CN108269249A - A kind of bolt detecting system and its implementation - Google Patents
A kind of bolt detecting system and its implementation Download PDFInfo
- Publication number
- CN108269249A CN108269249A CN201711307933.5A CN201711307933A CN108269249A CN 108269249 A CN108269249 A CN 108269249A CN 201711307933 A CN201711307933 A CN 201711307933A CN 108269249 A CN108269249 A CN 108269249A
- Authority
- CN
- China
- Prior art keywords
- bolt
- sound
- detecting system
- collection
- wavelet
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20064—Wavelet transform [DWT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
Abstract
The invention discloses a kind of bolt detecting system and its implementation, system includes training study module and detection module, method includes acquiring bolt percussion sound by bolt detecting system, sound is tapped to the bolt of acquisition and carries out feature extraction, and study is trained to the feature of extraction according to machine learning algorithm;The percussion sound of bolt to be detected is acquired by bolt detecting system, according to training study as a result, tapping sound to the bolt to be detected of acquisition carries out defects detection.The whole work process of the present invention reduces cost of labor and improves the reliability of detection without human intervention;The present invention also is able to tap a large amount of bolt sound progress feature extraction and training study, realizes the quantization of bolt defect index, can identify the various defects of bolt comprehensively, further improves the reliability of detection.It the composite can be widely applied to connector detection field.
Description
Technical field
The present invention relates to connector detection field, especially a kind of bolt detecting system and its implementation.
Background technology
It is well known that playing very important effect in civil engineering structure interior joint, node generation destruction can directly result in whole
A structure is destroyed, and consequence is extremely serious, main connecting elements of the bolt as steel structure node, and working condition should be by
The concern of Structural Engineer.And in recent years, the drop event of steel construction high-strength bolt fracture for several times, such as gantry occur for nuclear power plant
High-strength bolt brittle fracture occurs for frame, pumping plant roof truss and anti-whipping stent etc. and drop event, this is not only whole to steel construction
Body structure causes safely potential for adverse effects, and causes serious threat to personnel's personal safety below.Therefore, it is necessary to big
High-strength bolt carries out preventive inspection at formed steel construction node.
At present, service personnel can only use observation and hand hammer hammering method to carry out preventive inspection to bolt, but this
The method of artificial detection has the following problems:1st, often height is higher for large-scale steel structure, and the scale of construction is huge, and construction operation need to largely be taken
It tears scaffold open, expends huge financial resources, manpower and time cost, and exist compared with high safety risk, meanwhile, huge cost consumption
Such maintenance is caused to be difficult to, there are security risks.2nd, observation is only capable of finding whether high-strength bolt surface corrosion occurs
Or apparent deformation, the more difficult discovery of internal flaw, reliability are low.3rd, the sound that hand hammer hammering method is sent out by tapping bolt is sentenced
Breaking, whether it loosens, and compares the engineering experience for relying on construction worker, lacks quantizating index.
Invention content
In order to solve the above technical problems, first purpose of the present invention is:There is provided one kind at low cost, reliability Gao Bingneng
Carry out the bolt detecting system of quantification of targets.
Second object of the present invention is:There is provided it is a kind of it is at low cost, reliability is high and can carry out quantification of targets, bolt
The implementation method of detecting system.
First technical solution being taken of the present invention be:
A kind of bolt detecting system, including:
Training study module carries out feature extraction, and according to machine learning algorithm to extraction for tapping sound to bolt
Feature be trained study;
Detection module, for learning as a result, to bolt to be detected progress defects detection according to training.
Further, the trained study module includes:
Sound collection unit taps sound for acquiring defective and flawless sample bolt;
Characteristic extracting module carries out feature extraction for tapping sound collection of illustrative plates to the bolt of acquisition using wavelet moment algorithm;
Feature training module, for being trained identification to the feature of extraction using BP neural network.
Further, the characteristic extracting module includes:
Normalized unit is normalized for tapping sound collection of illustrative plates to the bolt of acquisition, obtains full simultaneously
The collection of illustrative plates of sufficient translation invariance and scaling invariance;
Map unit for the collection of illustrative plates after normalized to be mapped to polar coordinates, obtains the binary map on polar coordinates
Picture;
Structural unit for carrying out moment characteristics construction using wavelet function collection, obtains wavelet moment invariants, the small echo letter
Manifold ψa,b(r) it is:
Wherein, a represents broadening factor, a ∈ R+, b represents shift factor, and b ∈ R, ψ (r) represent morther wavelet, and r represents two-value
The polar diameter of image;
Characteristics of image generation unit, for according to obtained wavelet moment invariants, generating the local feature and the overall situation of collection of illustrative plates
Feature.
Further, the described function of the morther wavelet ψ (r) is cubic B-spline function.
Further, the feature training module includes:
Forward transfer element calculates for the input value to neuron each in BP neural network, obtains error letter
Number;
Unit is reversely adjusted, for judging whether obtained error function meets threshold requirement, if so, not processing;
Conversely, then correcting the input value of each neuron, and return to forward transfer element.
Further, the BP neural network includes input layer, hidden layer and output layer, the expression formula of the error function E
For:
Wherein, m is number of samples, and p is the number of output layer neuron, tksThe phase of s-th of neuron for k-th of sample
Hope output valve, yksThe real output value of s-th of neuron for k-th of sample.
Second technical solution being taken of the present invention be:
A kind of implementation method of bolt detecting system, includes the following steps:
Bolt is acquired by bolt detecting system and taps sound, tapping sound to the bolt of acquisition carries out feature extraction, and
Study is trained to the feature of extraction according to machine learning algorithm;
The percussion sound of bolt to be detected is acquired by bolt detecting system, according to training study as a result, to acquisition
Bolt to be detected taps sound and carries out defects detection.
Further, the bolt of described pair of acquisition taps sound and carries out feature extraction, and according to machine learning algorithm to extraction
Feature the step for being trained study, include the following steps:
Defective and flawless sample bolt is acquired by bolt detecting system and taps sound;
Sound collection of illustrative plates is tapped to the bolt of acquisition using wavelet moment algorithm and carries out feature extraction;
Identification is trained to the feature of extraction using BP neural network.
Further, the sound collection of illustrative plates that tapped using wavelet moment algorithm to the bolt of acquisition carries out feature extraction this step
Suddenly, include the following steps:
Sound collection of illustrative plates is tapped to the bolt of acquisition to be normalized, and is met translation invariance and scaling simultaneously not
The collection of illustrative plates of denaturation;
Collection of illustrative plates after normalized is mapped on polar coordinates, obtains the bianry image on polar coordinates;
Moment characteristics construction is carried out using wavelet function collection, obtains wavelet moment invariants, the wavelet function collection ψa,b(r) it is:
Wherein, a represents broadening factor, a ∈ R+, b represents shift factor, and b ∈ R, ψ (r) represent morther wavelet, and r represents two-value
The polar diameter of image;
According to obtained wavelet moment invariants, the local feature and global characteristics of collection of illustrative plates are generated.
Further, described the step for identification is trained to the feature of extraction using BP neural network, including following step
Suddenly:
The input value of neuron each in BP neural network is calculated, obtains error function;
Whether the error function for judging to obtain meets threshold requirement, if so, not processing;Conversely, then correct each god
Through member input value, and return the input value of neuron each in BP neural network is calculated, obtain error function this
Step, until obtained error function meets threshold requirement.
The advantageous effect of system of the present invention is:The trained study module of the system integration and detection module of the present invention is to spiral shell
The working condition of bolt is detected, and whole work process greatly reduces cost of labor and improve detection without human intervention
Reliability;In addition, this system can tap a large amount of bolt, sound carries out feature extraction and training learns, compared to existing
There is hand hammer hammering method, this system realizes the quantization of bolt defect index, can identify the various defects of bolt comprehensively, further
Improve the reliability of detection.
The beneficial effects of the method for the present invention is:The method of the present invention taps sound to bolt first and carries out feature extraction,
And study is trained to the feature of extraction according to machine learning algorithm, then according to training study as a result, to spiral shell to be detected
Bolt carries out defects detection, and whole work process greatly reduces cost of labor and improve the reliable of detection without human intervention
Property;In addition, this method can tap a large amount of bolt, sound carries out feature extraction and training learns, compared to existing hand hammer
Hammering method, this method realize the quantization of bolt defect index, can identify the various defects of bolt comprehensively, further improve
The reliability of detection.
Description of the drawings
Fig. 1 is a kind of overall structure block diagram of bolt detecting system of the present invention;
Fig. 2 is a kind of whole flow chart of steps of the implementation method of one bolt detecting system of embodiment;
Fig. 3 is the neuronal structure schematic diagram of one BP neural network of embodiment.
Specific embodiment
Reference Fig. 1, a kind of bolt detecting system, including:
Training study module carries out feature extraction, and according to machine learning algorithm to extraction for tapping sound to bolt
Feature be trained study;
Detection module, for learning as a result, to bolt to be detected progress defects detection according to training.
With reference to Fig. 1, preferred embodiment is further used as, the trained study module includes:
Sound collection unit taps sound for acquiring defective and flawless sample bolt;
Characteristic extracting module carries out feature extraction for tapping sound collection of illustrative plates to the bolt of acquisition using wavelet moment algorithm;
Feature training module, for being trained identification to the feature of extraction using BP neural network.
Wherein, the characteristics of wavelet moment is a kind of novel not bending moment, it combines not bending moment and wavelet transformation, wavelet moment is not only
Image overall feature can be extracted, can also extract the local feature of image, performance ratio Hu squares, Li squares and Zernike squares etc. its
His Moment Feature Extraction algorithm is good.
With reference to Fig. 1, preferred embodiment is further used as, the characteristic extracting module includes:
Normalized unit is normalized for tapping sound collection of illustrative plates to the bolt of acquisition, obtains full simultaneously
The collection of illustrative plates of sufficient translation invariance and scaling invariance;
Map unit for the collection of illustrative plates after normalized to be mapped to polar coordinates, obtains the binary map on polar coordinates
Picture;
Structural unit for carrying out moment characteristics construction using wavelet function collection, obtains wavelet moment invariants, the small echo letter
Manifold ψa,b(r) it is:
Wherein, a represents broadening factor, a ∈ R+, b represents shift factor, and b ∈ R, ψ (r) represent morther wavelet, and r represents two-value
The polar diameter of image;
Characteristics of image generation unit, for according to obtained wavelet moment invariants, generating the local feature and the overall situation of collection of illustrative plates
Feature.
Preferred embodiment is further used as, the described function of the morther wavelet ψ (r) is cubic B-spline function.
Wherein, B-spline curves curved surface has geometric invariance, convex closure, convexity-preserving, the reduction property that is deteriorated and local supportive
Etc. many advantageous properties.
With reference to Fig. 1, preferred embodiment is further used as, the feature training module includes:
Forward transfer element calculates for the input value to neuron each in BP neural network, obtains error letter
Number;
Unit is reversely adjusted, for judging whether obtained error function meets threshold requirement, if so, not processing;
Conversely, then correcting the input value of each neuron, and return to forward transfer element.
Wherein, BP (back propagation) neural network be 1986 by Rumelhart and McClelland headed by
Scientist propose concept, be a kind of multilayer feedforward neural network train according to error backpropagation algorithm, be at present answer
With widest neural network, there is arbitrarily complicated pattern classification ability and excellent multidimensional function mapping ability.
Preferred embodiment is further used as, the BP neural network includes input layer, hidden layer and output layer, described
The expression formula of error function E is:
Wherein, m is number of samples, and p is the number of output layer neuron, tksThe phase of s-th of neuron for k-th of sample
Hope output valve, yksThe real output value of s-th of neuron for k-th of sample.
A kind of system based on Fig. 1, implementation method of bolt detecting system of the present invention, includes the following steps:
Bolt is acquired by bolt detecting system and taps sound, tapping sound to the bolt of acquisition carries out feature extraction, and
Study is trained to the feature of extraction according to machine learning algorithm;
The percussion sound of bolt to be detected is acquired by bolt detecting system, according to training study as a result, to acquisition
Bolt to be detected taps sound and carries out defects detection.
It is further used as preferred embodiment, the bolt of described pair of acquisition taps sound and carries out feature extraction, and according to
The step for machine learning algorithm is trained study to the feature of extraction, includes the following steps:
Defective and flawless sample bolt is acquired by bolt detecting system and taps sound;
Sound collection of illustrative plates is tapped to the bolt of acquisition using wavelet moment algorithm and carries out feature extraction;
Identification is trained to the feature of extraction using BP neural network.
Be further used as preferred embodiment, it is described the bolt of acquisition is tapped using wavelet moment algorithm sound collection of illustrative plates into
The step for row feature extraction, include the following steps:
Sound collection of illustrative plates is tapped to the bolt of acquisition to be normalized, and is met translation invariance and scaling simultaneously not
The collection of illustrative plates of denaturation;
Collection of illustrative plates after normalized is mapped on polar coordinates, obtains the bianry image on polar coordinates;
Moment characteristics construction is carried out using wavelet function collection, obtains wavelet moment invariants, the wavelet function collection ψa,b(r) it is:
Wherein, a represents broadening factor, a ∈ R+, b represents shift factor, and b ∈ R, ψ (r) represent morther wavelet, and r represents two-value
The polar diameter of image;
According to obtained wavelet moment invariants, the local feature and global characteristics of collection of illustrative plates are generated.
Preferred embodiment is further used as, described be trained using BP neural network to the feature of extraction identifies this
One step, includes the following steps:
The input value of neuron each in BP neural network is calculated, obtains error function;
Whether the error function for judging to obtain meets threshold requirement, if so, not processing;Conversely, then correct each god
Through member input value, and return the input value of neuron each in BP neural network is calculated, obtain error function this
Step, until obtained error function meets threshold requirement.
The present invention is further explained and illustrated with specific embodiment with reference to the accompanying drawings of the specification.For of the invention real
The step number in example is applied, is set only for the purposes of illustrating explanation, the sequence between step does not do any restriction, implements
The execution sequence of each step in example can be adaptively adjusted according to the understanding of those skilled in the art.
Embodiment one
Due at present at large-scale steel structure node high-strength bolt carry out preventive inspection method only have observation and
Hand hammer hammering method, and this method there are of high cost, reliability it is low and lack quantizating index the shortcomings of, therefore, the present invention carries
Go out a kind of bolt detecting system and its implementation.The present invention taps sound to bolt by training study module first and carries out spy
Sign extraction, and study is trained to the feature of extraction according to machine learning algorithm, then by detection module to spiral shell to be detected
Bolt carries out defects detection, and whole work process greatly reduces cost of labor and improve the reliable of detection without human intervention
Property;In addition, the present invention can tap a large amount of bolt, sound carries out feature extraction and training learns, compared to existing hand hammer
Hammering method, the present invention realize the quantization of bolt defect index, can identify the various defects of bolt comprehensively, further improve
The reliability of detection.
With reference to Fig. 1, a kind of bolt detecting system of the present invention includes training study module and detection module.Training study module
Including sound collection unit, characteristic extracting module and feature training module, characteristic extracting module includes normalized unit, reflects
Unit, structural unit and characteristics of image generation unit are penetrated, feature training module includes forward transfer element and reversely adjusts unit.
Wherein, sound collection unit is sequentially connected normalized unit, map unit, structural unit, characteristics of image life
Into unit, forward transfer element, reversely adjust unit and detection module.
With reference to Fig. 2, a kind of specific steps flow of the implementation method of bolt detecting system of the present invention is as follows:
S1, it is used as training sample by a large amount of defective and flawless bolts percussion sound of bolt detecting system acquisition;
Wherein, the operation principle for tapping detection is:
The difference of one object vibration state shows as the sound sent out difference, physically this is because they are vibrated
Amplitude, frequency, duration and single vibration or complex vibration etc. difference.The material of these physical quantitys and vibrating object
Material, structure etc. are closely related.As a vibrational system, in single-frequency, the fundamental equation of mechanical oscillation is:F=
Zu, in formula:F is the driving force of mechanical oscillation;U is the vibration velocity of particle;Z is equivalent mechanical impedance, is expressed as:
In formula:M is equivalent mass;C is equivalent compliance;R is hindered for equivalent friction
Buddhist nun;I is electric current;ω is angular frequency.
By directly or indirectly measuring u, so that it is determined that characteristic the defects of examined workpiece, taps flaw detection and listens sound measurement actually
It is exactly indirect measurement u.The process of percussion be exactly in detected object excitation generate the process of mechanical oscillation, and sound and strike
The acquisition of hitter's sense is then the information collection in detection process, and testing staff divides the information of acquisition by the experience of oneself
Analysis judges that the feature extraction and result that obtained final conclusion just belongs in detection process judge.In order to improve this method
Accuracy, ease for use, many scholars are to the mode of oscillation deployment analysis of component.For any structure, always there are one
Or multiple natural frequencies, the variation of these natural frequencies follow the rule represented by following expression substantially:
The principal element for influencing component natural frequency includes:(E is Young's modulus to the bending stiffness EI of component, and EI is section
Second moment), the linear mass m of component, shearing rigidity Ks and rotary inertia ρ I (for density of material).These parameters
Frequency variation caused by slight change is approximately linear.When component partial existing defects and size or the difference of material
When, corresponding parameter EI, m, Ks and ρ I can generate variation, so as to cause the variation of frequency;
S2, the progress feature extraction of sound collection of illustrative plates, the square of wavelet moment algorithm tap the bolt of acquisition using wavelet moment algorithm
The general expression of feature can be defined as:Fpq=∫ ∫ f (r, θ) gp(r)ejqθRdrd θ, wherein, f (r, θ) is represented on polar coordinates
Two dimension bianry image, gp(r) radial component on polar coordinates, e are representedjpθRepresent the angle variables on polar coordinates;
Wherein, step S2 specifically includes following steps:
S21, the bolt percussion sound collection of illustrative plates of acquisition is normalized, is met translation invariance and contracting simultaneously
Put the collection of illustrative plates of invariance.First, the barycenter (x ', y ') of image f (r, θ) is obtained, then coordinate origin moves on to barycenter so that f '
(x, y)=f (x+x ', y+y ') obtains translation invariance;Then, artwork is zoomed in or out α times, definitionIts
In, m00Represent zeroth order square, β represents the size of desired image, obtains the scaling invariance of image;So at by Pan and Zoom
Image after reason is f ' ' (x, y)=f (α x+x ', α y+y ');
S22, the collection of illustrative plates after normalized is mapped on polar coordinates, obtains the bianry image on polar coordinates;
S23, moment characteristics construction is carried out using wavelet function collection, obtains wavelet moment invariants, the wavelet function collection isWherein, a (a ∈ R+) representing broadening factor, b (b ∈ R) represents shift factor, and ψ (r) represents morther wavelet;
The wavelet moment invariants that S24, basis obtain generate the local feature and global characteristics of collection of illustrative plates.
S3, identification is trained to the feature of extraction using BP neural network.BP neural network is made of neuron,
The specific structure is shown in FIG. 3 for neuron, wherein, xjRepresent j-th of neuron, they represent input, w in figureijRepresent jth
Connection weight between a neuron and i-th of neuron, uiRepresent the state of i-th of neuron,
θiRepresent the threshold value of i-th of neuron, yiRepresent the output of i-th of neuron, yi=f (ui) and next neuron is defeated
Enter, f () represents Sigmoid functions
S4, sound collection is carried out to bolt to be detected by bolt detecting system, according to training study as a result, to acquisition
Bolt sound carry out defect analysis, the structural rigidity of aeration level and bolt the defects of the bolt including bolt.
In conclusion a kind of bolt detecting system of the present invention and its implementation have the following advantages:
1), bolt detecting system of the invention effectively integrates all modules, and entire operation is completed in cooperation
Journey does not need to human intervention during operation process, automatically controls the orderly operation of each system in itself by system, greatly reduce people
Work cost and the reliability for improving detection.
2), the present invention can tap a large amount of bolt sound and carry out feature extraction and training study, compared to existing
Hand hammer hammering method, the present invention realize the quantization of bolt defect index, can identify the various defects of bolt comprehensively, further carry
The high reliability of detection.
3), the present invention taps sound collection of illustrative plates to the bolt of acquisition using wavelet moment algorithm and carries out feature extraction, can be simultaneously
Extract the local feature and global characteristics of sound collection of illustrative plates.
4), the present invention is trained identification using BP neural network to the feature of extraction, has arbitrarily complicated pattern point
Class ability and excellent multidimensional function mapping ability can cover all defect classification of bolt, improve the reliability of detection.
It is that the preferable of the present invention is implemented to be illustrated, but the present invention is not limited to the embodiment above, it is ripe
Various equivalent variations or replacement can also be made under the premise of without prejudice to spirit of the invention by knowing those skilled in the art, this
Equivalent deformation or replacement are all contained in the application claim limited range a bit.
Claims (10)
1. a kind of bolt detecting system, it is characterised in that:Including:
Training study module carries out feature extraction, and according to machine learning algorithm to the spy of extraction for tapping sound to bolt
Sign is trained study;
Detection module, for learning as a result, to bolt to be detected progress defects detection according to training.
2. a kind of bolt detecting system according to claim 1, it is characterised in that:The trained study module includes:
Sound collection unit taps sound for acquiring defective and flawless sample bolt;
Characteristic extracting module carries out feature extraction for tapping sound collection of illustrative plates to the bolt of acquisition using wavelet moment algorithm;
Feature training module, for being trained identification to the feature of extraction using BP neural network.
3. a kind of bolt detecting system according to claim 2, it is characterised in that:The characteristic extracting module includes:
Normalized unit is normalized for tapping sound collection of illustrative plates to the bolt of acquisition, is met simultaneously flat
The collection of illustrative plates of motion immovability and scaling invariance;
Map unit for the collection of illustrative plates after normalized to be mapped to polar coordinates, obtains the bianry image on polar coordinates;
Structural unit for carrying out moment characteristics construction using wavelet function collection, obtains wavelet moment invariants, the wavelet function collection
ψa,b(r) it is:
Wherein, a represents broadening factor, a ∈ R+, b represents shift factor, and b ∈ R, ψ (r) represent morther wavelet, and r represents bianry image
Polar diameter;
Characteristics of image generation unit, for according to obtained wavelet moment invariants, generating the local feature and global characteristics of collection of illustrative plates.
4. a kind of bolt detecting system according to claim 3, it is characterised in that:The described function of the morther wavelet ψ (r)
For cubic B-spline function.
5. a kind of bolt detecting system according to claim 2, it is characterised in that:The feature training module includes:
Forward transfer element calculates for the input value to neuron each in BP neural network, obtains error function;
Unit is reversely adjusted, for judging whether obtained error function meets threshold requirement, if so, not processing;Conversely,
The input value of each neuron is then corrected, and returns to forward transfer element.
6. a kind of bolt detecting system according to claim 5, it is characterised in that:The BP neural network includes input
Layer, hidden layer and output layer, the expression formula of the error function E are:
Wherein, m is number of samples, and p is the number of output layer neuron, tksThe expectation of s-th of neuron for k-th of sample is defeated
Go out value, yksThe real output value of s-th of neuron for k-th of sample.
7. a kind of implementation method of bolt detecting system, it is characterised in that:Include the following steps:
Bolt is acquired by bolt detecting system and taps sound, the bolt percussion sound progress feature extraction to acquisition, and according to
Machine learning algorithm is trained study to the feature of extraction;
The percussion sound of bolt to be detected is acquired by bolt detecting system, according to training study as a result, to the to be checked of acquisition
It surveys bolt and taps sound progress defects detection.
8. a kind of implementation method of bolt detecting system according to claim 7, it is characterised in that:The spiral shell of described pair of acquisition
Bolt taps sound and carries out feature extraction, and the step for be trained study to the feature of extraction according to machine learning algorithm, packet
Include following steps:
Defective and flawless sample bolt is acquired by bolt detecting system and taps sound;
Sound collection of illustrative plates is tapped to the bolt of acquisition using wavelet moment algorithm and carries out feature extraction;
Identification is trained to the feature of extraction using BP neural network.
9. a kind of implementation method of bolt detecting system according to claim 8, it is characterised in that:It is described to use wavelet moment
Algorithm taps the step for sound collection of illustrative plates carries out feature extraction to the bolt of acquisition, includes the following steps:
Sound collection of illustrative plates is tapped to the bolt of acquisition to be normalized, and is met translation invariance and scaling invariance simultaneously
Collection of illustrative plates;
Collection of illustrative plates after normalized is mapped on polar coordinates, obtains the bianry image on polar coordinates;
Moment characteristics construction is carried out using wavelet function collection, obtains wavelet moment invariants, the wavelet function collection ψa,b(r) it is:
Wherein, a represents broadening factor, a ∈ R+, b represents shift factor, and b ∈ R, ψ (r) represent morther wavelet, and r represents bianry image
Polar diameter;
According to obtained wavelet moment invariants, the local feature and global characteristics of collection of illustrative plates are generated.
10. a kind of implementation method of bolt detecting system according to claim 8, it is characterised in that:It is described refreshing using BP
The step for being trained identification to the feature of extraction through network, includes the following steps:
The input value of neuron each in BP neural network is calculated, obtains error function;
Whether the error function for judging to obtain meets threshold requirement, if so, not processing;Conversely, then correct each neuron
Input value, and return to the step for calculating to the input value of neuron each in BP neural network, obtain error function,
Until obtained error function meets threshold requirement.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711307933.5A CN108269249A (en) | 2017-12-11 | 2017-12-11 | A kind of bolt detecting system and its implementation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711307933.5A CN108269249A (en) | 2017-12-11 | 2017-12-11 | A kind of bolt detecting system and its implementation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108269249A true CN108269249A (en) | 2018-07-10 |
Family
ID=62772128
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711307933.5A Pending CN108269249A (en) | 2017-12-11 | 2017-12-11 | A kind of bolt detecting system and its implementation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108269249A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109236586A (en) * | 2018-08-08 | 2019-01-18 | 远景能源(江苏)有限公司 | Wind-driven generator bolt failure monitoring method based on sound collection |
CN109358605A (en) * | 2018-11-09 | 2019-02-19 | 电子科技大学 | Control system bearing calibration based on six rank B- spline wavelets neural networks |
CN110829885A (en) * | 2019-11-22 | 2020-02-21 | 温州大学 | Mechanical impedance matching control method of magnetostrictive precision driving device |
CN110935651A (en) * | 2019-11-28 | 2020-03-31 | 武汉科技大学 | Bolt sorting method and system |
CN112101301A (en) * | 2020-11-03 | 2020-12-18 | 武汉工程大学 | Good sound stability early warning method and device for screw water cooling unit and storage medium |
CN112507915A (en) * | 2020-12-15 | 2021-03-16 | 西安交通大学 | Method for identifying loosening state of bolt connection structure based on vibration response information |
CN115753059A (en) * | 2022-11-23 | 2023-03-07 | 郑州大学 | Device and method for detecting bolt looseness based on combination of hearing sense and vision sense |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101213589A (en) * | 2006-01-12 | 2008-07-02 | 松下电器产业株式会社 | Object sound analysis device, object sound analysis method, and object sound analysis program |
CN102590348A (en) * | 2012-01-19 | 2012-07-18 | 清华大学 | Method and system for detecting local impedance change of objects |
CN103439413A (en) * | 2013-08-07 | 2013-12-11 | 湘潭大学 | Acoustic emission signal analytical method for automatically identifying damage mode of thermal barrier coating |
CN105882687A (en) * | 2016-05-13 | 2016-08-24 | 南京雅信科技集团有限公司 | Method for analyzing categories of faults of point machines |
CN106841403A (en) * | 2017-01-23 | 2017-06-13 | 天津大学 | A kind of acoustics glass defect detection method based on neutral net |
-
2017
- 2017-12-11 CN CN201711307933.5A patent/CN108269249A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101213589A (en) * | 2006-01-12 | 2008-07-02 | 松下电器产业株式会社 | Object sound analysis device, object sound analysis method, and object sound analysis program |
CN102590348A (en) * | 2012-01-19 | 2012-07-18 | 清华大学 | Method and system for detecting local impedance change of objects |
CN103439413A (en) * | 2013-08-07 | 2013-12-11 | 湘潭大学 | Acoustic emission signal analytical method for automatically identifying damage mode of thermal barrier coating |
CN105882687A (en) * | 2016-05-13 | 2016-08-24 | 南京雅信科技集团有限公司 | Method for analyzing categories of faults of point machines |
CN106841403A (en) * | 2017-01-23 | 2017-06-13 | 天津大学 | A kind of acoustics glass defect detection method based on neutral net |
Non-Patent Citations (1)
Title |
---|
周晓敏,李应: "基于小波矩和BP网络的声音识别", 《计算机工程与应用》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109236586A (en) * | 2018-08-08 | 2019-01-18 | 远景能源(江苏)有限公司 | Wind-driven generator bolt failure monitoring method based on sound collection |
CN109358605A (en) * | 2018-11-09 | 2019-02-19 | 电子科技大学 | Control system bearing calibration based on six rank B- spline wavelets neural networks |
CN110829885A (en) * | 2019-11-22 | 2020-02-21 | 温州大学 | Mechanical impedance matching control method of magnetostrictive precision driving device |
US20210159386A1 (en) * | 2019-11-22 | 2021-05-27 | Wenzhou University | Matching control method for mechanical impedance of magnetostrictive precision transducer |
CN110829885B (en) * | 2019-11-22 | 2022-10-21 | 温州大学 | Mechanical impedance matching control method of magnetostrictive precision driving device |
US11778916B2 (en) * | 2019-11-22 | 2023-10-03 | Wenzhou University | Matching control method for mechanical impedance of magnetostrictive precision transducer |
CN110935651A (en) * | 2019-11-28 | 2020-03-31 | 武汉科技大学 | Bolt sorting method and system |
CN112101301A (en) * | 2020-11-03 | 2020-12-18 | 武汉工程大学 | Good sound stability early warning method and device for screw water cooling unit and storage medium |
CN112507915A (en) * | 2020-12-15 | 2021-03-16 | 西安交通大学 | Method for identifying loosening state of bolt connection structure based on vibration response information |
CN112507915B (en) * | 2020-12-15 | 2023-06-20 | 西安交通大学 | Bolt connection structure loosening state identification method based on vibration response information |
CN115753059A (en) * | 2022-11-23 | 2023-03-07 | 郑州大学 | Device and method for detecting bolt looseness based on combination of hearing sense and vision sense |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108269249A (en) | A kind of bolt detecting system and its implementation | |
CN105372087B (en) | Method for diagnosing faults based on multiple sensor signals analysis | |
CN110334764B (en) | Rotary machine intelligent fault diagnosis method based on integrated depth self-encoder | |
CN104483330B (en) | The dynamic assessment method of concrete surface crack real-time monitoring system and cracking risk | |
CN101751555B (en) | Deformation fingerprint identification method and system | |
CN109345507A (en) | A kind of dam image crack detection method based on transfer learning | |
Liu et al. | Data anomaly detection for structural health monitoring using a combination network of GANomaly and CNN | |
CN106841403A (en) | A kind of acoustics glass defect detection method based on neutral net | |
CN102158486A (en) | Method for rapidly detecting network invasion | |
CN104316167B (en) | Structural response method for detecting abnormality under a kind of low signal-to-noise ratio | |
CN109387565A (en) | A method of brake block internal flaw is detected by analysis voice signal | |
Teng et al. | Digital image correlation-based structural state detection through deep learning | |
Arbaoui et al. | Wavelet-based multiresolution analysis coupled with deep learning to efficiently monitor cracks in concrete | |
Moeinifard et al. | Lost vibration test data recovery using convolutional neural network: a case study | |
Mantawy et al. | Convolutional neural network based structural health monitoring for rocking bridge system by encoding time‐series into images | |
Naddaf-Sh et al. | Defect detection and classification in welding using deep learning and digital radiography | |
CN113570571A (en) | Industrial edge end power battery defect detection method and system | |
CN115791969A (en) | Jacket underwater crack detection system and method based on acoustic emission signals | |
Soleimani-Babakamali et al. | Zero-shot transfer learning for structural health monitoring using generative adversarial networks and spectral mapping | |
Fu et al. | Automatic bolt tightness detection using acoustic emission and deep learning | |
Mojtahedi et al. | A robust damage detection method developed for offshore jacket platforms using modified artificial immune system algorithm | |
Sony | Towards multiclass damage detection and localization using limited vibration measurements | |
CN106228132A (en) | Target identification method and Target Identification Unit | |
Dackermann | Vibration-based damage identification methods for civil engineering structures using artificial neural networks | |
Wang et al. | Comparative study of modal strain energy based damage localization methods for three-dimensional structure |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180710 |
|
RJ01 | Rejection of invention patent application after publication |