CN109115330A - A kind of abnormal sound recognition methods of light modulation motor device - Google Patents
A kind of abnormal sound recognition methods of light modulation motor device Download PDFInfo
- Publication number
- CN109115330A CN109115330A CN201810945473.7A CN201810945473A CN109115330A CN 109115330 A CN109115330 A CN 109115330A CN 201810945473 A CN201810945473 A CN 201810945473A CN 109115330 A CN109115330 A CN 109115330A
- Authority
- CN
- China
- Prior art keywords
- signal
- energy
- light modulation
- abnormal sound
- frequency band
- 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
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
It include: the vibration signal that light modulation motor device is obtained using acceleration transducer the invention discloses a kind of abnormal sound recognition methods of light modulation motor device；WAVELET PACKET DECOMPOSITION is carried out to vibration signal；Calculate the signal energy of each frequency band of WAVELET PACKET DECOMPOSITION, the signal energy of each frequency band of normalized；Analysis of time-domain characteristic is carried out to obtain multinomial time domain charactreristic parameter to vibration signal；The result of normalized and multinomial time domain charactreristic parameter are formed into total characteristic vector；Training BP neural network classifier；The total characteristic vector of the vibration signal of tested light modulation motor device is input in trained BP neural network classifier, the sound class of the light modulation motor device is exported.Recognition speed of the present invention is fast, and accuracy rate is high, high-efficient compared to ear recognition, there is practical application value.
Description
Technical field
The present invention relates to abnormal sound identification technology fields, and in particular to a kind of abnormal sound identification side of light modulation motor device
Method.
Background technique
Noise problem influences people to the subjective feeling of a product very much, therefore controlling noise is enterprise in manufacturing enterprise
The desired raising product competitiveness of industry pays close attention to direction.Automobile-used light dimming electric motor be on automobile essential important component it
One, producer will test its performance, parameter during production run, to ensure to meet associated specifications.Certain vapour
Vehicle vehicle lamp assembly factory finds in actual production: certain model car light light dimming electric motor device is less than rule to its noise testing at work
Fixed limit value 60dB (at 40cm), but human ear can hear that a kind of abnormal sound that beats such as " blocks " sound, and influencing should
Product sound quality, referred to as " abnormal sound " are to allow experienced worker's human ear to listen noise to screen for abnormal sound part enterprise practices well
Bad part out.But in producing line the number of motors of production in one day be human ear distinguish speed far away from, if things go on like this, not only efficiency
Lowly, irritated mood can be also brought to staff, cause to judge misalignment.
Automobile-used light dimming electric motor device is driven by the minitype permanent magnetism direct current generator that revolving speed is 13200r/min, will by worm screw
Power passes to nylon duplicate gear big end, then passes to another gear wheel by its small end, is cased with output shaft among gear wheel, leads to
Cross the forward-reverse that screw-driven realizes output shaft.Mechanical movement will necessarily bring vibration and noise.Sound generates essence
Vibration, extraneous has evoked the eigentone of itself.So research vibration is to detect the good approach of abnormal sound.
On the one hand mechanical noise derives from motor and gear itself operates, on the other hand from plastic gear engaged transmission
The shock of generation, friction, uneven and mesomerism.Occur that rotor altogether including the frictional force due to caused by worm and wormwheel defect
Vibration；The bearing of reinforcing rib that gear, axis defect transfer vibrations on light dimming electric motor shell is to be transmitted to cap and to external radiation
Noise.
Abnormal sound is that a part in noise failure is different from noise again.It more characterizes the degree of stability of equipment running status.
It is a kind of periodic knock or scrape that the abnormal sound of light dimming electric motor can be identified by human ear.The biography of one side and the inside
Dynamic device is related, and low module plastic gear is injection molding, and precision etc. can be not so good as metal gear, error easily occurs,
Cause unstable so as to cause abnormal sound in operational process；On the other hand lubricate it is not in place, assemble it is not in place be likely to cause it is different
Mail topic.
Therefore the abnormal sound recognition methods substitution artitificial ear for being badly in need of a kind of light modulation motor device at present listens identification to identify
Abnormal sound part out.
Summary of the invention
The technical problem to be solved by the present invention is to provide a kind of light modulation motor in view of the above shortcomings of the prior art
The abnormal sound recognition methods of device, the abnormal sound recognition methods of this light modulation motor device using after WAVELET PACKET DECOMPOSITION energy spectrum and
Time domain charactreristic parameter extracts total characteristic vector, imports in BP neural network and classifies, abnormal sound recognition result is obtained, to select
Abnormal sound part, recognition speed is fast, and accuracy rate is high, high-efficient compared to ear recognition.
To realize the above-mentioned technical purpose, the technical scheme adopted by the invention is as follows:
A kind of abnormal sound recognition methods of light modulation motor device, the following steps are included:
Step 1, the vibration signal that light modulation motor device is obtained using acceleration transducer；
Step 2: j layers of WAVELET PACKET DECOMPOSITION are carried out to vibration signal；
Step 3: calculating the signal energy of each frequency band of WAVELET PACKET DECOMPOSITION, the signal energy of each frequency band of normalized obtains
The energy ratio of gross energy shared by signal energy to each frequency band；
Step 4: analysis of time-domain characteristic is carried out to obtain multinomial time domain charactreristic parameter to vibration signal；
Step 5: the energy ratio of gross energy shared by the signal energy for each frequency band that step 3 obtains and step 4 being obtained more
Item time domain charactreristic parameter forms total characteristic vector；
Step 6: building BP neural network classifier；
Step 7: the vibration signal of test light modulation motor device of the multiple groups without abnormal sound and the automobile light-modulating electricity for having abnormal sound
The vibration signal of machine device extracts the total characteristic vector of every group of vibration signal as training sample respectively, training sample is imported
Classification learning is carried out in BP neural network classifier, obtains trained BP neural network classifier；
Step 8: the total characteristic vector of the vibration signal of tested light modulation motor device is input to trained BP
In neural network classifier, the sound class of the light modulation motor device is exported, the sound class includes no abnormal sound and has
Abnormal sound.
Technical solution as a further improvement of that present invention, the step 3 the following steps are included:
(1) it is equal to square of 2 norms of signal according to the energy of Parseval energy integral signal in the time domain, by
Step 2 obtains j layers of wavelet decomposition, there is 2jA frequency band, then the signal energy E of each frequency bandj.mAre as follows:
Wherein, p=N/2j, N is the sampling number of original signal, and m is jth layer WAVELET PACKET DECOMPOSITION coefficient, i.e. expression frequency band
Serial number；Sj,mFor the reconstruction signal of m-th of frequency band, xm,nFor the amplitude of reconstruction signal；
(2) gross energy is calculated, gross energy is the sum of the signal energy of each frequency band of WAVELET PACKET DECOMPOSITION, is indicated are as follows:
(3) signal energy of each frequency band of normalized, obtains the energy ratio of gross energy shared by the signal energy of each frequency band,
Construct normalized feature vector:
Technical solution as a further improvement of that present invention, the multinomial time domain charactreristic parameter in the step 4 includes wave crest
Factor and kurtosis.
Technical solution as a further improvement of that present invention, the signal energy of each frequency band obtained step 3 in the step 5
The multinomial time domain charactreristic parameter composition total characteristic vector that the energy ratio and step 4 for measuring shared gross energy obtain specifically: by step 3
Obtained normalized feature vector, crest factor and kurtosis composition total characteristic vector.
Training sample is imported BP nerve net in the step 7 by technical solution as a further improvement of that present invention
Classification learning is carried out in network classifier, obtains trained BP neural network classifier specifically: training sample is imported into BP mind
Through being instructed using the gradient descent method of momentum plus adjusting learning rate to BP neural network classifier in network classifier
Practice, obtains trained BP neural network classification.
The invention has the benefit that the present invention acquires the vibration signal of light modulation motor device, vibration signal is utilized
Energy spectrum and time domain charactreristic parameter after WAVELET PACKET DECOMPOSITION extract total characteristic vector, import in BP neural network and classify, obtain
To abnormal sound recognition result, to select the light modulation motor device there are abnormal sound, recognition speed is fast, and accuracy rate is high, compares people
Ear recognition efficiency is high, there is practical application value.
Detailed description of the invention
Fig. 1 is the abnormal sound identifying system block diagram of embodiment.
Fig. 2 is the vibration signal time domain waveform comparison diagram of normal motor and abnormal sound motor.
Fig. 3 is three layers of WAVELET PACKET DECOMPOSITION tree.
Fig. 4 is BP neural network illustraton of model.
Fig. 5 is each frequency band spectrogram of the vibration signal of normal motor and abnormal sound motor after three layers of WAVELET PACKET DECOMPOSITION.
Fig. 6 is training error and frequency of training curve graph.
Fig. 7 is the comparison diagram of the prediction data classification and real data classification of test data.
Specific embodiment
A specific embodiment of the invention is further illustrated below according to Fig. 1 to Fig. 7:
Sound speciality can reflect on vibration signal, and the present embodiment uses the vibration signal of light modulation motor device
Acquisition is to carry out abnormal sound identification.It is the abnormal sound identifying system block diagram of light modulation motor device referring to Fig. 1, Fig. 1.Including acceleration
Sensor, APM data collecting card and computer, APM data collecting card acquire light modulation motor by acceleration transducer and fill
The vibration signal set, and Correlation method for data processing is carried out by computer, to judge light modulation motor device with the presence or absence of different
Sound.Acceleration transducer selects piezoelectric acceleration transducer, the location arrangements reasonable consideration Vibration propagation of acceleration transducer
Direction and the direction for being easy to detect comprehensively consider thereon, will accelerate it is also to be ensured that testing acceleration sensor position is identical every time
Degree sensor is fixed by bolts on the fixture of the power interface of light modulation motor device, and being arranged so can be in power supply
When engagement, acceleration transducer i.e. be adjacent to motor mechanism, can guarantee when detecting acceleration transducer position immobilize and
Normal operating is not influenced, and this step operation is consistent with interface detecting step in production line, can be future in production line
It is convenient that upper addition detection device provides.
The present embodiment provides a kind of abnormal sound recognition methods of light modulation motor device, the following steps are included:
Step 1, the vibration signal that light modulation motor device is acquired using acceleration transducer:
Since turn frequency and highest meshing frequency of the motor of light modulation motor device are in 220Hz or so, the electricity of motor
Magnetic noise is up to 4000Hz or so, therefore sample frequency selects 10KHz, sampling time 3.5s.Sampling is obtaining no abnormal sound just
Normal motor (being hereafter referred to as normal motor) and have abnormal sound bad motor (hereafter general designation abnormal sound motor) vibration signal time domain wave
Shape is as shown in Figure 2.Wherein 2 (a) be normal motor vibration signal time domain waveform, 2 (b) be abnormal sound motor A vibration signal when
Domain waveform, 2 (c) be the vibration signal time domain waveform of abnormal sound motor B；It can be seen that carrying out the vibration letter of abnormal sound motor A from 2 (b)
Number there are obvious periodically impact, 3.5s occur 6 Secondary Shocks, the general 1.7Hz of frequency, the canine tooth with light modulation motor device altogether
It takes turns consistent with the speed of axis, can substantially speculate to go wrong and appear in gear wheel or output shaft.The vibration of abnormal sound motor B is believed
Number it then cannot make out obvious impulse period, but to carry out accurately abnormal sound differentiation to it in producing line cannot only see that time domain is believed
Number, the feature for being included to signal needs further to extract, and could preferably identify abnormal sound.Therefore it needs to continue to execute step below
Rapid 2.
Step 2: j layers of WAVELET PACKET DECOMPOSITION are carried out to vibration signal；
WAVELET PACKET DECOMPOSITION simultaneously decomposes high and low frequency, and the present embodiment carries out 3 layers of wavelet packet point to vibration signal
Solution, as shown in figure 3, being three layers of WAVELET PACKET DECOMPOSITION tree, S is vibration signal in Fig. 3；Signal obtains each frequency after WAVELET PACKET DECOMPOSITION
The sub-band of section, these sub-bands include large number of equipment running state information, and the Energy distribution of abnormal sound motor and normal motor must
It is so different, frequency band coefficient is showed in the form of energy variation, the feature letter of reflection motor abnormal sound can be extracted
Breath, therefore execute below step 3.
Step 3: calculating the signal energy of each frequency band of WAVELET PACKET DECOMPOSITION, the signal energy of each frequency band of normalized obtains
The energy ratio of gross energy shared by signal energy to each frequency band；
The step 3 specifically includes the following steps:
(1) it is equal to square of 2 norms of signal according to the energy of Parseval energy integral signal in the time domain, by
Step 2 obtains j layers of wavelet decomposition, there is 2jA frequency band, then the signal energy E of each frequency bandj.mAre as follows:
Wherein, p=N/2j, N is the sampling number of original vibration signal, and m is jth layer WAVELET PACKET DECOMPOSITION coefficient, that is, is indicated
The serial number of frequency band；Sj,mFor the reconstruction signal of m-th of frequency band, xm,nFor the amplitude of reconstruction signal；
(2) gross energy is calculated, gross energy is the sum of the signal energy of each frequency band of WAVELET PACKET DECOMPOSITION, is indicated are as follows:
(3) signal energy of each frequency band of normalized, obtains the energy ratio of gross energy shared by the signal energy of each frequency band,
Construct normalized feature vector:
Step 4: analysis of time-domain characteristic is carried out to obtain multinomial time domain charactreristic parameter to vibration signal；Multinomial temporal signatures
Parameter includes crest factor and kurtosis；
In order to increase the reliability of detection system, it is necessary to comprehensively consider the various features of signal, the ginseng of certain time-domain signals
Number can also characterize Vibration Condition very well.Crest factor Cf, kurtosis KuThis 2 time domain parameters can distinguish well beat with normally
State；Crest factorWhereinIt is signal peak, xrmsIt is virtual value, CfIt can be in sensitive signal acquisition
Impact ingredient；KurtosisWhereinIndicate the probability density of vibration acceleration x, x is vibration
Dynamic acceleration, μxIt is mean value, KuFor examining signal to deviate the degree of normal distribution, the amplitude distribution of vibration signal is close to normal state
Distribution, kurtosis index value Ku≈3；Energy spectrum and time domain charactreristic parameter are comprehensively considered herein.
Step 5: combining the normalized feature vector that step 3 obtains and the multinomial time domain charactreristic parameter that step 4 obtains
At total characteristic vector.
Step 6: building BP neural network classifier:
BP neural network (Back Propagation Feed-forward NN), the feed-forward type nerve of back propagation learning
Network.Referring to fig. 4, by input layer, hidden layer and output layer up of three-layer.So-called feed-forward type neural network structure, that is, handled
When sample, input that the output of preceding layer is next layer.Backpropagation is meant between reality output and anticipated output
Difference be error guiding, continuous rounding error carrys out the network weight between each network layer of reverse adjustment, until input layer with
Until network weight adjusting thresholds between first hidden layer.Its core is exactly to pass through transmission error backward, forward regulating networks
Parameter (weight and threshold value), to realize or approach desired input, output mapping relations.
In BP learning algorithm, learning rate parameter selection is critically important, and too small, convergence rate is slow, excessive, easily causes shake
It swings and even dissipates.In order to solve this problem, the method that variable step can be used, also known as adjusting learning rate.It can first use
One moderate learning rate is continued to decline by network stabilization, error, if increasing learning rate, network is still restrained, then continued
Increase learning rate, until network error no longer reduces, learning rate is reduced, so that network stabilization.
It can also be using the gradient descent method for having momentum.It is adjusted in weight and introduces a momentum term in formula, as follows:
Δwji(k+1)=mc Δ wji(k)+(1-mc)ηδjxi(4)；
In formula (4), wjiFor the connection weight of input layer to hidden layer；η ∈ (0,1) indicates learning rate or step-length；δjIt indicates
For the error of hidden layer；Input vector is xi, k is frequency of training, and mc is factor of momentum, mc ∈ [0,1].The momentum term phase of addition
When in damping term, in this way when correcting weight according to gradient descent method, it is contemplated that the experience of previous moment accumulation reduces network
For the sensibility of error surface local detail, the concussion trend in learning process is reduced, constringency performance is improved.
Step 7: the vibration signal of acquisition light modulation motor device of the multiple groups without abnormal sound and the automobile light-modulating electricity for having abnormal sound
The vibration signal of machine device extracts the total characteristic vector of every group of vibration signal as training sample respectively, training sample is imported
Classification learning is carried out in BP neural network classifier, obtains trained BP neural network classifier.
Step 8: the total characteristic vector of the vibration signal of tested light modulation motor device is input to trained BP
In neural network classifier, the sound class of the light modulation motor device is exported, the sound class includes no abnormal sound classification
And have abnormal sound classification.
The present embodiment distinguishes abnormal sound using analysis vibration signal to light dimming electric motor, acquires normal motor and abnormal sound motor
Vibration signal is operated, selects " db3 " wavelet basis to carry out 3 layers of WAVELET PACKET DECOMPOSITION signal.After three layers of WAVELET PACKET DECOMPOSITION, each frequency band system
Number (3,0)~(3,7) respectively correspond expression frequency range are as follows: 0~625Hz, 625~1250Hz, 1250~1875Hz, 1875~
2500Hz, 2500~3125Hz, 3125~3750Hz, 3750~4375Hz, 4375~5000Hz.As shown in figure 5, Fig. 5 (a) is
Each frequency band spectrogram of the vibration signal S1 of normal motor after three layers of WAVELET PACKET DECOMPOSITION, Fig. 5 (b) are that the vibration of abnormal sound motor is believed
Number each frequency band spectrogram of the S2 after three floor WAVELET PACKET DECOMPOSITION.
It is composed by the signal energy that matlab platform calculates each frequency band, after normalized, obtains the vibration of normal motor
Signal S1 normalized feature vector T1=[0.1970,0.0623,0.1124,0.0938,0.2025,0.1207,
0.1144,0.0969], the vibration signal S2 of abnormal sound motor normalized feature vector T2=[0.0902,0.0679,
0.1325,0.1143,0.3029,0.1078,0.109,0.0753].It can be seen that vibration signal S1 and vibration signal S2 is main
For difference in 0~625Hz and 1875~2500Hz, the vibration signal low-frequency range energy of normal motor is high.
In order to increase the reliability of detection system, it is necessary to comprehensively consider the various features of signal, the present embodiment comprehensively considers
Energy spectrum and time domain charactreristic parameter；The vibration signal of 5 groups of normal motors and the vibration signal of 5 groups of abnormal sound motors are tested, they
Energy spectra data (normalized feature vector), crest factor CfWith kurtosis KuThe total characteristic vector matrix of composition is as follows:
The energy spectrum of normal motor vibration signal and the energy spectrum of abnormal sound motor oscillating signal are found out from upper (5), (6) formula
There is more apparent difference.The vibration signal of normal motor concentrates on low-frequency range section, during the vibration signal of abnormal sound motor then concentrates on
High frequency also has larger difference both on time domain charactreristic parameter.Test 100 groups of normal motors operating vibration signal and 100 groups
The operating vibration signal of abnormal sound motor, the total characteristic vector of these vibration signals of extraction are normal by these as training sample
The total characteristic vector of the vibration signal of motor forms total characteristic vector matrix, the total characteristic Vector Groups of the vibration signal of abnormal sound motor
At total characteristic vector matrix, is imported in BP neural network classifier as training sample and carry out classification learning.
By repeatedly attempting, the gradient descent method that selection has momentum to add adjusting learning rate is trained network, and 3 layers
Network topology structure 10-18-4, node in hidden layer 18, last training error, which is converged in, have been learnt 1499 steps and just reaches the phase
Hope error, and time-consuming less than 10s, error with frequency of training curve as shown in fig. 6, final error is 0.00099634, miss by target
Difference is 0.001.
30 exemplars of artificial selection, wherein abnormal sound motor and each half of normal motor, trained knot is detected as test set
Fruit exports result and shows to belong to no abnormal sound classification for 1, and output result is -1, then belonging to has abnormal sound classification.Test result is as follows
Shown in Fig. 7, the practical identification and classification of human ear is compareed, other than No. 10 part judges misalignment, remaining all meets, for No. 10
Part, machine judgement is abnormal sound part, however not excluded that human ear has erroneous judgement situation.
Abnormal sound detection is the problem that present factory's emphasis needs to solve, and uses vibration to the abnormal sound problem of light dimming electric motor
Signal acquisition can effectively find out abnormal sound problem derived from elliptical gear.It is combined using energy spectrum with time domain charactreristic parameter
Extract the feature of signal, support vector machines carries out machine learning classification, can efficiently identify light modulation motor abnormal sound, keep away
Artificial neural network and other classifier requirement high-performance computers are exempted from, constantly training needs great amount of samples, obtains differentiating letter
The defects of number time is long, there is practical application value.
Protection scope of the present invention includes but is not limited to embodiment of above, and protection scope of the present invention is with claims
Subject to, replacement, deformation, the improvement that those skilled in the art that any pair of this technology is made is readily apparent that each fall within of the invention
Protection scope.
Claims (5)
1. a kind of abnormal sound recognition methods of light modulation motor device, it is characterised in that: the following steps are included:
Step 1, the vibration signal that light modulation motor device is obtained using acceleration transducer；
Step 2: j layers of WAVELET PACKET DECOMPOSITION are carried out to vibration signal；
Step 3: calculating the signal energy of each frequency band of WAVELET PACKET DECOMPOSITION, the signal energy of each frequency band of normalized obtains each
The energy ratio of gross energy shared by the signal energy of frequency band；
Step 4: analysis of time-domain characteristic is carried out to obtain multinomial time domain charactreristic parameter to vibration signal；
Step 5: by the energy ratio of gross energy shared by the signal energy for each frequency band that step 3 obtains and step 4 obtain it is multinomial when
Characteristic of field parameter forms total characteristic vector；
Step 6: building BP neural network classifier；
Step 7: the vibration signal of test light modulation motor device of the multiple groups without abnormal sound and the light modulation motor dress for having abnormal sound
The vibration signal set extracts the total characteristic vector of every group of vibration signal as training sample respectively, training sample is imported BP mind
Classification learning is carried out in network classifier, obtains trained BP neural network classifier；
Step 8: the total characteristic vector of the vibration signal of tested light modulation motor device is input to trained BP nerve
In network classifier, the sound class of the light modulation motor device is exported, the sound class includes no abnormal sound and has abnormal sound.
2. the abnormal sound recognition methods of light modulation motor device according to claim 1, it is characterised in that: the step 3
The following steps are included:
(1) it is equal to square of 2 norms of signal according to the energy of Parseval energy integral signal in the time domain, by step
2 obtain j layers of wavelet decomposition, there is 2jA frequency band, then the signal energy E of each frequency bandj.mAre as follows:
Wherein, p=N/2j, N is the sampling number of original signal, and m is jth layer WAVELET PACKET DECOMPOSITION coefficient, that is, indicates the sequence of frequency band
Number；Sj,mFor the reconstruction signal of m-th of frequency band, xm,nFor the amplitude of reconstruction signal；
(2) gross energy is calculated, gross energy is the sum of the signal energy of each frequency band of WAVELET PACKET DECOMPOSITION, is indicated are as follows:
(3) signal energy of each frequency band of normalized obtains the energy ratio of gross energy shared by the signal energy of each frequency band, construction
Normalized feature vector out:
3. the abnormal sound recognition methods of light modulation motor device according to claim 2, it is characterised in that: the step 4
In multinomial time domain charactreristic parameter include crest factor and kurtosis.
4. the abnormal sound recognition methods of light modulation motor device according to claim 3, it is characterised in that: the step 5
The multinomial temporal signatures ginseng that the energy ratio and step 4 of gross energy shared by the signal energy of the middle each frequency band for obtaining step 3 obtain
Array is at total characteristic vector specifically: normalized feature vector, crest factor and the kurtosis composition total characteristic for obtaining step 3
Vector.
5. the abnormal sound recognition methods of light modulation motor device according to claim 1, it is characterised in that: the step
Importing training sample in BP neural network classifier in 7 carries out classification learning, obtains trained BP neural network classification
Implement body are as follows: training sample is imported in BP neural network classifier, is declined using the gradient of momentum plus adjusting learning rate
Method is trained BP neural network classifier, obtains trained BP neural network classification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810945473.7A CN109115330A (en) | 2018-08-20 | 2018-08-20 | A kind of abnormal sound recognition methods of light modulation motor device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810945473.7A CN109115330A (en) | 2018-08-20 | 2018-08-20 | A kind of abnormal sound recognition methods of light modulation motor device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109115330A true CN109115330A (en) | 2019-01-01 |
Family
ID=64852820
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810945473.7A Pending CN109115330A (en) | 2018-08-20 | 2018-08-20 | A kind of abnormal sound recognition methods of light modulation motor device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109115330A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113155271A (en) * | 2020-01-23 | 2021-07-23 | 上海擎动信息科技有限公司 | Sound vibration detection method, system, terminal and medium |
| CN115856628A (en) * | 2023-02-28 | 2023-03-28 | 宁波慧声智创科技有限公司 | Micro-special motor acoustic quality detection method based on PSO-SVM detection model |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004340706A (en) * | 2003-05-15 | 2004-12-02 | Toshiba Mitsubishi-Electric Industrial System Corp | Apparatus for diagnosing instrument |
| CN104992714A (en) * | 2015-05-22 | 2015-10-21 | 株洲联诚集团有限责任公司 | Motor abnormal sound detection method |
-
2018
- 2018-08-20 CN CN201810945473.7A patent/CN109115330A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004340706A (en) * | 2003-05-15 | 2004-12-02 | Toshiba Mitsubishi-Electric Industrial System Corp | Apparatus for diagnosing instrument |
| CN104992714A (en) * | 2015-05-22 | 2015-10-21 | 株洲联诚集团有限责任公司 | Motor abnormal sound detection method |
Non-Patent Citations (4)
| Title |
|---|
| 丛爽: "《面向MATLAB工具箱的神经网络理论与应用》", 30 November 1998, 中国科学技术大学出版社 * |
| 张新 等: "某车用调光电机异音检测识别", 《微特电机》 * |
| 王煜 等: "优化神经网络在矿用电机轴承故障诊断的应用", 《安徽理工大学学报（自然科学版）》 * |
| 王腾蛟: "基于小波神经网络的异步电机故障诊断研究", 《中国优秀硕士学位论文全文数据库 工程科技II辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113155271A (en) * | 2020-01-23 | 2021-07-23 | 上海擎动信息科技有限公司 | Sound vibration detection method, system, terminal and medium |
| CN113155271B (en) * | 2020-01-23 | 2023-08-22 | 上海擎动信息科技有限公司 | Acoustic vibration detection method, system, terminal and medium |
| CN115856628A (en) * | 2023-02-28 | 2023-03-28 | 宁波慧声智创科技有限公司 | Micro-special motor acoustic quality detection method based on PSO-SVM detection model |
| CN115856628B (en) * | 2023-02-28 | 2023-06-27 | 宁波慧声智创科技有限公司 | Micro-special motor acoustic quality detection method based on PSO-SVM detection model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Kankar et al. | Rolling element bearing fault diagnosis using autocorrelation and continuous wavelet transform | |
| CN110070060B (en) | Fault diagnosis method for bearing equipment | |
| CN110161343B (en) | Non-invasive real-time dynamic monitoring method for external powered device of intelligent train | |
| CN107219457B (en) | Frame-type circuit breaker fault diagnosis and degree assessment method based on operation attachment electric current | |
| Wu et al. | Investigation of engine fault diagnosis using discrete wavelet transform and neural network | |
| CN109033632B (en) | Trend prediction method based on deep quantum neural network | |
| CN100485342C (en) | Integrated supporting vector machine mixed intelligent diagnosing method for mechanical fault | |
| CN107702922B (en) | Rolling bearing fault diagnosis method based on LCD and stacking automatic encoder | |
| CN102841131B (en) | Intelligent steel cord conveyer belt defect identification method and intelligent steel cord conveyer belt defect identification system | |
| CN109186964A (en) | Rotary machinery fault diagnosis method based on angle resampling and ROC-SVM | |
| CN104786101A (en) | Monitoring method for vertical milling cutting vibration | |
| CN109323754A (en) | A kind of train wheel polygon fault diagnosis detection method | |
| CN108549856A (en) | A kind of human action and road conditions recognition methods | |
| CN113077005A (en) | System and method for detecting abnormity based on LSTM self-encoder and normal signal data | |
| CN109115330A (en) | A kind of abnormal sound recognition methods of light modulation motor device | |
| CN108732421A (en) | The acquisition methods and device of the instantaneous frequency of bullet train dynamic response signal | |
| Min et al. | Novel pattern detection in children with autism spectrum disorder using iterative subspace identification | |
| Jiang et al. | Belt conveyor roller fault audio detection based on the wavelet neural network | |
| CN112508242A (en) | Method for constructing bearing fault location and classification model of wind power generator | |
| CN117972547B (en) | Fault early warning method, device and medium for key components of wind turbine generator | |
| Zhao et al. | A novel deep fuzzy clustering neural network model and its application in rolling bearing fault recognition | |
| Hwang et al. | Application of cepstrum and neural network to bearing fault detection | |
| Patel et al. | Fault diagnostics of rolling bearing based on improve time and frequency domain features using artificial neural networks | |
| Dai et al. | Acceleration-guided acoustic signal denoising framework based on learnable wavelet transform applied to slab track condition monitoring | |
| CN110032987B (en) | Surface electromyographic signal classification method based on cerebellar neural network model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190101 |
|
| RJ01 | Rejection of invention patent application after publication |