CN117907445A - Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network - Google Patents

Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network Download PDF

Info

Publication number
CN117907445A
CN117907445A CN202410049859.5A CN202410049859A CN117907445A CN 117907445 A CN117907445 A CN 117907445A CN 202410049859 A CN202410049859 A CN 202410049859A CN 117907445 A CN117907445 A CN 117907445A
Authority
CN
China
Prior art keywords
guided wave
damage
ultrasonic guided
space
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202410049859.5A
Other languages
Chinese (zh)
Other versions
CN117907445B (en
Inventor
武湛君
田童
杨雷
杨正岩
张佳奇
马书义
高东岳
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.)
Dalian University of Technology
Original Assignee
Dalian University of Technology
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 Dalian University of Technology filed Critical Dalian University of Technology
Priority to CN202410049859.5A priority Critical patent/CN117907445B/en
Publication of CN117907445A publication Critical patent/CN117907445A/en
Application granted granted Critical
Publication of CN117907445B publication Critical patent/CN117907445B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/04—Analysing solids
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4481—Neural networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • 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
    • G06N3/0464—Convolutional networks [CNN, ConvNet]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Pathology (AREA)
  • Biochemistry (AREA)
  • Analytical Chemistry (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Chemical & Material Sciences (AREA)
  • Immunology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Acoustics & Sound (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Signal Processing (AREA)
  • Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)

Abstract

The invention discloses a composite material stiffened plate damage identification method based on an ultrasonic guided wave and space-time hybrid network, which mainly comprises the steps of establishing a data set, establishing a damage identification model based on a space-time parallel hybrid network model, simultaneously learning time sequence characteristics and space characteristics of signals, learning the guided wave high-dimensional characteristics through a residual block, improving the information sensitivity of the model by utilizing a convolution attention module, and finally establishing a mapping relation between guided wave signals and damage coordinates. Test results show that the method can learn more guided wave signal characteristics, the generalization capability of the model can be better improved by adding the loss function, and the composite material damage identification technology combining deep learning and ultrasonic guided wave has certain advantages.

Description

Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network
Technical Field
The invention relates to the field of structural health monitoring, in particular to a method and a system for identifying damage to a reinforcing plate of a composite material.
Background
With the continuous advancement of technology, structural health monitoring technology has become a key tool in many fields. The technology can monitor the quality and damage condition of structural materials such as airplanes, railway tracks, bridges and the like in real time, and provides important basis for the use and maintenance of the structures. In particular in the field of aviation, carbon fiber reinforced composite materials are widely applied to main bearing parts of aircrafts due to the advantages of high specific strength and specific stiffness, excellent fatigue resistance, customizable performance design, easiness in integral molding and the like. However, composite structures are subject to external impacts, fatigue loads, and environmental erosion during aircraft service, which may lead to reduced load carrying capacity of the structure, material performance degradation, and potentially invisible damage. In order to ensure the safety of the structure, and simultaneously reduce the resource waste to the greatest extent, a timely and effective health monitoring technology becomes important.
The structural health monitoring system adopts various physical principles to detect damage, including ultrasonic guided wave, acoustic emission, fiber bragg grating, electromechanical impedance measurement and the like. Among these techniques, the ultrasonic guided wave technique is a particularly suitable method for lesion monitoring, since guided waves are affected by the traversing medium, which means that even invisible lesions can be detected. Conventional lesion recognition methods generally rely on techniques such as lesion factor and probability imaging, but have limited recognition accuracy due to the multi-modal complexity of the signals. For complex structures such as composite stiffened plates, determining the threshold of the damage factor is often difficult and accurate identification of the damage is challenging.
Disclosure of Invention
The application aims to provide an improved method for identifying damage to a reinforcing plate of a composite material, aiming at the problems existing in the prior art.
The method for identifying damage of the reinforcing plate of the composite material based on the ultrasonic guided wave and the space-time hybrid network is characterized by comprising the following steps of: the method comprises the following steps: acquiring ultrasonic guided wave signals arranged on the simulated damage positions of the composite material reinforcing plate based on a piezoelectric sensor network arranged on the composite material reinforcing plate to form ultrasonic guided wave simulated damage data simulating damage of the reinforcing plate; wherein the piezoelectric sensor network forms a path, the ultrasonic guided wave signal is a multi-frequency guided wave timing signal comprising a plurality of paths; labeling the simulated damage position of the simulated damage for the ultrasonic guided wave simulated damage data, establishing a database and dividing a training set and a testing set according to a proportion; establishing a space-time hybrid network model; the space-time hybrid network model comprises a parallel cyclic neural network and a convolutional neural network; training the recurrent neural network with the training set to obtain a first output comprising timing characteristics of the ultrasonic guided wave simulation damage data as global dependency information, and training the recurrent neural network with the training set to obtain a second output comprising local spatial characteristics of the ultrasonic guided wave simulation damage data; establishing a mapping relation between the ultrasonic guided wave simulation damage data and a simulation damage position at least based on a third output which is formed by adding the first output and the second output and comprises local space features and global space-time features; identifying damage of the composite material reinforcing plate based on the mapping relation; wherein, a sine time domain signal modulated by a Hanning window under a plurality of different center frequencies is adopted as excitation; the ultrasonic guided wave simulation damage data are ultrasonic guided wave simulation damage data obtained by fusing the signals with the different center frequencies; the data dimension corresponding to the ultrasonic wave guided wave data comprises a plurality of data channels with different center frequencies, guided wave multipath heights and guided wave signal lengths; the circulating neural network comprises a multi-layer gating circulating unit, the multi-layer gating circulating unit is used for generating the first output, the first output comprises a time sequence characteristic of a single path, and the circulating neural network is the second output of the multi-layer circulating neural network for generating local spatial characteristics; the second output includes inter-path guided wave features.
In some embodiments, the recurrent neural network includes a multi-layer gated loop unit for generating the first output including a timing characteristic for a single path, the convolutional neural network multi-layer convolutional neural network for generating the second output of local spatial characteristics; the second output includes inter-path guided wave features.
In some embodiments, batch normalization is performed after each convolution calculation of the convolutional neural network and an activation function is added after each layer of batch normalization to introduce non-linear factors.
In some embodiments, further comprising causing the third output to be used to learn high-dimensional characteristics of the ultrasonic guided wave simulation damage data through at least one layer of residual block.
In some embodiments, a convolution attention mechanism module is further included for enhancing model information sensitivity to enhance modeling capabilities of the convolution neural network model on data features.
In some embodiments, the piezoelectric sensor network includes a plurality of grids of four adjacent piezoelectric sensors, the simulated lesion is disposed on each of the grids in turn, ultrasonic guided wave signals are excited by the piezoelectric sensors in each of the grids in turn, and the ultrasonic guided wave simulated lesion data is received by the remaining piezoelectric sensors in the grids.
The composite material structure damage positioning method based on the ultrasonic guided wave and the space-time hybrid network has the advantages that: on one hand, the damage identification is performed by using a deep learning technology, the multi-mode complex characteristics of signals are not required to be considered, and the deep intrinsic characteristics of the data can be mined and learned. On the other hand, the space-time hybrid network can simultaneously capture global dependency information and local spatial characteristics of signals, and more fully learn signal abstract characteristics. In yet another aspect, the method of the present application is suitable for identifying the ultrasonic pilot signal after the multifrequency signal fusion to determine the lesion location, because the ultrasonic pilot signal after the multifrequency signal fusion contains more lesion information than a single frequency, the identification accuracy is higher. Other advantageous effects of the embodiments of the present application will be described in the detailed description section below.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present application, and other drawings may be obtained according to the drawings without inventive effort to those skilled in the art.
FIG. 1 is a schematic diagram of a test piece and a sensor network in an embodiment of a method for identifying damage to a composite stiffened plate based on an ultrasonic guided wave and space-time hybrid network according to the present application;
FIG. 2 is an excitation graph of ultrasonic guided wave signals at five frequencies in an embodiment of an identification method according to the present application;
FIG. 3 is a diagram of a network model of a lesion localization algorithm in an embodiment of an identification method according to the present application;
FIGS. 4A, 4B and 4C are graphs of activation functions in an embodiment of an identification method according to the present application, respectively;
FIG. 5 is an iterative graph of a loss function for a test set in an embodiment of an identification method according to the present application;
FIG. 6 is an iterative graph of evaluation indicators for a test set in an embodiment of an identification method according to the present application;
fig. 7A, 7B, 7C and 7D are graphs comparing the true value and the predicted result, respectively, of the lesion locations of the test set in an embodiment of the method according to the present application.
Detailed Description
In order to make the technical scheme and advantages of the present invention clearer, the technical scheme in the embodiment of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiment of the present invention.
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It should be understood that the terms "comprises" and "comprising," when used in this specification and the appended claims, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It is also to be understood that the terminology used in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations.
Various structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the drawings. The figures are not drawn to scale, wherein certain details are exaggerated for clarity of presentation and may have been omitted. The shapes of the various regions, layers and their relative sizes, positional relationships shown in the drawings are merely exemplary, may in practice deviate due to manufacturing tolerances or technical limitations, and one skilled in the art may additionally design regions/layers having different shapes, sizes, relative positions as actually required.
The method for identifying the damage of the reinforcing plate of the composite material based on the ultrasonic guided wave and space-time hybrid network comprises the following steps of:
step 1: and (3) pasting a piezoelectric sensor, performing a simulation damage experiment, and collecting ultrasonic guided wave signals.
This step includes, for example, attaching piezoelectric sensors to the composite structure to form a sensor network as shown in fig. 1. Every four adjacent sensors may form a sensor grid. Exciting ultrasonic guided wave signals in each sensor grid may employ a 5 period hanning window modulated sinusoidal time domain signal at a center frequency of 100, 150, 200, 250, 300 kHz as excitation, as shown in fig. 2. And sequentially pasting the cement at different positions on the test piece for simulating damage, receiving the response to the excitation ultrasonic guided wave signal by the residual sensor, taking the response as an ultrasonic guided wave simulated damage signal, and converting the ultrasonic guided wave simulated damage signal into ultrasonic guided wave simulated damage data.
Step 2: labeling data and establishing a database.
For example, the method comprises the steps of carrying out data annotation on the collected ultrasonic guided wave signals, namely, representing the damage position by using a coordinate pair (x, y), and carrying out the steps of: 3 to divide the training set and the test set. Herein (x, y) is also referred to as lesion coordinates.
Step 3: establishing a space-time hybrid network model, wherein the space-time hybrid network is based on a parallel convolutional neural network and a cyclic neural network, the convolutional neural network is excellent in processing static space data, spatial features and structures in the data can be effectively captured, and local information in the data can be extracted through a convolutional layer in the convolutional neural network; the recurrent neural network is specially used for processing time series data, and can establish time correlation. The space-time hybrid network can effectively process data containing both spatial and temporal information by combining a convolutional neural network and a cyclic neural network.
For example, a space-time hybrid network learning network model is established, and the network model structure is shown in fig. 3. The ultrasonic guided wave data after the multi-frequency signal fusion corresponds to a three-dimensional signal. The data dimension is recorded asFor example, is%) C, H, W respectively represent data channels, wherein the data channels comprise the five different frequencies, the guided wave multipath height and the guided wave signal length.
And taking the ultrasonic guided wave data fused by the multi-frequency signals as input data, carrying out global dependence information learning through a five-layer gating and circulating unit (GRU) network, and carrying out local feature learning through a five-layer convolutional neural network.
Wherein the number of neurons of the GRU network is 128, and the GRU network is obtainedAs a first output OP1, the GRU is shown in fig. 3, and specifically, the GRU is calculated as shown in formula (1).
；；；（1）。
Wherein,For the input of the current moment of time,The state output is hidden for the current time instant,It is the time to update the door that,Is a reset gate which is configured to reset the gate,For the candidate hidden state at the current time, t represents the current time step, and h (t-1) represents the hidden state output of the previous time step.In the form of a sigmoid activation function, as shown in figure 4A,The specific form of the activation function is shown in fig. 4B.，，Is a neural network weight matrix during training.
Five layers of CNNs through which the ultrasonic guided wave data after the multi-frequency signal fusion is used as input data pass, wherein the convolution kernel is as followsStride length isIs then subjected toAnd stride length ofIs obtained simultaneously with the maximum pooling operation of (a)As a second output OP2, the convolved form is seen in the CNN part of fig. 3. The convolution process is to multiply the original data values by the corresponding convolution kernel values and then sum the values as a result of the convolution calculation. And then the convolution kernel slides on the original data with a certain step length, and convolution calculation is carried out until the whole original data is slid. The calculation process of the two-dimensional convolution is shown in formula (2).
（2）
Where k is the size of the convolution kernel, f (i, j) is the raw data, and g (k-i, k-j) is the convolution kernel.
The guided wave signal is a timing signal comprising a plurality of paths, and is based on GRU to extract the timing characteristics of a single path; CNN-based is that the inter-path guided wave features are coupled. Doing so more fully learns the data characteristics. After being output by parallel space-time network, the first output OP1 and the second output OP2 are subjected to matrix addition to obtain a dimension ofA third output OP3 containing local spatial features and global spatio-temporal signal features.
Further, the third output OP3 may be subjected to learning by at least one residual block (RES), e.g. two residual blocks, each residual block having logic as shown in the RES part of fig. 3, learning the high-dimensional spatial features while preserving the low-dimensional signal features, resulting in a fourth output OP4. Wherein the linear projection layer is connected in the manner shown in formula (3).
（3）
Where y is the output after the residual block, and W 1、W2 is the convolution operation.
Batch Normalization (BN) was performed after each convolution calculation. Batch normalization can accelerate the training process, reduce internal covariate bias, improve performance and solve the problems of gradient disappearance and the like. The BN is a learning parameterThe batch normalization process in the present application is based on equation (4).
（4）
Wherein γ, β is a learnable parameter, the initialization setting is γ=1, β=0, μ β is the mean of the data features in the batch, σ β is the standard deviation, ε is a constant, and the denominator is prevented from being zero.
To avoid pure linear combinations, an activation function is added after batch normalization of each layer to introduce non-linear factors. The introduction LeakyReLU of the activation function into each residual block study described above is shown in equation (5), and the image is shown in fig. 4C.
（5）
The convolutional attention mechanism module (CBAM) aims to enhance the modeling ability of the Convolutional Neural Network (CNN) model on data features. The main problem it solves is how to automatically learn and focus on important features in the data to improve the representation capability of the CNN model. Its core idea is to introduce two key attention mechanisms: channel attention and spatial attention, such as the channel attention module and spatial attention module in section CBAM shown in fig. 3.
The channel attention objective is to perform weight distribution on different channels, namely different feature dimensions of the feature map, so that a model can automatically pay attention to which channels are more important for a specific task, and thus required features can be automatically learned by utilizing a convolution attention module, and a mapping relation with a damage position is conveniently established. The goal of spatial attention is to weight the features of the different spatial locations so that the model can automatically focus on important areas in the image. By combining channel attention and spatial attention CBAM, both important channels and important locations in the image can be focused on simultaneously, thereby improving the feature representation capabilities of the CNN model. And adding a convolution attention mechanism module to the space-time hybrid network learning network model to form the attention-enhanced space-time hybrid network learning network model.
After passing through the CBAM module, the fourth output OP4 is converted into a fifth output OP5 of one-dimensional signal characteristics, and in order to improve the generalization capability of the model, at least one Fully Connected Layer (FCL) may be provided to perform a classification task and a regression task, so that the fifth output OP5 obtains a sixth output OP6 through the at least one fully connected layer. And a dropout function with the loss rate of 0.3 can be designed on the full-connection layer, and finally, the mapping relation between the ultrasonic guided wave data fused by the multi-frequency signals and the damage coordinates (x, y) is established.
Absolute error Loss, also known as L1 Loss, is a Loss function that measures the difference between two vectors. The formula for L1 Loss for the two vectors x and y is shown as formula (6):
（6）
the true value and the corresponding predicted value of the ith sample are respectively represented, and N is the number of samples.
L1 Loss represents the sum of absolute differences between corresponding position elements in two vectors, so it is relatively sensitive to outliers, i.e. outliers, L1 Loss is used as a Loss function in many machine learning tasks, such as regression tasks, where the goal is to minimize the absolute difference between the predicted and actual values.
In the deep learning back propagation process, the optimizer is an optimization algorithm for solving the loss function, and guides each parameter of the loss function to update a proper value towards the correct direction, so that the parameter value of the updated loss function is continuously approximate to the global minimum value. The core idea of the optimizer is the gradient descent method, and the main parameters are gradient and learning rate. Some embodiments of the present application employ AdaGrad optimizers, which are an adaptive learning rate algorithm that monotonically decrease the step size by the square root of the sum of all gradient history squares, see equation (7). The learning rate of each dimension is adjusted according to the gradient value of the independent variable in each dimension, so that the problem that the fixed learning rate is difficult to adapt to all dimensions is avoided.
；；； （7）
Wherein the method comprises the steps ofIn order to calculate the gradient,In order to calculate the cumulative squared gradient,In order to multiply by element,The parameters are updated for the gradient and,For the global learning rate of the device,Small constant for numerical stabilization, default value of 1e-6, initial value，Is the number of iterations that are performed,Is an objective function.
Step 4: and (5) evaluating indexes.
In the regression task, the index for evaluating the performance of the model can help to know the fitting degree and the prediction accuracy of the model.
The invention takes the L1 Loss and R 2 scores as evaluation indexes. The L1 Loss calculates the absolute value of the prediction error of each sample, and calculates an average value, wherein the smaller the value is, the better the fitting effect of the model is. R, i.e., R-squared, is a score between 0.1 and 1.0, used to measure how well the model interprets the variance of the data, the closer to 1.0 the better the effect is, the following method is calculated:
（8）
where xi, yi represent the target variable, Representing the mean of the target variable. And R is equal to 1, the model perfect fitting data is represented by the formula, and N is a natural number.
Verification example:
to verify the effectiveness of the method of the present invention, a simulated damage test was performed on a carbon fiber composite reinforced wallboard having an overall size of 700 mm x 450 mm, containing 4T-shaped reinforcing ribs, as shown in fig. 1. 12 piezoelectric ceramic sensors (PZT) are arranged on the surface of the carbon fiber composite material between the ribs of the wallboard to form a PZT network, and the interval between the PZT is 160 mm multiplied by 130 mm. The PZT network is shown in FIG. 1, where each four PZTs form a grid, for example, a rectangular grid, each PZT being on one vertex of the rectangular grid, each PZT in turn being an excitation sensor, and the other three PZT being a receiving sensor. The PZT network designed by the present invention thus forms 72 signal paths from one to the next. The sampling frequency of the device is 12 MHz, and the sampling time is 330 . Grid points with the length of 2cm are drawn in the test area, the damage is simulated on each grid point of each grid in sequence, 5 different excitation frequency signals are excited, 375 damage points are acquired in total, and each acquired data dimension is 5 multiplied by 72 multiplied by 4000.
It should be understood that the piezoelectric sensor used in the sensor network of the present application is not limited to the piezoelectric ceramic sensor, but may be a piezoelectric sensor of other materials.
It should be understood that the mesh arrangement in the present application is not limited to the above-described size, shape, and the mesh arrangement may be employed as long as the object of the present application can be achieved.
The space-time hybrid network learning model is built through PyTorch frames, and the recognition method of the invention is realized by using two blocks NVIDIA GeForce RTX and 3090. The space-time hybrid network learning model, as shown in fig. 3, includes five layers of GRUs, five layers of CNNs, matrix addition, two layers of RES, one CBAM, and two layers of FCs.
Finally, adaGrad model parameters are optimized, the weight attenuation is 1e-5, and the initial learning rate is 1e-4. And gradually reducing the learning rate by adopting a cosine annealing algorithm. The dividing ratio of the training set and the test set is 7:3, the batch size is set to 16, and the total iteration is 80 epochs. The evaluation index and the Loss graph of the test set R are shown in fig. 5 and 6, respectively. After the iteration is completed, the test set has an L1 Loss value of 0.1195 and an R of 0.9084. The result of the ablation experiment is shown in table 1, and the result shows that the method can learn more guided wave signal characteristics.
Table 1:
Method of R2 Loss
Preferred embodiment of the application (cnn+gru+res+ CBAM +dropout): 0.9084 0.1195
without addition of RES 0.8857 0.1313
Does not add CBAM 0.8536 0.1503
Without adding Dropout 0.8988 0.1223
Single frequency 0.8710 0.1333
Performing complete training on the model by using all data of the training set, which is called Epoch, namely first generation training; one back-propagation parameter update of model weights is performed using a small fraction of samples in the training set, called Batch, i.e., a Batch of data; the number of samples for this small portion is the number of lot data, i.e., batch size.
The four test results are visualized as shown in fig. 7A, 7B, 7C and 7D, where circles represent actual positions and triangles represent predicted positions. The four test results show that the parallel network regression positioning algorithm based on the multi-frequency signal fusion is high in accuracy and visual, and the composite material damage identification method for deep learning has a wide application prospect.
It can be seen that combining ultrasonic guided wave technology with a deep learning algorithm to achieve more accurate identification of lesions in composite structures is a promising approach. The deep learning algorithm can learn the abstract intrinsic characteristics of the data without manually extracting damage factors or analyzing the multi-modal characteristics of the signals, so that the deep learning algorithm becomes an end-to-end learning mode, and the mapping relation between the guided wave signals and the damage information can be directly established, thereby improving the accuracy and the efficiency of damage identification.
In one embodiment of the invention, a terminal device is provided that includes a processor and a memory for storing a computer program including program instructions, the processor for executing the program instructions stored by the computer storage medium. The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processor, digital signal processor (DIGITAL SIGNAL Processor, DSP), application Specific Integrated Circuit (ASIC), off-the-shelf programmable gate array (field-programmable GATEARRAY, FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc., which are a computational core and a control core of the terminal adapted to implement one or more instructions, in particular adapted to load and execute one or more instructions to implement a corresponding method flow or a corresponding function; the processor provided by the embodiment of the invention can be used for executing the composite material stiffened plate damage identification method based on the ultrasonic guided wave and space-time hybrid network.
The present invention also provides a storage medium, in particular a computer readable storage medium (Memory), which is a Memory device in a terminal device, for storing programs and data. It will be appreciated that the computer readable storage medium herein may include both a built-in storage medium in the terminal device and an extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also stored in the memory space are one or more instructions, which may be one or more computer programs (including program code), adapted to be loaded and executed by the processor. The computer readable storage medium herein may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the respective steps of the above-described embodiments with respect to a method for identifying damage to a composite stiffener based on an ultrasonic guided wave and space-time hybrid network; one or more instructions in a computer-readable storage medium are loaded by a processor and perform the steps of the methods described above.
The foregoing is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art, who is within the scope of the present invention, should make equivalent substitutions or modifications according to the technical scheme of the present invention and the inventive concept thereof, and should be covered by the scope of the present invention.
The invention discloses a composite material stiffened plate damage identification method based on an ultrasonic guided wave and space-time hybrid network, which mainly comprises the steps of establishing a data set, establishing a damage identification model based on a space-time parallel hybrid network model, simultaneously learning time sequence characteristics and space characteristics of signals, learning the guided wave high-dimensional characteristics through a residual block, improving the information sensitivity of the model by utilizing a convolution attention module, and finally establishing a mapping relation between guided wave signals and damage coordinates. Test results show that the method can learn more guided wave signal characteristics, the generalization capability of the model can be better improved by adding the loss function, and the composite material damage identification technology combining deep learning and ultrasonic guided wave has certain advantages.

Claims (6)

1. The method for identifying damage of the reinforcing plate of the composite material based on the ultrasonic guided wave and the space-time hybrid network is characterized by comprising the following steps of: the method comprises the following steps:
Acquiring ultrasonic guided wave signals arranged on the simulated damage positions of the composite material reinforcing plate based on a piezoelectric sensor network arranged on the composite material reinforcing plate to form ultrasonic guided wave simulated damage data simulating damage of the reinforcing plate; wherein the piezoelectric sensor network forms a path, the ultrasonic guided wave signal is a multi-frequency guided wave timing signal comprising a plurality of paths;
Labeling the simulated damage position of the simulated damage for the ultrasonic guided wave simulated damage data, establishing a database and dividing a training set and a testing set according to a proportion;
Establishing a space-time hybrid network model; the space-time hybrid network model comprises a parallel cyclic neural network and a convolutional neural network;
training the recurrent neural network with the training set to obtain a first output comprising timing characteristics of the ultrasonic guided wave simulation damage data as global dependency information, and training the recurrent neural network with the training set to obtain a second output comprising local spatial characteristics of the ultrasonic guided wave simulation damage data;
Establishing a mapping relation between the ultrasonic guided wave simulation damage data and a simulation damage position at least based on a third output which is formed by adding the first output and the second output and comprises local space features and global space-time features; and
Identifying damage of the composite material reinforcing plate based on the mapping relation;
Wherein, a sine time domain signal modulated by a Hanning window under a plurality of different center frequencies is adopted as excitation; the ultrasonic guided wave simulation damage data are ultrasonic guided wave simulation damage data obtained by fusing the signals with the different center frequencies; the data dimension corresponding to the ultrasonic wave guided wave data comprises a plurality of data channels with different center frequencies, guided wave multipath heights and guided wave signal lengths;
The circulating neural network comprises a multi-layer gating circulating unit, the multi-layer gating circulating unit is used for generating the first output, the first output comprises a time sequence characteristic of a single path, and the circulating neural network is the second output of the multi-layer circulating neural network for generating local spatial characteristics; the second output includes inter-path guided wave features.
2. The method for identifying damage to the reinforcing plate of the composite material based on the ultrasonic guided wave and space-time hybrid network, which is characterized by comprising the following steps of: batch normalization is performed after each convolution calculation of the convolutional neural network and an activation function is added after each layer of batch normalization to introduce non-linear factors.
3. The method for identifying damage to the reinforcing plate of the composite material based on the ultrasonic guided wave and space-time hybrid network, which is characterized by comprising the following steps of: and the method further comprises the step of enabling the third output to be used for learning the high-dimensional characteristics of the ultrasonic guided wave simulation damage data through at least one layer of residual error block.
4. The method for identifying damage to the reinforcing plate of the composite material based on the ultrasonic guided wave and space-time hybrid network, which is characterized by comprising the following steps of: the convolution attention mechanism module is used for improving the sensitivity of the model information to enhance the modeling capability of the convolution neural network model on the data characteristics.
5. The method for identifying damage to the reinforcing plate of the composite material based on the ultrasonic guided wave and space-time hybrid network, which is characterized by comprising the following steps of: the piezoelectric sensor network comprises a plurality of grids formed by four adjacent piezoelectric sensors, the simulation damage is sequentially arranged on each grid, ultrasonic guided wave signals are sequentially excited by the piezoelectric sensors in each grid, and residual piezoelectric sensors in the grids are used for receiving the ultrasonic guided wave simulation damage data.
6. The method for identifying damage to the reinforcing plate of the composite material based on the ultrasonic guided wave and space-time hybrid network, which is characterized by comprising the following steps of: regression task with L1 Loss and R 2 score as evaluation index, wherein L1 Loss calculates absolute value of prediction error of each sample and calculates average value; r is used for measuring the interpretation degree of the model on the data variance, and the calculation method is as follows; Wherein y i,yi represents a target variable,/>Representing the mean of the target variable, R equals 1 to represent model perfect fit data, N to represent the number of samples.
CN202410049859.5A 2024-01-13 2024-01-13 Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network Active CN117907445B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202410049859.5A CN117907445B (en) 2024-01-13 2024-01-13 Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202410049859.5A CN117907445B (en) 2024-01-13 2024-01-13 Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network

Publications (2)

Publication Number Publication Date
CN117907445A true CN117907445A (en) 2024-04-19
CN117907445B CN117907445B (en) 2024-08-13

Family

ID=90687588

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202410049859.5A Active CN117907445B (en) 2024-01-13 2024-01-13 Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network

Country Status (1)

Country Link
CN (1) CN117907445B (en)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020156348A1 (en) * 2019-01-31 2020-08-06 青岛理工大学 Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network
KR20220011334A (en) * 2020-07-21 2022-01-28 한양대학교 산학협력단 System and method for the damage location detection of composite structures based on a convolutional neural network
CN115218802A (en) * 2022-06-23 2022-10-21 山东航天电子技术研究所 Spacecraft structure fatigue damage intelligent identification method based on optical fiber sensing and ultrasonic guided waves
CN115859074A (en) * 2022-10-27 2023-03-28 大连理工大学 Composite material stiffened plate damage identification method based on ultrasonic guided wave and deep learning
CN116028791A (en) * 2022-11-10 2023-04-28 北京工业大学 Steel strand stress identification method based on ultrasonic guided wave and convolutional neural network
CN116223635A (en) * 2023-03-17 2023-06-06 北京航空航天大学 Ultrasonic guided wave damage positioning imaging method based on convolution self-coding
CN116306231A (en) * 2023-02-06 2023-06-23 大连理工大学 Adhesive joint structure debonding damage identification method and device based on ultrasonic guided wave deep learning
US11709979B1 (en) * 2022-10-28 2023-07-25 Hefei University Of Technology Bridge damage identification method considering uncertainty
CN116908312A (en) * 2023-04-14 2023-10-20 大连理工大学 Variable-temperature environment damage positioning method, system and storage medium based on ultrasonic guided wave and generalized capsule network model

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2020156348A1 (en) * 2019-01-31 2020-08-06 青岛理工大学 Structural damage identification method based on ensemble empirical mode decomposition and convolution neural network
KR20220011334A (en) * 2020-07-21 2022-01-28 한양대학교 산학협력단 System and method for the damage location detection of composite structures based on a convolutional neural network
CN115218802A (en) * 2022-06-23 2022-10-21 山东航天电子技术研究所 Spacecraft structure fatigue damage intelligent identification method based on optical fiber sensing and ultrasonic guided waves
CN115859074A (en) * 2022-10-27 2023-03-28 大连理工大学 Composite material stiffened plate damage identification method based on ultrasonic guided wave and deep learning
US11709979B1 (en) * 2022-10-28 2023-07-25 Hefei University Of Technology Bridge damage identification method considering uncertainty
CN116028791A (en) * 2022-11-10 2023-04-28 北京工业大学 Steel strand stress identification method based on ultrasonic guided wave and convolutional neural network
CN116306231A (en) * 2023-02-06 2023-06-23 大连理工大学 Adhesive joint structure debonding damage identification method and device based on ultrasonic guided wave deep learning
CN116223635A (en) * 2023-03-17 2023-06-06 北京航空航天大学 Ultrasonic guided wave damage positioning imaging method based on convolution self-coding
CN116908312A (en) * 2023-04-14 2023-10-20 大连理工大学 Variable-temperature environment damage positioning method, system and storage medium based on ultrasonic guided wave and generalized capsule network model

Also Published As

Publication number Publication date
CN117907445B (en) 2024-08-13

Similar Documents

Publication Publication Date Title
Rautela et al. Combined two-level damage identification strategy using ultrasonic guided waves and physical knowledge assisted machine learning
Park et al. Concrete crack detection and quantification using deep learning and structured light
CN109947086B (en) Mechanical fault migration diagnosis method and system based on counterstudy
Gardner et al. Machine learning at the interface of structural health monitoring and non-destructive evaluation
Moore et al. Crack identification in a freely vibrating plate using Bayesian parameter estimation
CN114387201B (en) Cytopathic image auxiliary diagnosis system based on deep learning and reinforcement learning
de Rezende et al. Convolutional neural network and impedance-based SHM applied to damage detection
Zhang et al. Data anomaly detection for bridge SHM based on CNN combined with statistic features
CN115859074B (en) Ultrasonic guided wave and deep learning-based composite material stiffened plate damage identification method
Fritz et al. Fault diagnosis in structural health monitoring systems using signal processing and machine learning techniques
Lee et al. Channel pruning via gradient of mutual information for light-weight convolutional neural networks
CN116908312A (en) Variable-temperature environment damage positioning method, system and storage medium based on ultrasonic guided wave and generalized capsule network model
Ribeiro Junior et al. On the use of machine learning for damage assessment in composite structures: a review
US20240013044A1 (en) Method for characterizing a part through non-destructive inspection
CN117907445B (en) Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network
Hartmann et al. Hierarchical local model trees for design of experiments in the framework of ultrasonic structural health monitoring
CN110779477B (en) Acoustic method for identifying shape of object in real time
Junges et al. Convolutional autoencoders and CGANs for unsupervised structural damage localization
Jones et al. Heteroscedastic Gaussian processes for localising acoustic emission
Thiene et al. Optimal sensor placement for damage detection based on ultrasonic guided wave
CN110389176B (en) Fatigue damage diagnosis method for carbon fiber reinforced composite material structure
Moll et al. A rapid and velocity-independent damage localization approach for ultrasonic structural health monitoring
US20230314386A1 (en) Method for characterizing a part through non-destructive inspection
Zhang et al. Damage Detection of a Pressure Vessel with Smart Sensing and Deep Learning
Feng et al. Enhancing non-destructive testing in concrete structures: a GADF-CNN approach for defect detection

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant