by a trough
One crest, when being returned to a trough, it is judged as an acting section, calculates the ripple of speed waveform respectively in respective section
Dynamic parameter value, peak value and kurtosis time domain index.Wherein fluctuation parameters value is defined as：
B=| Mright-Mleft| (4)
M in formularightAnd MleftRepresent that single cylinder each does work the right endpoint of section waveform and the amplitude of left end point respectively.Fluctuation
Parameter value and peak index reflect fluctuation and the amplitude size distribution situation of difference cylinder transient speed respectively, and kurtosis index energy
Enough reflect the steep of transient speed probability distribution curve, transient speed waveform trend changes more to caused by misfire fault
It is sensitive.According to features above extracting method, to the following 3N dimensional feature vectors of transient speed waveform extracting in a working cycles：
F=[B P K], B=[Bi], P=[Pi], K=[Ki]；I=1,2 ... N (5)
Wherein P is peak index, and K is kurtosis index, and N is number of cylinders, and i is arranged according to work schedule.
The ignition order of engine is 1-3-4-2 used in present invention experiment.Engine is connected to dynamometer machine, single by cutting off
Cylinder oil circuit simulates single cylinder misfire fault, is tested under different engine speed and dynamometric power pattern.Experiment
Fault mode sets as shown in table 1.
Table 1. tests fault mode
The characteristic value defined according to formula (5), to rotating speed 2000rpm, normal and single cylinder catches fire (1 cylinder, 2 cylinders and 4 cylinders) respectively
Transient speed signals (transient speed waveform) under state have carried out feature extraction, and as a result as shown in table 2, feature 1~4 is right respectively
Should be done work the Instantaneous Speed Fluctuations parameter in section 1~4, and feature 5~6 corresponds to the transient speed peak value in acting section 1~2 respectively.
Cylinder character pair value adds italic to represent with runic in table 2, and adjacent cylinder character pair value is represented with runic in sequential.From table 2
In as can be seen that Cylinder correspond to fluctuation parameters value and significantly increased, and peak index substantially reduces, and kurtosis index increases.Sequential
The change of same trend also occurs for the corresponding index value of upper adjacent cylinder, but amplitude of variation diminishes.
The transient speed signals feature extraction result (part) of table 2.
Below by using theoretical principal component analysis (PCA) method is monitored based on multivariable, engine fire failure is entered
Row monitoring and diagnosis.Monitoring process for including d variable, its data may be defined as Xp∈Rm×d(R is characterized data set, m
The number of value sample is characterized, d is the number (value is the Characteristic Number of 3N dimensional feature vectors) of monitored parameterses, and a job follows
Ring includes two samples, and one group of characteristic value is tried to achieve by a sample.PCA can be by XpIt is decomposed into vectorial tiAnd piApposition and residual
Complementary submatrix EpSum：
T in formulaiIt is the score vector for including information between sample, piIt is the load vector of information between comprising variable, subscript
I represents feature number, and T is transposition.When carrying out process monitoring, it is primarily based on normal historical data and carries out PCA analyses, establishes normal
The principal component model of data, then by the data projection newly obtained during monitoring to two orthogonal subspaces, and respectively
Calculate T2With SPE monitoring indexes.When monitoring index deviates normal value, it can be determined that failure occurs.T2Monitoring index is defined as：
T2(k)=t (k)TΛ-1t(k) (7)
Λ in formula-1The diagonal matrix that the characteristic value related to the principal component of reservation is formed, t (k) be corresponding feature to
Moment matrix.T2The upper control limit UCL of monitoring index is defined as：
Wherein r be retain principal component number, r according to retain number of components account for original number of components 70~95% (for example,
85%) standard, which is configured, (to be carried out characteristic value solution to eigenvalue matrix, is arranged according to the size of required characteristic value
Sequence, the feature value vector corresponding to larger characteristic value that former ranks remain is principal component), m is sample number, Fα(r,
M-r it is) that confidence level is α, the F that the free degree is r and m-r is distributed.
SPE monitoring indexes are defined as：
P in formularIt is the matrix that preceding r load vector is formed in principal component model, r is the principal component number retained.SPE is monitored
The upper control limit Q of indexαIt is defined as：
WhereinR is the principal component number retained, and d is monitored parameterses number, and λ is characterized vectorial (subscript
I numbers for variable, and subscript j is characterized value numbering).CαThe normal distribution for being α for confidence level.
The experiment fault mode of the table of comparisons 1, engine single cylinder misfire fault is diagnosed using the inventive method.Press
Illuminated (5) is extracted the transient speed feature value vector under normal and malfunction, and T is calculated according to formula (7) and formula (9)2With
The monitoring figure of SPE monitoring indexes, the monitoring figure of wherein pattern 1 are as shown in Figure 5.Dotted line is calculated by normal condition in Fig. 5
Monitoring index threshold value (being upper control limit) under the conditions of confidence level 95%.Sample shown in Fig. 5 is before the 65th sample
It is a cylinder misfire fault state since the 66th sample for normal condition.From figure 5 it can be seen that in addition to indivedual samples, T2
With SPE monitoring indexes value before the moment occurs for failure, under monitoring index threshold value.And since occurring the moment for failure,
Monitoring index value greatly exceed monitoring index threshold value.By calculating, the fault detect rate in table 1 under each pattern is
100%.In order to verify the rate of false alarm of institute's extracting method of the present invention, using the sample training under normal condition and test,
Under the conditions of confidence level 95%, its T2Rate of false alarm with SPE monitoring indexes is respectively 0.33% and 0.79%, and this shows, the present invention
Method can be effectively monitored under the conditions of low rate of false alarm to single cylinder misfire fault.
In Figure 5, T is worked as2When exceeding monitoring index threshold value with SPE monitoring indexes value, (the 66th sample is referred specifically to the moment
Originally corresponding characteristic value) is to T2The contribution margin of index is calculated (realizes that Cylinder positions, reason using SPE contribution plot
By upper and feasible), obtained contribution plot is as shown in Figure 6.As can be seen from Figure 6 when misfire fault occurs for 1 cylinder, 1 cylinder institute
Corresponding characteristic value monitoring index T to caused by failure2Abnormal contribution amount is maximum, next to that the phase of 3 cylinders adjacent in sequential
Close contribution amount.Therefore, analyzed by failure occurring the monitoring index contribution plot at moment, can accurately position and catch fire
Cylinder.
Claims (9)
1. the engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis, it is characterised in that：Including
Following steps：
1) engine rotational speed signal is divided into the continuous signal that length is L at equal intervals, then every segment length to division is L's
Continuous signal carries out short time discrete Fourier transform, obtains the time-frequency distributions of tach signal；
2) according to the time-frequency distributions of tach signal, instantaneous frequency is calculated using the time-frequency distributions crestal line extracting method based on cost function
Rate；The transient speed of engine is calculated according to instantaneous frequency；
The computational methods of the instantaneous frequency specifically include following steps：
2.1) cost function being shown below is built：
CFk=-| S (k, a (k)) |2+|a(k)-a(k-1)|2, k=2,3 ..., n
Wherein k is the index value on time shaft, and a (k) is the frequency values at k moment, and n is the maximum of index value, and S is the corresponding moment
Short-time Fourier is composed；
2.2) by minimizing cost function, searching spectral peak modulus value is larger on a timeline and changes smooth local maximum
As instantaneous frequency values；
3) transient speed waveform is divided according to difference cylinder acting section, speed waveform is calculated respectively in respective section
Fluctuation parameters value, peak value and kurtosis, wherein fluctuation parameters value is defined as：
B=| Mright-Mleft|
Wherein MrightAnd MleftThe single cylinder each amplitude of section speed waveform right endpoint and the amplitude of left end point are represented respectively；
So, the characteristic based on following 3N dimensional feature vectors F is obtained by transient speed waveform extracting：
F=[B P K], B=[Bi], P=[Pi], K=[Ki]；I=1,2 ... N
Wherein P is peak value, and K is kurtosis, and N is number of cylinders, and i is arranged according to cylinder operation sequential；
4) characteristic obtained according to step 3), misfire fault monitoring and diagnosis are carried out using principal component analysis.
2. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 1
Method, it is characterised in that：The acquisition methods of the engine rotational speed signal comprise the following steps：To the original rotating speed of engine of collection
Signal removes recurrent pulses type interference signal by LPF, then removes as caused by engine compresses top dead centre hypodontia
Abnormal signal.
3. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 2
Method, it is characterised in that：The original tach signal of engine is exported by the speed probe for being installed on engine flywheel Pan Chu
Voltage signal.
4. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 2
Method, it is characterised in that：The LPF uses exponent number, and for 50~150 low pass filter, the cut-off frequency of the wave filter is
3000~5000Hz.
5. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 2
Method, it is characterised in that：The minimizing technology of the abnormal signal specifically includes following steps：Tach signal after LPF is entered
Row peak value searching, when the data points between adjacent peak are less than given threshold, remove all numbers between the adjacent peak
Strong point.
6. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 1
Method, it is characterised in that：The value of the L is 4000~6000.
7. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 1
Method, it is characterised in that：The short time discrete Fourier transform is carried out according to the following formula：
<mrow>
<msub>
<mi>STFT</mi>
<mi>x</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>,</mo>
<mi>f</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msubsup>
<mo>&Integral;</mo>
<mrow>
<mo>-</mo>
<mi>&infin;</mi>
</mrow>
<mi>&infin;</mi>
</msubsup>
<mi>x</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>)</mo>
</mrow>
<mi>h</mi>
<mrow>
<mo>(</mo>
<mi>t</mi>
<mo>-</mo>
<mi>&tau;</mi>
<mo>)</mo>
</mrow>
<msup>
<mi>e</mi>
<mrow>
<mo>-</mo>
<mi>j</mi>
<mn>2</mn>
<mi>&pi;</mi>
<mi>f</mi>
<mi>t</mi>
</mrow>
</msup>
<mi>d</mi>
<mi>t</mi>
</mrow>
Wherein x (t) is the continuous signal that length is L, and h (t) is a window function along time shaft sliding time, and τ is the time
Shift factor, t represent signal time, and f represents signal frequency.
8. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 1
Method, it is characterised in that：The transient speed is calculated according to the following formula：
Sn=60Cffs/zNfft
Wherein SnFor transient speed, CfFor instantaneous frequency, z is the Flywheel disc gear ring number of teeth, fsFor sample rate, NfftFor Fourier transformation
Length.
9. the engine fire fault diagnosis side based on short time discrete Fourier transform and principal component analysis according to claim 1
Method, it is characterised in that：The step 4) specifically includes following steps：
4.1) it is X by characteristic definition datap∈Rm×d, R is characterized data set, and m is the number of sample, and d is of monitored parameterses
Number；By XpIt is decomposed into vectorial tiAnd piApposition and remaining matrix EpSum：
<mrow>
<msub>
<mi>X</mi>
<mi>p</mi>
</msub>
<mo>=</mo>
<msub>
<mi>&Sigma;t</mi>
<mi>i</mi>
</msub>
<msubsup>
<mi>p</mi>
<mi>i</mi>
<mi>T</mi>
</msubsup>
<mo>+</mo>
<msub>
<mi>E</mi>
<mi>p</mi>
</msub>
</mrow>
Wherein tiIt is the score vector for including information between sample, piIt is the load vector of information between comprising variable；
4.2) according to characteristic decomposition result, the monitoring index T of each sample is calculated2Or SPE value and monitoring index are corresponding
Upper control limit；
T2It is defined as：T2(k)=t (k)TΛ-1t(k)
Λ in formula-1It is the diagonal matrix that the characteristic value related to the principal component of reservation is formed, t (k) is corresponding characteristic vector square
Battle array；
SPE is defined as：
P in formularIt is the matrix that preceding r load vector is formed in principal component model, r is the principal component number retained；
4.3) as monitoring index T2Or SPE value is when exceeding upper control limit, to the characteristic of each cylinder of corresponding sample to prison
Survey index T2Or SPE contribution margin is calculated, and according to contribution margin size, the cylinder that misfire fault occurs is positioned.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610317426.9A CN105865794B (en) | 2016-05-12 | 2016-05-12 | Engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610317426.9A CN105865794B (en) | 2016-05-12 | 2016-05-12 | Engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105865794A CN105865794A (en) | 2016-08-17 |
| CN105865794B true CN105865794B (en) | 2018-02-02 |
Family
ID=56631036
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610317426.9A Expired - Fee Related CN105865794B (en) | 2016-05-12 | 2016-05-12 | Engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105865794B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10208697B2 (en) * | 2016-10-26 | 2019-02-19 | GM Global Technology Operations LLC | Detection of irregularities in engine cylinder firing |
| CN106844922A (en) * | 2017-01-11 | 2017-06-13 | 重庆邮电大学 | The engine fire fault diagnosis method with manifold learning is estimated based on cylinder pressure |
| CN108388839A (en) * | 2018-01-26 | 2018-08-10 | 电子科技大学 | A kind of strong fluctuation of speed feature extracting method based on second order sync extraction transformation |
| CN109030010A (en) * | 2018-06-28 | 2018-12-18 | 潍柴动力股份有限公司 | One kind being used for engine fire detection method and device |
| CN108692947A (en) * | 2018-06-28 | 2018-10-23 | 潍柴动力股份有限公司 | One kind being used for engine fire detection method and device |
| CN109030009A (en) * | 2018-06-28 | 2018-12-18 | 潍柴动力股份有限公司 | One kind being used for engine fire detection method and device |
| CN111120094B (en) * | 2019-11-29 | 2021-02-23 | 潍柴动力股份有限公司 | Engine fire detection method and device, storage medium and terminal |
| WO2021195948A1 (en) * | 2020-03-31 | 2021-10-07 | 潍柴动力股份有限公司 | Engine misfire detection method and apparatus, device, storage medium, and program |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007303426A (en) * | 2006-05-12 | 2007-11-22 | Hitachi Ltd | Diagnosis device of internal combustion engine |
| CN104504200A (en) * | 2014-12-23 | 2015-04-08 | 华中科技大学 | Trend curve diagram display method used for online vibration monitoring of rotary mechanism |
| CN104880247A (en) * | 2015-05-04 | 2015-09-02 | 华中科技大学 | Combined alarming method for on-line monitoring system for rotary machinery |
| CN105021403A (en) * | 2015-07-17 | 2015-11-04 | 中国科学院自动化研究所 | Aero-engine starting system fault diagnosis method and fault simulation system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2965915B1 (en) * | 2010-10-11 | 2013-08-02 | Snecma | SYSTEM FOR MONITORING AN AIRCRAFT ENGINE TEST BENCH |
-
2016
- 2016-05-12 CN CN201610317426.9A patent/CN105865794B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2007303426A (en) * | 2006-05-12 | 2007-11-22 | Hitachi Ltd | Diagnosis device of internal combustion engine |
| CN104504200A (en) * | 2014-12-23 | 2015-04-08 | 华中科技大学 | Trend curve diagram display method used for online vibration monitoring of rotary mechanism |
| CN104880247A (en) * | 2015-05-04 | 2015-09-02 | 华中科技大学 | Combined alarming method for on-line monitoring system for rotary machinery |
| CN105021403A (en) * | 2015-07-17 | 2015-11-04 | 中国科学院自动化研究所 | Aero-engine starting system fault diagnosis method and fault simulation system |
Non-Patent Citations (3)
| Title |
|---|
| 基于瞬时转速统计分析的内燃机失火检测;胡重庆;《内燃机学报》;20090930;第27卷(第5期);第446-449页左栏 * |
| 基于瞬时频率估计的旋转机械阶比跟踪;郭瑜;《机械工程学报》;20030331;第39卷(第3期);第31-33页 * |
| 基于自适应短时Chirp_Fourier变换的瞬时转速估计及应用;赵明;《机械工程学报》;20150731;第51卷(第14期);第9页左栏 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105865794A (en) | 2016-08-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105865794B (en) | Engine fire fault diagnosis method based on short time discrete Fourier transform and principal component analysis | |
| EP1586880B1 (en) | Knock detecting apparatus and method for internal combustion engine | |
| US5487008A (en) | Method and system for detecting the misfire of a reciprocating internal combustion engine in frequency domain | |
| KR100557667B1 (en) | Process for detecting a misfire in an internal combustion engine and system for carrying out said process | |
| US7530261B2 (en) | Fourier-based misfire detection strategy | |
| US11591981B2 (en) | System and method for detecting malfunctioning turbo-diesel cylinders | |
| US5804711A (en) | Pattern recognition method and system for determining a misfire condition in a reciprocating engine | |
| US20040260454A1 (en) | Vibro-acoustic engine diagnostic system | |
| CN208125377U (en) | A kind of motor rolling bearing failure diagnosis system | |
| US6314802B1 (en) | Optimal engine speed compensation method used in misfire detection | |
| Varbanets et al. | Improvement of diagnosing methods of the diesel engine functioning under operating conditions | |
| Delvecchio et al. | On the use of cyclostationary indicators in IC engine quality control by cold tests | |
| KR102063974B1 (en) | Engine misfire diagnosis system and method using discrete Fourier transform and analysis of fluctuation of engine speed with linear component removed | |
| CN109946081A (en) | A kind of method for diagnosing faults under variable speed when rolling bearing skidding | |
| CN113431680B (en) | Full-period acquisition and instantaneous speed calculation method of engine monitoring signals based on tooth pulse of tooth-lacking fluted disc | |
| Hu et al. | Multivariate statistical analysis strategy for multiple misfire detection in internal combustion engines | |
| EP2581588A1 (en) | Method of fuel quality determination | |
| Loughlin et al. | Conditional moments analysis of transients with application to helicopter fault data | |
| JPH04232828A (en) | Cylinder-performance analyzing method for internal combustion engine | |
| CN105910828B (en) | A kind of circulation polar diagram diagnostic method of multi-cylinder diesel engine combustion failure | |
| Jin et al. | A vibration-based approach for diesel engine fault diagnosis | |
| Abadi et al. | Single and multiple misfire detection in internal combustion engines using vold-kalman filter order-tracking | |
| Watzenig et al. | Engine state monitoring and fault diagnosis of large marine diesel engines. | |
| Siano et al. | A nonlinear black-box modeling method for knock detection in spark-ignition engines | |
| US20060074543A1 (en) | Method for detecting combustion failure by filtering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180202 Termination date: 20190512 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |