Signal processing method and system for ultrasonic detection of internal defects of building concrete

Abstract

The application discloses a building concrete internal defect ultrasonic detection signal processing method and system, belongs to the material nondestructive testing technical field, and the method comprises the following steps: configuring a full matrix capture array to collect multi-channel original time domain data; performing a time domain flip operation on the data and mapping to the full matrix capture array to perform a synchronous retransmission algorithm; extracting sum frequency and difference frequency amplitudes of specific frequency components in the matrix, and constructing a nonlinear amplitude coordinate mapping table; performing statistical analysis on a full-field energy focusing distribution matrix, performing a difference operation with the sum frequency component amplitudes to generate a defect response coordinate set; and feeding back the coordinate set to the full matrix capture array to trigger supplementary scanning and update the matrix. The application adopts dual-frequency nonlinear modulation and time reversal focusing technology, can effectively suppress aggregate scattering noise interference, and improves the three-dimensional imaging precision and detection signal-to-noise ratio of concrete micro defects.

Description

Technical Field

[0001] This invention belongs to the field of nondestructive testing technology for materials, and particularly relates to a method and system for processing ultrasonic detection signals of internal defects in building concrete. Background Technology

[0002] Concrete structures are widely used in modern engineering fields such as bridges, tunnels, and high-rise buildings. They are the core foundational components of building engineering. During the pouring, construction, and service operation stages, they are easily affected by material ratios, construction techniques, environmental loads, and long-term fatigue, which can easily lead to hidden defects such as cracks, voids, and looseness, directly threatening structural safety and service life. Ultrasonic testing has become the mainstream technology for non-destructive testing of internal defects in concrete due to its advantages such as non-invasiveness, ease of operation, and suitability for on-site testing. Related technologies have formed a mature application system. For example, the invention patent application with publication number CN119936213A discloses a non-destructive testing system and method for concrete structure defects, which represents a typical technical path in the current industry: This patent deploys ultrasonic detectors, emits ultrasonic signals, and collects multiple sets of raw echo data. After data preprocessing, the data is input into a real-time defect identification model built based on MP-FA-GRU. Multi-path feature extraction and frequency domain attention and gated loop fusion modules are used to complete defect classification and performance verification. This is a representative technology for defect identification based on deep learning.

[0003] Existing ultrasonic testing technologies, exemplified by this patent, have significant technical shortcomings, making it difficult to meet the demands for high-precision and high-reliability testing. Firstly, signal processing relies on conventional filtering and denoising preprocessing, which, while suppressing noise from concrete aggregate scattering, easily filters out weak response signals from minute defects, resulting in the loss of crucial information and reducing the detection rate of minor defects. Secondly, the core technology focuses on signal feature classification using deep learning models, only outputting defect category labels and failing to achieve precise coordinate positioning and geometric imaging of defects within the three-dimensional space of concrete, thus unable to intuitively present the true distribution of defects. Thirdly, it does not address the nonlinear acoustic characteristics of concrete, making it difficult to separate defect signals from aggregate scattering background noise from a physical perspective, resulting in a low signal-to-noise ratio and a high false alarm rate, easily misjudging linear scatterers like aggregates as defects. Fourthly, it employs a global uniform scanning and calculation mode without an intelligent focusing detection mechanism, leading to a large amount of redundant calculation in defect-free areas, making it difficult to balance detection efficiency and imaging accuracy, and unable to simultaneously achieve rapid screening over a wide area and fine imaging of minute defects. Summary of the Invention

[0004] To address the aforementioned problems, the present invention aims to provide a method and system for processing ultrasonic detection signals of internal defects in building concrete. By employing dual-frequency nonlinear modulation and time-reversal focusing technology, it can effectively suppress aggregate scattering noise interference and improve the three-dimensional imaging accuracy and detection signal-to-noise ratio of minor concrete defects.

[0005] The above objectives can be achieved through the following approach: The method for processing ultrasonic detection signals of internal defects in building concrete includes the following steps: S1. Drive the dual-source broadband excitation probe to emit a combined pulse wave to the concrete structure under test, and configure a full matrix acquisition array to collect multi-channel raw time domain data. The combined pulse wave includes a first frequency and a second frequency. S2. Perform time-domain flipping operation on the multi-channel raw time-domain data to construct a time-reversed retransmission signal set, and map it to the full matrix capture array to perform synchronous retransmission calculation and generate a full-field energy focusing distribution matrix; S3. Based on the full-field energy focusing distribution matrix, perform specific frequency component extraction with the sum and difference of the first and second frequencies as the center frequency, extract the sum frequency component amplitude and the difference frequency component amplitude, and construct a nonlinear amplitude coordinate mapping table. S4. Based on the statistical analysis of the full-field energy focusing distribution matrix, generate the aggregate scattering background reference value, and perform a difference operation between the sum frequency component amplitude and the aggregate scattering background reference value to generate a set of defect response coordinates. S5. Feed back the defect response coordinate set to the full matrix capture array, trigger supplementary scanning for the defect response coordinate set, update the full field energy focusing distribution matrix, and output the three-dimensional reconstruction model of the defect.

[0006] Preferably, in step S1, configuring the full matrix acquisition array to acquire multi-channel raw time-domain data includes: The first and second transducer crystals in the synchronously driven dual-source broadband excitation probe are used to emit combined pulse waves to the concrete structure under test. Configure a full matrix capture array to perform line-by-line cyclic scanning, acquire signals from each channel of the concrete structure under test in real time, and construct multi-channel raw time-domain data.

[0007] Preferably, in S2, generating the full-field energy focusing distribution matrix includes: Read the raw time-domain data from multiple channels, perform linear reverse rearrangement based on the end of the time axis to the beginning, and construct a time-reversed retransmission signal set; A discretized spatial pixel grid is established for the concrete structure under test. Based on the discretized spatial pixel grid and the full matrix capture array, spatial coherence superposition is performed to generate a full-field energy focusing distribution matrix.

[0008] Preferably, spatial coherence is performed based on a discretized spatial pixel grid and a full-matrix capture array, specifically as follows: The acoustic path transit time is calculated using a full matrix capture array and a discretized spatial pixel grid, and a transit time index table is constructed. Based on the transit time index table, the channel amplitude is extracted from the time-reversed retransmission signal set, and the channel amplitude is accumulated and superimposed to generate the full-field energy focusing distribution matrix.

[0009] Preferably, in step S3, constructing the nonlinear amplitude coordinate mapping table includes: The time-domain reconstructed waveforms of the grid nodes are extracted based on the full-field energy focusing distribution matrix, and a fast Fourier transform is performed to generate a full-field spectrum data matrix. The spectral index interval is calculated with the sum and difference of the first and second frequencies as the center frequency, and the sum frequency component amplitude and difference frequency component amplitude of the spectral index interval are locked in the full-field spectral data matrix. The complex modulus value is calculated based on the sum-frequency component amplitude and the difference-frequency component amplitude. The complex modulus value is used as the nonlinear amplitude and associated with the spatial coordinates of the grid nodes to construct a nonlinear amplitude coordinate mapping table.

[0010] Preferably, after generating the full-field energy focusing distribution matrix, the full-field energy focusing distribution matrix is ​​used as the full-band energy reference, and the normalized ratio calculation is performed on the nonlinear amplitude coordinate mapping table to generate a nonlinear modulation coefficient distribution map.

[0011] Preferably, in step S4, generating the defect response coordinate set includes: Perform global amplitude histogram statistics on the full-field energy focusing distribution matrix to construct the probability density function curve; Perform parameter fitting iteration on the probability density function curve, and extract the fitting mean parameter to generate the aggregate scattering background baseline value; The sum-frequency component amplitude is numerically subtracted from the aggregate scattering background reference value. If the result is positive, the matrix position index is extracted to generate a set of defect response coordinates.

[0012] Preferably, the fitted mean parameters are extracted to generate the aggregate scattering background baseline value, specifically as follows: The global maximum coordinates are searched based on the probability density function curve to determine the initial center parameters, which drive the preset Gaussian distribution model. Least square fitting iteration is performed on the probability density function curve to obtain a convergent Gaussian distribution model. The baseline value of aggregate scattering background is obtained by extracting the fitting mean parameters based on the convergent Gaussian distribution model.

[0013] Preferably, in step S5, the output defect 3D reconstruction model includes: Traverse the defect response coordinate set to extract the defect response coordinates, and expand the pixel grid unit outward from the defect response coordinate to construct a local supplementary scan window; The full-matrix capture array is driven to perform focused resampling on spatial pixels within the local supplementary scan window at a sampling interval of half the spacing of the discretized spatial pixel grid, and then overwrites the full-field energy focusing distribution matrix. Based on the nonlinear modulation coefficient distribution map, a mask filter is applied to the full-field energy focusing distribution matrix to extract the defect response coordinate data for triangulation meshing, and a three-dimensional reconstruction model of the defect is output.

[0014] An ultrasonic testing signal processing system for internal defects in building concrete, used to implement the above-mentioned method, includes: The dual-frequency combined pulse acquisition module is used to drive the dual-source broadband excitation probe to emit combined pulse waves to the concrete structure under test. It is configured with a full matrix acquisition array to acquire multi-channel raw time-domain data. The combined pulse wave includes a first frequency and a second frequency. The time-reversal focusing calculation module is used to perform time-domain reversal operations on multi-channel raw time-domain data, construct a time-reversal retransmission signal set, and map it to the full matrix capture array to perform synchronous retransmission calculations and generate a full-field energy focusing distribution matrix. The nonlinear amplitude coordinate mapping module is used to extract specific frequency components based on the full-field energy focusing distribution matrix, with the sum and difference of the first and second frequencies as the center frequency, extract the sum frequency component amplitude and the difference frequency component amplitude, and construct a nonlinear amplitude coordinate mapping table. The aggregate background differential screening module is used to generate aggregate scattering background benchmark values ​​based on statistical analysis of the full-field energy focusing distribution matrix, and to perform differential operation between the sum frequency component amplitude and the aggregate scattering background benchmark values ​​to generate a set of defect response coordinates. The defect 3D reconstruction output module is used to feed back the defect response coordinate set to the full matrix capture array, trigger supplementary scanning for the defect response coordinate set, update the full field energy focusing distribution matrix, and output the defect 3D reconstruction model.

[0015] The present invention has the following advantages: This invention, by actively exciting and demodulating the nonlinear frequency components unique to defects, can effectively separate weak defect signals from strong aggregate scattering background noise from a physical mechanism perspective, thereby improving the signal-to-noise ratio of detection and enhancing the ability to identify and detect early and minute defects inside concrete. This invention constructs and utilizes a data-driven nonlinear spatial matched filter to achieve direct, high-contrast imaging from signal features to the spatial location of defects. This imaging mechanism based on differences in material physical properties, compared to traditional methods relying on acoustic impedance differences, can distinguish between real defects and normal structural inhomogeneities, thereby reducing the false alarm rate and improving the reliability of diagnostic results. The intelligent closed-loop detection process of "global initial screening and local fine inspection" established in this invention can focus high-precision detection resources on potential defect areas. This intelligent resource scheduling not only improves the overall detection efficiency and avoids redundant calculations for defect-free areas, but also ensures the accuracy of detailed characterization of key defect areas, thus resolving the contradiction between high efficiency and high precision in traditional methods. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the method of the present invention; Figure 2 This is a comparison diagram of the acoustic spectrum response of the complete region and the defective region in Embodiment 1 of the present invention; Figure 3 This is the histogram of the full-field nonlinear amplitude probability density distribution and the fitted graph of the aggregate scattering background in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the system of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0018] Example 1: As Figure 1 As shown, the ultrasonic signal processing method for internal defects in building concrete includes the following steps: S1. Drive the dual-source broadband excitation probe to emit a combined pulse wave to the concrete structure under test, and configure a full matrix acquisition array to collect multi-channel raw time domain data. The combined pulse wave includes a first frequency and a second frequency. S2. Perform time-domain flipping operation on the multi-channel raw time-domain data to construct a time-reversed retransmission signal set, and map it to the full matrix capture array to perform synchronous retransmission calculation and generate a full-field energy focusing distribution matrix; S3. Based on the full-field energy focusing distribution matrix, perform specific frequency component extraction with the sum and difference of the first and second frequencies as the center frequency, extract the sum frequency component amplitude and the difference frequency component amplitude, and construct a nonlinear amplitude coordinate mapping table. S4. Based on the statistical analysis of the full-field energy focusing distribution matrix, generate the aggregate scattering background reference value, and perform a difference operation between the sum frequency component amplitude and the aggregate scattering background reference value to generate a set of defect response coordinates. S5. Feed back the defect response coordinate set to the full matrix capture array, trigger supplementary scanning for the defect response coordinate set, update the full field energy focusing distribution matrix, and output the three-dimensional reconstruction model of the defect.

[0019] In S1, configuring the full matrix capture array to acquire multi-channel raw time-domain data includes: The first and second transducer crystals in the synchronously driven dual-source broadband excitation probe are used to emit combined pulse waves to the concrete structure under test. During the drive and launch phase, a multi-channel arbitrary waveform generator or high-voltage power amplifier is used as the excitation source, independently connected to the first and second transducer crystals of the dual-source broadband excitation probe. To generate a nonlinear modulation effect within the concrete medium, the excitation source needs to generate two pulse signals with different frequencies but strictly synchronized timing. The generation of this combined pulse wave is not a simple voltage superposition, but rather achieved through spatial coupling of the physical sound field. The construction logic of its time-domain excitation function follows the principle of linear superposition, but induces a nonlinear response during medium propagation. The waveform characteristics of this combined pulse wave are defined using the following dual-frequency excitation signal model: ; in, The amplitude of the combined pulse wave excitation signal at time t; Representing the first frequency, this parameter is selected based on the principle of low-frequency carrier, and is usually set in the frequency band where the scattering noise of concrete aggregate is relatively small, such as 20kHz-50kHz, in order to ensure the penetration depth of sound waves in thick concrete components. This represents the second frequency. The principle for selecting this parameter is high-frequency modulation wave. It is usually set to 3 to 5 times the first frequency, such as 100kHz-200kHz, to improve the sensitivity to microcracks and intermodulate with low-frequency waves. and These represent the voltage amplitudes of two frequency components, respectively. By adjusting the ratio between the two, the distribution of mixing energy can be controlled. and This represents the initial phase, which is usually set to 0 to achieve synchronous excitation. This represents a time-domain window function used to limit the duration of a pulse, prevent sidelobe leakage, and improve time resolution.

[0020] For example, in one application scenario, when testing a C30 strength concrete shear wall, the average sound velocity of the concrete was first measured to be 4000 m / s. Based on this, a first frequency was set. The frequency was set to 50kHz to ensure the wavelength was sufficient to bypass coarse aggregate with a diameter of approximately 20mm; a second frequency was then set. The excitation voltage amplitude is 150 kHz to induce a nonlinear breathing effect in potential microcracks. and Both are set to 200V. When the excitation command is activated, the dual-source probes simultaneously radiate two beams of sound into the interior of the wall. The two beams of sound overlap on the propagation path, forming a combined pulse wave sound field with specific spectral characteristics.

[0021] Configure a full matrix capture array to perform line-by-line cyclic scanning, acquire signals from each channel of the concrete structure under test in real time, and construct multi-channel raw time-domain data.

[0022] During the data acquisition phase, a full matrix acquisition array (FMC) controller is used to perform strict timing control. This FMC array consists of N independent piezoelectric crystal elements. The controller sequentially controls the i-th element as the transmitter and simultaneously controls all N elements as receivers, recording the reflected echoes. The data generated in this process is constructed as a three-dimensional data tensor or a two-dimensional signal matrix set. To clarify the physical meaning and storage structure of the data, a multi-channel raw time-domain data matrix is ​​defined. as follows: ; in, represents the full matrix data set at sampling time t; N represents the total number of array elements in the full matrix acquisition array, which determines the array's aperture size and imaging resolution; Let represent the original time-domain echo signal transmitted by the i-th array element and received by the j-th array element. This signal not only includes the linear echo reflected by the defect but also implicitly contains the nonlinear intermodulation component excited by the combined pulse wave. The value of t ranges from... ,in To determine the maximum recording duration, it must be greater than the time required for the sound wave to travel one round trip through the farthest sound path in the structure under test.

[0023] To ensure signal integrity, the sampling frequency of the analog-to-digital converter... The Nyquist sampling theorem must be followed, and the detection requirements for nonlinear higher harmonics must be considered. The sampling frequency... The optimal setting is 5 times or more of the highest frequency component.

[0024] For example, in one application example, a linear full-matrix capture array with 64 elements is used, i.e., N=64. To capture sum-frequency components and their harmonics up to 200kHz, the sampling frequency is... Set to 50MHz. Set maximum recording duration. 200 microseconds, which means for each channel The system will record 10,000 sampling points. During the scanning process, the first array element emits a combined wave, which is simultaneously received by array elements 1 to 64, generating the first row of data in the matrix. Then, the second array element emits the wave, which is again received by array elements 1 to 64, generating the second row of data. This process repeats until the 64th array element has emitted its wave. Ultimately, a multi-channel raw time-domain data volume with dimensions of 64×64×10000 is constructed. This data volume completely records all acoustic scattering information within the concrete, providing a complete information source for time reversal and nonlinear feature extraction.

[0025] In S2, the generation of the full-field energy focusing distribution matrix includes: Read the raw time-domain data from multiple channels, perform linear reverse rearrangement based on the end of the time axis to the beginning, and construct a time-reversed retransmission signal set; Unlike traditional physical retransmission, this method performs time-series flipping on the acquired data in computer memory. First, it retrieves the raw time-domain data from multiple channels in memory. The length of its sampling time axis is determined to be .for Perform the following time-domain flipping operation: ; in, The time-reversed retransmission signal representing the i-th transmit channel (corresponding to the i-th array element transmit) and the j-th receive channel (corresponding to the j-th array element receive) is the cornerstone of constructing the virtual retransmission field; Indicates to Perform time domain flipping (time reversal).

[0026] According to the time-reversal invariance of the wave equation, if the received scattered signal is reversed in time and used as a virtual "retransmission" source, the sound wave will automatically trace back along the original path and the energy will be spatiotemporally focused at the scattering source.

[0027] For example, in one application example, suppose the full matrix capture array contains 64 elements, the sampling frequency is 50MHz, and the maximum record length in a single acquisition is... The time interval is 200 microseconds. For the channel where element 1 transmits and element 32 receives, it is assumed that in the original data, there is a strong echo peak formed by defect reflection at the 1500th sampling point. After performing linear reordering, this peak is "moved" to the 8500th sampling point in the new time-reversed retransmission signal set. This operation prepares a virtual wave source with "adaptive focusing" characteristics for the synchronous retransmission calculation.

[0028] A discretized spatial pixel grid is established for the concrete structure under test. Based on the discretized spatial pixel grid and the full matrix capture array, spatial coherence superposition is performed to generate a full-field energy focusing distribution matrix.

[0029] First, define the ROI (Region of Interest) for the two-dimensional or three-dimensional imaging of the concrete structure to be measured, and divide it into a dense, discretized spatial pixel grid. Each grid node... This is considered a potential focusing target point. Subsequently, for any pixel grid point P in space, its full-field energy focusing amplitude is calculated. This calculation is based on the principle of Total Focusing Method (TFM), which involves delaying and superimposing the signals from all N×N transmit and receive paths in the entire matrix. This core computation is strictly defined by the following formula: ; in, Representing coordinates The energy focusing amplitude at the discrete spatial pixel grid point; the larger the value, the higher the probability that there is a defect at that point. This indicates a traversal of the entire matrix data; and These represent the physical horizontal coordinates of the i-th transmitting element and the j-th receiving element, respectively. The longitudinal wave velocity of ultrasound in the tested concrete medium is represented by this parameter, which is obtained through calibration experiments and typically ranges from 3500 m / s to 4500 m / s. (Molecular part) Represents the total sound path, which is the geometric path length of "transmitting array element i → pixel point P → receiving array element j"; This represents the extraction operation, which involves extracting the corresponding instantaneous amplitude from the time-reversed retransmission signal set based on the calculated sound path time index.

[0030] If pixel P does indeed have a defect, then after specific delay compensation, the signal phases of all sound wave paths passing through that point will be perfectly aligned. At this point, performing a superposition operation will cause constructive interference between the signals from each channel, resulting in an amplitude... The amplitude increases dramatically; conversely, if point P is defect-free, the phases of the signals in each channel are random, and after superposition, destructive interference occurs, with the amplitude approaching the background noise level.

[0031] For example, in one application example, the area to be measured is set as a 500mm × 500mm C35 concrete section, and the discretization mesh accuracy is set to 1mm × 1mm. Calibration of sound velocity. The coordinates are now calculated as follows: The focusing energy of each pixel grid point. Calculate the emitted sound path: from element 1 to point... distance ; Calculate the received sound path: from point Distance to the 10th array element ; Calculate the total transit time: ;Amplitude extraction: In the time-reversed retransmission signal set, the first... In the channel, the reading time The corresponding voltage amplitude. Repeat the above calculation for all 64×64=4096 paths, and sum the extracted 4096 voltage amplitudes. The final sum is the pixel value. The values ​​in the full-field energy focusing distribution matrix. By traversing all pixels in the entire field, a complete matrix image can be generated.

[0032] Generating the full-field energy focusing distribution matrix includes: The acoustic path transit time is calculated using a full matrix capture array and a discretized spatial pixel grid, and a transit time index table is constructed. The purpose of this step is to pre-calculate the acoustic wave propagation time for each transmit-receive pair in the full matrix acquisition array for each spatial pixel within the imaging region, and convert these times into memory address indices to accelerate the subsequent image synthesis process. First, a unified Cartesian coordinate system is established, and the full matrix acquisition array is positioned on the surface at z=0. The full matrix acquisition array is defined to contain N array elements, with the coordinates of the i-th transmit element being... The coordinates of the j-th receiving element are Simultaneously, a discretized spatial pixel grid is defined inside the concrete structure to be measured, containing M pixels, where the coordinates of any p-th pixel are... Next, based on the linear propagation model of sound waves in a uniform medium, the propagation of the sound wave from the transmitting element i to pixel p, and then its reflection back to the receiving element, is calculated. Total sound path transit time The calculation process is defined by the following formula: ; in, denoted by the path transit time of the i-th transmitting channel and the j-th receiving channel for spatial pixel p, which is the theoretical flight time required for the sound wave to travel along the path if there is a point scatterer at pixel p. Represents the horizontal and vertical coordinates of the currently calculated discretized spatial pixel grid points. and These represent the center position coordinates of the i-th transmitting element and the j-th receiving element, respectively, derived from the probe's factory physical parameters. The equivalent propagation velocity of ultrasound waves in the concrete under test is obtained through on-site calibration experiments, such as using standard test blocks or by back-calculation through penetration tests on a known thickness area of ​​the concrete member. The determination of this value must take into account the concrete grade and density; typically, the longitudinal wave velocity of C30-C50 concrete is between 3500 m / s and 4500 m / s.

[0033] To adapt to the discrete storage characteristics of digital signals, the calculated continuous time... Convert to discrete sampling point index and all The index value under the combination is stored in a dimension. A three-dimensional array, namely the transit time index table. The index conversion formula is: ; in, The data acquisition sampling frequency, This indicates the rounding operation.

[0034] For example, in one application example, suppose a 64-element full-matrix capture array is used in a single detection, with an element spacing of 2 mm. The on-site calibrated sound velocity in concrete... sampling frequency For a pixel P located deep in space, calculate the path for transmission by element 1 and reception by element 32. Transmission path: ; Receiving path: Total sound interval: Crossing time: Index mapping: The value "5335" is then stored in the corresponding transit time index table. The position indicates that in subsequent imaging, this channel should directly read the data from the 5335th sampling point.

[0035] Based on the transit time index table, the channel amplitude is extracted from the time-reversed retransmission signal set, and the channel amplitude is accumulated and superimposed to generate the full-field energy focusing distribution matrix.

[0036] A full-field energy focusing distribution matrix I, with the same size as the discretized spatial pixel grid, is initialized, and all elements are set to zero. Then, a parallel computing architecture is used to traverse each pixel p in the space. For each pixel, the corresponding storage address in the time-reversed retransmission signal set is quickly indexed according to the transit time index table, and the instantaneous amplitude values ​​of all N×N transmit-receive channels are read in parallel. Finally, a coherent accumulation operation is performed on all extracted amplitude values. The physical essence of this operation is to simulate the interference effect of sound waves at a certain point: if pixel p does indeed have a defect, the signals extracted from each channel according to the index table are highly consistent in phase, and the amplitude is significantly enhanced after accumulation; if there is no defect at this point, the phase of the extracted signals is random, and they cancel each other out after accumulation, tending towards zero. This accumulation and superposition process is described by the following formula: ; in, Generate the value of pixel p in the full-field energy focusing distribution matrix, which represents the acoustic focusing energy intensity at that location. It is the first in the constructed time-reversal retransmission signal set. Each channel data vector. Square brackets are used to represent array addressing operations, that is, directly reading the vector at index 0. Data points.

[0037] For example, in one application example, suppose there is indeed a 5mm diameter hole defect at pixel P. When performing cumulative stacking: for channels... The index table points to point 5335. Because the time reversal operation has aligned the echo signal peak to the theoretical propagation time, therefore... The voltage value read is the peak value, for example, 0.8V. For the channel... The index table points to another moment, which also corresponds to the echo peak value, with a read voltage of 0.75V. The summation of the read values ​​for all 64 × 64 = 4096 channels is then performed. Ultimately, the full-field energy focusing distribution matrix has an extremely high value at pixel P, while in the surrounding defect-free region, due to phase distortion, the accumulated value may only be around 50. This huge numerical difference creates a high-contrast defect image.

[0038] In S3, constructing the nonlinear magnitude coordinate mapping table includes: The time-domain reconstructed waveforms of the grid nodes are extracted based on the full-field energy focusing distribution matrix, and a fast Fourier transform is performed to generate a full-field spectrum data matrix. For any grid node in the full-field energy focusing distribution matrix Its time-domain reconstructed waveform It is the coherent superposition of all channel signals from the full matrix capture array after delay alignment. Subsequently, a Fast Fourier Transform (FFT) is performed on this reconstructed time-domain waveform to transform the signal from the time domain to the frequency domain. The aim is to reveal the frequency component distribution hidden in the time-domain waveform, particularly the nonlinear frequency characteristics induced by defects. Its calculation process is described by the following Discrete Fourier Transform formula: ; in, The complex value of the u-th frequency point in the frequency domain response sequence at grid node Z represents the value of the u-th frequency point. This represents the discrete sampling sequence of the time-domain reconstructed waveform at grid node Z. n represents the index of the time-domain sampling point, ranging from 0 to L-1. L represents the length of the signal involved in the transformation, depending on the sampling frequency. The focusing time window length is determined. For example, to obtain sufficient frequency resolution, data from 50 cycles before and after the focusing moment is typically extracted. u represents the spectral line index in the frequency domain, ranging from 0 to L-1. J represents the imaginary unit. By performing the above operations on every grid node in the entire field, a three-dimensional full-field spectrum data matrix is ​​constructed, with dimensions of... ,in This represents the size of the spatial grid.

[0039] For example, in one application example, assume the sampling frequency For coordinates The time-domain reconstructed waveform was extracted from the grid nodes with a length of L=1024 points. After performing FFT, a spectral sequence containing 512 effective frequency points was generated. The first point represents the DC component, and the 512th point represents the Nyquist frequency. This sequence fully records the response intensity of this spatial location to each frequency component.

[0040] The spectral index interval is calculated with the sum and difference of the first and second frequencies as the center frequency, and the sum frequency component amplitude and difference frequency component amplitude of the spectral index interval are locked in the full-field spectral data matrix. Due to nonlinear acoustic effects, when the first frequency Second frequency When sound waves interact at a defect, a spectral shift occurs, producing a sum of frequencies. , and difference frequency Components. First, calculate the digital index range corresponding to these two characteristic frequencies in the spectral sequence. Considering the frequency drift of actual hardware and the spectral leakage effect of signal processing, it is not possible to lock onto a single frequency point, but rather to define a spectral index interval with a certain bandwidth B. The index calculation formula is as follows: ; ; in, Representing the target center frequency, each value is [value to be filled in]. and . L1 represents the sampling frequency. L1 represents the number of points in the FFT transform, which is derived from the signal truncation length. The integer value representing the center spectral line index corresponding to the target frequency. The index number represents the half-bandwidth index. Its value is determined based on the probe's bandwidth characteristics and the main lobe width of the window function, typically using 3 to 5 frequency points to ensure coverage of the characteristic peak's energy main lobe. After locking the index interval, the maximum modulus or integral energy within that interval is extracted from the full-field spectrum data matrix, defined as the sum-frequency component amplitude, respectively. With difference frequency component amplitude .

[0041] For example, in one application example, a first frequency is set. Second frequency The difference frequency target is... , and frequency target Given sampling frequency FFT point count L1 = 1024. Difference frequency center index: Sum-frequency center index: Set half bandwidth The index ranges will be locked respectively. and In the grid nodes In the spectrum, search for the maximum amplitude value between indices 8 and 12. Assuming the maximum value of 0.05V is obtained at index 10, then... Similarly, the maximum value of 0.08V is obtained between indices 18 and 22. .

[0042] The complex modulus value is calculated based on the sum-frequency component amplitude and the difference-frequency component amplitude. The complex modulus value is used as the nonlinear amplitude and associated with the spatial coordinates of the grid nodes to construct a nonlinear amplitude coordinate mapping table.

[0043] To comprehensively evaluate the nonlinear response intensity at a specific point, a nonlinear complex modulus is introduced. The sum-frequency component amplitude is then considered... With difference frequency component amplitude Considering the two orthogonal dimensions describing the nonlinear intensity, a unified scalar index is calculated through vector synthesis. This approach effectively avoids missed detections caused by the influence of standing waves or interference on single frequency components. Complex modulus value. The calculation follows the Euclidean norm principle, and its formula is as follows: ; in, Representing coordinates The value of the complex modulus at that point is the final nonlinear amplitude. and These represent the sum frequency and difference frequency component amplitudes extracted from this coordinate point, respectively. and This represents the weighting coefficient, and its value depends on the difference in sensitivity between the defect type and the sum frequency and difference frequency. In general detection models lacking prior knowledge, it is typically set to... That is, they are treated with equal weight; however, for specific closed cracks, the difference frequency penetration can be appropriately increased due to its stronger penetrating power. The weight.

[0044] After the calculation is completed, a nonlinear amplitude coordinate mapping table is created containing three columns of data: the first column is the horizontal coordinate (x), the second column is the vertical coordinate (z), and the third column is the calculated nonlinear amplitude. This table provides data support for defect screening. For example... Figure 2 As shown, this illustrates the physical phenomenon that under dual-frequency excitation, the defect region significantly excites difference frequency and sum frequency components due to nonlinear modulation effects, while the intact region retains only the fundamental frequency component.

[0045] For example, in one application example, it is known that the extraction is... , Set weights Perform complex modulus calculation: The value "0.0943" is then used as a node. The nonlinear intensity characteristics are recorded in the corresponding coordinate row of the mapping table. Compared to using the difference frequency or sum frequency alone, this complex modulus value more comprehensively characterizes the degree of nonlinear activity at that location.

[0046] After generating the full-field energy focusing distribution matrix, the full-field energy focusing distribution matrix is ​​used as the full-band energy reference. The normalized ratio is calculated on the nonlinear amplitude coordinate mapping table to generate the nonlinear modulation coefficient distribution map.

[0047] The purpose of this step is to construct an evaluation index that reflects the local nonlinear conversion efficiency of the material, rather than relying solely on the absolute intensity of the signal. First, the full-field energy focusing distribution matrix is ​​read, denoted as... , representing the total sound field intensity including fundamental frequency energy, and a nonlinear amplitude coordinate mapping table, denoted as , representing the nonlinear intensity of the sum frequency and difference frequency. Subsequently, for each spatial grid node within the imaging region... The normalized ratio is calculated using the following formula: ; in, Represents the corresponding coordinates in the nonlinear modulation coefficient distribution spectrum. The higher the value, the stronger the ability of that location to convert incident sound energy into nonlinear harmonics, which is a typical characteristic of the "breathing effect" of microcracks. It originates from the complex modulus value in the nonlinear amplitude coordinate mapping table. It originates from the focusing amplitude in the full-field energy focusing distribution matrix. Physically, it approximately represents the total reflected energy at that point. The noise floor stability factor is determined based on the electronic thermal noise level when there is no signal input, and is usually taken as 0.1% to 1% of the maximum value of the full-field energy focusing distribution matrix. This represents the scaling factor used for numerical dynamic range adjustment. Through calculation, a distribution map of nonlinear modulation coefficients with the same mesh size as the original was generated. In this map, although linear strong scatterers such as coarse aggregate... The median value is very high, but its The total reflected energy is very low, therefore the NM value is low; while closed cracks, although... It may be relatively weak, but its nonlinear conversion efficiency is high, and the molecular... The denominator is relatively large, so the NM value increases significantly.

[0048] For example, in one application scenario, suppose there is a 20mm diameter granite aggregate in the concrete structure being tested. In the full-field energy focusing distribution matrix, due to the large difference in acoustic impedance between the aggregate and the cement matrix, this location exhibits strong reflection, and the measured... In the nonlinear amplitude coordinate mapping table, since the aggregate mainly exhibits linear elastic scattering, its sum frequency / difference frequency components are extremely weak, and the measured values ​​are... Set the stability factor. proportionality coefficient Perform normalized ratio calculation: The calculation results show that although the echo at this location is very strong, its nonlinear modulation coefficient is only 0.01, which can be used to determine that it is a "non-defect background".

[0049] In S4, the set of defect response coordinates includes: Perform global amplitude histogram statistics on the full-field energy focusing distribution matrix to construct the probability density function curve; First, read the full-field energy focusing distribution matrix and iterate through the focusing amplitude of all valid grid nodes in the matrix. Then, define a statistical interval for the amplitude. Typically, this covers the range from 0 to the maximum value of the matrix, and this range is uniformly divided into K statistical sub-intervals, each with a width of . Then, the number of grid nodes falling into each sub-interval was counted. And calculate its frequency probability. To construct a continuous probability density function curve, the following normalization calculation needs to be performed: ; in, The probability density function curve represents the amplitude The ordinate value at the location; This represents the number of pixels whose amplitude falls within the k-th sub-interval in the full-field energy focusing distribution matrix. Represents the total number of grid nodes participating in the statistics; The group interval representing the histogram is typically determined based on the Friedman-Diaconis criterion, i.e. ,in The interquartile range is used to ensure the histogram is neither too smooth nor too fragmented. Finally, all discrete... Points are smoothly connected by interpolation to construct a probability density function curve describing the distribution of the total scattering intensity.

[0050] For example, in one application example, consider a 500×500 pixel full-field energy focusing distribution matrix. The matrix amplitude ranges from 0V to 5V. The group spacing is set after calculating the interquartile range. ,Will The system was divided into 500 intervals. Statistical analysis revealed that there are 12,500 pixels within the 0.2V to 0.21V interval. Therefore, the probability density value for this interval is: By calculating each of the 500 intervals, a curve exhibiting a "long-tail distribution" characteristic was plotted: the main peak is located in the low-amplitude region, and the long tail extends to the high-amplitude region.

[0051] Perform parameter fitting iteration on the probability density function curve, and extract the fitting mean parameter to generate the aggregate scattering background baseline value; Since the random scattering noise of aggregates in concrete typically follows a Gaussian or Rayleigh distribution statistically, a Gaussian distribution is chosen as the fitting model. The Gaussian distribution model function is defined. The main peak of the probability density function curve is approximated using the least squares method or the maximum likelihood estimation method. The goal of the fitting is to find the optimal mean parameters. and standard deviation parameter After fitting is complete, extract the fitting mean parameters. And combined with standard deviation parameter Generate aggregate scattering background reference value The calculation formula is as follows: ; in, The reference value for aggregate scattering background is the critical voltage threshold that distinguishes background from defects; The mean background noise obtained from the fitting represents the average scattering level of the concrete material. The standard deviation of the background noise obtained from the fitting represents the range of fluctuation in scattering intensity; This represents the confidence level coefficient.

[0052] For example, in one application scenario, the main peak of the probability density function curve was identified as being located near 0.2V. A Gaussian model was used to iterate and fit the main peak; after the algorithm converged, the fitted mean parameters were obtained. , standard deviation of fit Set the confidence coefficient. Calculate the reference value for aggregate scattering background: This means that any signal with an amplitude below 0.4V, no matter how much it looks like a defect, is statistically highly likely to be just ordinary aggregate scattering and should be discarded.

[0053] The sum-frequency component amplitude is numerically subtracted from the aggregate scattering background reference value. If the result is positive, the matrix position index is extracted to generate a set of defect response coordinates.

[0054] While the full-field energy focusing distribution matrix provides a background statistical benchmark, the sum-frequency components are often more sensitive to microcracks. Therefore, using the benchmark value calculated from the full-field linear background to gate nonlinear and linear components is an effective fusion criterion. For each spatial coordinate in the mapping table Perform the following logical operations: ; like If the point is determined to be a valid defect response point, its spatial index is then established. Add to defect response coordinate set ;like If the value is not specified, then the point is considered background noise and is ignored. (This is the set of parameters.) It adopts a sparse matrix or coordinate list storage format, which only records the geometric position of valid points, greatly compressing the amount of data and providing guidance for supplementary scanning.

[0055] For example, in one application example, suppose the aggregate scattering background reference value has been determined as In the nonlinear amplitude coordinate mapping table: coordinate point A Its sum-frequency component amplitude is 0.85V. Calculation: Result: Point A was determined to be a defect, coordinates It is stored in the defect response coordinate set. Coordinate point B Its sum-frequency component amplitude is 0.35V. Calculation: Result: Point B was identified as background noise and its coordinates were discarded. Ultimately, the output is a set of hundreds of coordinate points that spatially outline the skeleton of the defect.

[0056] Extracting the fitted mean parameters to generate aggregate scattering background baseline values ​​includes: The global maximum coordinates are searched based on the probability density function curve to determine the initial center parameters, which drive the preset Gaussian distribution model. Least square fitting iteration is performed on the probability density function curve to obtain a convergent Gaussian distribution model. First, a peak search algorithm is performed on the probability density function curve. This algorithm iterates through every data point on the curve, compares adjacent values, identifies the point with the highest probability density value, and records its x-coordinate as the initial mean. Record its ordinate as the initial peak value. The purpose is to provide a "cold start" parameter for the iterative algorithm that approximates the true solution, preventing the iteration from getting trapped in local optima. Subsequently, a Gaussian distribution model is used as the objective function. This model physically describes the distribution of random speckle noise generated by a large number of tiny scatterers very well. The model formula is defined as follows: ; in, The representative model in amplitude The theoretical probability density value at that location. v represents the independent variable, namely the amplitude voltage in the full-field energy focusing distribution matrix. The mean parameter representing the distribution has the physical meaning of the average intensity center of aggregate scattering noise. The standard deviation parameter represents the distribution, and its physical meaning is the dispersion or fluctuation width of the noise intensity. A represents the peak amplitude parameter of the distribution. To obtain the optimal... and Iterative calculations are performed using the nonlinear least squares method. The goal of the iteration is to minimize the sum of squared residuals S, i.e.: ; in, These are the values ​​on the probability density function curve obtained from actual statistics. The iteration terminates when the sum of squared residuals S converges to the preset error threshold or reaches the maximum number of iterations. The model parameters at this point are the parameters of the convergent Gaussian distribution model.

[0057] For example, in one application scenario, suppose that after statistically analyzing the test data of a C30 concrete wall, the generated probability density function curve exhibits a typical unimodal skewed distribution. First, the highest point of the curve is found to be at a voltage value v = 0.25V, corresponding to a probability density of 4. Based on this, the initial parameters are set as follows: initial mean... initial peak Initial standard deviation The estimated value is 0.42 times the half-width at half-maximum (WHM). Then, least squares fitting iterations are initiated. In the first iteration, there is a deviation between the model curve and the actual curve, with a residual sum of squares (S) of 0.5. After 15 iterations, the parameters are continuously corrected, and the residual sum of squares eventually decreases to 0.0001, at which point the algorithm is considered converged. The output parameters at this point are: , This set of parameters describes the statistical regularity of the background noise inside the concrete structure.

[0058] The baseline value of aggregate scattering background is obtained by extracting the fitting mean parameters based on the convergent Gaussian distribution model.

[0059] Extract the fitting mean parameters from the convergent Gaussian distribution model. and the fitting standard deviation parameter In physical acoustics, the mean... This represents the area of ​​highest concentration of background noise. However, to ensure that the aggregate scattering background baseline value can effectively cover the vast majority of random noise fluctuations, the mean alone cannot be used; a confidence interval is usually set in conjunction with the standard deviation. Therefore, the aggregate scattering background baseline value is calculated using the following formula. : ; in, This represents the final aggregate scattering background reference value. It is derived from the mean of the converged Gaussian distribution. It is derived from the standard deviation of the converged Gaussian distribution. This represents the noise suppression coefficient, and its value is determined based on statistical hypothesis testing theory, typically ranging from 3 to 6. For example, taking... Based on the Gaussian distribution characteristics, theoretically 99.7% of the background noise can be eliminated; taking... At this time, 99.99% of noise can be eliminated, retaining only signals with a very high probability of being abnormal. Users can also fine-tune this through the human-computer interface based on their tolerance for false alarms. For example... Figure 3 As shown, the statistical distribution law of random scattering noise of aggregates inside concrete is demonstrated, as well as the significant difference between the adaptive background benchmark threshold determined by Gaussian fitting and the defect response signal.

[0060] For example, in one application example, after the iteration converges, the extracted fitting mean parameters are... Fitting standard deviation parameter If the current inspection task is extremely sensitive to minute defects and a small number of false alarms are permissible, the technicians will set the noise suppression coefficient to [value missing]. Perform the calculation: Ultimately, 0.383V was locked as the baseline value for aggregate scattering background. In the differential operation, any signal point with a sum-frequency component amplitude less than 0.383V was considered background noise and filtered out; only signal points with a amplitude greater than 0.383V were identified as potential defect responses. This dynamic threshold setting based on a statistical model is more robust than the fixed threshold method and can adapt to the differences in noise levels of concrete with different mix proportions.

[0061] In S5, the output defect 3D reconstruction model includes: Traverse the defect response coordinate set to extract the defect response coordinates, and expand the pixel grid unit outward from the defect response coordinate to construct a local supplementary scan window; First, the defect response coordinate set is read. Since the coordinate points in this set are usually discrete and discontinuous, spatial neighborhood expansion of these points is needed to obtain the complete shape of the defect. For any defect response coordinate in the set... With this coordinate as the geometric center, extend outwards along the horizontal and vertical directions respectively. Each pixel is a grid unit. This expansion operation actually constructs a bounding box region of a rectangle or cuboid. The expanded region's extent... The definition is as follows: ; in, Represents the spatial coverage of the constructed local supplementary scan window; Represents the central defect response coordinates extracted from the set; The initial grid spacing, representing the discrete spatial pixel grid, is derived from the initial imaging resolution setting; This represents the number of pixel grid units that expand outwards, and its value depends on the wavelength of the ultrasound wave. And the point spread function, in order to ensure coverage of the diffraction signal at the defect edge, is usually set to This ensures that the extended radius is at least equal to the wavelength of a shear wave, for example, a value of 2 to 5 grid cells.

[0062] A Boolean union operation is performed on the windows generated from all coordinate points in the set to merge overlapping areas, ultimately generating several independent local supplementary scan windows as the target areas for fine-grained scanning.

[0063] For example, in one application example, assume the initial discretized spatial pixel grid spacing is... Extract a coordinate point from the defect response coordinate set. Set the number of expansions. The local supplementary scan window range corresponding to this point is: x-axis from 98mm to 102mm, z-axis from 48mm to 52mm. If there is a neighboring point in the set... Its window range is Merge these two windows to form a larger combined window. , as the target of resampling.

[0064] The full-matrix capture array is driven to perform focused resampling on spatial pixels within the local supplementary scan window at a sampling interval of half the spacing of the discretized spatial pixel grid, and then overwrites the full-field energy focusing distribution matrix. By increasing the spatial sampling rate to capture edge details of minute defects, the data processing module of the full-matrix capture array is controlled to no longer perform calculations across the entire field, but instead performs full-focusing calculations only on the spatial range within a local supplementary scanning window. Crucially, the focused calculation grid no longer uses the initial spacing. Instead, it uses a finer sampling interval. The calculation formula is as follows: ; This operation follows the spatial Nyquist sampling theorem in signal processing, utilizing oversampling to reduce the picket fence effect and improve the accuracy of pinpointing defect peaks. For these high-density new pixels, the path transit time is recalculated, and amplitudes are extracted from the raw data acquired by the full-matrix capture array and coherently superimposed. After calculation, these high-resolution focused amplitudes are overwritten or interpolated to update the corresponding positions in the full-field energy focusing distribution matrix. If the full-field matrix resolution does not support direct writing, only the pixel values ​​in the corresponding region are updated, or a local high-resolution submatrix is ​​constructed in memory to replace the original low-resolution data block.

[0065] For example, in one application example, the initial spacing Within the local window, adjust the sampling interval to... Originally, there were only 5 pixels (98, 99, 100, 101, 102) within the x-axis range of 98mm to 102mm. After resampling, 9 points (98.0, 98.5, 99.0, ..., 102.0) will be calculated within this range. By precisely focusing on these 9 points using the full matrix data, it was found that the amplitude was originally 0.8V at 100mm, while a higher amplitude of 0.95V was measured at 100.5mm. Updating the matrix with this more accurate data improved the clarity of the defect edges.

[0066] Based on the nonlinear modulation coefficient distribution map, a mask filter is applied to the full-field energy focusing distribution matrix to extract the defect response coordinate data for triangulation meshing, and a three-dimensional reconstruction model of the defect is output.

[0067] To eliminate pseudo-defects that have high energy but weak nonlinearity, a nonlinear modulation coefficient distribution map is used as a confidence mask. This is applied to each pixel in the full-field energy focusing distribution matrix. Perform the weighted filtering operation using the following formula: ; in, This represents the final image amplitude after filtering; Represents the amplitude in the updated full-field energy focusing distribution matrix; The coefficient values ​​represent the corresponding points in the nonlinear modulation coefficient distribution map; This represents the nonlinear confidence threshold, typically derived from experimental calibration. For example, a value of 0.2 indicates that only signals with a nonlinear modulation intensity exceeding 20% ​​are considered true cracks. After masking filtering, all... The spatial coordinates of the points are used to form a clean defect point cloud. Then, a triangulation and meshing algorithm is applied. This algorithm, based on the spatial topological relationships of the point cloud, connects discrete coordinate points into a series of non-overlapping triangular patches, constructing closed or open geometric surfaces. Finally, this geometric surface is output as a 3D defect reconstruction model, which can be directly imported into CAD software or used for volume calculations.

[0068] For example, in one application scenario, during a certain detection, the resampled full-field matrix contains two high-energy regions: region A and region B. Referring to the nonlinear modulation coefficient distribution spectrum, region A corresponds to a coefficient of 0.05; region B corresponds to a coefficient of 0.45. A threshold is then set. After performing mask filtering, the amplitude of region A is forcibly set to 0, while the amplitude of region B is preserved. The coordinates of all points within region B are extracted, and triangulation is performed. The algorithm automatically connects these points, generating a long, uneven triangular mesh model that recreates the three-dimensional morphology of the crack inside the concrete, while interference A does not appear in the final model.

[0069] Example 2: As Figure 4 As shown, the ultrasonic testing signal processing system for internal defects in building concrete is used to implement the method in Example 1, including: The dual-frequency combined pulse acquisition module is used to drive the dual-source broadband excitation probe to emit combined pulse waves to the concrete structure under test. It is configured with a full matrix acquisition array to acquire multi-channel raw time-domain data. The combined pulse wave includes a first frequency and a second frequency. The time-reversal focusing calculation module is used to perform time-domain reversal operations on multi-channel raw time-domain data, construct a time-reversal retransmission signal set, and map it to the full matrix capture array to perform synchronous retransmission calculations and generate a full-field energy focusing distribution matrix. The nonlinear amplitude coordinate mapping module is used to extract specific frequency components based on the full-field energy focusing distribution matrix, with the sum and difference of the first and second frequencies as the center frequency, extract the sum frequency component amplitude and the difference frequency component amplitude, and construct a nonlinear amplitude coordinate mapping table. The aggregate background differential screening module is used to generate aggregate scattering background benchmark values ​​based on statistical analysis of the full-field energy focusing distribution matrix, and to perform differential operation between the sum frequency component amplitude and the aggregate scattering background benchmark values ​​to generate a set of defect response coordinates. The defect 3D reconstruction output module is used to feed back the defect response coordinate set to the full matrix capture array, trigger supplementary scanning for the defect response coordinate set, update the full field energy focusing distribution matrix, and output the defect 3D reconstruction model.

[0070] It should be noted that the functional division and information interaction between the various modules described above are logical, but in terms of physical implementation, they can be integrated on the same software platform or deployed in a distributed manner. The connections between them represent data flow and control flow, aiming to collaboratively achieve the objectives of this invention. The above are merely exemplary embodiments of this invention and should not be construed as limiting the scope of protection of this invention.

Claims

1. A method for processing ultrasonic detection signals of internal defects in building concrete, characterized by the following steps: include: S1. Drive the dual-source broadband excitation probe to emit a combined pulse wave to the concrete structure under test, and configure a full matrix acquisition array to collect multi-channel raw time domain data. The combined pulse wave includes a first frequency and a second frequency. S2. Perform time-domain flipping operation on the multi-channel raw time-domain data to construct a time-reversed retransmission signal set, and map it to the full matrix capture array to perform synchronous retransmission calculation and generate a full-field energy focusing distribution matrix; S3. Based on the full-field energy focusing distribution matrix, perform specific frequency component extraction with the sum and difference of the first and second frequencies as the center frequency, extract the sum frequency component amplitude and the difference frequency component amplitude, and construct a nonlinear amplitude coordinate mapping table. S4. Based on the statistical analysis of the full-field energy focusing distribution matrix, generate the aggregate scattering background reference value, and perform a difference operation between the sum frequency component amplitude and the aggregate scattering background reference value to generate a set of defect response coordinates. S5. Feed back the defect response coordinate set to the full matrix capture array, trigger supplementary scanning for the defect response coordinate set, update the full field energy focusing distribution matrix, and output the three-dimensional reconstruction model of the defect.

2. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 1, characterized in that, In S1, configuring the full matrix acquisition array to acquire multi-channel raw time-domain data includes: The first and second transducer crystals in the synchronously driven dual-source broadband excitation probe are used to emit combined pulse waves to the concrete structure under test. Configure a full matrix capture array to perform line-by-line cyclic scanning, acquire signals from each channel of the concrete structure under test in real time, and construct multi-channel raw time-domain data.

3. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 1, characterized in that, In S2, generating the full-field energy focusing distribution matrix includes: Read the raw time-domain data from multiple channels, perform linear reverse rearrangement based on the end of the time axis to the beginning, and construct a time-reversed retransmission signal set; A discretized spatial pixel grid is established for the concrete structure under test. Based on the discretized spatial pixel grid and the full matrix capture array, spatial coherence superposition is performed to generate a full-field energy focusing distribution matrix.

4. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 3, characterized in that, Spatial coherent superposition is performed based on a discretized spatial pixel grid and a full-matrix capture array, specifically: The acoustic path transit time is calculated using a full matrix capture array and a discretized spatial pixel grid, and a transit time index table is constructed. Based on the transit time index table, the channel amplitude is extracted from the time-reversed retransmission signal set, and the channel amplitude is accumulated and superimposed to generate the full-field energy focusing distribution matrix.

5. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 1, characterized in that, In S3, constructing the nonlinear amplitude coordinate mapping table includes: The time-domain reconstructed waveforms of the grid nodes are extracted based on the full-field energy focusing distribution matrix, and a fast Fourier transform is performed to generate a full-field spectrum data matrix. The spectral index interval is calculated with the sum and difference of the first and second frequencies as the center frequency, and the sum frequency component amplitude and difference frequency component amplitude of the spectral index interval are locked in the full-field spectral data matrix. The complex modulus value is calculated based on the sum-frequency component amplitude and the difference-frequency component amplitude. The complex modulus value is used as the nonlinear amplitude and associated with the spatial coordinates of the grid nodes to construct a nonlinear amplitude coordinate mapping table.

6. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 3, characterized in that, After generating the full-field energy focusing distribution matrix, the full-field energy focusing distribution matrix is ​​used as the full-band energy reference. The normalized ratio is calculated on the nonlinear amplitude coordinate mapping table to generate the nonlinear modulation coefficient distribution map.

7. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 1, characterized in that, In S4, generating the defect response coordinate set includes: Perform global amplitude histogram statistics on the full-field energy focusing distribution matrix to construct the probability density function curve; Perform parameter fitting iteration on the probability density function curve, and extract the fitting mean parameter to generate the aggregate scattering background baseline value; The sum-frequency component amplitude is numerically subtracted from the aggregate scattering background reference value. If the result is positive, the matrix position index is extracted to generate a set of defect response coordinates.

8. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 7, characterized in that, Extracting the fitted mean parameters to generate the aggregate scattering background baseline value, specifically: The global maximum coordinates are searched based on the probability density function curve to determine the initial center parameters, which drive the preset Gaussian distribution model. Least square fitting iteration is performed on the probability density function curve to obtain a convergent Gaussian distribution model. The baseline value of aggregate scattering background is obtained by extracting the fitting mean parameters based on the convergent Gaussian distribution model.

9. The ultrasonic signal processing method for detecting internal defects in building concrete according to claim 6, characterized in that, In S5, the output defect 3D reconstruction model includes: Traverse the defect response coordinate set to extract the defect response coordinates, and expand the pixel grid unit outward from the defect response coordinate to construct a local supplementary scan window; The full-matrix capture array is driven to perform focused resampling on spatial pixels within the local supplementary scan window at a sampling interval of half the spacing of the discretized spatial pixel grid, and then overwrites the full-field energy focusing distribution matrix. Based on the nonlinear modulation coefficient distribution map, a mask filter is applied to the full-field energy focusing distribution matrix to extract the defect response coordinate data for triangulation meshing, and a three-dimensional reconstruction model of the defect is output.

10. An ultrasonic testing signal processing system for internal defects in building concrete, used to implement the method described in any one of claims 1-9, characterized in that, include: The dual-frequency combined pulse acquisition module is used to drive the dual-source broadband excitation probe to emit combined pulse waves to the concrete structure under test. It is configured with a full matrix acquisition array to acquire multi-channel raw time-domain data. The combined pulse wave includes a first frequency and a second frequency. The time-reversal focusing calculation module is used to perform time-domain reversal operations on multi-channel raw time-domain data, construct a time-reversal retransmission signal set, and map it to the full matrix capture array to perform synchronous retransmission calculations and generate a full-field energy focusing distribution matrix. The nonlinear amplitude coordinate mapping module is used to extract specific frequency components based on the full-field energy focusing distribution matrix, with the sum and difference of the first and second frequencies as the center frequency, extract the sum frequency component amplitude and the difference frequency component amplitude, and construct a nonlinear amplitude coordinate mapping table. The aggregate background differential screening module is used to generate aggregate scattering background benchmark values ​​based on statistical analysis of the full-field energy focusing distribution matrix, and to perform differential operation between the sum frequency component amplitude and the aggregate scattering background benchmark values ​​to generate a set of defect response coordinates. The defect 3D reconstruction output module is used to feed back the defect response coordinate set to the full matrix capture array, trigger supplementary scanning for the defect response coordinate set, update the full field energy focusing distribution matrix, and output the defect 3D reconstruction model.

Images