CN110208375A - A kind of detection method and terminal device of anchor rod anchored defect - Google Patents

A kind of detection method and terminal device of anchor rod anchored defect Download PDF

Info

Publication number
CN110208375A
CN110208375A CN201910511203.XA CN201910511203A CN110208375A CN 110208375 A CN110208375 A CN 110208375A CN 201910511203 A CN201910511203 A CN 201910511203A CN 110208375 A CN110208375 A CN 110208375A
Authority
CN
China
Prior art keywords
layer
signal
node
echo
anchor rod
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
Application number
CN201910511203.XA
Other languages
Chinese (zh)
Other versions
CN110208375B (en
Inventor
孙晓云
林童
王明明
宫世杰
闫志勋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shijiazhuang Tiedao University
Original Assignee
Shijiazhuang Tiedao University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shijiazhuang Tiedao University filed Critical Shijiazhuang Tiedao University
Priority to CN201910511203.XA priority Critical patent/CN110208375B/en
Publication of CN110208375A publication Critical patent/CN110208375A/en
Application granted granted Critical
Publication of CN110208375B publication Critical patent/CN110208375B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • G01N29/04—Analysing solids
    • 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
    • G01N29/22—Details, e.g. general constructional or apparatus details
    • G01N29/24—Probes
    • G01N29/2412—Probes using the magnetostrictive properties of the material to be examined, e.g. electromagnetic acoustic transducers [EMAT]
    • 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
    • G01N29/4481—Neural networks
    • 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
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2291/00—Indexing codes associated with group G01N29/00
    • G01N2291/02—Indexing codes associated with the analysed material
    • G01N2291/023—Solids
    • G01N2291/0234—Metals, e.g. steel
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2291/00—Indexing codes associated with group G01N29/00
    • G01N2291/02—Indexing codes associated with the analysed material
    • G01N2291/028—Material parameters
    • G01N2291/0289—Internal structure, e.g. defects, grain size, texture
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2291/00—Indexing codes associated with group G01N29/00
    • G01N2291/04—Wave modes and trajectories
    • G01N2291/044—Internal reflections (echoes), e.g. on walls or defects
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2291/00—Indexing codes associated with group G01N29/00
    • G01N2291/26—Scanned objects
    • G01N2291/269—Various geometry objects

Landscapes

  • Physics & Mathematics (AREA)
  • Analytical Chemistry (AREA)
  • Immunology (AREA)
  • General Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Biochemistry (AREA)
  • Pathology (AREA)
  • Signal Processing (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Acoustics & Sound (AREA)
  • Electromagnetism (AREA)
  • Mathematical Physics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)

Abstract

The present invention provides the detection method and terminal device of a kind of anchor rod anchored defect, method includes: acquisition echo-signal；The echo-signal is decomposed using variation mode decomposition method, the corresponding intrinsic modal components of each layer after being decomposed；Echo signal by the corresponding intrinsic modal components noise reduction process of each layer, after obtaining noise reduction；The echo signal is inputted into target nerve network, obtains anchor rod anchored defects detection result.Echo-signal of the present invention to acquisition, using based on the intrinsic modal components for improving each layer after variation mode decomposition method decomposes echo-signal, noise reduction is carried out to the intrinsic modal components of each layer, to input value of the intrinsic modal components as structural self-organizing Elman neural network after noise reduction, Detection of Bolt Bonding Integrity detection is carried out using structural self-organizing Elman neural network, anchor rod anchored defect, and anchor rod anchored defects detection precision with higher can quickly be detected.

Description

A kind of detection method and terminal device of anchor rod anchored defect
Technical field
The invention belongs to the detection methods and terminal of technical field of quality detection more particularly to a kind of anchor rod anchored defect to set It is standby.
Background technique
Anchor pole is mostly steel nail in engineering, and anchoring is mostly by coagulation local soil type at and being fixed on outside anchor pole.Due to Anchor rod anchored present application environment is more severe, will lead to Detection of Bolt Bonding Integrity and is damaged, anchor rod anchored impaired or corrosion Serious accident is likely to result in when serious.How quick and precisely to carry out detection to anchor rod anchored internal soundness seems especially heavy It wants.
At present since anchor rod anchored use environment has concealment, standard can not be carried out with the presence or absence of defect to anchor rod anchored Really detection is big to anchor rod anchored detection difficulty.
Summary of the invention
In view of this, the embodiment of the invention provides the detection method and terminal device of a kind of anchor rod anchored defect, with solution The problem of certainly anchor rod anchored defect can not accurately being detected at present.
The first aspect of the embodiment of the present invention provides a kind of detection method of anchor rod anchored defect, comprising:
Echo-signal is obtained, the echo-signal is acquisition for detecting the initial information of the anchor rod anchored defect；
The echo-signal is decomposed using variation mode decomposition method, the corresponding intrinsic mode of target of each layer after being decomposed Component；
Echo signal by the intrinsic modal components noise reduction process of the corresponding target of each layer, after obtaining noise reduction；
The echo signal is inputted into target nerve network, obtains anchor rod anchored defects detection result.
The second aspect of the embodiment of the present invention provides a kind of terminal device, including memory, processor and is stored in In the memory and the computer program that can run on the processor, when the processor executes the computer program The step of realizing the detection method of anchor rod anchored defect as described above.
The third aspect of the embodiment of the present invention provides a kind of computer readable storage medium, the computer-readable storage Media storage has computer program, and the inspection of anchor rod anchored defect as described above is realized when the computer program is executed by processor The step of survey method.
The present invention decomposes the echo-signal of acquisition using based on improvement variation mode decomposition method to echo-signal The intrinsic modal components of each layer after being decomposed carry out noise reduction to the intrinsic modal components of each layer, to the eigen mode after noise reduction Input value of the state component as target nerve network carries out Detection of Bolt Bonding Integrity detection using target nerve network, can basis Different detection cases complete structural adaptation, can quickly detect anchor rod anchored defect, and with higher anchor rod anchored Defects detection precision.
Detailed description of the invention
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to embodiment or description of the prior art Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description is only of the invention some Embodiment for those of ordinary skill in the art without any creative labor, can also be according to these Attached drawing obtains other attached drawings.
Fig. 1 is the flow diagram of the detection method for the anchor rod anchored defect that one embodiment of the present of invention provides；
Fig. 2 is the schematic diagram for the echo-signal that one embodiment of the present of invention provides；
Fig. 3 is the noise reduction effect schematic diagram for the Threshold Denoising method that one embodiment of the present of invention provides；
Fig. 4 is the echo letter after variation mode decomposition method and noise reduction process that one embodiment of the present of invention provides Number exemplary diagram；
Fig. 5 is the exemplary diagram of the training process of the neural network for the building that one embodiment of the present of invention provides；
Fig. 6 is the topology example figure of the neural network for the building that one embodiment of the present of invention provides；
Fig. 7 is the topology example figure for the node split that one embodiment of the present of invention provides；
Fig. 8 is the echo wave signal acquisition device that one embodiment of the present of invention provides and anchor rod anchored topology example figure；
Fig. 9 is the topology example figure for the anchor rod anchored front end cavity blemish that one embodiment of the present of invention provides；
Figure 10 is the topology example figure for the anchor rod anchored rear end cavity blemish that one embodiment of the present of invention provides；
Figure 11 is the topology example figure for anchor rod anchored double cavity blemish that one embodiment of the present of invention provides；
Figure 12 is the structural schematic diagram of the detection device for the anchor rod anchored defect that one embodiment of the present of invention provides；
Figure 13 is the schematic diagram for the terminal device that one embodiment of the present of invention provides.
Wherein: 1, anchor pole；2, it anchors；3, permanent magnet；4, yoke；5, coil；6, defective locations.
Specific embodiment
In being described below, for illustration and not for limitation, the tool of such as particular system structure, technology etc is proposed Body details, to understand thoroughly the embodiment of the present invention.However, it will be clear to one skilled in the art that there is no these specific The present invention also may be implemented in the other embodiments of details.In other situations, it omits to well-known system, device, electricity The detailed description of road and method, in case unnecessary details interferes description of the invention.
Description and claims of this specification and term " includes " and other any deformations in above-mentioned attached drawing are Refer to " including but not limited to ", it is intended that cover and non-exclusive include.Such as the process, method comprising a series of steps or units Or system, product or equipment are not limited to listed step or unit, but optionally further comprising the step of not listing Or unit, or optionally further comprising other step or units intrinsic for these process, methods, product or equipment.In addition, art Language " first ", " second " and " third " etc. is for distinguishing different objects, not for description particular order.
In order to illustrate technical solutions according to the invention, the following is a description of specific embodiments.
Embodiment 1:
Fig. 1 shows the implementation flow chart of the detection method of anchor rod anchored defect provided by one embodiment of the invention, is Convenient for explanation, only parts related to embodiments of the present invention are shown, details are as follows:
As shown in Figure 1, a kind of detection method of anchor rod anchored defect provided by the embodiment of the present invention, comprising:
S101, obtains echo-signal, and the echo-signal is acquisition for detecting the initial of the anchor rod anchored defect Information.
S102 decomposes the echo-signal using variation mode decomposition method, the corresponding target sheet of each layer after being decomposed Levy modal components.
S103, the echo signal by the intrinsic modal components noise reduction process of the corresponding target of each layer, after obtaining noise reduction.
The echo signal is inputted target nerve network, obtains anchor rod anchored defects detection result by S104.
In an embodiment of the present invention, S102 includes:
S201 is initially layered the echo-signal, and the initial intrinsic mode of this layer is cyclically updated based on every layer Component obtains the intrinsic modal components of target of current layer after intrinsic modal components in the updated meet first condition of convergence.
S202, the intrinsic modal components of target based on the current layer, the weighted kurtosis of each layer after calculating this secondary clearing Value.
S203 stops being layered if the weighted kurtosis value of each layer after this secondary clearing meets the first preset condition, obtains most Whole hierarchy number is the number of plies after last layering.
S204 continues to be layered if the weighted kurtosis value of each layer after this secondary clearing is unsatisfactory for the first preset condition, until The weighted kurtosis value of each layer after layering meets the first preset condition.
Wherein, first preset condition are as follows:
Wherein,Sgn () is symbol letter Number guarantees that output signal and the phase of original signal are consistent as far as possible；KwFor weighted kurtosis value；R is the index of C；D is kurtosis value；μ For the mean value of the echo-signal f (t)；σ is the standard deviation of the echo-signal f (t)；T is the time；T is echo-signal f (t) Length；uk(t) be current hierarchical when kth layer the intrinsic modal components of target；For uk(t) mean value；For the equal of f (t) Value；ζ is preset threshold；C is the cross-correlation coefficient between two signals.
Wherein, the intrinsic modal components of the targetForIt is obtained by inverse Fourier transform:
Wherein:N is cycle-index；k For kth layer；It indicatesFourier transformation, be kth layer by n+1 circulation gained；For the echo Value of the signal f (t) after Fourier transformation；α is constant；It indicatesFourier transformation, be d layer by n Obtained by secondary circulation, wherein d ≠ k；ω is frequency；For kth layer and be (n+1)th time circulation when mode centre frequency； Indicate Lagrangian λn(t) Fourier transformation, for n-th circulation gained；τ is time constant；K is current hierarchical number；
As an example, when K=5, the intrinsic modal components d difference value for calculating first layer is 2,3,4,5.Calculate the second layer Intrinsic modal components d difference value be 1,3,4,5.
First condition of convergence are as follows:
Wherein: ε=10-7。
In the present embodiment, if hierarchy number is K=3,3 initial intrinsic modal components is updated respectively, are made Initial intrinsic modal components meet first condition of convergence, then stop updating, and export initial intrinsic when meeting first condition of convergence Echo signal is divided into 2 layers, and save K=if every layer of weighted kurtosis value meets the first preset condition at this time by modal components Initial intrinsic modal components when 2 first conditions of convergence of satisfaction exported when 2 are the intrinsic modal components of target.
As an example, initial hierarchy number K=2, primary iteration frequency n=0 are set；
(1) the intrinsic modal components of the target of current layer are determined:
1. defining intrinsic mode component function:
Variation mode decomposition (VMD) is carried out to echo-signal f (t) to decompose, and defines intrinsic mode function (Intrinsic Mode Function, IMF) its expression formula are as follows:
Wherein, uk(t) AM/FM amplitude modulation/frequency modulation signal i.e. the intrinsic modal components of target,The phase of signal, Ak(t) it is Instantaneous amplitude；K is the kth layer of current hierarchical number.
2. the restricted model of intrinsic modal components:
To each layer intrinsic mode function (Intrinsic Mode Function, IMF) progress Hilbert transform, and with IndexMixing, the bandwidth of intrinsic mode function is estimated by the squared norm of gradient, obtains its restricted model:
Wherein,It is then to indicate to seek t partial derivative, ukIt (t) is the intrinsic modal components of target, ωk(t) in indicating all Frequency of heart, f (t) indicate echo-signal；K is current hierarchical number；ωkFor center frequency；δ (t) is impulse function；H is imaginary number list Position.
3. determining the intrinsic modal components of target of current layer:
Introduce Lagrangian λ (t) and secondary penalty factor α, wherein penalty factor α=2000 pass through alternating direction Multiplier method solves the Solve problems of Lagrangian " saddle point ", i.e., alternately updates λn+1、Seek constraint variation problem most Excellent solution.ω-ω is replaced by with ω using after Parseval/Plancherel Fourier's equilong transformationk, and obtained result is turned It is changed to the form of negative frequency interval integral, modal components more new formula and centre frequency more new formula is obtained, it is intrinsic to update each layer Modal components obtain the intrinsic modal components of target of current layer.
Wherein:N is cycle-index；k For kth layer；It indicatesFourier transformation, be kth layer by n+1 circulation gained；For the echo Value of the signal f (t) after Fourier transformation；α is constant；It indicatesFourier transformation, be d layer by n times Gained is recycled, wherein d ≠ k；ω is frequency；For kth layer and be (n+1)th time circulation when mode centre frequency；Table Show Lagrangian λn(t) Fourier transformation, for n-th circulation gained；τ is time constant；K is current hierarchical number.
It calculatesLater, the intrinsic modal components u of target is obtained by inverse Fourier transformk(t)。
Until meetingDiscrimination precision ε=10-7, circulation terminates, and exports K A component, otherwise, n=n+1 continues iteration.
(2) Decomposition order is determined:
1. calculating each layer kurtosis value when current hierarchical.
Wherein, D is kurtosis value；μ is the mean value of the echo-signal f (t)；σ is the standard deviation of the echo-signal f (t).
2. calculating the cross-correlation coefficient of each layer intrinsic modal components and original signal:
Wherein, T is the length of echo-signal f (t)；uk(t) be current hierarchical when kth layer the intrinsic modal components of target；For uk(t) mean value；For the mean value of f (t)；ζ is preset threshold.
According to Cauchy-Schwarz inequality | C |≤1:
Work as uk(t)=f (t) when, C=1；Work as uk(t)=- f (t) when, C=-1, phase phase difference 180 degree.It is known as C > 0 Two signals are positively correlated, and as C < 0, referred to as two signals are negatively correlated.
3. determining hierarchy number:
The weighted kurtosis value of each layer when calculating current hierarchical, given threshold 1, if the weighted kurtosis of each layer after decomposing Minimum value is less than threshold value, then stops decomposing, then K value is K-1；On the contrary then K=K+1 continues to decompose, and stops item until meeting Part.
Kw=sgn (C) D | C |r< ζ；
Wherein, KwFor weighted kurtosis value；R is the index of C, takes positive real number, adjusts the index of input signal and output signal, R usually takes 1；D is kurtosis value；ζ is preset threshold=1.
As verifying, as shown in Fig. 2, extracting one group of data, each layer intrinsic modal components (IMF) weighted kurtosis value is calculated such as Shown in table 1, finally occur component of the weighted kurtosis value less than 1 in K=6, therefore K takes 5.
1 each layer of IMF component weighted kurtosis value of table
In an embodiment of the present invention, S103 includes:
Noise reduction process is carried out using the intrinsic modal components of target of the wavelet thresholding method to each layer, the target letter after obtaining noise reduction Number.
As an example: 1. to the intrinsic modal components of target (IMF) of each layer after decomposition, it is sym4 and true for choosing wavelet basis Determining Decomposition order is 5, carries out wavelet transformation to the intrinsic modal components of each layer target and obtains low-frequency wavelet coefficients and high frequency wavelet system Number.
2. selecting minimax threshold value, using hard threshold function, threshold process is carried out to the high-frequency wavelet coefficient of decomposition.
3. finally high-frequency wavelet coefficient carries out wavelet reconstruction with treated by low-frequency wavelet coefficients, each layer eigen mode is obtained Signal after state component noise reduction.To treated, each intrinsic modal components of layer target (IMF) are recombinated, and complete noise reduction.
As verifying, noise reduction comparing result is as shown in table 2 below:
2 noise reduction Comparative result of table
As known from Table 2, the method for the present invention has preferable noise reduction effect.To Noise reducing of data, comparing result such as Fig. 2, Fig. 3 and Shown in Fig. 4, it can be seen that from the waveform of Fig. 2 there are much noise signal in original signal, reflected signal characteristic is difficult to sentence It is disconnected.As can be seen from Figure 3 wavelet de-noising waveform is smooth, but loses bottom end echo.
As shown in figure 5, in an embodiment of the present invention, before S104 further include:
S1101 obtains the neural network of building, as shown in Figure 6.
S1102 obtains sample set, and carries out variation mode decomposition and noise reduction process to the sample in the sample set, obtains Obtain sample signal collection.
One group of sample signal that sample signal is concentrated is inputted the neural network of the building, obtains this training by S1103 The output of hidden layer in the process.
S1104 deletes the node for meeting the second preset condition, it is pre- that division meets third based on the output of the hidden layer The neural network if node of condition, after obtaining this training.
Whether S1105, the neural network after determining this training meet second condition of convergence.
If meeting second condition of convergence, the neural network after this training is the target nerve network.
If being unsatisfactory for second condition of convergence, after being trained by Elman neural network weight more new formula to this Neural network weight and bias be updated；And it is concentrated from the sample signal and chooses one group of untrained sample signal This updated neural network is inputted, is trained next time, the neural network after training meets second condition of convergence.
In the present embodiment, S1101 includes:
The intrinsic modal components of target to the echo signal after noise reduction, that is, after noise reduction carry out three layers of WAVELET PACKET DECOMPOSITION, The result of decomposition is indicated into (i.e. Wavelet Packet Energy Spectrum) according to energy mode, the energy of each frequency band is extracted and is normalized, structure The anchor pole defect characteristic value for being 8 at dimension, the input of the defect recognition as network.
As an example: the Detection of Bolt Bonding Integrity signal in 560 groups randomly selected after J group noise reduction is FJ(t), to FJ(t) M-th of signal of kth layer is used after carrying out three layers of WAVELET PACKET DECOMPOSITIONIt indicates,Length be λ, WAVELET PACKET DECOMPOSITION energy It is as follows to measure formula:
Wherein,For the decomposition energy of m-th of signal in J group kth layer；K indicates the quantity of decomposition layer, k= 3；M indicates the position of subband, that is, m-th of signal of third layer；J indicates the Detection of Bolt Bonding Integrity signal after J group noise reduction.From Conservation of energy principle can see that
Wherein, E [FJIt (t)] is the gross energy of the Detection of Bolt Bonding Integrity signal after J group noise reduction.
Standardization is to construct defect characteristic vector matrix:
E (m) is defect characteristic vector matrix.
Then input of the anchor pole defect characteristic value matrix that our available dimensions are 8 as network.
Initialize installation is carried out to Elman neural network parameter, the structure and parameter that network is arranged is as follows, input layer Number is set as 8, and hidden layer start node number is set as 18, and output layer node number is set as 4, and network cut threshold value is c =1.70, increase threshold value d=1.60, e-learning rate is 0.08, and hidden layer biases β1=0.1, output layer biases β2=0.08, a To accept layer self feed back factor a=0.1, each layer weight of network is the random number less than 1.
In an embodiment of the present invention, the sample set in S1102 are as follows:
640 groups of data are acquired to four class defective anchor bars respectively, after the completion of above-mentioned noise reduction operation, data are carried out three layers small Wave packet decomposes, and extracts 8 dimension flaw indication characteristic values, neural network of the 560 groups of characteristic value data collection randomly selected as building Training sample, the neural network of building is trained.
In an embodiment of the present invention, S1103 includes:
The parameters such as each layer output valve of network are obtained, including
X (p)=f (W1xc(p)+W2(u(p-1))+β1)；
xc(p)=x (p-1)+axc(p-1)；
Y (p)=g (W3x(p)+β2)；
Wherein,
Wherein, f (x) is hidden layer activation primitive, and g (x) is output layer activation primitive, W1For hidden layer and accept between layer Connection weight matrix, W2The connection weight matrix between hidden layer and input layer, W3The connection weight between output layer and hidden layer Matrix, x (p) are the output of hidden layer pth group, and y (p) is the output of the pth group of network, xc(p) the pth group output to accept layer；u (p-1) it is inputted for -1 group of pth (previous group) of network, i.e. -1 group characteristic of pth of anchor pole defect characteristic signal data collection；x (p-1) it is exported for -1 group of hidden layer pth (previous group), xc(p-1) -1 group output of pth to accept layer；β1It is inclined for hidden layer output It sets；β2It exports and biases for output layer；A is the self feed back factor for accepting layer, value a=0.1；xh inFor the input of hidden layer；xo in For the input of output layer.
In an embodiment of the present invention, S1104 includes:
S11041 obtains the contribution degree of each node of every layer of hidden layer, Current neural based on the output of every layer of hidden layer The trimming threshold value of network and the growth threshold value of Current Situation of Neural Network.
The contribution degree of each node of every layer of hidden layer is compared, deletion of node by S11042 with the trimming threshold value Contribution degree be less than trimming threshold value hidden layer node undertaking node layer corresponding with the hidden layer node.
The contribution degree of each node of every layer of hidden layer is compared, split vertexes by S11043 with the growth threshold value Contribution degree be greater than increase threshold value hidden layer node undertaking the node layer corresponding with hidden layer node, obtain this train after Neural network.
In an embodiment of the present invention, the contribution degree of each node of the hidden layer are as follows:
Wherein, SConjFor the contribution degree of the node j of hidden layer；For hidden layer node j and output layer node i it Between connection weight；zj(p) the pth group output valve of the node j of hidden layer；S is characterized the s group data of data set；N is characterized number According to total group of number of collection；M is the node number of output layer；
The trimming threshold value of the Current Situation of Neural Network are as follows:
Wherein,SACon is the average contribution degree of hidden layer；Pth is repairing for Current Situation of Neural Network Cut threshold value；N is the node number of hidden layer；C is trimming constant, for the constant greater than 1；
The growth threshold value of the Current Situation of Neural Network are as follows:
Gth=dSACon；
Wherein, Gth is the growth threshold value of Current Situation of Neural Network；D is growth constant, for the constant greater than 1.
As shown in fig. 6, in an embodiment of the present invention, S11042 includes:
According to
Wherein, j is to be deleted hidden layer node, jcNode layer is accepted to be deleted,For hidden layer j node and input Connection weight between layer q node,For output layer i-node and hidden layer j node connection weight,For hidden layer j node with Accept connection weight between layer l node, wz,jc 1For hidden layer z node and accept layer jcConnection weight between node.
The contribution degree of deletion of node is less than the hidden layer node undertaking layer section corresponding with the hidden layer node of trimming threshold value Point.
As shown in fig. 7, in an embodiment of the present invention, S11043 includes: basis
B=1-a a is random number between 0 to 1；
Wherein, hidden layer j ' and k node are the node after hidden layer node j division, wjq 2It is saved for hidden layer j before division Connection weight between point and input layer q node, wj′q 2Connect between hidden layer j ' node and input layer q node after increasing for network Meet weight, wkq 2For the connection weight between new growth hidden layer k node and input layer q, a, b are node split coefficient.
L=1,2 ... n and l ≠ jc′,kc；N is to accept node layer number；
L=1,2 ... n and l ≠ jc′,kc；N is to accept node layer number；
Z=1,2 ... n；N is hidden layer node number；
Z=1,2 ... n；N is hidden layer node number；
Wherein, layer j is acceptedc′And kcNode is to accept node layer jcNode after division, wjl 1It is hidden before increasing for network structure Connection weight between the node of j containing layer and undertaking layer l node, wj′l 1Hidden layer j ' node and undertaking layer after increasing for network node Connection weight between l node, wkl 1For hidden layer node k after growth and accept the connection weight between layer l node；wz,jc 1For Hidden layer z node and undertaking layer j before network increasescConnection weight between node, wz,jc′ 1For hidden layer node z after growth With undertaking node layer jc′Between connection weight, wz,kc 1For hidden layer node z after growth and accept node layer kcBetween connection weight.
After network node division increases, division changes hidden layer node j ' and the output of k node is,
M is output layer node number；
M is output layer node number；
Wherein, xj′It (p) is the pth group output of the new split vertexes j ' of hidden layer after division, xk(p) new for hidden layer after division The pth group of split vertexes k exports, yiFor the output of output layer node i, yi expFor the desired output of output layer node i, eiNetwork The error of training output layer node i, xkFor the output of hidden layer node k, xj' it is hidden layer node j ' output, wik 3After division Connection weight between output layer node i and the new split vertexes k of hidden layer,It is new for output layer node i after division and hidden layer The connection weight of split vertexes j '；For the connection weight for dividing preceding output layer node i and hidden layer node j；wj′l 1For division The connection weight of the new split vertexes j ' of hidden layer and undertaking layer l afterwards；For the new split vertexes k of hidden layer after division and accept layer The connection weight of l；For connection weight between hidden layer node j ' and input layer q after division；After division Connection weight between hidden layer node k and input layer q；xc,l(p) the pth group output to accept node layer l；uq(p-1) it is - 1 group input of pth of network.
The contribution degree of split vertexes is greater than the hidden layer node undertaking layer section corresponding with the hidden layer node for increasing threshold value Point, the neural network after obtaining this training.
In an embodiment of the present invention, S1105 includes:
Calculate network error:
Wherein, MSE is network error；M is the total number of element in network output matrix in an iteration, yiIt is defeated for network Node layer i reality output out,For network output layer node i desired output；M is network output layer node number.
Judge whether network error is less than circulation and jumps out error, the knot when satisfaction jumps out condition namely second condition of convergence Beam network circulation, saves network structure and other parameters.
MSE < ψ, wherein ψ is that circulation jumps out error.
When error is unsatisfactory for loop stop conditions, then it is updated according to weight, bias of the following formula to network.
Wherein
G'(x)=1；
f'j(x)=fj(x)(1-fj(x))；
Wherein, Δ wijThe 3 connection weight renewal amount between output layer i-node and hidden layer j node,For network Connection weight renewal amount between hidden layer j node and input layer q node,For hidden layer j node and accept layer l node it Between connection weight renewal amount；xc,lIt (p) is the pth group output for accepting layer l node, E is network back transfer error function；η is mind Learning rate through network, fj' (x) is hidden layer activation primitive fj(x) to the partial derivative of hidden layer j node output, g ' (x) is defeated The derivative of layer activation primitive out；Δ w is the right value update moment matrix of network；Represent network back transfer error function To the local derviation of weight matrix；For the right value update factor of output layer node i；xj(p) it is exported for the pth group of hidden layer node j； M is output layer number of nodes；N is node in hidden layer；For the right value update factor of hidden layer node j；uq(p-1) output layer The input of P-1 group；R is input layer number；X is exported for the pth group to hidden layer node jj(p) ask about Connection weight between hidden layer node j and undertaking node layer lLocal derviation；For the pth of network output layer node i Group desired output；yiIt (p) is the pth group reality output of network output layer node i；It g'(is) to output layer activation primitive to defeated Layer output out is differentiated；fj' () be hidden layer activation primitive differentiate to the output of hidden layer node j；Y is that network once changes For the output matrix of process network；yexpFor the desired output matrix of network an iteration process network；T is to Matrix Calculating transposition.
As verifying, target nerve network of the invention (namely structural self-organizing Elman neural network) is to anchor pole anchor Gu defects detection result and trimming-growth BP neural network defects detection result and fixed structure Elman neural network defect are examined The results are shown in Table 3 for survey.
3 structural self-organizing of table and fixed structure Elman neural network testing result
As shown in Table 3, structural self-organizing Elman neural network can complete structural self-organizing adjustment in the training process, There is higher defects detection precision relative to fixed structure Elman neural network and trimming-growth BP neural network.Using knot Structure self-organizing Elman neural network, network structure can adaptively be adjusted according to different detection cases, and method is to anchor rod anchored matter Amount detection adaptive faculty is strong, detection recognition correct rate is high, the non-destructive testing suitable for the anchoring of complex engineering environment lower bolt.
As shown in figure 8, in an embodiment of the present invention, the echo-signal is the acquisition of preset echo wave signal acquisition device For detecting the initial information of the anchor rod anchored defect；
Wherein, the echo wave signal acquisition device includes:
Signal generation apparatus, magnetic field generation device and coil 5；The coil 5 is wrapped on anchor pole 1, and anchoring 2 is arranged in anchor Outside bar 1, the magnetic field generation device is set on the anchor pole, and the coil is set in the magnetic field generation device, institute Signal generation apparatus is stated to be electrically connected with the coil；
The pumping signal that the signal generation apparatus issues makes in the anchor pole under the action of magnetic field generation device Supersonic guide-wave is inspired, is echo-signal by the anchor rod anchored reflected supersonic guide-wave.
In the present embodiment, echo wave signal acquisition device further includes power amplifier and duplexer, power amplifier difference It is connected with signal generation apparatus and duplexer, duplexer is connected with coil 5.
In the present embodiment, magnetic field generation device includes permanent magnet 3 and yoke 4, and permanent magnet 3 is in for central axes with anchor pole 1 It is uniformly and symmetrically distributed.
Pumping signal is transmitted to signal amplifier by signal generation apparatus, and signal amplifier will be believed by amplified excitation It number is transmitted to duplexer, duplexer will be transmitted to coil by amplified pumping signal, in the magnetic that magnetic field generation device generates Under the action of, magnetostrictive effect occurs being detected inside anchor pole, and then supersonic guide-wave is inspired in anchor pole；In anchor pole The signal returned in anchoring makes coil generate induced voltage, as echo-signal under the influence of a magnetic field.
In the present embodiment, signal generation apparatus uses Tektronix AFG3052, and signal amplifier uses AE TECHRON 7224, duplexer use RITEC RDX-EM2DIPLEXER PRE-AMPLIFIER.
In the present embodiment, echo wave signal acquisition device is the self excitation and self receiving formula device based on magnetostriction mechanism.
As shown in figs. 9-11, in the present embodiment, anchor rod anchored defect specifically includes that front end cavity blemish, rear end cavity Defect and double cavity blemish, front end cavity blemish are defective locations 6 in front end.
It should be understood that the size of the serial number of each step is not meant that the order of the execution order in above-described embodiment, each process Execution sequence should be determined by its function and internal logic, the implementation process without coping with the embodiment of the present invention constitutes any limit It is fixed.
Embodiment 2:
As shown in figure 12, the detection device 100 for the anchor rod anchored defect that one embodiment of the present of invention provides, for executing The detection device 100 of method and step in embodiment corresponding to Fig. 1, anchor rod anchored defect is connected with duplexer comprising:
Signal acquisition module 110, for obtaining echo-signal, the echo-signal is acquisition for detecting the anchor pole Anchor the initial information of defect；
First computing module 120, it is each after being decomposed for decomposing the echo-signal using variation mode decomposition method The corresponding intrinsic modal components of target of layer；
Second computing module 130 is used for by the intrinsic modal components noise reduction process of the corresponding target of each layer, after obtaining noise reduction Echo signal；
Detection module 140 obtains anchor rod anchored defects detection for the echo signal to be inputted target nerve network As a result.
In an embodiment of the present invention, the first computing module 120 includes:
First computing unit for being initially layered to the echo-signal, and is cyclically updated this layer based on every layer Initial intrinsic modal components obtain the target sheet of current layer after intrinsic modal components in the updated meet first condition of convergence Levy modal components；
Second computing unit, it is each after calculating this secondary clearing for the intrinsic modal components of target based on the current layer The weighted kurtosis value of layer；
First judging unit stops if the weighted kurtosis value for each layer after this secondary clearing meets the first preset condition It is only layered, obtaining final hierarchy number is the number of plies after last layering；
Second judgment unit, if the weighted kurtosis value for each layer after this secondary clearing is unsatisfactory for the first preset condition, Continue to be layered, the weighted kurtosis value of each layer after layering meets the first preset condition；
Wherein, first preset condition are as follows:
Kw=sgn (C) D | C |r<ζ；
Wherein,Sgn () is symbol letter Number guarantees that output signal and the phase of original signal are consistent as far as possible；KwFor weighted kurtosis value；R is the index of C；D is kurtosis value；μ For the mean value of the echo-signal f (t)；σ is the standard deviation of the echo-signal f (t)；T is the time；T is echo-signal f (t) Length；uk(t) be current hierarchical when kth layer the intrinsic modal components of target；For uk(t) mean value；For the equal of f (t) Value；ζ is preset threshold；C is the cross-correlation coefficient between two signals.
Wherein, the intrinsic modal components of the targetForIt is obtained by inverse Fourier transform:
Wherein:N is cycle-index；k For kth layer；It indicatesFourier transformation, be kth layer by n+1 circulation gained；For the echo Value of the signal f (t) after Fourier transformation；α is constant；It indicatesFourier transformation, be d layer by n times Gained is recycled, wherein d ≠ k；ω is frequency；For kth layer and be (n+1)th time circulation when mode centre frequency；Table Show Lagrangian λn(t) Fourier transformation, for n-th circulation gained；τ is time constant；K is current hierarchical number.
First condition of convergence are as follows:
Wherein: ε=10-7。
In an embodiment of the present invention, the second computing module 130 includes:
Noise reduction process is carried out using the intrinsic modal components of target of the wavelet thresholding method to each layer, the target letter after obtaining noise reduction Number.
In an embodiment of the present invention, it is connect with detection module 140 further include:
Network module is constructed, for obtaining the neural network of building；
Sample acquisition module, for obtaining sample set, and to the sample in the sample set carry out variation mode decomposition and Noise reduction process obtains sample signal collection；
First output module, one group of sample signal for concentrating sample signal input the neural network of the building, Obtain the output of hidden layer in this training process；
Second output module deletes the node for meeting the second preset condition, division for the output based on the hidden layer The node for meeting third preset condition, the neural network after obtaining this training；
Judgment module, for determining whether the neural network after this training meets second condition of convergence；
If meeting second condition of convergence, the neural network after this training is the target nerve network；
If being unsatisfactory for second condition of convergence, after being trained by Elman neural network weight more new formula to this Neural network weight and bias be updated；And it is concentrated from the sample signal and chooses one group of untrained sample signal This updated neural network is inputted, is trained next time, the neural network after training meets second condition of convergence.
In an embodiment of the present invention, the second output module includes:
Third computing unit obtains the contribution of each node of every layer of hidden layer for the output based on every layer of hidden layer Degree, the trimming threshold value of Current Situation of Neural Network and the growth threshold value of Current Situation of Neural Network；
Deletion of node unit, for comparing the contribution degree of each node of every layer of hidden layer and the trimming threshold value Compared with the contribution degree of deletion of node is less than the hidden layer node undertaking node layer corresponding with the hidden layer node of trimming threshold value；
Split vertexes unit, for comparing the contribution degree of each node of every layer of hidden layer and the growth threshold value Compared with the contribution degree of split vertexes is greater than the hidden layer node undertaking node layer corresponding with the hidden layer node for increasing threshold value, obtains Neural network after obtaining this training.
In an embodiment of the present invention, the contribution degree of each node of hidden layer are as follows:
Wherein, SConjFor the contribution degree of the node j of hidden layer；For hidden layer node j and output layer node i it Between connection weight；zj(p) the pth group output valve of the node j of hidden layer；S is characterized the s group data of data set；N is characterized number According to total group of number of collection；M is the node number of output layer；
The trimming threshold value of the Current Situation of Neural Network are as follows:
Wherein,SACon is the average contribution degree of hidden layer；Pth is repairing for Current Situation of Neural Network Cut threshold value；N is the node number of hidden layer；C is trimming constant；
The growth threshold value of the Current Situation of Neural Network are as follows:
Gth=dSACon；
Wherein, Gth is the growth threshold value of Current Situation of Neural Network；D is growth constant.
In an embodiment of the present invention, second condition of convergence are as follows:
MSE<ψ；
Wherein,MSE is network error；M is output matrix after the completion of network an iteration Element total number, yiFor network output layer node i reality output, yi expFor network output layer node i desired output；M is output The number of node layer；ψ is that circulation jumps out error.
In an embodiment of the present invention, echo-signal is the described for detecting of preset echo wave signal acquisition device acquisition The initial information of anchor rod anchored defect；
Wherein, the echo wave signal acquisition device includes:
Signal generation apparatus, magnetic field generation device and coil；The coil is wrapped on anchor pole, the magnetic field generation device It is set on the anchor pole, and the coil is set in the magnetic field generation device, the signal generation apparatus and the line Circle electrical connection；
The pumping signal that the signal generation apparatus issues makes in the anchor pole under the action of magnetic field generation device Supersonic guide-wave is inspired, is echo-signal by the anchor rod anchored reflected supersonic guide-wave.
It is apparent to those skilled in the art that for convenience and simplicity of description, only with above-mentioned each function The division progress of module can according to need and for example, in practical application by above-mentioned function distribution by different function moulds Block is completed, i.e., the internal structure of the detection device of the described anchor rod anchored defect is divided into different functional modules, more than completing The all or part of function of description.Each functional module in embodiment can integrate in one processing unit, be also possible to Each unit physically exists alone, and can also be integrated in one unit with two or more units, above-mentioned integrated module Both it can take the form of hardware realization, can also realize in the form of software functional units.In addition, the tool of each functional module Body title is also only for convenience of distinguishing each other, the protection scope being not intended to limit this application.Above-mentioned anchor rod anchored defect The specific work process of module in detection device, can be no longer superfluous herein with reference to the corresponding process in preceding method embodiment 1 It states.
Embodiment 3:
Figure 13 is the schematic diagram for the terminal device that one embodiment of the invention provides.As shown in figure 13, the terminal of the embodiment Equipment 136 includes: processor 1360, memory 1361 and is stored in the memory 1361 and can be in the processor The computer program 1362 run on 1360.The processor 1360 realizes such as embodiment when executing the computer program 1362 Step in each embodiment described in 1, such as step S101 to S104 shown in FIG. 1.Alternatively, the processor 1360 executes The function of each module/unit in each system embodiment as described in example 2 above, example are realized when the computer program 1362 The function of module 110 to 140 as shown in figure 12.
The terminal device 136 refers to the terminal with data-handling capacity, including but not limited to computer, work station, Server, the smart phone more even haveing excellent performance, palm PC, tablet computer, personal digital assistant (PDA), intelligence TV (Smart TV) etc..Operating system is generally fitted on terminal device, including but not limited to: Windows operating system, LINUX operating system, Android (Android) operating system, Symbian operating system, Windows mobile operating system, with And iOS operating system etc..The specific example of terminal device 136 is enumerated in detail above, and those skilled in the art will be appreciated that It arrives, terminal device is not limited to above-mentioned enumerate example.
The terminal device may include, but be not limited only to, processor 1360, memory 1361.Those skilled in the art can To understand, Figure 13 is only the example of terminal device 136, does not constitute the restriction to terminal device 136, may include than diagram More or fewer components perhaps combine certain components or different components, such as the terminal device 136 can also wrap Include input-output equipment, network access equipment, bus etc..
Alleged processor 1360 can be central processing unit (Central Processing Unit, CPU), can be with It is other general processors, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit (Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field- Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic, Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor Deng.
The memory 1361 can be the internal storage unit of the terminal device 136, such as terminal device 136 is hard Disk or memory.The memory 1361 is also possible to the External memory equipment of the terminal device 136, such as the terminal device The plug-in type hard disk being equipped on 136, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash Card) etc..Further, the memory 1361 can also both include the end The internal storage unit of end equipment 136 also includes External memory equipment.The memory 1361 is for storing the computer journey Other programs and data needed for sequence and the terminal device 136.The memory 1361 can be also used for temporarily storing The data that has exported or will export.
Embodiment 4:
The embodiment of the invention also provides a kind of computer readable storage medium, computer-readable recording medium storage has meter Calculation machine program is realized the step in each embodiment as described in example 1 above, such as is schemed when computer program is executed by processor Step S101 shown in 1 to step S104.Alternatively, realizing when the computer program is executed by processor such as institute in embodiment 2 The function of each module/unit in each system embodiment stated, such as the function of module 110 to 140 shown in Figure 12.
The computer program can be stored in a computer readable storage medium, and the computer program is by processor When execution, it can be achieved that the step of above-mentioned each embodiment of the method.Wherein, the computer program includes computer program code, The computer program code can be source code form, object identification code form, executable file or certain intermediate forms etc..Institute State computer-readable medium may include: can carry the computer program code any entity or device, recording medium, USB flash disk, mobile hard disk, magnetic disk, CD, computer storage, read-only memory (ROM, Read-Only Memory), arbitrary access Memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium etc..
In the above-described embodiments, it all emphasizes particularly on different fields to the description of each embodiment, embodiment 1 to 4 can in any combination, group The new embodiment formed after conjunction is also within the scope of protection of this application.There is no the portion for being described in detail or recording in some embodiment Point, it may refer to the associated description of other embodiments.
Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure Member and algorithm steps can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions are actually It is implemented in hardware or software, the specific application and design constraint depending on technical solution.Professional technician Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed The scope of the present invention.
In embodiment provided by the present invention, it should be understood that disclosed terminal device and method can pass through it Its mode is realized.For example, system described above/terminal device embodiment is only schematical, for example, the module Or the division of unit, only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple lists Member or component can be combined or can be integrated into another system, or some features can be ignored or not executed.Another point, Shown or discussed mutual coupling or direct-coupling or communication connection can be through some interfaces, device or unit INDIRECT COUPLING or communication connection, can be electrical property, mechanical or other forms.
Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations；Although referring to aforementioned reality Applying example, invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each Technical solution documented by embodiment is modified or equivalent replacement of some of the technical features；And these are modified Or replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all It is included within protection scope of the present invention.

Claims (10)

1. a kind of detection method of anchor rod anchored defect characterized by comprising
Echo-signal is obtained, the echo-signal is acquisition for detecting the initial information of the anchor rod anchored defect；
The echo-signal is decomposed using variation mode decomposition method, the corresponding intrinsic mode of target of each layer is divided after being decomposed Amount；
Echo signal by the intrinsic modal components noise reduction process of the corresponding target of each layer, after obtaining noise reduction；
The echo signal is inputted into target nerve network, obtains anchor rod anchored defects detection result.
2. the detection method of anchor rod anchored defect as described in claim 1, which is characterized in that described to utilize variation mode decomposition Method decomposes the echo-signal, and the corresponding intrinsic modal components of each layer include: after being decomposed
The echo-signal is initially layered, and is cyclically updated the initial intrinsic modal components of this layer based on every layer, more After intrinsic modal components after new meet first condition of convergence, the intrinsic modal components of target of current layer are obtained；
The intrinsic modal components of target based on the current layer, the weighted kurtosis value of each layer after calculating this secondary clearing；
If the weighted kurtosis value of each layer after this secondary clearing meets the first preset condition, stops being layered, obtain final hierarchy number For the number of plies after last layering；
If the weighted kurtosis value of each layer after this secondary clearing is unsatisfactory for the first preset condition, continue to be layered, after layering The weighted kurtosis value of each layer meets the first preset condition；
Wherein, the intrinsic modal components of the targetForIt is obtained by inverse Fourier transform:
Wherein:N is cycle-index；K is kth Layer；It indicatesFourier transformation, be kth layer by n+1 circulation gained；For the echo-signal f (t) value after Fourier transformation；α is constant；It indicatesFourier transformation, for d layer by n times circulation Gained, wherein d ≠ k；ω is frequency；For kth layer and be (n+1)th time circulation when mode centre frequency；It indicates to draw Ge Lang operator λn(t) Fourier transformation, for n-th circulation gained；τ is time constant；K is current hierarchical number；
First condition of convergence are as follows:
Wherein: ε=10-7；
First preset condition are as follows:
Kw=sgn (C) D | C |r< ζ；
Wherein,Sgn () is sign function, is protected It demonstrate,proves output signal and the phase of original signal is consistent as far as possible；KwFor weighted kurtosis value；R is the index of C；D is kurtosis value；μ is institute State the mean value of echo-signal f (t)；σ is the standard deviation of the echo-signal f (t)；T is the time；T is the length of echo-signal f (t) Degree；uk(t) be current hierarchical when kth layer the intrinsic modal components of target；For uk(t) mean value；For the mean value of f (t)；ζ For preset threshold；C is the cross-correlation coefficient between two signals.
3. the detection method of anchor rod anchored defect as described in claim 1, which is characterized in that described that each layer is corresponding intrinsic Modal components noise reduction process, the echo signal after obtaining noise reduction, comprising:
Noise reduction process is carried out using intrinsic modal components of the wavelet thresholding method to each layer, the echo signal after obtaining noise reduction.
4. the detection method of anchor rod anchored defect as described in claim 1, which is characterized in that defeated by the echo signal Before entering target nerve network, further includes:
Obtain the neural network of building；
Sample set is obtained, and variation mode decomposition and noise reduction process are carried out to the sample in the sample set, obtains sample signal Collection；
The neural network that one group of sample signal that sample signal is concentrated is inputted to the building is obtained in this training process and is implied The output of layer；
Based on the output of the hidden layer, the node for meeting the second preset condition is deleted, division meets the section of third preset condition Point, the neural network after obtaining this training；
Whether the neural network after determining this training meets second condition of convergence；
If meeting second condition of convergence, the neural network after this training is the target nerve network；
If being unsatisfactory for second condition of convergence, by Elman neural network weight more new formula to the mind after this training Weight and bias through network are updated；And it is concentrated from the sample signal and chooses one group of untrained sample signal input This updated neural network, is trained next time, and the neural network after training meets second condition of convergence.
5. the detection method of anchor rod anchored defect as claimed in claim 4, which is characterized in that described based on the hidden layer The node for meeting the second preset condition is deleted in output, and division meets the node of third preset condition, the mind after obtaining this training Include: through network
Based on the output of every layer of hidden layer, the contribution degree of each node of every layer of hidden layer, the trimming of Current Situation of Neural Network are obtained The growth threshold value of threshold value and Current Situation of Neural Network；
The contribution degree of each node of every layer of hidden layer is compared with the trimming threshold value, the contribution degree of deletion of node is less than Trim the hidden layer node undertaking node layer corresponding with the hidden layer node of threshold value；
The contribution degree of each node of every layer of hidden layer is compared with the growth threshold value, the contribution degree of split vertexes is greater than The hidden layer node undertaking node layer corresponding with the hidden layer node for increasing threshold value, the neural network after obtaining this training.
6. the detection method of anchor rod anchored defect as claimed in claim 5, which is characterized in that each node of the hidden layer Contribution degree are as follows:
Wherein, SConjFor the contribution degree of the node j of hidden layer；wij 3Connect between the node j of hidden layer and the node i of output layer Connect weight；zj(p) the pth group output valve of the node j of hidden layer；S is characterized the s group data of data set；N is characterized data set Total group of number；M is the node number of output layer；
The trimming threshold value of the Current Situation of Neural Network are as follows:
Wherein,SACon is the average contribution degree of hidden layer；Pth is the trimming threshold of Current Situation of Neural Network Value；N is the node number of hidden layer；C is trimming constant；
The growth threshold value of the Current Situation of Neural Network are as follows:
Gth=dSACon；
Wherein, Gth is the growth threshold value of Current Situation of Neural Network；D is growth constant.
7. the detection method of anchor rod anchored defect as claimed in claim 6, which is characterized in that second condition of convergence are as follows:
MSE < ψ；
Wherein,MSE is network error；M is element in network output matrix in an iteration Total number, yiFor network output layer node i reality output, yi expFor network output layer node i desired output；M is output node layer Number；ψ is that circulation jumps out error.
8. the detection method of the anchor rod anchored defect as described in claim 1-7 any one, which is characterized in that the echo letter It number is the acquisition of preset echo wave signal acquisition device for detecting the initial information of the anchor rod anchored defect；
Wherein, the echo wave signal acquisition device includes:
Signal generation apparatus, magnetic field generation device and coil；The coil is wrapped on anchor pole, the magnetic field generation device setting In on the anchor pole, and the coil is set in the magnetic field generation device, the signal generation apparatus and coil electricity Connection；
The pumping signal that the signal generation apparatus issues makes to excite in the anchor pole under the action of magnetic field generation device Supersonic guide-wave out is echo-signal by the anchor rod anchored reflected supersonic guide-wave.
9. a kind of terminal device, which is characterized in that in the memory and can be in institute including memory, processor and storage The computer program run on processor is stated, the processor realizes such as claim 1 to 8 times when executing the computer program The step of detection method of one anchor rod anchored defect.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer journey Sequence realizes the detection of the anchor rod anchored defect as described in any one of claim 1 to 8 when the computer program is executed by processor The step of method.
CN201910511203.XA 2019-06-13 2019-06-13 Detection method for anchor rod anchoring defect and terminal equipment Active CN110208375B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910511203.XA CN110208375B (en) 2019-06-13 2019-06-13 Detection method for anchor rod anchoring defect and terminal equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910511203.XA CN110208375B (en) 2019-06-13 2019-06-13 Detection method for anchor rod anchoring defect and terminal equipment

Publications (2)

Publication Number Publication Date
CN110208375A true CN110208375A (en) 2019-09-06
CN110208375B CN110208375B (en) 2021-12-14

Family

ID=67792519

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910511203.XA Active CN110208375B (en) 2019-06-13 2019-06-13 Detection method for anchor rod anchoring defect and terminal equipment

Country Status (1)

Country Link
CN (1) CN110208375B (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110749650A (en) * 2019-10-16 2020-02-04 四川大学 Anchor rod anchoring compactness grade evaluation method based on support vector machine
CN110826598A (en) * 2019-10-16 2020-02-21 四川大学 Anchor rod anchoring compactness grade evaluation method based on convolutional neural network
CN110849968A (en) * 2019-11-05 2020-02-28 东南大学 Crane main beam damage acoustic emission nondestructive detection method based on self-adaptive optimization VMD
CN111723701A (en) * 2020-06-08 2020-09-29 西安交通大学 Underwater target identification method
CN114200016A (en) * 2021-10-18 2022-03-18 中国科学院武汉岩土力学研究所 Double-channel nondestructive detection method for rock anchor rod and related equipment
CN114354969A (en) * 2021-09-23 2022-04-15 山东开泰抛丸机械股份有限公司 Speed measuring device for testing shot projection speed of shot blasting machine
CN117091648A (en) * 2023-07-27 2023-11-21 石家庄铁道大学 Air-ground integrated construction ecological environment monitoring device and visual processing method

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106053602A (en) * 2016-05-04 2016-10-26 石家庄铁道大学 A self-closed rock bolt nondestructive testing method based on a magnetostrictive effect
CN106778809A (en) * 2016-11-23 2017-05-31 安徽理工大学 A kind of blind source separation method based on improvement chicken group's algorithm
US20170284974A1 (en) * 2016-03-31 2017-10-05 General Electric Company System for sensing and locating delamination
CN108845352A (en) * 2018-06-27 2018-11-20 吉林大学 Desert Denoising of Seismic Data method based on VMD approximate entropy and multi-layer perception (MLP)
CN109446928A (en) * 2018-10-10 2019-03-08 南京航空航天大学 A kind of signal de-noising method based on variation mode decomposition and least mean-square error sef-adapting filter
CN109633431A (en) * 2019-01-09 2019-04-16 国网江苏省电力有限公司电力科学研究院 The load ratio bridging switch fault recognition method extracted based on vibration signal characteristics

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20170284974A1 (en) * 2016-03-31 2017-10-05 General Electric Company System for sensing and locating delamination
CN106053602A (en) * 2016-05-04 2016-10-26 石家庄铁道大学 A self-closed rock bolt nondestructive testing method based on a magnetostrictive effect
CN106778809A (en) * 2016-11-23 2017-05-31 安徽理工大学 A kind of blind source separation method based on improvement chicken group's algorithm
CN108845352A (en) * 2018-06-27 2018-11-20 吉林大学 Desert Denoising of Seismic Data method based on VMD approximate entropy and multi-layer perception (MLP)
CN109446928A (en) * 2018-10-10 2019-03-08 南京航空航天大学 A kind of signal de-noising method based on variation mode decomposition and least mean-square error sef-adapting filter
CN109633431A (en) * 2019-01-09 2019-04-16 国网江苏省电力有限公司电力科学研究院 The load ratio bridging switch fault recognition method extracted based on vibration signal characteristics

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
XIAOYUN SUN等: "Pruning Elman neural network and its application in bolt defects classification", 《INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS》 *
蒲子玺等: "基于峭度准则VMD及平稳小波的轴承故障诊断", 《机械设计与研究》 *
谭继勇等: "冲击信号的随机共振自适应检测方法", 《机械工程学报》 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110749650A (en) * 2019-10-16 2020-02-04 四川大学 Anchor rod anchoring compactness grade evaluation method based on support vector machine
CN110826598A (en) * 2019-10-16 2020-02-21 四川大学 Anchor rod anchoring compactness grade evaluation method based on convolutional neural network
CN110849968A (en) * 2019-11-05 2020-02-28 东南大学 Crane main beam damage acoustic emission nondestructive detection method based on self-adaptive optimization VMD
CN111723701A (en) * 2020-06-08 2020-09-29 西安交通大学 Underwater target identification method
CN114354969A (en) * 2021-09-23 2022-04-15 山东开泰抛丸机械股份有限公司 Speed measuring device for testing shot projection speed of shot blasting machine
CN114200016A (en) * 2021-10-18 2022-03-18 中国科学院武汉岩土力学研究所 Double-channel nondestructive detection method for rock anchor rod and related equipment
CN114200016B (en) * 2021-10-18 2024-04-16 中国科学院武汉岩土力学研究所 Double-channel nondestructive testing method and related equipment for rock anchor rod
CN117091648A (en) * 2023-07-27 2023-11-21 石家庄铁道大学 Air-ground integrated construction ecological environment monitoring device and visual processing method

Also Published As

Publication number Publication date
CN110208375B (en) 2021-12-14

Similar Documents

Publication Publication Date Title
CN110208375A (en) A kind of detection method and terminal device of anchor rod anchored defect
EP3832553A1 (en) Method for identifying energy of micro-energy device on basis of bp neural network
Colburn et al. State estimation in wall-bounded flow systems. Part 3. The ensemble Kalman filter
CN103049792B (en) Deep-neural-network distinguish pre-training
Wan et al. Dual Kalman filtering methods for nonlinear prediction, smoothing and estimation
KR101958674B1 (en) Actually-measured marine environment data assimilation method based on sequence recursive filtering three-dimensional variation
CN110174698B (en) Elastic impedance inversion method and system based on orientation Fourier coefficients
CN105142177A (en) Complex neural network channel prediction method
CN101221213A (en) Analogue circuit fault diagnosis neural network method based on particle swarm algorithm
CN112884059B (en) Small sample radar working mode classification method fusing priori knowledge
CN104052612A (en) Method and system for fault identification and positioning of telecommunication service
CN109388778A (en) A kind of iteration volume point Unscented kalman filtering method
CN111541255B (en) Low-frequency oscillation mode identification method and system based on dynamic system
CN104614647A (en) Complex wavelet transform partial discharge location test method and device
Kim et al. Statistical estimation of extreme roll responses in short crested irregular head seas
CN110909453A (en) EEMD-based power transmission line icing grade prediction method
CN110570114A (en) Power quality disturbance identification method, system and medium based on variable cycle neural network
CN110837111A (en) Seismic data interpolation method and system
CN113740671A (en) Fault arc identification method based on VMD and ELM
CN107145732A (en) A kind of method based on the rapid solving target list station Electromagnetic Scattering Characteristics for improving CBFM
CN105891826A (en) Airborne radar fast maximum posteriori imaging method
CN109654383A (en) A kind of leak degree of water supply line determines method and system
CN109657196B (en) MOM-KA mixing method based on airspace strong coupling path acceleration
CN116975742A (en) Partial discharge pattern recognition method, apparatus, device, and storage medium
CN110518588B (en) Filtering method, device, equipment and power electronic device

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