Steel-concrete joint surface defect identification method based on air coupling ultrasonic guided wave

Abstract

The invention discloses a steel-concrete joint surface defect identification method based on air coupling ultrasonic guided waves, which comprises the following steps: S1, selecting an ultrasonic guided wave mode and a probe frequency, and determining an incident angle theta of an air coupling excitation probe; s2, constructing a detection system according to the parameters in the step S1; s3, constructing a data set, wherein the acquisition mode of the sample of the ultrasonic guided wave signal in the data set comprises the following steps of: performing actual detection by adopting the detection system to obtain the sample of the ultrasonic guided wave signal; by solving the air coupling ultrasonic three-dimensional finite element calculation model, obtaining a sample of an ultrasonic guided wave signal under the condition that interference factors and joint surface defects exist; s4, training the recognition model by using the data set; the input of the recognition model is the time domain feature and the frequency domain feature of the ultrasonic guided wave signal, and the output of the recognition model is a defect probability value of a measuring point corresponding to the ultrasonic guided wave signal; and S5, identifying the ultrasonic guided wave signal of each measuring point by adopting the trained identification model so as to calculate the defect probability of each measuring point.

Description

Technical Field

[0001] The invention relates to the field of nondestructive testing of buildings, and in particular to a method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves. Background Art

[0002] Steel-concrete composite structures such as steel tube concrete, steel plate concrete, and steel-concrete composite beams give full play to the mechanical properties of steel and concrete materials, improve the stiffness, bearing capacity, and seismic toughness of the structure, and have been widely used in high-rise and super-high-rise buildings, large-span bridges, cross-sea tunnels and other structures. During the construction process, affected by the construction technology, personnel quality, and complex environment, the steel-concrete interface is prone to defects such as looseness, voids, and inclusions, which will seriously reduce the mechanical and seismic performance of the steel-concrete composite structure. Therefore, accurate identification of defects at the construction site is crucial to evaluating the safety performance of the entire structural system.

[0003] At present, there are two methods for detecting defects on the interface between steel and concrete: ultrasonic method and ultrasonic phased array imaging method. The invention patent "A device and method for detecting the bonding quality of concrete steel plate" (application number: CN201811492889.4) introduces a method of exciting and receiving ultrasonic waves on the concrete side using an air-coupled probe, and judging the bonding quality of the concrete and steel plate interface by the waveform phase. However, this method is detected on the concrete side, and the wavelength of the excited ultrasonic wave is large, the defect recognition ability is weak, and the defect size cannot be quantified. Regarding the ultrasonic phased array imaging method for detecting defects on the steel-concrete interface, the invention patents "A method for detecting the quality of the interface between new and old concrete for strengthening concrete structures" (application number: CN202010628895.9) and "An ultrasonic phased array system and steel tube concrete construction method and process" (application number: CN202210434404.6) introduce the ultrasonic phased array imaging method based on ultrasonic images to judge the construction quality of the steel-concrete interface, but the ultrasonic phased array equipment used in this method mostly uses low-frequency ultrasonic waves as the detection carrier, with a large wavelength and poor defect recognition accuracy. At the same time, both the ultrasonic method and the ultrasonic phased array imaging method are based on the characteristics of a single longitudinal wave signal, such as propagation velocity and signal amplitude for qualitative defect detection. They are easily interfered by reflected signals from steel bars, anchor bolts, steel ribs, etc. in steel-concrete composite structures, resulting in misjudgment of defects. They have poor detection accuracy and low intelligence level, and cannot meet the needs of refined evaluation of the mechanical properties of steel-concrete composite structures, limiting their application in engineering. Summary of the invention

[0004] In view of the above technical problems, the present invention provides a method for identifying defects on the steel-concrete interface based on air-coupled ultrasonic guided waves, which establishes a database of ultrasonic guided wave signals to conduct deep learning of signal features, thereby realizing intelligent identification of defects on the steel-concrete interface. To achieve the above objectives, the technical solutions of the present invention include:

[0005] A method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves comprises the following steps:

[0006] S1: Select the mode and probe frequency of the ultrasonic guided wave, and determine the incident angle θ of the air-coupled excitation probe;

[0007] S2: constructing a detection system according to the parameters determined in step S1;

[0008] S3: constructing a data set, wherein the ultrasonic guided wave signal samples in the data set are obtained by: using the detection system to perform actual detection to obtain ultrasonic guided wave signal samples; and obtaining ultrasonic guided wave signal samples in the presence of interference factors and bonding surface defects by solving an air-coupled ultrasonic three-dimensional finite element calculation model.

[0009] S4: training a recognition model using the data set; the input of the recognition model is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, and the output is the defect probability value of the ultrasonic guided wave signal corresponding to the measuring point;

[0010] S5: The trained recognition model is used to recognize the ultrasonic guided wave signal of each measuring point to calculate the defect probability of each measuring point.

[0011] A further improvement of the present invention is that the step S1 specifically includes: based on the sound propagation theory in the multi-layer dielectric structure and the cyclic flexibility matrix algorithm, establishing the ultrasonic guided wave dispersion equation in the steel and concrete composite structure, solving its phase velocity and group velocity dispersion curves, selecting the mode and probe frequency of the ultrasonic guided wave used for detection, and calculating the incident angle θ of the air-coupled excitation probe according to Snell's law.

[0012] A further improvement of the present invention is that step S2 specifically includes: connecting a transmitting module, a receiving module and a deep learning module, building a detection device using a dual air-coupled ultrasonic probe with one transmitting and one receiving oblique incidence, adjusting the distances from the excitation probe and the receiving probe to the steel plate to be H, and the incident angles to be θ, turning on the power supply to adjust the emission voltage of the excitation unit and the gain of the signal amplifier until an ultrasonic guided wave signal with a maximum signal-to-noise ratio is obtained.

[0013] A further improvement of the present invention is that in step S3, in the process of using the detection system to perform actual detection to obtain samples of ultrasonic guided wave signals, ultrasonic guided wave signals are batch-collected in the laboratory and the engineering site using the data acquisition unit of the detection system, and are stored in the deep learning module after being processed by signal filtering and lightweight units.

[0014] A further improvement of the present invention is that, in step S3, the types of interference factors include: steel bars, anchor bolts, and stiffening ribs.

[0015] A further improvement of the present invention is that in step S3, the process of constructing the data set also includes data preprocessing and labeling. During the data preprocessing and labeling process, the time domain characteristics and frequency domain characteristics of each ultrasonic guided wave signal sample in the data set are extracted, interference signals and defect signals are labeled, and interfering guided wave modes are eliminated through wavelet filtering and two-dimensional Fourier transform technology; the data set is divided into a training set, a verification set and a test set.

[0016] A further improvement of the present invention is that in step S4, the recognition model includes a deep Boltzmann machine, adopts a multiple hidden layer structure mode, the input is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, the output is the recognition feature, and the formula for calculating the defect probability value based on the output recognition feature is:

[0017]

[0018] Where x1 and x2 are the calculated values ​​of normal and defect recognition features in the network structure respectively;

[0019] During the training process, the domain invariant features of the guided wave time-frequency image are extracted by maximizing the domain recognition error and minimizing the probability distribution distance to calculate the defect probability value, and two-dimensional imaging is performed to identify defects on the steel-concrete interface.

[0020] A further improvement of the present invention is that in step S4, the recognition model includes a convolutional neural network, the input is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, and the output is the defect probability value.

[0021] A further improvement of the present invention is that, in step S4, the ultrasonic guided wave signal is subjected to time-frequency signal conversion, signal component decomposition, and frequency band decomposition to obtain the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal.

[0022] The beneficial effects of the present invention include:

[0023] 1) The database of ultrasonic guided wave signals is established through laboratory specimens, finite element calculations and engineering site detection, and the ultrasonic guided wave signals have a wide coverage;

[0024] 2) Compared with traditional ultrasonic testing, the air-coupled ultrasonic probe does not require the addition of coupling agent, making on-site testing easy and efficient;

[0025] 3) The deep learning algorithm is used to accurately identify interference signals and defect signals such as steel bars, anchor bolts, and steel ribs, realizing quantitative evaluation of defects on the steel-concrete interface with high defect detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a steel-concrete joint surface defect identification device based on air-coupled ultrasonic guided waves in the present invention;

[0027] Figure 2 It is a schematic diagram of air-coupled ultrasonic guided wave propagation in a steel and concrete structure in the present invention;

[0028] Figure 3 This is a schematic diagram of the principle of ultrasonic guided wave defect recognition in the present invention;

[0029] Figure 4 The figure is a flow chart of the method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves in the present invention. DETAILED DESCRIPTION

[0030] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0031] See also Figure 1 An embodiment of the present invention provides an air-coupled ultrasonic guided wave intelligent detection device, which is used to implement the steel-concrete joint surface defect identification method based on air-coupled ultrasonic guided waves of the present invention. The system includes an ultrasonic guided wave transmitting module 1, an ultrasonic guided wave receiving module 2 and a guided wave signal deep learning module 3.

[0032] The ultrasonic guided wave transmitting module 1 comprises: a power supply 11, an excitation unit 12, a signal amplifier 13, an air-coupled excitation probe 14 and a signal synchronization unit 15. The power supply 11 supplies power to the excitation unit 12, and the excitation unit 12 is used to generate a driving signal, which drives the air-coupled excitation probe 14 after being amplified by the signal amplifier 13.

[0033] The ultrasonic guided wave receiving module 2 includes: an air-coupled receiving probe 21, a signal amplifier 22, a data acquisition unit 23, a signal filtering and lightweight unit 24 and a sound absorption unit 25. The air-coupled receiving probe 21 is used to receive the ultrasonic guided wave signal, and the received signal is amplified by the signal amplifier 22 and processed by the data acquisition unit 23. The data acquisition unit 23 is also connected to the excitation unit 12 through the signal synchronization unit 15 to ensure signal synchronization. A sound absorption unit 25 is provided between the ultrasonic guided wave transmitting module 1 and the ultrasonic guided wave receiving module 2 to prevent direct sound wave signal interference between the two.

[0034] The waveguide signal deep learning module 3 includes: a database 31, a preprocessing unit 32, a labeling unit 33, a deep learning unit 34, a damage probability value prediction unit 35 and a display unit 36. The connection relationship is as follows: Figure 1 The waveguide signal deep learning module 3 may use a commercially available deep learning server, and the deep learning unit 34 inside it may use a dedicated deep learning processor (NPU) or a deep learning model implemented by software and executed by a general-purpose CPU or GPU.

[0035] In this embodiment, the air-coupled ultrasonic probes used are divided into two types: one is a planar probe with three specifications of center frequencies of 100kHz, 200kHz and 400kHz, and diameters of 32mm, 24mm and 16mm respectively; the other is a focused probe with three specifications of center frequencies of 100kHz, 200kHz and 400kHz, and focal lengths of 40mm, 38mm and 36mm respectively. The excitation probe 14 is a planar probe, and the receiving probe 21 can select a planar probe or a focused probe according to different scenarios.

[0036] In this embodiment, the sound absorbing unit 25 may be made of sound absorbing foam, a porous structure or an acoustic metamaterial, and may absorb direct waves from the excitation probe to the receiving probe in the air.

[0037] In this embodiment, the database 31 includes a training set, a validation set and a test set, including defect-free signals and damaged signals, and the data sources include experimental specimen testing, finite element simulation and engineering site inspection.

[0038] In this embodiment, the deep learning unit 34 utilizes an unsupervised domain adaptive network to perform training and verification based on the characteristics of the waveguide signal, extracts the domain invariant features of the waveguide time-frequency image by maximizing the domain recognition error and minimizing the probability distribution distance, and calculates the defect probability value to perform two-dimensional imaging to identify steel-concrete interface defects.

[0039] See also Figure 2 to Figure 4A method for identifying defects on a steel-concrete interface based on air-coupled ultrasonic guided waves according to an embodiment of the present invention comprises the following steps:

[0040] S1: Select the ultrasonic guided wave mode and probe frequency, and determine the incident angle θ of the air-coupled excitation probe.

[0041] This step specifically includes: based on the sound propagation theory in multi-layer dielectric structures and the cyclic flexibility matrix algorithm, establishing the dispersion equation of ultrasonic guided waves in steel and concrete composite structures, solving its phase velocity and group velocity dispersion curves, selecting the mode and probe frequency of the ultrasonic guided waves used for detection, and calculating the incident angle θ of the air-coupled excitation probe according to Snell's law.

[0042] In the specific implementation process, step S1 includes:

[0043] S11: The steel-concrete structure is regarded as a multi-layer plate structure of air-steel plate-concrete plate-air. Based on the wave equation of the single-layer medium state space, the cyclic recursive algorithm of the stiffness matrix of the single-layer medium, the spring model and the displacement and stress continuity on the interface between layers, the overall flexibility matrix of the multi-layer plate structure is calculated:

[0044]

[0045] Where: S M is the overall flexibility matrix of the m-layer plate structure, S M-1 is the overall flexibility matrix of the upper m-1 layers, S m is the flexibility matrix of the mth layer; is the matrix S M The second-order submatrix in the upper left corner, is the matrix S M The second-order submatrix in the upper right corner, is the matrix S M The second-order submatrix in the lower left corner, is the matrix S M The second-order submatrix in the lower right corner;

[0046] S12: Based on the zero stress boundary condition, the dispersion equation of the waveguide in the multi-layer dielectric structure is obtained:

[0047] det[S M (k x ,ω) -1 ]=0

[0048] Where: k x represents the wave number in the propagation direction x, ω represents the angular frequency, and det[·] represents the determinant of the matrix;

[0049] S13: The phase velocity dispersion curve of ultrasonic guided waves in steel-concrete structures is obtained by iteratively solving the dispersion equation. The types and thicknesses of the steel plates and concrete in the steel-concrete structures are measured respectively. Low-frequency mode ultrasonic guided waves are selected for steel-concrete interface defect detection. The incident angle θ of this mode is calculated based on Snell's law as:

[0050] θ=arc -1 (C a / C p )

[0051] Among them C p is the phase velocity of ultrasonic guided wave propagation, C a is the speed of sound waves in air;

[0052] S14: According to the frequency-thickness product of the selected ultrasonic guided wave mode, the thickness of the steel plate and the concrete, the wave structure of the ultrasonic guided wave at different frequencies is calculated, the corresponding frequency f is determined, the center frequency f0 of the air-coupled probe is selected as the probe frequency, and the frequency f is within the working frequency range of the air-coupled probe. In this embodiment, the optimal ultrasonic guided wave mode and frequency are selected by theoretically analyzing and calculating the dispersion curve and wave structure to perform steel-concrete interface defect detection, and the defect recognition sensitivity is high.

[0053] S2: Construct a detection system according to the parameters determined in step S1.

[0054] This step specifically includes: connecting the transmitting module 1, the receiving module 2 and the deep learning module 3, building a detection device using a dual air-coupled ultrasonic probe with one transmitting and one receiving oblique incidence, adjusting the distance between the excitation probe 14 and the receiving probe 21 (both the excitation probe 14 and the receiving probe 21 are air-coupled ultrasonic probes) and the steel plate to be H, and the incident angle to be θ, turning on the power to adjust the transmitting voltage of the excitation unit and the gain of the signal amplifier until an ultrasonic guided wave signal with a maximum signal-to-noise ratio is obtained.

[0055] S3: constructing a data set, wherein the method for obtaining samples of ultrasonic guided wave signals in the data set includes:

[0056] (1) A sample of an ultrasonic guided wave signal obtained by actual detection using the detection system;

[0057] (2) Obtain samples of ultrasonic guided wave signals in the presence of interference factors and joint surface defects by solving the air-coupled ultrasonic three-dimensional finite element calculation model. In the second acquisition method, the types of interference factors include: steel bars, anchor bolts, and stiffening ribs. Various interference factors are preset in the air-coupled ultrasonic three-dimensional finite element calculation model. By solving the sound field of the air-coupled ultrasonic three-dimensional finite element calculation model, the leakage wave signal of the ultrasonic guided wave in the air can be extracted, and its form is consistent with the measured ultrasonic guided wave signal, so it is used as a sample of the ultrasonic guided wave signal.

[0058] In step S3, in the process of using the detection system to perform actual detection to obtain samples of ultrasonic guided wave signals, ultrasonic guided wave signals are batch collected in the laboratory and the engineering site using the data acquisition unit of the detection system, and are processed by the signal filtering and lightweight unit 24 and stored in the deep learning module 3.

[0059] In step S3, the process of constructing the data set also includes data preprocessing and annotation. In the process of data preprocessing and annotation, the time domain characteristics and frequency domain characteristics of each ultrasonic guided wave signal sample in the data set are extracted, the interference signal and defect signal are annotated, and the interfering guided wave mode is eliminated by wavelet filtering and two-dimensional Fourier transform technology; the data set is divided into a training set, a verification set and a test set. In the process of eliminating the interfering guided wave mode, other ultrasonic guided wave modes are eliminated according to the ultrasonic guided wave mode determined in step S1. In some embodiments, the two-dimensional Fourier transform technology can transform the ultrasonic guided wave signal in the time-space domain into a frequency-wavenumber domain image, and different guided wave modes can be screened and eliminated in combination with the dispersion curve.

[0060] In the specific implementation process, step S3 is implemented based on the database management system MySQL. The process includes: converting the ultrasonic guided wave signal into a signal matrix, encoding and setting indexes based on the database management system MySQL, establishing an ultrasonic guided wave signal database 31, and dividing it into a training set, a validation set, and a test set for database performance testing and optimization.

[0061] S4: using the data set to train the recognition model; the input of the recognition model is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, and the output is the defect probability value of the ultrasonic guided wave signal corresponding to the measuring point.

[0062] In some embodiments, the recognition model includes a deep Boltzmann machine, which adopts a multiple hidden layer structure mode, the input is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, and the output is the recognition feature. The formula for calculating the defect probability value based on the output recognition feature is:

[0063]

[0064] Among them, x1 and x2 are the calculated values ​​of normal and defect recognition features in the network structure respectively. The SoftMax formula indicates that the more defect features the ultrasonic guided wave signal has, the closer the probability value of the defect is to 1, and vice versa. The embedded software has the functions of focused imaging and automatic calculation of defect area. The calculation accuracy of defect area can reach 0.1mm2, and the detection results can be quickly given at the engineering site.

[0065] During the training process, the domain invariant features of the guided wave time-frequency image are extracted by maximizing the domain recognition error and minimizing the probability distribution distance to calculate the defect probability value, and two-dimensional imaging is performed to identify defects on the steel-concrete interface.

[0066] The Deep Boltzmann Machine is an unsupervised deep learning model; its data does not need to be labeled. However, the labeled data can provide an accurate reference during the verification and model fine-tuning process, which can effectively improve the accuracy of the final model.

[0067] In other embodiments, the recognition model includes a convolutional neural network, the input is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, and the output is the defect probability value.

[0068] The ultrasonic guided wave signal is subjected to time-frequency signal conversion, signal component decomposition, and frequency band decomposition to obtain the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal.

[0069] S5: The trained recognition model is used to recognize the ultrasonic guided wave signal of each measuring point to calculate the defect probability of each measuring point. Two-dimensional imaging is performed based on the defect probability value and the embedded software in the display unit 35 to realize intelligent recognition of steel-concrete interface defects and automatically calculate the area of ​​the defects.

[0070] In this embodiment, the ultrasonic guided waves in the plate-like structure are formed by multiple reciprocating reflections of transverse waves and longitudinal waves in the plate, and further superposition interference and geometric dispersion. It is usually required that the wavelength of the sound wave is similar to or greater than the thickness of the structure. When the ultrasonic guided waves propagate in the plate-like structure, each particle in the plate participates in the vibration, which can effectively identify hidden defects such as looseness, cracking, and delamination in the structure.

[0071] In this embodiment, when the ultrasonic guided wave encounters defects or areas with changed impedance values, reflection, scattering, and diffraction may occur. The characteristics of different signals and mode conversion phenomena are analyzed based on a deep learning algorithm to accurately identify interference signals and defect signals at steel bars, anchor bolts, steel ribs, etc.

[0072] In this embodiment, in step S1, the ultrasonic guided wave mode for detection should be selected based on the principles of weak dispersion effect, small number of guided wave modes, and large out-of-plane displacement, taking into account factors such as frequency range, strength of ultrasonic guided wave leakage signal in the air, mode conversion and signal attenuation.

[0073] In this embodiment, the samples of ultrasonic guided wave signals include reflection signals from steel bars, anchor bolts, stiffening ribs, etc., and also include ultrasonic guided wave signals from normal locations and defects of different locations and sizes.

[0074] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves, characterized in that: The following steps are involved: S1: Select the mode and probe frequency of the ultrasonic guided wave, and determine the incident angle θ of the air-coupled excitation probe; S2: constructing a detection system according to the parameters determined in step S1; S3: constructing a data set, wherein the method for obtaining the ultrasonic guided wave signal samples in the data set includes: using the detection system to perform actual detection to obtain the ultrasonic guided wave signal samples; By solving the air-coupled ultrasonic three-dimensional finite element calculation model, samples of ultrasonic guided wave signals in the presence of interference factors and bonding surface defects are obtained; S4: training a recognition model using the data set; the input of the recognition model is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, and the output is the defect probability value of the ultrasonic guided wave signal corresponding to the measuring point; S5: The trained recognition model is used to recognize the ultrasonic guided wave signal of each measuring point to calculate the defect probability of each measuring point.

2. According to claim 1, a method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves is characterized in that: The step S1 specifically includes: based on the sound propagation theory in the multilayer dielectric structure and the cyclic flexibility matrix algorithm, establishing the dispersion equation of ultrasonic guided waves in the steel and concrete composite structure, solving its phase velocity and group velocity dispersion curves, selecting the mode and probe frequency of the ultrasonic guided waves used for detection, and calculating the incident angle θ of the air-coupled excitation probe according to Snell's law.

3. The method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves according to claim 1 is characterized in that: Step S2 specifically includes: connecting the transmitting module, the receiving module and the deep learning module, building a detection device using a dual air-coupled ultrasonic probe with one transmitting and one receiving oblique incidence, adjusting the distances from the excitation probe and the receiving probe to the steel plate to H, and the incident angles to θ, turning on the power to adjust the transmitting voltage of the excitation unit and the gain of the signal amplifier until an ultrasonic guided wave signal with a maximum signal-to-noise ratio is obtained.

4. The method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves according to claim 3 is characterized in that: In step S3, in the process of using the detection system to perform actual detection to obtain samples of ultrasonic guided wave signals, ultrasonic guided wave signals are collected in batches in the laboratory and the engineering site using the data acquisition unit of the detection system, and are stored in the deep learning module after being processed by signal filtering and lightweight units.

5. The method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves according to claim 3 is characterized in that: In step S3, the types of interference factors include: steel bars, anchor bolts, and stiffening ribs.

6. The method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves according to claim 1 is characterized in that: In step S3, the process of constructing the data set also includes data preprocessing and labeling. During the data preprocessing and labeling process, the time domain characteristics and frequency domain characteristics of each ultrasonic guided wave signal sample in the data set are extracted, the interference signal and defect signal are labeled, and the interfering guided wave mode is eliminated through wavelet filtering and two-dimensional Fourier transform technology; the data set is divided into a training set, a validation set and a test set.

7. The method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves according to claim 1 is characterized in that: In step S4, the recognition model includes a deep Boltzmann machine, adopts a multiple hidden layer structure mode, inputs the time domain features and frequency domain features of the ultrasonic guided wave signal, and outputs the recognition features. The formula for calculating the defect probability value based on the output recognition features is: Where x1 and x2 are the calculated values ​​of normal and defect recognition features in the network structure respectively; During the training process, the domain invariant features of the guided wave time-frequency image are extracted by maximizing the domain recognition error and minimizing the probability distribution distance to calculate the defect probability value, and two-dimensional imaging is performed to identify defects on the steel-concrete interface.

8. The method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves according to claim 1 is characterized in that: In step S4, the recognition model includes a convolutional neural network, the input is the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal, and the output is the defect probability value.

9. A method for identifying defects on steel-concrete joint surfaces based on air-coupled ultrasonic guided waves according to claim 7 or 8, characterized in that: In step S4, the ultrasonic guided wave signal is subjected to time-frequency signal conversion, signal component decomposition, and frequency band decomposition to obtain the time domain characteristics and frequency domain characteristics of the ultrasonic guided wave signal.

Images