[go: up one dir, main page]

CN119395160B - Mamba model-based carbon fiber reinforced composite material damage detection method - Google Patents

Mamba model-based carbon fiber reinforced composite material damage detection method

Info

Publication number
CN119395160B
CN119395160B CN202411459600.4A CN202411459600A CN119395160B CN 119395160 B CN119395160 B CN 119395160B CN 202411459600 A CN202411459600 A CN 202411459600A CN 119395160 B CN119395160 B CN 119395160B
Authority
CN
China
Prior art keywords
guided wave
wave signal
mamba
damage
carbon fiber
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.)
Active
Application number
CN202411459600.4A
Other languages
Chinese (zh)
Other versions
CN119395160A (en
Inventor
张延兵
马向东
贺申
徐中原
吴肖
丁小平
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.)
Special Equipment Safety Supervision Inspection Institute of Jiangsu Province
Original Assignee
Special Equipment Safety Supervision Inspection Institute of Jiangsu Province
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 Special Equipment Safety Supervision Inspection Institute of Jiangsu Province filed Critical Special Equipment Safety Supervision Inspection Institute of Jiangsu Province
Priority to CN202411459600.4A priority Critical patent/CN119395160B/en
Publication of CN119395160A publication Critical patent/CN119395160A/en
Application granted granted Critical
Publication of CN119395160B publication Critical patent/CN119395160B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N29/00—Investigating or analysing materials by the use of ultrasonic, sonic or infrasonic waves; Visualisation of the interior of objects by transmitting ultrasonic or sonic waves through the object
    • G01N29/44—Processing the detected response signal, e.g. electronic circuits specially adapted therefor
    • G01N29/4409—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison
    • G01N29/4418—Processing the detected response signal, e.g. electronic circuits specially adapted therefor by comparison with a model, e.g. best-fit, regression analysis
    • 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 OR CALCULATING; 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
    • G06F18/2131—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on a transform domain processing, e.g. wavelet transform
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/2431—Multiple classes
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2123/00—Data types
    • G06F2123/02—Data types in the time domain, e.g. time-series data
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/12—Classification; Matching

Landscapes

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

Abstract

本发明公开了基于Mamba模型的碳纤维增强复合材料损伤检测方法，涉及超声波导波检测技术领域。该方法包括：S1：在存在损伤的碳纤维增强符合材料上采集导波信号集；S2：通过导波信号处理分支提取导波信号集的时间特征序列；S3：通过连续小波变换分支提取导波信号集的谱特征；S4：融合导波信号集的时间特征与谱特征，全面获取导波信号中的多维损伤特征，对整体碳纤维增强复合材料进行缺陷检测与定位。通过将Mamba的时间特征，连续小波变换和卷积神经网络提取的时域谱特征融合计算，在无基线导波信号下实现碳纤维增强复合材料的无基线导波损伤检测与定位。

This invention discloses a damage detection method for carbon fiber reinforced composite materials based on the Mamba model, belonging to the field of ultrasonic guided wave detection technology. The method includes: S1: acquiring guided wave signal sets on damaged carbon fiber reinforced composite materials; S2: extracting the temporal feature sequence of the guided wave signal sets through guided wave signal processing; S3: extracting the spectral features of the guided wave signal sets through continuous wavelet transform; S4: fusing the temporal and spectral features of the guided wave signal sets to comprehensively obtain multidimensional damage features in the guided wave signals, and performing defect detection and localization on the overall carbon fiber reinforced composite material. By fusing the temporal features of Mamba, continuous wavelet transform, and time-domain spectral features extracted by convolutional neural networks, baseline-free guided wave damage detection and localization of carbon fiber reinforced composite materials is achieved without baseline guided wave signals.

Description

Mamba model-based carbon fiber reinforced composite material damage detection method
Technical Field
The invention relates to the technical field of ultrasonic guided wave detection, in particular to a method for detecting damage of a carbon fiber reinforced composite material based on Mamba model.
Background
Compared with metal materials, the carbon fiber reinforced composite material has the remarkable advantages of light weight, high specific strength, high specific rigidity, corrosion resistance and the like, and is widely applied to the fields of aerospace and various industries. Due to time-varying load, low-speed impact and severe operating environment, the carbon fiber reinforced composite material is easy to generate invisible damage such as layering and debonding, and the safety of equipment and personnel is seriously threatened. Therefore, the research on nondestructive testing and evaluation technology of the carbon fiber reinforced polymer composite material has important significance for the economic and reliable operation of aerospace and engineering equipment.
The structural health monitoring technology based on the guided wave is widely considered as an ideal choice for detecting structural damage of the platy composite material due to the advantages of long propagation distance, small attenuation, low cost, high sensitivity to various damage and the like. Guided wave detection is typically achieved by a sensor network consisting of piezoelectric transducers, in which a transmitter excites guided waves in the structure being examined and a receiver collects the corresponding signals for subsequent analysis and processing. The characteristics of the guided wave change due to the presence of damage, and may be represented by linear characteristics such as amplitude attenuation and speed change. Different characteristic information corresponds to different damage indexes, and a special detection algorithm is designed to realize accurate damage positioning and assessment.
The traditional damage identification method lacks an effective signal characteristic extraction technology, and a test signal and a baseline signal (a signal from a healthy structure) are usually required to be compared or background noise is subtracted to extract a damage scattering signal, so that information such as damage flight time, gravitational wave energy attenuation and the like is obtained. The damage index extraction method generally only uses single change of the guided wave characteristic, and cannot accurately construct a high-order nonlinear mapping relation between the guided wave signal and the damage position.
Disclosure of Invention
Based on the above-mentioned shortcomings of the prior art, the present invention aims to provide a method for detecting damage to a carbon fiber reinforced composite material based on Mamba model, so as to solve the above-mentioned technical problems.
In order to achieve the purpose, the invention provides the following technical scheme that the damage detection method for the carbon fiber reinforced composite material based on Mamba model comprises the following steps:
s1, collecting a guided wave signal set on a damaged carbon fiber reinforced composite material;
S2, extracting a time feature sequence of a guided wave signal set through a guided wave signal processing branch;
s3, extracting spectral features of the guided wave signal set through continuous wavelet transformation branches;
and S4, fusing the time characteristics and the spectrum characteristics of the guided wave signal set, comprehensively obtaining multidimensional damage characteristics in the guided wave signal, and detecting and positioning the defects of the whole carbon fiber reinforced composite material.
The invention is further arranged in step S1 using four piezoelectric transducers arranged in a dual-transmit dual-receive fashion for collecting a limited set of guided wave signals.
The present invention is further configured such that, in step S2, the time-series feature sequence of the guided wave signal set is extracted through the guided wave signal processing branch, and the long-series data is processed using the state space model Mamba:
deriving time points t=0, 1,2 output from the state space model, starting at time t=0;
h t denotes a state quantity at time t, x t denotes an input control quantity at time t, y t denotes a system output at time t, a denotes a state transition matrix, B denotes an influence of the input control quantity x t on the state quantity h t, C denotes an influence of the state quantity h t on the system output y t, B and C are embodied as weight matrices of the input control quantity x t and the state quantity h t, the input control quantity is mapped to a system output quantity, Representing the average value of the weight matrix, and using the average value smoothing process of a group of weight matrices at different time points to reduce noise influence;
Extracting the state equation when t=k:
introducing convolution x to carry out convolution operation:
wherein, the The transfer function matrix between the input control quantity x t and the system output y t consists of a transfer matrix and a weight matrix, and finally an operation result y is output;
Taking the guided wave signal set as input data f (1)∈RB×D, taking D as the sequence length of the guided wave signal set, carrying out average pooling on the data to obtain a dimension reduction feature f (2)∈RB×D′, segmenting the sequence f (2) into f (3)∈RB×D″, inputting a state space model Mamba for feature extraction, and f (4)＝M(f(3), wherein M (-) represents that based on the state space model Mamba, inputting segmented sequence data into the state space model Mamba for feature extraction, and obtaining a serialization feature f (4)∈RB×D″;
stacking and linearizing the serialized features:
f 1 (4)＝Reshape(f(4));fGW＝Linear(f1 (4)), wherein, For the stacked serialized features, reshape (·) represents stacking operation on the features, so that feature dimensions are correctly docked with thread processing, f GW is the serialized features after Linear processing, linear (·) represents Linear processing on the serialized features through a Linear layer, and a feature vector with a fixed size is output.
The invention is further configured that, in step S3, a continuous wavelet transformation branch is used to perform time-frequency transformation on the guided wave signal set, average pooling is performed on the signals, a dimension reduction feature f (2)∈RB×D′ is obtained, continuous wavelet transformation is performed on the feature, and a feature f CWT (1)∈RB×DCWT is output:
fCWT (1)＝CWT(a,b)[f(2)];
wherein CWT (·) represents a continuous wavelet transform of the signal, t represents a time variable in the signal L s (t), ψ (t) is a wavelet mother function, ψ * (t) represents a complex conjugate of ψ (t), a is a scale parameter, controlling wavelet width, b is a shift parameter, determining the position of the wavelet in time;
Obtaining the output f CWT (1) of wavelet transformation, carrying out average pooling dimension reduction and neural network convolution, and obtaining the spectral characteristic output of the final wavelet transformation branch
The cross entropy loss function more suitable for classification tasks is used in the continuous wavelet transform branches:
wherein, the For the cross entropy loss function, Y represents the true probability that the signal feature belongs to a certain class, f θ (·) represents the predictive score of the input sample of a certain class, and S (·) represents the Softmax activation function.
The invention is further configured that, in step S4, the features of the two branches are fused by cascade operation, and the damage classification is performed by using multiple linear layers:
fFusion＝Concat(fGW,fCWT);
f1 Fusion＝Linear1(fFusion);
f2 Fusion＝Linear2(f1 Fusion);
fResult＝Linear3(f2 Fusion);
Wherein f Fusion is a feature after fusing two branches for a cascade operation, concat (·) represents a join operation, As a feature of the serialization after the linear processing,For spectral feature output of the wavelet transform branch, D1 represents the feature dimension size of the guided wave branch, set to 1000, D2 represents the feature dimension size of the spectrogram of the continuous wavelet transform branch, set to 384, linear i (·) represents the linear layer sequence number,For the serialized feature after the ith linear layer processing, f Result∈RB×N represents the output damage classification result, and N represents the number of damage categories.
The invention provides a method for detecting damage of a carbon fiber reinforced composite material based on Mamba model, which comprises the steps of collecting a guided wave signal set on a damaged carbon fiber reinforced composite material, extracting a time feature sequence of the guided wave signal set through a guided wave signal processing branch, extracting a spectrum feature of the guided wave signal set through a continuous wavelet transformation branch, and comprehensively obtaining multidimensional damage features in the guided wave signal by fusing the time feature and the spectrum feature of the guided wave signal set, wherein the method has the beneficial effects that the defect detection and the positioning are carried out on the whole carbon fiber reinforced composite material, and the generated beneficial effects comprise:
The proposed detection method integrates a state space model Mamba, continuous wavelet transformation and a convolutional neural network, and realizes efficient parallel computation. The time characteristics are extracted from the long-distance guided wave sequence by Mamba method, so that the computational complexity is lower. The continuous wavelet transformation enhances the extraction capacity and robustness of signal characteristics, the spectrogram of the continuous wavelet transformation is not a time characteristic image, the convolutional neural network is used for extracting time-frequency characteristics from the guided wave signals after the continuous wavelet transformation, the time characteristics obtained by Mamba are effectively combined with the convolutional neural network, the multi-scale damage characteristics are comprehensively obtained, the deeper relation information in the guided wave signals is extracted, the guided wave signal information is adaptively analyzed, the characteristic information related to damage is accurately extracted, and the high-precision detection of the damage of the composite material is realized. The method realizes damage detection under the conditions of limited guided wave signal sets and no baseline signal, directly uses the guided wave signal sets of the complete time sequence as input, expands engineering application of guided waves in the baseline-free environment, and improves the application range of ultrasonic guided wave nondestructive detection.
The foregoing description is only an overview of the present application, and is intended to be implemented in accordance with the teachings of the present application in order that the same may be more clearly understood and to make the same and other objects, features and advantages of the present application more readily apparent.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. In the drawings:
FIG. 1 is a flow chart of a method for detecting damage to a carbon fiber reinforced composite based on Mamba model, according to an example embodiment of the present invention;
FIG. 2 is a diagram illustrating a guided wave signal processing branch time feature extraction according to an exemplary embodiment of the present invention;
fig. 3 is a schematic diagram of a model of a continuous wavelet transform branch shown in an exemplary embodiment of the present invention.
Detailed Description
Further advantages and effects of the present invention will become readily apparent to those skilled in the art from the disclosure herein, by referring to the accompanying drawings and the preferred embodiments. The invention may be practiced or carried out in other embodiments that depart from the specific details, and the details of the present description may be modified or varied from the spirit and scope of the present invention. It should be understood that the preferred embodiments are presented by way of illustration only and not by way of limitation.
It should be noted that the illustrations provided in the following embodiments merely illustrate the basic concept of the present invention by way of illustration, and only the components related to the present invention are shown in the drawings and are not drawn according to the number, shape and size of the components in actual implementation, and the form, number and proportion of the components in actual implementation may be arbitrarily changed, and the layout of the components may be more complicated.
In the following description, numerous details are set forth in order to provide a more thorough explanation of embodiments of the present invention, it will be apparent, however, to one skilled in the art that embodiments of the present invention may be practiced without these specific details, in other embodiments, well-known structures and devices are shown in block diagram form, rather than in detail, in order to avoid obscuring the embodiments of the present invention.
As shown in fig. 1, the method for detecting damage to the carbon fiber reinforced composite material based on Mamba model comprises the following steps:
s1, collecting a guided wave signal set on a damaged carbon fiber reinforced composite material;
S2, extracting a time feature sequence of a guided wave signal set through a guided wave signal processing branch;
s3, extracting spectral features of the guided wave signal set through continuous wavelet transformation branches;
and S4, fusing the time characteristics and the spectrum characteristics of the guided wave signal set, comprehensively obtaining multidimensional damage characteristics in the guided wave signal, and detecting and positioning the defects of the whole carbon fiber reinforced composite material.
Specifically, in step S1, first, four piezoelectric transducers are arranged on a composite material plate in a dual-transmitting and dual-receiving mode, an excitation signal is generated by an arbitrary waveform generator, the excitation signal is applied to the excitation piezoelectric transducers through an ultrasonic signal preamplifier, and an oscilloscope collects ultrasonic signals of the reception piezoelectric transducers. The ultrasonic signals constitute a guided wave signal set.
As shown in fig. 2, specifically, in step S2, a time characteristic sequence of the guided wave signal set is extracted, and the output of the state space model at different time points is derived from the time t=0, so as to obtain a state equation at the time point. By introducing convolution parameters, convolutions are introduced into the computation while parallelizing the convolution operation. And carrying out average pooling on the input guided wave signal set, reducing the data dimension, and obtaining the dimension reduction characteristic. Segmenting the dimensionality reduction data, inputting Mamba models to perform feature extraction to obtain serialization features, and finally performing stacking and linearization processing on the serialization features to obtain guided wave signal set features.
The guided wave signal set characteristic extraction stage specifically comprises the following steps:
starting from time t=0, the time points t=0, 1,2 output from the state space model are derived.
H t denotes a state quantity at time t, x t denotes an input control quantity at time t, y t denotes a system output at time t, a denotes a state transition matrix, B denotes an influence of the input control quantity x t on the state quantity h t, C denotes an influence of the state quantity h t on the system output y t, B and C are embodied as weight matrices of the input control quantity x t and the state quantity h t, the input control quantity is mapped to a system output quantity,Representing the average value of the weight matrix, and using the average value smoothing process of a group of weight matrices at different time points to reduce noise influence.
Extracting the state equation when t=k:
Extracting the state equation when t=k:
introducing convolution x to carry out convolution operation:
wherein, the The transfer function matrix between the input control quantity x t and the system output y t consists of a transfer matrix and a weight matrix, and finally an operation result y is output;
Taking the guided wave signal set as input data f (1)∈RB×D, D represents the sequence length of the guided wave signal set,
Carrying out average pooling on data to obtain a dimension reduction feature f (2)∈RB×D′, segmenting a sequence f (2) into f (3)∈RB ×D″, inputting a state space model Mamba to carry out feature extraction, namely f (4)＝M(f(3), wherein M (·) represents a state space model Mamba, inputting segmented sequence data into a state space model Mamba to carry out feature extraction, and obtaining a serialization feature f (4)∈RB×D″;
stacking and linearizing the serialized features:
f 1 (4)＝Reshape(f(4));fGW＝Linear(f1 (4)), wherein, For the stacked serialized features, reshape (·) represents stacking operation on the features, so that feature dimensions are correctly docked with thread processing, f GW is the serialized features after Linear processing, linear (·) represents Linear processing on the serialized features through a Linear layer, and a feature vector with a fixed size is output.
As shown in fig. 3, specifically, in step S3, the spectral features of the guided wave signal set are extracted through the continuous wavelet transformation branch, the guided wave signal set is averaged and pooled to obtain the dimensionality reduction feature of the guided wave signal set, the continuous wavelet transformation is performed on the dimensionality reduction guided wave signal set to obtain the spectral features of the guided wave signal set, the dimensionality reduction is performed on the spectral features to obtain the dimensionality reduction spectral features, then convolution and signal segmentation are performed, and the signal features of the continuous wavelet transformation spectral features in the time domain are extracted.
The extraction stage of the time domain spectrum characteristics of the guided wave signal set is specifically as follows:
The signals are averaged and pooled to obtain dimension-reducing characteristics f (2)∈RB×D′, the characteristics are subjected to continuous wavelet transformation, and the characteristics are output
fCWT (1)＝CWT(a,b)[f(2)];
Wherein CWT (·) represents a continuous wavelet transform of the signal, t represents a time variable in the signal L s (t), ψ (t) is a wavelet mother function, ψ * (t) represents a complex conjugate of ψ (t), a is a scale parameter, controlling wavelet width, b is a shift parameter, determining the position of the wavelet in time;
obtaining the output f CWT (1) of wavelet transformation, carrying out average pooling dimensionality reduction and neural network convolution, outputting the spectral characteristics of the dimensions (16, 24), carrying out signal segmentation, setting the dimensions 384, and obtaining the spectral characteristic output of the final wavelet transformation branch
The cross entropy loss function more suitable for classification tasks is used in the continuous wavelet transform branches:
wherein, the For the cross entropy loss function, Y represents the true probability that the signal feature belongs to a certain class, f θ (·) represents the predictive score of the input sample of a certain class, and S (·) represents the Softmax activation function.
In step S4, the time features and the spectrum features of the guided wave signal set are fused, the multidimensional damage features in the guided wave signal are comprehensively obtained, and the defect detection and the location of the whole carbon fiber reinforced composite material are performed.
Specifically, the characteristics of the guided wave signal processing branch and the continuous wavelet transformation branch are fused through cascading operation, three-layer linear layer operation is carried out on the fused guided wave signal set characteristics, the damage classification of the carbon fiber reinforced composite material is obtained, and the damage detection and positioning are completed.
The time characteristic and the spectrum characteristic of the guided wave signal set are fused, and the defect detection and positioning stage of the whole carbon fiber reinforced composite material is specifically as follows:
fFusion＝Concat(fGW,fCWT);
f1 Fusion＝Linear1(fFusion);
f2 Fusion＝Linear2(f1 Fusion);
fResult＝Linear3(f2 Fusion);
Wherein f Fusion is a feature after fusing two branches for a cascade operation, concat (·) represents a join operation, As a feature of the serialization after the linear processing,For spectral feature output of the wavelet transform branch, D1 represents the feature dimension size of the guided wave branch, set to 1000, D2 represents the feature dimension size of the spectrogram of the continuous wavelet transform branch, set to 384, linear i (·) represents the linear layer sequence number,For the serialized feature after the ith linear layer processing, f Result∈RB×N represents the output damage classification result, and N represents the number of damage categories.
When the traditional detection technology is used for detecting, the guided wave signal set of the healthy carbon fiber reinforced composite material is used as a baseline signal, and the neural network model is trained through a large number of healthy guided wave signal sets and damage guided wave signal sets, so that damage detection and positioning of the carbon fiber reinforced composite material are performed, and damage detection and positioning of the carbon fiber reinforced composite material cannot be performed in a baseline-free guided wave signal or limited guided wave data set.
The time characteristics of the guided wave signal set are extracted by using Mamba models in the prior art, and seven different areas are randomly selected. A lesion site is placed in the center of each zone, and there are five lesion sites for each zone. Four sets of guided wave signal set data are generated for each lesion scene, for a total of 140 guided wave signal data sets.
Compared with the prior art, the damage detection is carried out by using a carbon fiber reinforced composite material guided wave method based on Mamba model, 80% of data set is used for training, 10% of data set is used for verification, 10% of data set is used for testing, the verification precision is 100%, the test precision is 100%, and the detection time is 84.92 seconds.
In this embodiment, the extracted guided wave signal sets are all guided wave signals collected under the condition that a defect exists on the carbon fiber reinforced composite material, and no guided wave signal exists in a healthy state, that is, baseline-free guided wave damage detection and positioning of the carbon fiber reinforced composite material based on the Mamba model are performed under the condition that a baseline signal is not needed. Therefore, a specific formula process is adopted to extract the time characteristic sequence of the guided wave signal set. By introducing convolution parameters, convolutions are introduced into the computation while parallelizing the convolution operation. And carrying out average pooling on the input guided wave signal set, and reducing the data dimension. Inputting Mamba a model to perform feature extraction, stacking and linearizing the serialized features, and obtaining guided wave signal set features. And simultaneously, extracting the characteristics of the guided wave signal set spectrum characteristics on the time domain through continuous wavelet transformation and a convolutional neural network, finally fusing the time characteristics and the spectrum characteristics of the guided wave signal set, comprehensively acquiring multidimensional damage characteristics in the guided wave signal, and carrying out defect detection and positioning on the whole carbon fiber reinforced composite material. Note that the depth characteristics and overall information of the impairment information in the guided wave signal without baseline cannot be obtained using other formulas.
The state space model Mamba is higher in efficiency in processing the long-sequence guided wave signal set and lower in computational complexity, which is very beneficial to actual damage detection engineering. The method extracts time features from long-distance guided wave sequences by Mamba method. The continuous wavelet transformation enhances the extraction capacity and robustness of signal characteristics, the convolutional neural network is used for extracting time-frequency characteristics from the guided wave signals after the continuous wavelet transformation, the time characteristics obtained by Mamba are effectively combined with the convolutional neural network, multi-scale damage characteristics are comprehensively obtained, deeper relation information in the guided wave signals is extracted, guided wave signal information is adaptively analyzed, characteristic information related to damage is accurately extracted, and high-precision detection of composite material damage is realized. The method realizes damage detection under the conditions of limited guided wave signal sets and no baseline signal, directly uses the guided wave signal sets of the complete time sequence as input, expands engineering application of guided waves in the baseline-free environment, and improves the application range of ultrasonic guided wave nondestructive detection.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions described in accordance with embodiments of the present application are produced in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from one website site, computer, server, or data center to another website site, computer, server, or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. that contains one or more sets of available media. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium may be a solid state disk.
It should be understood that the term "and/or" is merely an association relationship describing the associated object, and means that three relationships may exist, for example, a and/or B, and may mean that a exists alone, while a and B exist alone, and B exists alone, wherein a and B may be singular or plural. In addition, the character "/" herein generally indicates that the associated object is an "or" relationship, but may also indicate an "and/or" relationship, and may be understood by referring to the context.
In the present application, "at least one" means one or more, and "a plurality" means two or more. "at least one of" or the like means any combination of these items, including any combination of single item(s) or plural items(s). For example, at least one (a, b, or c) of a, b, c, a-b, a-c, b-c, or a-b-c may be represented, wherein a, b, c may be single or plural.
It should be understood that, in various embodiments of the present application, the sequence numbers of the foregoing processes do not mean the order of execution, and the order of execution of the processes should be determined by the functions and internal logic thereof, and should not constitute any limitation on the implementation process of the embodiments of the present application.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described systems, apparatuses and units may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In the several embodiments provided by the present application, it should be understood that the disclosed system may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present application. The storage medium includes a U disk, a removable hard disk, a read-only memory (ROM), a random access memory (random access memory, RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
The foregoing is merely illustrative of the present application, and the present application is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (3)

1. The damage detection method for the carbon fiber reinforced composite material based on Mamba model is characterized by comprising the following steps:
s1, collecting a guided wave signal set on a carbon fiber reinforced composite material with damage;
In the step S2, the time characteristic sequence of the guided wave signal set is extracted through the guided wave signal processing branch, and long sequence data is processed by using a state space model Mamba:
deriving time points t=0, 1,2 output from the state space model, starting at time t=0;
;
;
;
;
;
;
h t denotes a state quantity at time t, x t denotes an input control quantity at time t, y t denotes a system output at time t, a denotes a state transition matrix, B denotes an influence of the input control quantity x t on the state quantity h t, C denotes an influence of the state quantity h t on the system output y t, B and C are embodied as weight matrices of the input control quantity x t and the state quantity h t, the input control quantity is mapped to a system output quantity, Representing the average value of the matrix, and using a group of weight matrixes to smooth the average value at different time points so as to reduce noise influence;
Extracting the state equation when t=k:
;
introducing convolution x to carry out convolution operation:
;
;
wherein, the The transfer function matrix between the input control quantity x t and the system output y t consists of a transfer matrix and a weight matrix, and finally an operation result y is output;
using the guided wave signal set as input data D represents the sequence length of the guided wave signal set, and the data is subjected to average pooling to obtain the dimension reduction characteristicFor sequencesSegmenting intoThe input state space model Mamba performs feature extraction: Wherein M (-) represents the sequence data based on the state space model Mamba, and the segmented sequence data is input into the state space model Mamba for feature extraction to obtain the serialization features ;
Stacking and linearizing the serialized features:
; , wherein, For stacked serialized features, reshape (·) represents stacking operations on the features, proper interfacing of feature dimensions with thread processing,Linear (·) represents Linear processing of the serialized features by a Linear layer, outputting feature vectors of a fixed size;
s3, extracting spectral features of the guided wave signal set through continuous wavelet transformation branches;
S4, fusing the time characteristics and the spectrum characteristics of the guided wave signal set, comprehensively acquiring multidimensional damage characteristics in the guided wave signal, detecting and positioning defects of the whole carbon fiber reinforced composite material, fusing the characteristics of two branches through cascade operation, and adopting a plurality of linear layers to carry out damage classification:
;
;
;
;
wherein, the The features after the two branches are fused for cascading operations,The connection operation is represented by a number of steps,As a feature of the serialization after the linear processing,For spectral feature output of the wavelet transform branches, D1 represents the feature dimension size of the guided wave branch, set to 1000, D2 represents the spectral feature dimension size of the continuous wavelet transform branch, set to 384,A linear layer sequence number is indicated,To pass through the firstThe characteristics of the serialization after the linear layer processing,And the output damage classification result is represented, and N represents the number of damage categories.
2. The method for detecting damage to a carbon fiber reinforced composite material based on Mamba model according to claim 1, wherein in step S1, four piezoelectric transducers are used to collect a limited set of guided wave signals, arranged in a dual-transmit dual-receive format.
3. The method for detecting damage to a carbon fiber reinforced composite material based on Mamba model as claimed in claim 1, wherein in step S3, a continuous wavelet transformation branch is used to perform time-frequency transformation on a guided wave signal set, and average pooling is performed on the signal to obtain a dimension-reducing featurePerforming continuous wavelet transformation on the features to output the features:
;
;
;
Wherein CWT (·) represents a continuous wavelet transform of the signal and t represents the signalIs used as a time variable in the time-variable system,Is a wavelet mother-of-wave function,Representation ofA is a scale parameter, controlling the width of the wavelet, b is a translation parameter, determining the position of the wavelet in time;
Acquiring output of wavelet transform Carrying out average pooling dimensionality reduction and neural network convolution to obtain spectrum characteristic output of a final wavelet transformation branch;
The cross entropy loss function more suitable for classification tasks is used in the continuous wavelet transform branches:
;
wherein, the For the cross entropy loss function, Y represents the true probability that the signal feature belongs to a certain class,Representing the predictive score of an input sample of a certain class,Representing the Softmax activation function.
CN202411459600.4A 2024-10-18 2024-10-18 Mamba model-based carbon fiber reinforced composite material damage detection method Active CN119395160B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202411459600.4A CN119395160B (en) 2024-10-18 2024-10-18 Mamba model-based carbon fiber reinforced composite material damage detection method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202411459600.4A CN119395160B (en) 2024-10-18 2024-10-18 Mamba model-based carbon fiber reinforced composite material damage detection method

Publications (2)

Publication Number Publication Date
CN119395160A CN119395160A (en) 2025-02-07
CN119395160B true CN119395160B (en) 2025-11-07

Family

ID=94427365

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202411459600.4A Active CN119395160B (en) 2024-10-18 2024-10-18 Mamba model-based carbon fiber reinforced composite material damage detection method

Country Status (1)

Country Link
CN (1) CN119395160B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116465974A (en) * 2023-02-27 2023-07-21 江苏省特种设备安全监督检验研究院 A Method of Detection and Location of Ultrasonic Guided Wave Defects Based on Singular Value Decomposition
CN117434144A (en) * 2023-09-22 2024-01-23 中国计量科学研究院 An ultrasonic guided wave damage detection and in-situ calibration method

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110389176B (en) * 2019-07-10 2020-09-18 华中科技大学 Fatigue damage diagnosis method for carbon fiber reinforced composite material structure
US20230409901A1 (en) * 2022-05-20 2023-12-21 Salesforce, Inc. Systems and methods for time series forecasting
CN117907445B (en) * 2024-01-13 2024-08-13 大连理工大学 Composite material stiffened plate damage identification method based on ultrasonic guided wave and space-time hybrid network
CN118731170A (en) * 2024-07-25 2024-10-01 山东交通学院 A method and system for detecting internal defects of carbon fiber composite materials without reference

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116465974A (en) * 2023-02-27 2023-07-21 江苏省特种设备安全监督检验研究院 A Method of Detection and Location of Ultrasonic Guided Wave Defects Based on Singular Value Decomposition
CN117434144A (en) * 2023-09-22 2024-01-23 中国计量科学研究院 An ultrasonic guided wave damage detection and in-situ calibration method

Also Published As

Publication number Publication date
CN119395160A (en) 2025-02-07

Similar Documents

Publication Publication Date Title
Pan et al. Intelligent fault identification for industrial automation system via multi-scale convolutional generative adversarial network with partially labeled samples
Li et al. Acoustic emission wave classification for rail crack monitoring based on synchrosqueezed wavelet transform and multi-branch convolutional neural network
Chen et al. Bearing fault diagnosis base on multi-scale CNN and LSTM model
Oliveira et al. Ultrasound-based identification of damage in wind turbine blades using novelty detection
Pan et al. A deep learning network via shunt-wound restricted Boltzmann machines using raw data for fault detection
Zhang et al. Data anomaly detection for bridge SHM based on CNN combined with statistic features
CN112507915A (en) Method for identifying loosening state of bolt connection structure based on vibration response information
Barbosh et al. Acoustic emission-based damage localization using wavelet-assisted deep learning
CN112668527A (en) Ultrasonic guided wave semi-supervised imaging detection method
CN114925724B (en) Mechanical equipment fault diagnosis method, device and storage medium
CN112668526B (en) Bolt group loosening positioning monitoring method based on deep learning and piezoelectric active sensing
CN113092112A (en) Bearing composite fault diagnosis method based on EEMD multi-feature fusion
Tavares et al. Damage detection in lightweight structures using artificial intelligence techniques
Vejdannik et al. Signal processing for NDE
CN117949539B (en) Engineering building strength detection system and detection method thereof
CN119150140A (en) Structural damage detection system based on asymmetric multitask learning
Li et al. Feature extraction methods of ship-radiated noise: From single feature of multi-scale dispersion Lempel-Ziv complexity to mixed double features
Ma et al. Guided wave-based real-time damage detection in composite structures: a gramian angular field image coding lightweight network approach
Hu et al. A novel rolling bearing fault diagnosis method based on parameter optimization variational mode decomposition with feature weighted reconstruction and multi-target attention convolutional neural networks under small samples
Inderyas et al. Deep learning-based acoustic emission signal filtration model in reinforced concrete
Tian et al. Two-stage residual networks for damage identification and location of stiffened composite panel based on guided waves
CN119395160B (en) Mamba model-based carbon fiber reinforced composite material damage detection method
CN112528849B (en) Structure health monitoring method based on inner product matrix and deep learning
CN119466988B (en) Goaf monitoring method and goaf monitoring system for coal mine
CN115452957B (en) Small sample metal damage identification method based on attention prototype network

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