CN109946389A - Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network - Google Patents
Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network Download PDFInfo
- Publication number
- CN109946389A CN109946389A CN201910099951.1A CN201910099951A CN109946389A CN 109946389 A CN109946389 A CN 109946389A CN 201910099951 A CN201910099951 A CN 201910099951A CN 109946389 A CN109946389 A CN 109946389A
- Authority
- CN
- China
- Prior art keywords
- layer
- convolutional neural
- obtains
- training
- neural networks
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000006378 damage Effects 0.000 title claims abstract description 43
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 34
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 title claims abstract description 19
- 238000012549 training Methods 0.000 claims description 34
- 230000001133 acceleration Effects 0.000 claims description 15
- 238000012545 processing Methods 0.000 claims description 15
- 230000000644 propagated effect Effects 0.000 claims description 15
- 238000001514 detection method Methods 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 12
- 238000005516 engineering process Methods 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 8
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 6
- 230000004044 response Effects 0.000 claims description 6
- 230000003595 spectral effect Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 abstract description 7
- 230000036541 health Effects 0.000 abstract description 6
- 230000008901 benefit Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 19
- 238000013528 artificial neural network Methods 0.000 description 6
- 230000001066 destructive effect Effects 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 5
- 238000000605 extraction Methods 0.000 description 3
- 230000003044 adaptive effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000009022 nonlinear effect Effects 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
- G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
- G01N29/46—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by spectral analysis, e.g. Fourier analysis or wavelet analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Signal Processing (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Biochemistry (AREA)
- Pathology (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Immunology (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Complex Calculations (AREA)
- Saccharide Compounds (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
- Image Analysis (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a structural damage identification method based on ensemble empirical mode decomposition and a convolutional neural network, which belongs to the technical field of structural health monitoring. The method reduces the influence of noise on the identification result of the structural damage position, improves the identification accuracy rate, improves the identification efficiency, and has remarkable advantages in real-time structural health monitoring.
Description
Technical field
The invention belongs to structural health monitoring technology fields, more particularly to one kind to be based on overall experience mode decomposition and convolution
The Damage Detection of Structures of neural network.
Background technique
Since heavy construction structure is closely bound up with people's lives, Latent destruction present in structure and damage, all
Heavy losses may be caused to people's production and life, thus, the monitoring structural health conditions of heavy construction structure have important
Meaning.Nearly two during the last ten years, and domestic and foreign scholars propose a large amount of Structural Damage Identification, these methods are broadly divided into base
In the method for the method and data-driven of model, wherein the response data that the method for data-driven directly utilizes sensor to survey
Damage Assessment Method is carried out, the finite element model of structure is not needed, receives the favor of lot of domestic and foreign scholar, has been obtained quickly
Development.
In the method for data-driven, machine learning techniques cause more and more concerns, and method includes Bayesian network
Network, fuzzy logic, support vector machines (Support vector machine, SVM) and artificial neural network (Artificial
Neural network, ANN) etc..Wherein, ANN has been applied in Damage Assessment Method field for many years, and damage characteristic is usual
It is used as their input, as a kind of adaptive mode identification technology, does not need to provide discriminant function in advance, pass through
The study mechanism of itself can automatically form required decision region, and can make full use of sample information.But ANN needs
It is artificial to extract characteristic parameter, it is easy to cause that the representativeness of feature is poor, accuracy rate is low, calculating is time-consuming, can not carry out real-time monitoring etc.
Problem.Since convolutional neural networks (Convolution neural network, CNN) extracts feature without artificial, has phase
Document report is closed to be used in Damage Assessment Method, can be directly using acceleration responsive data as input vector, it will
Feature extraction and damage classifying carry out real-time non-destructive tests and positioning to structure as a whole.However, current identification side
Method is usually directly using original vibration response signal as network inputs vector, since vector dimension is larger, computational efficiency compared with
It is low, moreover, often containing noise and the information unrelated with structural damage, the accuracy rate of non-destructive tests in actual measurement vibration response signal
It is not high.
Further, existing Structural Damage Identification is typically based on Hilbert-Huang transform (Hilbert- at present
Huang Transform, HHT), HHT is a kind of new Time-Frequency Analysis Method with adaptive ability, by empirical mode decomposition
Two parts composition is converted with Hilbert, specific function form is not needed and data is decomposed, overcome traditional time frequency analysis side
The defect of non-stationary, nonlinear properties is indicated in method with meaningless harmonic component.But the Empirical Mode as the core HHT
State decomposition can still have the problems such as modal overlap and end effect, and wherein modal overlap phenomenon is especially prominent.
Thus, the efficiency and accuracy rate of Damage Assessment Method how are improved, is a current technical problem urgently to be solved.
Summary of the invention
The present invention for the above technical issues, proposes a kind of based on overall experience mode decomposition and convolutional neural networks
Damage Detection of Structures, the acceleration responsive signal of selecting structure use overall experience mode decomposition as analysis object
Technology is decomposed and is reconstructed to acceleration responsive signal, while using one-dimensional convolutional neural networks (One-dimensional
Convolution neural network, 1D CNN) feature extraction and damage classifying are carried out to reconstruction signal, to be tied
The recognition result of structure damage position had both reduced influence of the noise to damage location identification of structures result, and it is accurate to improve identification
Rate, and improve recognition efficiency.
In order to achieve the above object, the technical solution adopted by the present invention are as follows:
Damage Detection of Structures based on overall experience mode decomposition and convolutional neural networks, comprising the following steps:
Step 1: measuring structure acceleration response signal x (t)；
Step 2: acceleration responsive x (t) being decomposed using overall experience mode decomposition technology, is obtained a series of intrinsic
Mode function IMF；
Step 3: Fourier transformation being carried out to the intrinsic mode function IMF that step 2 obtains and obtains its spectral characteristic, selection contains
There is the intrinsic mode function IMF of natural frequency of structures to be reconstructed, obtains reconstruction signal z (t)；
Step 4: the reconstruction signal z (t) that step 3 obtains is standardized, standardization reconstruction signal Z (t) is obtained,
Standardization reconstruction signal Z (t) is divided into the certain subsignal of multiple lengths, chooses the part signal in subsignal as training sample
This, remainder signal is as test sample；
Step 5: the training sample that step 4 obtains being inputted in one-dimensional convolutional neural networks model and is trained, was trained
Using cross entropy loss function as loss function in journey, using Adam optimization algorithm as optimization algorithm；
Step 6: the test sample that step 4 is obtained is input in the trained one-dimensional convolutional neural networks model of step 5,
Obtain the recognition result of damage position.
Preferably, being standardized place to the reconstruction signal z (t) using standardization formula in the step 4
Reason, the expression formula of the standardization formula are as follows:
In formula, Z (t) is standardization reconstruction signal；Z (t) is reconstruction signal；μ is the average value of reconstruction signal；σ is reconstruct
The variance of signal.
Preferably, the one-dimensional convolutional neural networks model includes sequentially connected input layer, in the step 5
One convolutional layer, first batch normalization layer, the first pond layer, the second convolutional layer, second batch normalize layer, the second pond layer,
First full articulamentum, the second full articulamentum and output layer；Wherein, first convolutional layer and the second convolutional layer contain 16 big
The pond window size of the small convolution kernel for being 10 × 1, first pond layer and the second pond layer is 2 × 1.
Preferably, the training of the one-dimensional convolutional neural networks model includes forward direction alternately in the step 5
Trained and backpropagation training is propagated, until the output knot of the output result and setting of network convergence or propagated forward training
Error amount between fruit reaches specified stopping criterion for iteration；The propagated forward training step includes: the instruction for obtaining step 4
Practice sample to input from the input layer, first convolutional layer handles the training sample of input to obtain first processing spy
Sign figure, the first processing feature figure is sent to the first batch normalization layer and is normalized, after normalized
It is sent to first pond layer and carries out pond processing, obtain initial sample characteristic pattern；The initial sample characteristic pattern is sent to
Second convolutional layer is handled, and secondary treatment characteristic pattern is obtained, and the secondary treatment characteristic pattern is sent to the second batch
Amount normalization layer is normalized, and second pond layer is sent to after normalized and carries out pond processing, obtains two
Secondary sampling characteristic pattern；The double sampling characteristic pattern is sequentially delivered at the described first full articulamentum and the second full articulamentum
Reason, is sent to the output layer after processing, the output layer to treated, classified and exports damage by double sampling characteristic pattern
Hurt the class vector of position；The backpropagation training step includes: by the output result and setting of propagated forward training
Output result error amount according to the propagated forward training reversed passback, obtain every layer of error amount, it is excellent using Adam
Change the network parameter that algorithm adjusts the one-dimensional convolutional neural networks model.
Compared with prior art, the advantages and positive effects of the present invention are:
Damage Detection of Structures provided by the invention based on overall experience mode decomposition and convolutional neural networks chooses knot
The acceleration responsive signal of structure carries out acceleration responsive signal as analysis object, and using overall experience mode decomposition technology
It decomposes and reconstructs, only extract information related with structural damage in reconstruct, the noise unrelated with structural damage can be reduced
It influences, improves the accuracy rate of non-destructive tests；Meanwhile using one-dimensional convolutional neural networks to reconstruction signal carry out feature extraction and
The efficiency of damage classifying, non-destructive tests is higher.Thus, it is provided by the invention to be based on overall experience mode decomposition and convolutional Neural net
The Damage Detection of Structures of network has significant advantage in real-time architecture health monitoring.
Detailed description of the invention
Fig. 1 is the structural damage based on overall experience mode decomposition and convolutional neural networks provided by the embodiment of the present invention
The flow diagram of method of identification；
Fig. 2 is the structural schematic diagram of one-dimensional convolutional neural networks provided by the embodiment of the present invention；
In upper figure, 1, input layer；2, the first convolutional layer；3, the first batch normalization layer；4, the first pond layer；5, volume Two
Lamination；6, the second batch normalization layer；7, the second pond layer；8, the first full articulamentum；9, the second full articulamentum；10, output layer.
Specific embodiment
The technical scheme in the embodiments of the invention will be clearly and completely described below, it is clear that described implementation
Example is only a part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, this field is common
Technical staff's every other embodiment obtained without making creative work belongs to the model that the present invention protects
It encloses.
As shown in Figure 1, the embodiment of the invention provides a kind of based on overall experience mode decomposition and convolutional neural networks
Damage Detection of Structures, comprising the following steps:
Step 1: measuring structure acceleration response signal x (t).
Step 2: acceleration responsive x (t) being decomposed using overall experience mode decomposition technology, is obtained a series of intrinsic
Mode function IMF.It should be noted that the overall experience mode decomposition technology is the prior art, referring to " base in this step
Contributed in the internal combustion engine radiated noise of MEEMD " (Zheng Xu etc., journal of Zhejiang university (engineering version), the 5th phase of volume 46, page 954-
Page 960).Decomposed that specific step is as follows to acceleration responsive x (t) using overall experience mode decomposition technology:
(1) the white noise signal ω (t) that one group of equal length is added to acceleration responsive x (t), obtains:
Y (t)=x (t)+ω (t) (1)
(2) the signal y (t) after adding white noise to step (1) is decomposed, the specific steps are as follows:
1. determining all Local Extremums of signal y (t), all Local modulus maximas are connected with cubic spline line
Coenvelope line is formed, all local minizing points are connected to form lower envelope line.
2. the average value of coenvelope line and lower envelope line is denoted as m1(t), then it finds out:
h1(t)=y (t)-m1(t) (2)。
3. if 2. h that step acquires1(t) meet the condition of intrinsic mode function IMF, then h1It (t) is first of y (t)
IMF component c1(t)；If h1(t) it is unsatisfactory for the condition of intrinsic mode function IMF, then uses h1(t) replace y (t) repeat step 1.-
3. then having:
h1k(t)=h1(k-1)-m1k(t) (3)
In formula (3), k be step 1.-number of repetition 3., m1k(t) for kth time repeat step 1. -2. when the upper packet that obtains
The average value of winding thread and lower envelope line；
Work as h1k(t) stop repeating when meeting intrinsic mode function IMF, remember first IMF component c of y (t)1(t)=h1k
(t).It should be noted that intrinsic mode function IMF need to meet two conditions: first is that extreme point quantity and zeroaxial quantity are entire
Equal or difference one in data area, second is that the envelope that all maximum points and minimum point are formed at any point is flat
Mean value is equal to zero.
4. isolating first IMF component c from y (t) using formula (4)1(t), first residual volume r is obtained1(t), formula
(4) expression formula is as follows:
r1(t)=y (t)-c1(t) (4)
With first residual volume r1(t) replace y (t) repeat step 1. -3., obtain second IMF component of y (t), be denoted as
c2(t)。
5. repeat step 1. -4., obtain the n IMF component of y (t), then have:
rn(t)=rn-1(t)-cn(t) (5)
In formula (5), rnIt (t) is n-th of residual volume, rn-1It (t) is (n-1)th residual volume, cn(t) n-th of IMF for being y (t)
Component；
As n-th of residual volume rn(t) become a monotonic function, it cannot be again from the IMF component for wherein extracting the condition that meets
When, circulation terminates.
(3) step (1) and step (2) are repeated, M white noise signal ω (t), the white noise added every time are added to x (t) altogether
Acoustical signal ω (t) is different, takes the average value of M IMF component set as final intrinsic mode function IMF, then final
The weight expression of intrinsic mode function IMF are as follows:
In formula (6), cziIt (t) is i-th final of IMF component, cimIt (t) is the of the m time addition white noise signal acquisition
I IMF component, M are total addition number of white noise signal.
Step 3: Fourier transformation being carried out to the intrinsic mode function IMF that step 2 obtains and obtains its spectral characteristic, selection contains
There is the intrinsic mode function IMF of natural frequency of structures to be reconstructed, obtains reconstruction signal z (t).
Step 4: the reconstruction signal z (t) that step 3 obtains being standardized using standardization formula, is obtained
It standardizes reconstruction signal Z (t), the expression formula of the standardization formula are as follows:
In formula (7), Z (t) is standardization reconstruction signal；Z (t) is reconstruction signal；μ is the average value of reconstruction signal；σ attaches most importance to
The variance of structure signal；
Standardization reconstruction signal Z (t) is divided into the certain subsignal of multiple lengths, the part signal chosen in subsignal is made
For training sample, remainder signal is as test sample.
Step 5: the training sample that step 4 obtains being inputted in one-dimensional convolutional neural networks model and is trained, was trained
Using cross entropy loss function as loss function in journey, using Adam optimization algorithm as optimization algorithm；
As shown in Fig. 2, one-dimensional convolutional neural networks model includes sequentially connected input layer 1, the first convolution in this step
Layer the 2, first batch normalization layer 3, the first pond layer 4, the second batch of convolutional layer 5, second normalize layer 6, the second pond layer 7,
First full articulamentum 8, the second full articulamentum 9 and output layer 10；Wherein, the first convolutional layer 2 and the second convolutional layer 5 contain 16
The pond window size of the convolution kernel that size is 10 × 1, the first pond layer 4 and the second pond layer 7 is 2 × 1；
The training of one-dimensional convolutional neural networks model includes propagated forward training and backpropagation training alternately, directly
Reach specified iteration to the error amount between the output result of the output result and setting of network convergence or propagated forward training
Termination condition；
Propagated forward training step includes: to input the training sample that step 4 obtains from input layer 1, the first convolutional layer 2
The training sample of input is handled to obtain first processing feature figure, first processing feature figure is sent to the first batch and normalizes
Layer 3 is normalized, and the first pond layer 4 is sent to after normalized and carries out pond processing, obtains initial sample feature
Figure；Initial sample characteristic pattern is sent to the second convolutional layer 5 and is handled, and obtains secondary treatment characteristic pattern, secondary treatment characteristic pattern
It is sent to the second batch normalization layer 6 to be normalized, the second pond layer 7 is sent to after normalized and carries out pond Hua Chu
Reason, obtains double sampling characteristic pattern；Double sampling characteristic pattern be sequentially delivered to the first full articulamentum 8 and the second full articulamentum 9 into
Row processing, is sent to output layer 10 after processing, output layer 10 to treated, classified and exports damage by double sampling characteristic pattern
Hurt the class vector of position；
Backpropagation training step includes: the error amount of the output result by the output result of propagated forward training and setting
According to the reversed passback of propagated forward training, every layer of error amount is obtained, adjusts one-dimensional convolutional Neural using Adam optimization algorithm
The network parameter of network model.
Step 6: the test sample that step 4 is obtained is input in the trained one-dimensional convolutional neural networks model of step 5,
Obtain the recognition result of damage position.
The above-mentioned Damage Detection of Structures based on overall experience mode decomposition and convolutional neural networks provided by the invention, choosing
Take the acceleration responsive signal of structure as analysis object, and using overall experience mode decomposition technology to acceleration responsive signal
It is decomposed and is reconstructed, only extracted information related with structural damage in reconstruct, unrelated with structural damage make an uproar can be reduced
The influence of sound improves the accuracy rate of non-destructive tests；Meanwhile feature is carried out to reconstruction signal using one-dimensional convolutional neural networks and is mentioned
It takes and damage classifying, the efficiency of non-destructive tests is higher.Thus, it is provided by the invention above-mentioned based on overall experience mode decomposition and volume
The Damage Detection of Structures of product neural network, has significant advantage in real-time architecture health monitoring.
Claims (4)
1. the Damage Detection of Structures based on overall experience mode decomposition and convolutional neural networks, which is characterized in that including following
Step:
Step 1: measuring structure acceleration response signal x (t)；
Step 2: acceleration responsive x (t) being decomposed using overall experience mode decomposition technology, obtains a series of natural mode of vibration
Function IMF；
Step 3: Fourier transformation being carried out to the intrinsic mode function IMF that step 2 obtains and obtains its spectral characteristic, selection contains knot
The intrinsic mode function IMF of the structure natural frequency of vibration is reconstructed, and obtains reconstruction signal z (t)；
Step 4: the reconstruction signal z (t) that step 3 obtains being standardized, standardization reconstruction signal Z (t) is obtained, will mark
Standardization reconstruction signal Z (t) is divided into the certain subsignal of multiple lengths, chooses part signal in subsignal as training sample,
Remainder signal is as test sample；
Step 5: the training sample that step 4 obtains being inputted in one-dimensional convolutional neural networks model and is trained, in training process
Using cross entropy loss function as loss function, using Adam optimization algorithm as optimization algorithm；
Step 6: the test sample that step 4 is obtained is input in the trained one-dimensional convolutional neural networks model of step 5, is obtained
The recognition result of damage position.
2. the Damage Detection of Structures according to claim 1 based on overall experience mode decomposition and convolutional neural networks,
It is characterized by: being standardized using standardization formula to the reconstruction signal z (t), institute in the step 4
State the expression formula of standardization formula are as follows:
In formula, Z (t) is standardization reconstruction signal；Z (t) is reconstruction signal；μ is the average value of reconstruction signal；σ is reconstruction signal
Variance.
3. the Damage Detection of Structures according to claim 1 based on overall experience mode decomposition and convolutional neural networks,
It is characterized by: the one-dimensional convolutional neural networks model includes sequentially connected input layer, the first convolution in the step 5
Layer, first batch normalization layer, the first pond layer, the second convolutional layer, second batch normalization layer, the second pond layer, first entirely
Articulamentum, the second full articulamentum and output layer；Wherein, it is 10 that first convolutional layer and the second convolutional layer, which contain 16 sizes,
The pond window size of × 1 convolution kernel, first pond layer and the second pond layer is 2 × 1.
4. the Damage Detection of Structures according to claim 3 based on overall experience mode decomposition and convolutional neural networks,
It is characterized in that, the training of the one-dimensional convolutional neural networks model includes propagated forward instruction alternately in the step 5
Experienced and backpropagation training, until between the output result of the output result and setting of network convergence or propagated forward training
Error amount reach specified stopping criterion for iteration；
The propagated forward training step includes: to input the training sample that step 4 obtains from the input layer, and described first
Convolutional layer handles the training sample of input to obtain first processing feature figure, and the first processing feature figure is sent to described
First batch normalization layer is normalized, and first pond layer is sent to after normalized and carries out pond processing,
Obtain initial sample characteristic pattern；The initial sample characteristic pattern is sent to second convolutional layer and is handled, and obtains secondary place
Characteristic pattern is managed, the secondary treatment characteristic pattern is sent to the second batch normalization layer and is normalized, at normalization
It is sent to second pond layer after reason and carries out pond processing, obtains double sampling characteristic pattern；The double sampling characteristic pattern according to
It is secondary to be sent to the described first full articulamentum and the second full articulamentum is handled, the output layer is sent to after processing, it is described defeated
Layer to treated, classified and exports the class vector of damage position by double sampling characteristic pattern out；
The backpropagation training step includes: the mistake of the output result by the output result of propagated forward training and setting
Difference obtains every layer of error amount according to the reversed passback of propagated forward training, using described in the adjustment of Adam optimization algorithm
The network parameter of one-dimensional convolutional neural networks model.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910099951.1A CN109946389B (en) | 2019-01-31 | 2019-01-31 | Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network |
| AU2020214409A AU2020214409B2 (en) | 2019-01-31 | 2020-01-21 | Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network |
| PCT/CN2020/073357 WO2020156348A1 (en) | 2019-01-31 | 2020-01-21 | Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910099951.1A CN109946389B (en) | 2019-01-31 | 2019-01-31 | Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109946389A true CN109946389A (en) | 2019-06-28 |
| CN109946389B CN109946389B (en) | 2020-12-25 |
Family
ID=67007470
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910099951.1A Active CN109946389B (en) | 2019-01-31 | 2019-01-31 | Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network |
Country Status (3)
| Country | Link |
|---|---|
| CN (1) | CN109946389B (en) |
| AU (1) | AU2020214409B2 (en) |
| WO (1) | WO2020156348A1 (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472272A (en) * | 2019-07-01 | 2019-11-19 | 广东工业大学 | A kind of Structural Damage Identification based on multi-parameter and convolutional neural networks |
| CN110490249A (en) * | 2019-08-16 | 2019-11-22 | 哈尔滨工业大学 | Structural Damage Identification based on attribute class relations and few sample meta learning |
| CN111220921A (en) * | 2020-01-08 | 2020-06-02 | 重庆邮电大学 | Lithium battery capacity estimation method based on improved convolution-long-and-short-term memory neural network |
| CN111241725A (en) * | 2019-12-30 | 2020-06-05 | 浙江大学 | Structure response reconstruction method for generating countermeasure network based on conditions |
| WO2020156348A1 (en) * | 2019-01-31 | 2020-08-06 | 青岛理工大学 | Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network |
| CN111595541A (en) * | 2020-04-27 | 2020-08-28 | 河海大学 | Multi-dimensional structure damage identification method based on mass vibration transmissibility data convolutional neural network processing |
| CN111753776A (en) * | 2020-06-29 | 2020-10-09 | 重庆交通大学 | Structural damage identification method based on echo state and multi-scale convolution combined model |
| CN111931983A (en) * | 2020-07-07 | 2020-11-13 | 湖南大学 | Precipitation prediction method and system |
| CN112182490A (en) * | 2020-09-01 | 2021-01-05 | 华中科技大学 | Reactor state diagnosis method and system |
| CN112434624A (en) * | 2020-11-30 | 2021-03-02 | 广州大学 | Structural health monitoring data distortion detection method, system, device and storage medium |
| CN112525467A (en) * | 2020-11-26 | 2021-03-19 | 山东科技大学 | Impact damage area identification method and device suitable for cantilever beam |
| CN112712060A (en) * | 2021-01-15 | 2021-04-27 | 国网浙江省电力有限公司电力科学研究院 | Automatic high-voltage post porcelain insulator acoustic signal identification method |
| CN113159282A (en) * | 2021-03-29 | 2021-07-23 | 东南大学 | Intelligent updating method for bridge modal parameters based on cross modal confidence criterion matrix |
| CN114002334A (en) * | 2021-09-29 | 2022-02-01 | 西安交通大学 | Structural damage acoustic emission signal identification method and device and storage medium |
| CN114062511A (en) * | 2021-10-24 | 2022-02-18 | 北京化工大学 | Single-sensor-based intelligent acoustic emission identification method for early damage of aircraft engine |
| CN114358214A (en) * | 2022-03-11 | 2022-04-15 | 中国科学院上海高等研究院 | Gait adaptive identification method and device, storage medium and terminal |
| CN114462457A (en) * | 2022-04-11 | 2022-05-10 | 自然资源部第一海洋研究所 | Ship underwater noise deep learning identification method based on intrinsic probability density function |
| CN115169409A (en) * | 2022-07-18 | 2022-10-11 | 四川省公路规划勘察设计研究院有限公司 | Bridge structure natural vibration frequency identification and early warning method and device based on sliding window |
| CN116244803A (en) * | 2023-02-22 | 2023-06-09 | 合肥工业大学 | Bridge internal response reconstruction method based on physical and data dual drive |
Families Citing this family (37)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111896625B (en) * | 2020-08-17 | 2023-07-14 | 中南大学 | Rail damage real-time monitoring method and monitoring system thereof |
| CN112364710B (en) * | 2020-10-20 | 2024-04-05 | 西安理工大学 | Plant electric signal classification and identification method based on deep learning algorithm |
| CN112284736B (en) * | 2020-10-23 | 2022-02-01 | 天津大学 | Convolutional neural network fault diagnosis method based on multi-channel attention module |
| CN112381770B (en) * | 2020-11-03 | 2024-04-05 | 西安交通大学 | Wear surface three-dimensional morphology measurement method based on fusion convolutional neural network |
| CN112348255B (en) * | 2020-11-06 | 2024-04-09 | 湖南大学 | Ultra-short-term wind power prediction method based on wavelet time-frequency imaging |
| CN112465027A (en) * | 2020-11-27 | 2021-03-09 | 株洲时代电子技术有限公司 | Steel rail damage detection method |
| CN112465225A (en) * | 2020-11-27 | 2021-03-09 | 云南电网有限责任公司电力科学研究院 | Wind power prediction method based on secondary modal decomposition and cascade deep learning |
| CN112528820A (en) * | 2020-12-03 | 2021-03-19 | 重庆邮智机器人研究院有限公司 | Motor imagery electroencephalogram signal identification method based on EMD and DCNN |
| CN112528819B (en) * | 2020-12-05 | 2023-01-20 | 西安电子科技大学 | P300 electroencephalogram signal classification method based on convolutional neural network |
| CN112580239B (en) * | 2020-12-15 | 2022-07-05 | 中南大学 | Substructure response reconstruction method and system, and storage medium |
| CN112613105B (en) * | 2020-12-25 | 2024-03-08 | 扬州大学 | Reinforced concrete beam damage state identification method based on acoustic emission parameter and genetic algorithm optimization BP neural network |
| CN112699792B (en) * | 2020-12-29 | 2024-04-05 | 长安大学 | Machine tool bearing state identification method based on sine and cosine algorithm optimized convolutional neural network |
| CN112668527B (en) * | 2020-12-31 | 2023-04-25 | 华南理工大学 | Ultrasonic guided wave semi-supervised imaging detection method |
| CN113158769A (en) * | 2021-03-03 | 2021-07-23 | 安徽大学 | CEEMDAN and FastICA-based electromechanical device bearing vibration signal denoising method |
| CN113159179B (en) * | 2021-04-22 | 2023-04-18 | 中车株洲电力机车有限公司 | Subway and subway bogie running state identification method and system |
| CN113155973B (en) * | 2021-05-05 | 2023-05-16 | 温州大学 | Beam damage identification method based on self-adaptive singular value decomposition |
| CN113378426B (en) * | 2021-05-05 | 2023-07-14 | 温州大学 | Mechanical blade damage identification method based on information fusion |
| CN114330636A (en) * | 2021-06-04 | 2022-04-12 | 北京蜜蜂工物科技有限公司 | Non-intrusive power load decomposition and identification method and device, electronic equipment and storage medium |
| CN113591604A (en) * | 2021-07-09 | 2021-11-02 | 浙江工业大学 | Wind driven generator power quality disturbance state monitoring method based on noise label reconstruction |
| CN113743297B (en) * | 2021-09-03 | 2024-08-06 | 重庆大学 | Storage tank dome displacement data restoration method and device based on deep learning |
| CN113742983A (en) * | 2021-10-09 | 2021-12-03 | 福州大学 | Long-span structural damage identification method based on depth self-encoder neural network |
| CN113837141B (en) * | 2021-10-12 | 2023-10-27 | 国网山东省电力公司电力科学研究院 | Signal extraction method and device for resisting mouse repeller interference |
| CN114077849B (en) * | 2021-11-17 | 2024-04-05 | 西安交通大学 | Engine health state identification method based on component level fusion |
| CN114324580A (en) * | 2021-12-03 | 2022-04-12 | 西安交通大学 | Intelligent knocking detection method and system for structural defects |
| CN114386452B (en) * | 2021-12-06 | 2024-04-05 | 西安交通大学 | Nuclear power circulating water pump sun gear fault detection method |
| CN114169374B (en) * | 2021-12-10 | 2024-02-20 | 湖南工商大学 | Cable-stayed bridge stay cable damage identification method and electronic equipment |
| CN114266280B (en) * | 2021-12-29 | 2023-10-13 | 西安交通大学 | Bolt loosening state identification method based on improved convolutional neural network |
| CN114510762B (en) * | 2022-01-26 | 2024-06-28 | 中铁第四勘察设计院集团有限公司 | Structural damage identification method and system based on time sequence model coefficient sensitivity |
| CN114611560B (en) * | 2022-03-18 | 2024-09-06 | 西安电子科技大学 | SSVEP electroencephalogram signal classification method based on convolutional neural network |
| CN114417681B (en) * | 2022-04-01 | 2022-06-17 | 中国工程物理研究院流体物理研究所 | Two-dimensional structure deformation monitoring method and device based on dynamic decision and neural network |
| CN114821415B (en) * | 2022-04-24 | 2022-11-29 | 中国水利水电科学研究院 | Video micro-vibration amplification calibration and real-time extraction method based on machine learning |
| CN114818839B (en) * | 2022-07-01 | 2022-09-16 | 之江实验室 | Deep learning-based optical fiber sensing underwater acoustic signal identification method and device |
| CN116208356B (en) * | 2022-10-27 | 2023-09-29 | 浙江大学 | Virtual currency mining flow detection method based on deep learning |
| CN116559809B (en) * | 2023-03-28 | 2024-06-14 | 南京桂瑞得信息科技有限公司 | Radar radiation source individual identification method based on multi-source fusion network |
| CN116756515B (en) * | 2023-07-07 | 2024-02-23 | 广州大学 | Artificial seismic wave generation method based on deep convolution generation countermeasure network |
| CN117907445B (en) * | 2024-01-13 | 2024-08-13 | 大连理工大学 | Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network |
| CN118312836B (en) * | 2024-06-07 | 2024-08-13 | 山东科技大学 | Method and system for detecting bolt connection state between rigid cage guides |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104502126A (en) * | 2014-12-28 | 2015-04-08 | 华东交通大学 | Modal intervals-based high-speed train bogie fault diagnosis method |
| US20150164413A1 (en) * | 2013-12-13 | 2015-06-18 | National Chung Shan Institute Of Science And Technology | Method of creating anesthetic consciousness index with artificial neural network |
| CN106874957A (en) * | 2017-02-27 | 2017-06-20 | 苏州大学 | A kind of Fault Diagnosis of Roller Bearings |
| CN107123033A (en) * | 2017-05-04 | 2017-09-01 | 北京科技大学 | A kind of garment coordination method based on depth convolutional neural networks |
| CN107329933A (en) * | 2017-07-14 | 2017-11-07 | 北京知觉科技有限公司 | Fault detection method and device based on Fibre Optical Sensor vibration signal |
| CN107505396A (en) * | 2017-09-12 | 2017-12-22 | 燕山大学 | A kind of structural damage on-line real time monitoring method and system |
| CN107826919A (en) * | 2017-10-20 | 2018-03-23 | 中国矿业大学 | A kind of lifting system critical component multimode health monitoring device and monitoring method |
| CN108717869A (en) * | 2018-05-03 | 2018-10-30 | 中国石油大学(华东) | Diabetic retinopathy diagnosis aid system based on convolutional neural networks |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150293143A1 (en) * | 2013-07-21 | 2015-10-15 | National Chiao Tung University | Feature Extraction from Human Gaiting Patterns using Principal Component Analysis and Multivariate Empirical Mode Decomposition |
| CN103543026B (en) * | 2013-10-23 | 2015-07-01 | 青岛理工大学 | Structural damage identification method based on vibration transfer rate function and support vector machine |
| CN106096605B (en) * | 2016-06-02 | 2019-03-19 | 史方 | A kind of image obscuring area detection method and device based on deep learning |
| CN107328868B (en) * | 2017-06-27 | 2019-06-07 | 福州大学 | A kind of Analysis of Acoustic Emission Signal method of quick identification ceramic coating failure type |
| US11630040B2 (en) * | 2017-07-11 | 2023-04-18 | Qatar University | Real-time structural damage detection by convolutional neural networks |
| CN109166100A (en) * | 2018-07-24 | 2019-01-08 | 中南大学 | Multi-task learning method for cell count based on convolutional neural networks |
| CN109145961B (en) * | 2018-07-30 | 2021-10-22 | 上海交通大学 | Pattern recognition method and system for unstructured partial discharge data |
| CN109946389B (en) * | 2019-01-31 | 2020-12-25 | 青岛理工大学 | Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network |
-
2019
- 2019-01-31 CN CN201910099951.1A patent/CN109946389B/en active Active
-
2020
- 2020-01-21 AU AU2020214409A patent/AU2020214409B2/en active Active
- 2020-01-21 WO PCT/CN2020/073357 patent/WO2020156348A1/en active Application Filing
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150164413A1 (en) * | 2013-12-13 | 2015-06-18 | National Chung Shan Institute Of Science And Technology | Method of creating anesthetic consciousness index with artificial neural network |
| CN104502126A (en) * | 2014-12-28 | 2015-04-08 | 华东交通大学 | Modal intervals-based high-speed train bogie fault diagnosis method |
| CN106874957A (en) * | 2017-02-27 | 2017-06-20 | 苏州大学 | A kind of Fault Diagnosis of Roller Bearings |
| CN107123033A (en) * | 2017-05-04 | 2017-09-01 | 北京科技大学 | A kind of garment coordination method based on depth convolutional neural networks |
| CN107329933A (en) * | 2017-07-14 | 2017-11-07 | 北京知觉科技有限公司 | Fault detection method and device based on Fibre Optical Sensor vibration signal |
| CN107505396A (en) * | 2017-09-12 | 2017-12-22 | 燕山大学 | A kind of structural damage on-line real time monitoring method and system |
| CN107826919A (en) * | 2017-10-20 | 2018-03-23 | 中国矿业大学 | A kind of lifting system critical component multimode health monitoring device and monitoring method |
| CN108717869A (en) * | 2018-05-03 | 2018-10-30 | 中国石油大学(华东) | Diabetic retinopathy diagnosis aid system based on convolutional neural networks |
Non-Patent Citations (2)
| Title |
|---|
| 李雪松等: "基于卷积神经网络的桥梁损伤识别方法应用研究", 《青海大学学报》 * |
| 程磊等: "基于EMD方法与RBF神经网络的结构损伤检测", 《地震工程与工程振动》 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020156348A1 (en) * | 2019-01-31 | 2020-08-06 | 青岛理工大学 | Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network |
| CN110472272A (en) * | 2019-07-01 | 2019-11-19 | 广东工业大学 | A kind of Structural Damage Identification based on multi-parameter and convolutional neural networks |
| CN110472272B (en) * | 2019-07-01 | 2023-04-07 | 广东工业大学 | Structural damage identification method based on multi-parameter and convolutional neural network |
| CN110490249A (en) * | 2019-08-16 | 2019-11-22 | 哈尔滨工业大学 | Structural Damage Identification based on attribute class relations and few sample meta learning |
| CN111241725A (en) * | 2019-12-30 | 2020-06-05 | 浙江大学 | Structure response reconstruction method for generating countermeasure network based on conditions |
| CN111220921A (en) * | 2020-01-08 | 2020-06-02 | 重庆邮电大学 | Lithium battery capacity estimation method based on improved convolution-long-and-short-term memory neural network |
| CN111595541A (en) * | 2020-04-27 | 2020-08-28 | 河海大学 | Multi-dimensional structure damage identification method based on mass vibration transmissibility data convolutional neural network processing |
| CN111753776B (en) * | 2020-06-29 | 2022-05-10 | 重庆交通大学 | Structural damage identification method based on echo state and multi-scale convolution combined model |
| CN111753776A (en) * | 2020-06-29 | 2020-10-09 | 重庆交通大学 | Structural damage identification method based on echo state and multi-scale convolution combined model |
| CN111931983B (en) * | 2020-07-07 | 2023-09-19 | 湖南大学 | Precipitation prediction method and system |
| CN111931983A (en) * | 2020-07-07 | 2020-11-13 | 湖南大学 | Precipitation prediction method and system |
| CN112182490A (en) * | 2020-09-01 | 2021-01-05 | 华中科技大学 | Reactor state diagnosis method and system |
| CN112182490B (en) * | 2020-09-01 | 2024-02-02 | 华中科技大学 | Reactor state diagnosis method and system |
| CN112525467A (en) * | 2020-11-26 | 2021-03-19 | 山东科技大学 | Impact damage area identification method and device suitable for cantilever beam |
| CN112434624A (en) * | 2020-11-30 | 2021-03-02 | 广州大学 | Structural health monitoring data distortion detection method, system, device and storage medium |
| CN112712060A (en) * | 2021-01-15 | 2021-04-27 | 国网浙江省电力有限公司电力科学研究院 | Automatic high-voltage post porcelain insulator acoustic signal identification method |
| CN113159282B (en) * | 2021-03-29 | 2022-04-22 | 东南大学 | Intelligent updating method for bridge modal parameters based on cross modal confidence criterion matrix |
| CN113159282A (en) * | 2021-03-29 | 2021-07-23 | 东南大学 | Intelligent updating method for bridge modal parameters based on cross modal confidence criterion matrix |
| CN114002334A (en) * | 2021-09-29 | 2022-02-01 | 西安交通大学 | Structural damage acoustic emission signal identification method and device and storage medium |
| CN114062511A (en) * | 2021-10-24 | 2022-02-18 | 北京化工大学 | Single-sensor-based intelligent acoustic emission identification method for early damage of aircraft engine |
| CN114358214A (en) * | 2022-03-11 | 2022-04-15 | 中国科学院上海高等研究院 | Gait adaptive identification method and device, storage medium and terminal |
| CN114358214B (en) * | 2022-03-11 | 2022-07-22 | 中国科学院上海高等研究院 | Gait adaptive recognition method and device, storage medium and terminal |
| CN114462457A (en) * | 2022-04-11 | 2022-05-10 | 自然资源部第一海洋研究所 | Ship underwater noise deep learning identification method based on intrinsic probability density function |
| CN115169409A (en) * | 2022-07-18 | 2022-10-11 | 四川省公路规划勘察设计研究院有限公司 | Bridge structure natural vibration frequency identification and early warning method and device based on sliding window |
| CN116244803A (en) * | 2023-02-22 | 2023-06-09 | 合肥工业大学 | Bridge internal response reconstruction method based on physical and data dual drive |
| CN116244803B (en) * | 2023-02-22 | 2023-09-12 | 合肥工业大学 | Bridge internal response reconstruction method based on physical and data dual drive |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020156348A1 (en) | 2020-08-06 |
| AU2020214409A1 (en) | 2021-03-04 |
| CN109946389B (en) | 2020-12-25 |
| AU2020214409B2 (en) | 2022-07-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109946389A (en) | Structural damage identification method based on ensemble empirical mode decomposition and convolutional neural network | |
| CN112418013A (en) | Complex working condition bearing fault diagnosis method based on meta-learning under small sample | |
| CN108256556A (en) | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network | |
| CN109800795A (en) | A kind of fruit and vegetable recognition method and system | |
| CN106017876A (en) | Wheel set bearing fault diagnosis method based on equally-weighted local feature sparse filter network | |
| CN104167207A (en) | Equipment sound identification method based on transformer substation patrol inspection robot | |
| CN110070102B (en) | Method for establishing sequence-to-sequence model for identifying power quality disturbance type | |
| CN103115789A (en) | Second generation small-wave support vector machine assessment method for damage and remaining life of metal structure | |
| CN111680788A (en) | Equipment fault diagnosis method based on deep learning | |
| CN112163472A (en) | Rolling bearing diagnosis method based on multi-view feature fusion | |
| Zhu et al. | Complex disturbances identification: A novel PQDs decomposition and modeling method | |
| CN117150359A (en) | Small sample fault diagnosis method, system, device and medium based on model independent element learning | |
| Zhao et al. | Bearing fault diagnosis based on mel frequency cepstrum coefficient and deformable space-frequency attention network | |
| CN116753471A (en) | Water supply pipeline leakage multi-domain feature extraction and fusion identification method | |
| CN114722886A (en) | Knowledge distillation-based crankshaft internal defect detection method and detection equipment | |
| CN117056678A (en) | Machine pump equipment operation fault diagnosis method and device based on small sample | |
| CN116399592A (en) | Bearing fault diagnosis method based on channel attention dual-path feature extraction | |
| GB2625165A (en) | Method and system for fault diagnosis of rolling bearing | |
| CN116484176A (en) | Bearing fault diagnosis method, system and storage medium based on ultra-wavelet | |
| CN116910648A (en) | GIS equipment partial discharge spectrum analysis method, system and medium | |
| CN116340812A (en) | Transformer partial discharge fault mode identification method and system | |
| Orozco-García et al. | A study on the recognition of patterns of infant cry for the identification of deafness in just born babies with neural networks | |
| Wang et al. | Research on fault diagnosis method of wind turbine bearing based on deep belief network | |
| CN114624027A (en) | Bearing fault diagnosis method based on multi-input CNN | |
| Jiang et al. | VMD and HMM based rolling bearing fault diagnosis |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |