CN115458088A - Impact positioning and energy detection method and system based on convolutional neural network - Google Patents

Impact positioning and energy detection method and system based on convolutional neural network Download PDF

Info

Publication number
CN115458088A
CN115458088A CN202211114481.XA CN202211114481A CN115458088A CN 115458088 A CN115458088 A CN 115458088A CN 202211114481 A CN202211114481 A CN 202211114481A CN 115458088 A CN115458088 A CN 115458088A
Authority
CN
China
Prior art keywords
impact
energy
area
positioning
neural network
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
Application number
CN202211114481.XA
Other languages
Chinese (zh)
Inventor
王奕首
王明华
吴迪
卿新林
孙虎
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.)
Xiamen University
Original Assignee
Xiamen 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 Xiamen University filed Critical Xiamen University
Priority to CN202211114481.XA priority Critical patent/CN115458088A/en
Publication of CN115458088A publication Critical patent/CN115458088A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C60/00—Computational materials science, i.e. ICT specially adapted for investigating the physical or chemical properties of materials or phenomena associated with their design, synthesis, processing, characterisation or utilisation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N3/00—Investigating strength properties of solid materials by application of mechanical stress
    • G01N3/02—Details
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N3/00—Investigating strength properties of solid materials by application of mechanical stress
    • G01N3/30—Investigating strength properties of solid materials by application of mechanical stress by applying a single impulsive force, e.g. by falling weight
    • G01N3/303—Investigating strength properties of solid materials by application of mechanical stress by applying a single impulsive force, e.g. by falling weight generated only by free-falling weight
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • 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
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N2203/00—Investigating strength properties of solid materials by application of mechanical stress
    • G01N2203/02—Details not specific for a particular testing method
    • G01N2203/06—Indicating or recording means; Sensing means
    • G01N2203/067—Parameter measured for estimating the property
    • G01N2203/0676—Force, weight, load, energy, speed or acceleration
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/26—Composites

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Immunology (AREA)
  • Chemical & Material Sciences (AREA)
  • Pathology (AREA)
  • Software Systems (AREA)
  • Biochemistry (AREA)
  • General Engineering & Computer Science (AREA)
  • Analytical Chemistry (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Geometry (AREA)
  • Computer Hardware Design (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • Mathematical Physics (AREA)
  • Investigating Strength Of Materials By Application Of Mechanical Stress (AREA)

Abstract

The invention relates to the technical field of structural health monitoring, in particular to an impact location and energy detection method and system based on a convolutional neural network, wherein the impact location and energy detection method adopts a convolutional neural network model to perform rough location on an impact area, and accurately locates a specific impact position in the impact area through a DTW-based centroid weighting algorithm on the basis of the rough location; and the impact energy is also characterized by the energy of the impact response signal. The method has accurate positioning, high impact inversion efficiency and simple operation, not only can avoid the influence of the structural complexity on the positioning precision, but also can realize the effective control of the impact energy estimation error. The maintenance cost is reduced, and meanwhile, the positioning and detecting efficiency is effectively improved.

Description

Impact positioning and energy detection method and system based on convolutional neural network
Technical Field
The invention relates to the technical field of structural health monitoring, in particular to an impact positioning and energy detection method and system based on a convolutional neural network.
Background
The composite material has the advantages of high specific strength, large specific modulus, designable structure and the like, and has obvious advantages in the aspects of light-weight structure manufacture, structure operation and maintenance economy, advanced structure design and the like. And thus is widely used in the aerospace field. However, the composite material structure is inevitably impacted in the manufacturing, service and maintenance processes to cause various almost invisible damages inside the composite material, so that the bearing performance of the structure is degraded, and the in-service safety of aerospace equipment is seriously threatened. Therefore, the on-line impact monitoring of the composite material structure to ensure the safety of the equipment is an urgent engineering problem to be solved.
Conventional non-destructive testing techniques currently used for impact damage testing include ultrasonic scanning, eddy current, thermal imaging, and the like. These methods usually require point-by-point and area-by-area scanning measurements of the structure, and for large structures, the time cost and economic cost are high, and only off-line monitoring is possible, which makes it difficult to meet the on-board on-line monitoring requirements. The existing impact positioning method aiming at the complex composite material structure generally adopts a positioning method based on stress wave propagation speed, relates to time domain feature extraction of stress waves and speed compensation in different propagation directions, and has the problems of large workload and large influence on precision by the structure. The method based on the system model also relates to the processes of solving the transfer function and constructing the system model through the impulse response, and has the problems of low efficiency due to the fact that a large amount of mathematical calculations and parameter adjustments are needed to make the model convergent.
In addition, impact energy is used as another characteristic parameter for characterizing the impact damage of the composite material structure, and the detection of the impact energy is crucial to the evaluation of the structural state of the composite material. The prior art generally adopts the time history of establishing a response model to invert the impact load to estimate the impact energy, but for a rib-dense structure, the system identification model is quite complex.
Therefore, based on the requirements of the short plate and the engineering application in the prior art, a method for impact positioning and energy detection is urgently needed to perform online monitoring on the impact damage of the composite material structure more accurately, simply and effectively.
Disclosure of Invention
In order to solve the defects of inaccurate positioning, low impact inversion efficiency and complex operation of the impact positioning and energy detection method for the composite material structure in the prior art, the invention provides an impact positioning and energy detection method based on a convolutional neural network, which comprises the following steps:
and a region processing step, namely distributing M sensors for receiving external impact signals on the composite material structure to be detected, and dividing the surface of the composite material structure to be detected into at least M regions.
A sample acquisition step, wherein impact tests are respectively carried out on a plurality of training points and marking points of each area so as to acquire a first sample database and a second sample database which contain impact signals; and by a fixed energy E 0 Impacting the surface of the composite material structure and acquiring fixed energy E according to impact signals received by a plurality of sensors 0 Characteristic value under impact E s 。
A model training step, namely constructing a 1D-CNN neural network with input of (l, M) and output of (M, 1), wherein l is the signal length of each sensor; and dividing the first sample database into a training set and a testing set, and inputting the training set and the testing set into the 1D-CNN neural network for training and verification to obtain an impact event monitoring model.
A coarse positioning step, wherein if a certain impact event occurs, a plurality of sensors receive an impact signal S impact (ii) a Will impact signal S impact Inputting the data into an impact event monitoring model, and determining an impact area where the impact event occurs according to an output result to finish coarse positioning。
A fine positioning step, namely calculating an impact signal S according to a DTW algorithm based on the impact area determined by the coarse positioning impact DTW distance L from all mark points in the impact area in the second sample database i (ii) a The obtained L is i And calculating the impact position coordinates (x, y) in the impact area as the weight factors of the weighted centroid positioning algorithm to complete fine positioning.
An energy detection step of detecting an impact signal S received by sensors in all the regions impact Obtaining an energy characterizing value E s ', the energy E of the impact is obtained by calculation according to the following formula est ：
Figure BDA0003844887030000031
In an embodiment, in the area processing step, area numbering is performed on each of the areas to obtain a signal tag; in the sample acquisition step, the impact signals in each area are respectively mapped into corresponding signal labels for storage.
In an embodiment, in the sample acquiring step, the sample acquiring step of the first sample database includes: selecting N (N is more than or equal to 1) training points in each region to respectively carry out impact tests to obtain corresponding impact signals, and mapping the impact signals on each mark in each region into corresponding signal labels to be stored in a first sample database;
the sample acquisition step of the second sample database comprises the following steps: b marking points are selected in each area to respectively carry out impact tests to obtain corresponding impact signals, and then the impact signals obtained by the impact tests of the marking points in each area are used as samples to be stored.
In one embodiment, in the model training step, the samples in the first sample database are divided into a training set and a testing set according to a ratio of 7: 3.
In one embodiment, in the fine positioning step, the resulting L is used i As weighted centroid localizationThe calculation of the impact position coordinates (x, y) in the impact region by the weighting factors of the algorithm specifically comprises the following formula:
Figure BDA0003844887030000032
Figure BDA0003844887030000033
wherein, W i Representing the weighting coefficients of all the marking points in the impact area; x is the number of i ，y i Respectively representing the coordinate positions of all the mark points in the impact area; and x and y are the impact position coordinates of the impact event.
In one embodiment, the sample acquiring step acquires the fixed energy E by the following formula 0 Characteristic value under impact E s ：
Figure BDA0003844887030000041
Figure BDA0003844887030000042
Wherein f (t) is an expression of an impact signal obtained by a sensor, and E is an energy value corresponding to the impact signal; n is the number of sensors arranged on the structure of the composite material to be detected, E i The energy value of the impact response signal corresponding to the ith sensor;
in the energy detection step, an energy characteristic value E at the occurrence of an impact event is obtained by the following formula s ′：
Figure BDA0003844887030000043
Wherein E is i ' is the energy value of the impact response signal corresponding to the ith sensor at the time of the impact event.
In one embodiment, the method further comprises an energy correction coefficient obtaining step, wherein the fixed energy E is carried out 0 After the test of impacting the surface of the composite material structure, selecting one area as a reference area, and acquiring an energy characterization value E of the reference area R Calculating an energy correction factor R for each region relative to the reference region i The formula is as follows:
Figure BDA0003844887030000044
Figure BDA0003844887030000045
wherein n' is the number of sensors in the reference area, E i′ The energy value of the impact response signal is corresponding to the ith' sensor in the reference area.
In one embodiment, the method further comprises an energy correction step, wherein the impact area determined based on the coarse positioning is corrected according to the energy correction coefficient R i Acquiring an energy correction coefficient R of the impact area relative to a reference area, and acquiring a corrected impact energy value E through the following formula impact ：
E impact ＝E est ·R。
The invention also provides an impact positioning and energy detection system based on the convolutional neural network, which comprises a region processing module, a detection module and a control module, wherein the region processing module is used for distributing a plurality of sensors for receiving external impact signals on the composite material structure to be detected and dividing the surface of the composite material structure to be detected into at least M regions;
the sample acquisition module is used for respectively carrying out impact tests on a plurality of training points and marking points of each area so as to acquire a first sample database and a second sample database which contain impact signals; and by a fixed energy E 0 Impacting the surface of the composite material structure and acquiring fixed energy E according to impact signals received by a plurality of sensors 0 Characteristic value E under impact s ；
The model training module is used for constructing a 1D-CNN neural network with input of (1, M) and output of (M, 1), wherein l is the signal length of each sensor; dividing the first sample database into a training set and a testing set, and inputting the training set and the testing set into a 1D-CNN neural network for training and verification to obtain an impact event monitoring model;
a coarse positioning module, wherein if a certain impact event occurs, a plurality of sensors receive an impact signal S impact (ii) a For generating an impact signal S impact Inputting the impact event into an impact event monitoring model, and determining an impact area where the impact event occurs according to an output result to finish coarse positioning;
a fine positioning module for calculating the impact signal S according to the DTW algorithm based on the impact region determined by the coarse positioning impact DTW distance L between the first sample database and all the mark points in the impact area in the second sample database i (ii) a Subjecting the obtained L to i Calculating impact position coordinates (x, y) in the impact area as a weight factor of a weighted centroid positioning algorithm to complete fine positioning;
an energy detection module for detecting the impact signal S received by the sensors in all the regions impact Obtaining an energy characterizing value E s ', the energy E of the impact is calculated according to the following formula est ：
Figure BDA0003844887030000051
In one embodiment, the energy correction module is further included for fixing the energy E 0 After the test of impacting the surface of the composite material structure, selecting one area as a reference area, and acquiring an energy characterization value E of the reference area R Calculating an energy correction factor R for each region relative to the reference region i ；
Based on the impact area determined by the coarse positioning, correcting the coefficient R according to the energy i Acquiring an energy correction coefficient R of the impact area relative to a reference area, and acquiring a corrected impact energy value E through the following formula impact ：
E impact ＝E est ·R。
Based on the above, compared with the prior art, the impact location and energy detection method based on the convolutional neural network provided by the invention adopts the convolutional neural network model to perform rough location on the impact area, and accurately locates the specific impact position in the impact area through the DTW-based centroid weighting algorithm on the basis of the rough location; and the impact energy is also characterized by the energy of the response signal. The impact energy estimation method can avoid the influence of the structural complexity on the positioning precision and can also realize the effective control of the impact energy estimation error. The method effectively improves the positioning and detecting efficiency while reducing the maintenance cost.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts; in the following description, the drawings are illustrated in a schematic view, and the drawings are not intended to limit the present invention.
FIG. 1 is a flow chart illustrating the steps of a method for impact location and energy detection according to the present invention;
FIG. 2 is a flow chart of a coarse positioning step;
FIG. 3 is a schematic diagram of a DTW calculation process;
FIG. 4 is a flow chart of a fine positioning step;
FIG. 5 is a flow chart of the energy detection steps;
FIG. 6 is a flowchart of the energy detection step and the energy correction step;
FIG. 7 is a schematic view of an impact location and energy detection system provided by the present invention;
FIG. 8 is a schematic structural view of a zone treatment step for a composite stiffened panel;
FIG. 9 is a diagram illustrating the training results of the 1D-CNN model;
FIG. 10 is a selection of locations of marker points within a region;
FIG. 11 is a schematic diagram of the coarse positioning result;
FIG. 12 is a schematic illustration of the fine positioning results;
FIG. 13 is a diagram illustrating comparison of impact energy estimation errors before and after zone correction compensation.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments; the technical features devised in the different embodiments of the invention described below can be combined with each other as long as they do not conflict with each other; all other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the description of the present invention, it should be noted that all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs, and should not be construed as limiting the present invention; it will be further understood that terms used herein should be interpreted as having a meaning that is consistent with their meaning in the context of this specification and the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The method aims to solve the problems that in the prior art, the positioning result is greatly influenced by the structure, the system modeling workload is large, and the impact inversion efficiency is low by aiming at a complex composite material structure impact positioning and impact energy detection method. Referring to fig. 1, the present invention provides an impact localization and energy detection method based on convolutional neural network, which at least includes the following steps:
and a region processing step, namely distributing M sensors for receiving external impact signals on the composite material structure to be detected, and dividing the surface of the composite material structure to be detected into at least M regions.
It should be noted that the sensors may be piezoelectric sensors, acoustic emission sensors, and the like, and the number of the sensors may be set according to actual requirements, which is not limited herein. Preferably, the surface of the composite structure can be divided into M regions of equal or substantially equal area for ease of subsequent calculation and analysis.
A sample acquisition step, wherein impact tests are respectively carried out on a plurality of training points and marking points of each area so as to acquire a first sample database and a second sample database which contain impact signals; and by a fixed energy E 0 Impacting the surface of the composite material structure and acquiring fixed energy E according to impact signals received by a plurality of sensors 0 Characteristic value under impact E s 。
In specific implementation, the sample acquisition step of the first sample database is as follows: selecting N (N is more than or equal to 1) training points in each region to respectively carry out impact tests to obtain corresponding impact signals, and mapping the impact signals on the training points in each region into corresponding signal labels to be stored in a first sample database. The mapping relationship of the first sample database can be represented by referring to the following table:
Figure BDA0003844887030000081
Figure BDA0003844887030000091
the sample acquisition step of the second sample database comprises the following steps: b marking points are selected in each area to respectively carry out impact tests to obtain corresponding impact signals, and then the impact signals obtained by the impact tests of the marking points in each area are used as samples to be stored. The mapping relationship of the second sample database can be expressed by referring to the following table:
Figure BDA0003844887030000092
a model training step, namely constructing a 1D-CNN neural network with input of (l, M) and output of (M, 1), wherein l is the signal length of each sensor, M is the number of the sensors, and M is the number of areas divided by the composite material structure; and dividing the first sample database into a training set and a testing set, and inputting the training set and the testing set into the 1D-CNN neural network for training and verification to obtain an impact event monitoring model. It should be noted that, the training and verification method for the convolutional neural network is more conventional in the prior art, and details are not repeated herein. Preferably, the samples in the first sample database can be divided into a training set and a test set according to the ratio of 7.
A coarse positioning step, wherein if a certain impact event occurs, a plurality of sensors receive an impact signal S impact (ii) a Will impact signal S impact Inputting the impact event into an impact event monitoring model, and determining an impact area where the impact event occurs according to an output result so as to finish coarse positioning; that is, the signal output by the neural network is labeled as the impact region where the impact event occurs.
A fine positioning step, referring to fig. 4, based on the impact region determined by the coarse positioning, calculating an impact signal S according to the DTW algorithm impact DTW distance L from all mark points in the impact area in the second sample database i (ii) a Subjecting the obtained L to i And calculating the impact position coordinates (x, y) in the impact area as the weight factors of the weighted centroid positioning algorithm to complete fine positioning.
In specific implementation, this embodiment obtains the weighted centroid location algorithm (WCA for short) based on the coarse locationAnd an accurate impact position in the impact area is obtained, and the method has the characteristics of simple operation and strong adaptability. The specific steps are that a rectangular coordinate system is constructed for each region, and then an impact signal S is calculated through a DTW algorithm impact DTW distance L from all mark points in the impact area in the second sample database i . In this embodiment, a DTW (Dynamic Time Warping) algorithm is applied to the fine positioning of the impulse positioning, and mainly compares the similarity of the Time sequences of two impulse signals, that is, calculates the euclidean distance between the two Time sequences. The specific calculation process is shown in fig. 3, taking time sequence X and time sequence Y as an example, on the premise of aligning time sequence X and time sequence Y, calculating a distance matrix between each point of the two sequences, and then finding the minimum value of the sum of path distances calculated according to the DTW path mapping relationship in the path consisting of p 1-pk, where the path distance is DTW distance L i Wherein the expression is: l is a radical of an alcohol i ＝DTW(S impact ，S i ) I =1 to B; wherein L is i Indicating the impact signal S generated by the impact location impact Impact signal S with reference mark point i DTW distance between; s impact An impact signal representing an actual impact location; s. the i An impact signal representing the ith marker point within the identified impact region.
Finally, the obtained DTW distance L is further used i The weight factor used as the weighted centroid locating algorithm is used for calculating the impact position coordinates (x, y) in the impact area, and specifically includes the following formula:
Figure BDA0003844887030000101
Figure BDA0003844887030000102
wherein, W i Representing the weighting coefficients of all the marking points in the impact area; x is a radical of a fluorine atom i ，y i Respectively representing the coordinate positions of all the marking points in the impact area; x and y are the impact position coordinates of the impact event。
By the above-mentioned pair DTW distance L i The impact position coordinate obtained by calculation after weighting has higher positioning precision and can accurately position the impact position.
Step of energy detection, please refer to fig. 5, according to the impact signal S received by the sensors in all the regions impact Obtaining an energy characterization value E s ', the energy E of the impact is calculated according to the following formula est ：
Figure BDA0003844887030000111
In the formula, E 0 And E s The obtaining of (a) can be obtained by the sample obtaining step, namely: passing a fixed energy E in advance 0 Impacting the surface of the composite material structure and acquiring fixed energy E according to impact signals received by a plurality of sensors 0 Characteristic value E under impact s The fixed energy E is obtained by the following formula 0 Characteristic value under impact E s ：
Figure BDA0003844887030000112
Figure BDA0003844887030000113
Wherein f (t) is an expression of an impact signal obtained by a sensor, and E is an energy value corresponding to the impact signal; n is the number of sensors arranged on the composite material structure to be detected, E i The energy value of the impact response signal corresponding to the ith sensor; e i Can be obtained by calculation according to the formula of E.
In the formula, the energy characteristic value E when the impact event occurs is obtained by the following formula s ′：
Figure BDA0003844887030000114
Wherein, E i ' is the energy value of the impact response signal corresponding to the ith sensor at the time of the impact event.
Finally, E is 0 、E s 、E s ' substitution into the above-mentioned energy E for this impact est The impact energy of the impact area can be effectively estimated by the formula (2).
In order to generalize the above method to other impact regions, the present invention further provides a region compensation correction method for impact energy, please refer to fig. 6, which specifically includes an energy correction coefficient obtaining step for performing a fixed energy E 0 After the test of impacting the surface of the composite material structure, selecting one area as a reference area, and acquiring an energy characterization value E of the reference area R Calculating an energy correction factor R for each region relative to the reference region i The formula is as follows:
Figure BDA0003844887030000121
Figure BDA0003844887030000122
wherein n' is the number of sensors in the reference area, E i′ The energy value of the impact response signal is corresponding to the ith sensor in the reference region.
An energy correction step of correcting the coefficient R according to the energy based on the impact area determined by the coarse positioning i Acquiring an energy correction coefficient R of the impact area relative to a reference area, and acquiring a corrected impact energy value E through the following formula impact ：
E impact ＝E est ·R。
In the method for detecting the energy, the impact response signal is adopted to represent the impact energy, and the impact response signal is corrected according to the regional characteristics, so that an accurate impact energy value is obtained.
The invention also provides an impact positioning and energy detection system based on the convolutional neural network, and please refer to fig. 7, which includes a region processing module, configured to distribute M sensors for receiving external impact signals on a composite material structure to be detected, and divide the surface of the composite material structure to be detected into at least M regions;
the sample acquisition module is used for respectively carrying out impact tests on a plurality of training points and marking points of each area so as to acquire a first sample database and a second sample database which contain impact signals; and by fixing the energy E 0 Impacting the surface of the composite material structure and acquiring fixed energy E according to impact signals received by a plurality of sensors 0 Characteristic value E under impact s ；
The model training module is used for constructing a 1D-CNN neural network with input of (l, M) and output of (M, 1), wherein l is the signal length of each sensor; dividing the first sample database into a training set and a testing set, and inputting the training set and the testing set into a 1D-CNN neural network for training and verification to obtain an impact event monitoring model;
a coarse positioning module, wherein if a certain impact event occurs, a plurality of sensors receive an impact signal S impact (ii) a For generating an impact signal S impact Inputting the impact event into an impact event monitoring model, and determining an impact area where the impact event occurs according to an output result so as to finish coarse positioning;
a fine positioning module for calculating the impact signal S according to the DTW algorithm based on the impact region determined by the coarse positioning impact DTW distance L between the first sample database and all the mark points in the impact area in the second sample database i (ii) a Subjecting the obtained L to i Calculating impact position coordinates (x, y) in the impact area as a weight factor of a weighted centroid positioning algorithm to complete fine positioning;
an energy detection module for detecting impact signals S received by the sensors in all regions impact Obtaining an energy characterization value E s ', the energy E of the impact is calculated according to the following formula est ：
Figure BDA0003844887030000131
Preferably, an energy correction module is also included for fixing the energy E 0 After the test of impacting the surface of the composite material structure, selecting one area as a reference area, and acquiring an energy characterization value E of the reference area R Calculating an energy correction factor R for each region relative to the reference region i ；
Based on the impact area determined by coarse positioning, the coefficient R is corrected according to energy i Acquiring an energy correction coefficient R of the impact area relative to a reference area, and acquiring a corrected impact energy value E through the following formula impact ：
E impact ＝E est ·R。
In order to better explain the impact location and energy detection method based on the convolutional neural network, the implementation process of the method is described in detail by taking a specific composite material stiffened plate as an example:
(1) Dividing a monitoring area (360 mm multiplied by 360 mm) of a composite material stiffened plate with the size of 480mm multiplied by 480mm into rectangular areas with 16 areas equal to each other, as shown in figure 8, and numbering by letters A-P; piezoelectric sensors are arranged at the intersections of four boundaries of the monitoring area.
(2) A first sample database containing 16 x 10 sets of impulse signals was constructed by selecting 10 training points per area for 10 random impulse tests and using the area number as the signal label of the signal.
A rectangular coordinate system is constructed in each region, 5 marking points are selected for carrying out an impact test, and a second sample database for precise positioning is constructed, as shown in fig. 10, 5 marking points which are uniformly distributed in the region are selected, wherein the marking points are respectively No.1, no.2, no.3, no.4 and No.5.
(3) And constructing a 1D-CNN neural network with inputs of (1600, 4) and outputs of (16, 1), wherein 1600 is the signal length of each sensor, 4 is the number of the sensors, and 16 is the number of areas divided by the monitored structure. And (3) dividing the samples in the first sample database into a training set and a testing set according to the standard of 7, and respectively using the training set and the testing set for the model generation training and the model training effect verification process of the built 1D-CNN network. As shown in fig. 9, it can be seen that as the training period increases, the loss function of the model gradually decreases, the accuracy gradually increases, the model converges after approximately 35 training periods, and the prediction accuracy reaches 100%.
(3) Impact tests were randomly performed 5 times per area to verify the reliability of the identification of the impact area. Namely, according to the coarse positioning step provided by the invention, the impact area identification is carried out by adopting the trained impact event monitoring model. The results of the coarse positioning are shown in fig. 11, with the abscissa representing the actual impact region and the ordinate representing the identified impact position. According to the graph 11, the method provided by the invention can realize effective identification of the impact area, the accuracy is as high as 98.75% (79/80), and the only identification error is that the impact of the B area is identified as the F area.
(4) According to the precise positioning step provided by the invention, the impact position of the impact area is identified by adopting a DTW algorithm and a weighted centroid positioning algorithm. The precise positioning result is shown in fig. 12, and from the structural distribution of the positioning result, the impact position identified by the centroid weighted positioning algorithm is very close to the actual position, which further proves the effectiveness and accuracy of the positioning method provided by the present invention.
(5) As shown in fig. 8, the correction coefficient of each area was calculated by selecting a free fall impact of a rubber ball having a height of 40mm for each area with reference to the area G. Then, the rubber ball with the height of 80mm is subjected to free-falling body impact in the area G, and impact energy detection and correction are carried out according to the energy detection steps provided by the invention. The energy detection results are shown in fig. 13, and it can be seen from fig. 13 that the average error of the impact energy estimation before the correction and compensation is 61.46%, and the estimation error of the individual regions, such as a, N, I, and P, is even more than 60%. After correction and compensation, the overall average error is only 8.95%, although the average error after compensation in individual regions is slightly larger than the average error before compensation, the overall energy estimation method after error compensation has more stable estimation results, and can continuously output results with high reliability. Therefore, the energy detection step provided by the invention has the advantages of high stability and lower average error, which is more significant for practical application scenarios.
It should also be noted that the method and system provided by the invention are not only suitable for composite material flat plates, but also suitable for complex structures with dense ribs, such as fuselage panels, wing skins and the like of aircrafts. When being applied to aircraft structure and maintaining, can inject maintenance work in limited within range, can provide certain priori knowledge for combined material's damage condition through energy estimation moreover, when reducing the maintenance cost, improve maintenance efficiency.
In summary, compared with the prior art, the impulse positioning and energy detection method and system based on the convolutional neural network have the characteristics of accurate positioning, high impulse inversion efficiency and simplicity in operation. The method can not only avoid the influence of the structural complexity on the positioning precision, but also realize the effective control of the impact energy estimation error. The method has the advantages that the maintenance cost is reduced, the positioning and detecting efficiency is effectively improved, and the method has wide application prospect.
In addition, it will be appreciated by those skilled in the art that, although there may be many problems with the prior art, each embodiment or aspect of the present invention may be improved only in one or several respects, without necessarily simultaneously solving all the technical problems listed in the prior art or in the background. It will be understood by those skilled in the art that nothing in a claim should be taken as a limitation on that claim.
Although terms such as sensor, training point, marking point, first sample database, second sample database, impulse signal, DTW algorithm, 1D-CNN neural network, weighted centroid locating algorithm, etc. are used more herein, the possibility of using other terms is not excluded. These terms are used merely to more conveniently describe and explain the nature of the present invention; they are to be construed as being without limitation to any additional limitations that may be imposed by the spirit of the present invention; the terms "first," "second," and the like in the description and in the claims, and in the drawings, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims (10)

1. An impact positioning and energy detection method based on a convolutional neural network is characterized by comprising the following steps:
the method comprises the following steps of area processing, wherein M sensors for receiving external impact signals are distributed on a composite material structure to be detected, and the surface of the composite material structure to be detected is divided into at least M areas;
a sample acquisition step, wherein impact tests are respectively carried out on a plurality of training points and marking points of each area so as to acquire a first sample database and a second sample database which contain impact signals; and by fixing the energy E 0 Impacting the surface of the composite material structure and acquiring fixed energy E according to impact signals received by a plurality of sensors 0 Characteristic value E under impact s ；
A model training step, namely constructing a 1D-CNN neural network with input of (l, M) and output of (M, 1), wherein l is the signal length of each sensor; dividing the first sample database into a training set and a testing set, and inputting the training set and the testing set into a 1D-CNN neural network for training and verification to obtain an impact event monitoring model;
a coarse positioning step, wherein if a certain impact event occurs, a plurality of sensors are usedReceiving an impact signal S impact (ii) a Will impact signal S impact Inputting the impact event into an impact event monitoring model, and determining an impact area where the impact event occurs according to an output result to finish coarse positioning;
a fine positioning step, namely calculating an impact signal S according to a DTW algorithm based on the impact area determined by the coarse positioning impact DTW distance L from all mark points in the impact area in the second sample database i (ii) a Subjecting the obtained L to i Calculating impact position coordinates (x, y) in the impact area as a weight factor of a weighted centroid positioning algorithm to complete fine positioning;
an energy detection step of detecting an impact signal S received by sensors in all the regions impact Obtaining an energy characterization value E s ', the energy E of the impact is calculated according to the following formula est ：
Figure FDA0003844887020000011
2. The convolutional neural network-based impulse positioning and energy detection method of claim 1, wherein: in the area processing step, area numbering is carried out on each area to obtain a signal label; in the sample acquisition step, the impact signals in each area are respectively mapped into corresponding signal labels for storage.
3. The convolutional neural network-based impulse positioning and energy detection method as claimed in claim 2, wherein in the sample acquiring step, the sample acquiring step of the first sample database is: selecting N (N is more than or equal to 1) training points in each region to respectively carry out impact tests to obtain corresponding impact signals, and mapping the impact signals on the training points in each region into corresponding signal labels to be stored in a first sample database;
the sample acquisition step of the second sample database comprises the following steps: b marking points are selected in each area to respectively carry out impact tests to obtain corresponding impact signals, and then the impact signals obtained by the impact tests of the marking points in each area are used as samples to be stored.
4. The convolutional neural network-based impulse positioning and energy detection method of claim 1, wherein: in the model training step, samples in the first sample database are divided into a training set and a test set according to the proportion of 7.
5. The convolutional neural network-based impulse positioning and energy detection method of claim 1, wherein in the fine positioning step, the obtained L is measured i The calculation of the impact position coordinates (x, y) in the impact region as the weighting factors of the weighted centroid locating algorithm specifically includes the following formula:
Figure FDA0003844887020000021
Figure FDA0003844887020000022
wherein, W i Weighting coefficients representing all marker points within the impact area; x is a radical of a fluorine atom i ,y i Respectively representing the coordinate positions of all the mark points in the impact area; and x and y are the impact position coordinates of the impact event.
6. The convolutional neural network-based impulse positioning and energy detection method of claim 1, wherein in the sample acquiring step, the fixed energy E is acquired by the following formula 0 Characteristic value E under impact s ：
Figure FDA0003844887020000031
Figure FDA0003844887020000032
Wherein f (t) is an expression of an impact signal obtained by a sensor, and E is an energy value corresponding to the impact signal; n is the number of sensors arranged on the structure of the composite material to be detected, E i The energy value of the impact response signal corresponding to the ith sensor;
in the energy detection step, an energy characteristic value E at the occurrence of an impact event is obtained by the following formula s ′：
Figure FDA0003844887020000033
Wherein E is i ' is the energy value of the impact response signal corresponding to the ith sensor at the time of the impact event.
7. The convolutional neural network-based impulse positioning and energy detecting method of claim 6, further comprising an energy correction coefficient obtaining step of performing a fixed energy E 0 After the test of the impact composite material structure surface, selecting one area as a reference area, and acquiring an energy characterization value E of the reference area R Calculating an energy correction factor R for each region relative to the reference region i The formula is as follows:
Figure FDA0003844887020000034
Figure FDA0003844887020000035
wherein n' is the number of sensors in the reference area, E i′ For the ith' sensor in the reference regionCorresponding to the energy value of the impulse response signal.
8. The convolutional neural network-based impulse positioning and energy detecting method as claimed in claim 7, further comprising an energy correction step of determining an impulse region based on the coarse positioning according to an energy correction coefficient R i Acquiring an energy correction coefficient R of the impact area relative to a reference area, and acquiring a corrected impact energy value E through the following formula impact ：
E impact ＝E est ·R。
9. The utility model provides an impact location and energy detection system based on convolutional neural network which characterized in that: comprises that
The area processing module is used for distributing a plurality of sensors for receiving external impact signals on the composite material structure to be detected and dividing the surface of the composite material structure to be detected into at least M areas;
the sample acquisition module is used for respectively carrying out impact tests on a plurality of training points and marking points of each region so as to acquire a first sample database and a second sample database which contain impact signals; and by fixing the energy E 0 Impacting the surface of the composite material structure and acquiring fixed energy E according to impact signals received by a plurality of sensors 0 Characteristic value under impact E s ；
The model training module is used for constructing a 1D-CNN neural network with input of (l, M) and output of (M, 1), wherein l is the signal length of each sensor; dividing the first sample database into a training set and a testing set, and inputting the training set and the testing set into a 1D-CNN neural network for training and verification to obtain an impact event monitoring model;
a coarse positioning module, wherein if a certain impact event occurs, a plurality of sensors receive an impact signal S impact (ii) a For generating an impact signal S impact Inputting the impact event into an impact event monitoring model, and determining an impact area where the impact event occurs according to an output result to finish coarse positioning;
fine settingA bit module for calculating an impact signal S according to a DTW algorithm based on the impact region determined by the coarse positioning impact DTW distance L between the first sample database and all the mark points in the impact area in the second sample database i (ii) a Subjecting the obtained L to i Calculating impact position coordinates (x, y) in the impact area as a weight factor of a weighted centroid positioning algorithm to complete fine positioning;
an energy detection module for detecting the impact signal S received by the sensors in all the regions impact Obtaining an energy characterizing value E s ', the energy E of the impact is calculated according to the following formula est ：
Figure FDA0003844887020000041
10. The convolutional neural network-based impulse location and energy detection device of claim 9, wherein: also comprises an energy correction module used for fixing the energy E 0 After the test of the impact composite material structure surface, selecting one area as a reference area, and acquiring an energy characterization value E of the reference area R Calculating an energy correction factor R for each region relative to the reference region i ；
Based on the impact area determined by the coarse positioning, correcting the coefficient R according to the energy i Acquiring an energy correction coefficient R of the impact area relative to a reference area, and acquiring a corrected impact energy value E through the following formula impact ：
E impact ＝E est ·R。
CN202211114481.XA 2022-09-14 2022-09-14 Impact positioning and energy detection method and system based on convolutional neural network Pending CN115458088A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211114481.XA CN115458088A (en) 2022-09-14 2022-09-14 Impact positioning and energy detection method and system based on convolutional neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211114481.XA CN115458088A (en) 2022-09-14 2022-09-14 Impact positioning and energy detection method and system based on convolutional neural network

Publications (1)

Publication Number Publication Date
CN115458088A true CN115458088A (en) 2022-12-09

Family

ID=84302490

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211114481.XA Pending CN115458088A (en) 2022-09-14 2022-09-14 Impact positioning and energy detection method and system based on convolutional neural network

Country Status (1)

Country Link
CN (1) CN115458088A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116428129A (en) * 2023-06-13 2023-07-14 山东大学 Fan blade impact positioning method and system based on attention mixing neural network

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116428129A (en) * 2023-06-13 2023-07-14 山东大学 Fan blade impact positioning method and system based on attention mixing neural network
CN116428129B (en) * 2023-06-13 2023-09-01 山东大学 Fan blade impact positioning method and system based on attention mixing neural network

Similar Documents

Publication Publication Date Title
US20200271756A1 (en) System and method for identification of an airborne object
CN104155650B (en) A kind of method for tracking target based on entropy weight method point mark quality evaluation
CN107247259B (en) K distribution sea clutter shape parameter estimation method based on neural network
CN113191087B (en) Navigation type depth measurement data profile sound velocity correction method combining historical profile sound velocity and actually measured surface layer sound velocity
CN111680870B (en) Comprehensive evaluation method for quality of target motion trail
CN105844055B (en) It is fissioned-is polymerize the damage monitoring method of probabilistic model based on guided wave dynamic contract-enhanced
CN104463876B (en) Adaptive-filtering-based rapid multi-circle detection method for image under complex background
CN111308912B (en) Credibility evaluation method of anti-radiation unmanned aerial vehicle guidance semi-physical simulation system
CN110879927B (en) Sea clutter amplitude statistical distribution on-site modeling method for sea target detection
CN109632963A (en) It is a kind of based on when invariant features signal building structural damage four-dimensional imaging method
CN115458088A (en) Impact positioning and energy detection method and system based on convolutional neural network
CN111598170A (en) Crack detection probability evaluation method considering model selection uncertainty
CN110780128A (en) Sensitive equipment voltage sag fault probability evaluation method
CN112083413A (en) Radar wave stealth weapon equipment maintenance test method
CN116542912A (en) Flexible body bridge vibration detection model with multi-target visual tracking function and application
CN115326935A (en) Impact positioning method based on convolutional neural network and centroid weighting, readable storage medium and device
CN115497581A (en) Impact positioning and energy detection method and system based on dynamic time warping
Wei et al. Modified reconstruction algorithm for probabilistic inspection of damage based on damaged virtual sensing paths
CN114383834B (en) Ocean engineering structure micro damage judging method
CN112729742B (en) Impact positioning method based on probability imaging
CN114266776A (en) Digital image correlation method applying composite crack displacement field function
CN110850366B (en) Positioning method based on received signal strength under mixed line-of-sight and non-line-of-sight environment
Wang et al. Identification of engine foreign object impact based on acoustic emission and radical basis function neural network
CN108983205A (en) A kind of first value-acquiring method of filtering when measuring unreliable
CN112381792A (en) Radar wave-absorbing coating/electromagnetic shielding film damage intelligent imaging online detection method based on deep learning

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