A method for comprehensive detection of concrete strength based on rebound method and ultrasonic method

Abstract

This invention relates to the field of non-destructive testing technology in building engineering, and discloses a comprehensive method for testing concrete strength based on the rebound method and ultrasonic method. The method includes: acquiring the full-wavelength ultrasonic signal of the test area and calculating the propagation velocity; performing a rebound impact in the test area to obtain the rebound value and acceleration sequence, analyzing and defining the impact contact time and calculating the initial energy absorption ratio; mapping the impact contact time to a cutoff frequency, and performing frequency band integration on the full-wavelength ultrasonic signal to obtain the dispersion attenuation coefficient; comparing the dispersion attenuation coefficient with a scattering threshold, and establishing the initial energy absorption ratio or pure mortar phase characteristics as a mechanical feature set based on the results; converging the propagation velocity, rebound value, impact contact time, dispersion attenuation coefficient, and mechanical feature set to construct an initial comprehensive state matrix; extracting orthogonal principal component latent variables from this matrix and performing quantized superposition to output the concrete compressive strength. This invention eliminates parameter multicollinearity and improves detection accuracy by establishing a physical correlation between mechanical impact and the ultrasonic frequency domain.

Description

Technical Field

[0001] This invention relates to the field of non-destructive testing technology in building engineering, specifically a comprehensive method for testing concrete strength based on rebound and ultrasonic methods. Background Technology

[0002] Concrete compressive strength is a core indicator for the quality acceptance of engineering structures. Currently, non-destructive testing (NDT) technology is widely used in the appraisal of existing buildings and the random inspection of new construction projects. While the rebound method alone primarily reflects the hardness characteristics of the concrete surface, the ultrasonic method alone focuses on reflecting the internal density and mechanical properties of the concrete. To overcome the physical limitations of single testing methods, a combined rebound and ultrasonic method is often used in engineering practice. This method simultaneously collects the ultrasonic wave propagation velocity and rebound value in the same test area and establishes a multivariate regression empirical model between the sound velocity, rebound value, and concrete compressive strength, thereby achieving a comprehensive estimation and evaluation of concrete strength.

[0003] However, existing integrated testing methods still have shortcomings in terms of calculation accuracy and environmental adaptability under complex working conditions. Conventional methods typically treat surface rebound parameters and internal ultrasonic parameters as independent variables, applying fixed empirical formulas without establishing a deep physical correlation between mechanical impact response and ultrasonic frequency domain energy distribution. This results in low underlying fusion between multi-source heterogeneous data. Furthermore, existing testing models lack adaptive discrimination mechanisms for the internal moisture state and aggregate distribution of materials. When the test area is damp or the impact point happens to encounter randomly accumulated coarse aggregates, the collected single surface parameters are prone to deviating from the true baseline. In addition, parameters such as sound velocity, hardness, and frequency domain characteristics obtained from ultrasonic testing and rebound testing have a certain degree of overlap in mechanical mechanisms. Directly inputting these highly correlated features into conventional regression models can easily lead to severe multicollinearity problems, thereby compromising the mathematical stability of the strength calculation model and causing significant deviations in the final strength prediction values. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a comprehensive concrete strength testing method based on rebound and ultrasonic methods. This method solves the problems of low physical correlation of multi-source data, susceptibility to interference from environmental humidity and surface roughness, and limited strength calculation accuracy due to multicollinearity among multiple parameters in existing testing technologies.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a comprehensive method for detecting concrete strength based on the rebound method and the ultrasonic method, comprising acquiring the full waveform signal of the ultrasonic field in the test area and calculating the propagation speed of sound by combining the transducer spacing; In the test area, projectile impacts are performed to obtain rebound values ​​and acceleration sequences. The acceleration sequences are analyzed to determine the impact contact time and the initial energy absorption ratio is calculated. The impact contact time is mapped to the cutoff frequency, and the dispersion attenuation coefficient is calculated by performing a frequency band integral on the full waveform ultrasonic signal based on the cutoff frequency. The dispersion attenuation coefficient is compared with the scattering threshold, and the initial energy absorption ratio or pure mortar phase characteristics are established as the mechanical feature set based on the comparison results. The propagation speed of sound, rebound value, impact contact time, dispersion attenuation coefficient, and mechanical feature set are combined to construct a standardized initial integrated state matrix; The orthogonal principal component latent variables are extracted from the initial comprehensive state matrix and quantified and superimposed to output the concrete compressive strength.

[0006] Furthermore, the technical principle of analyzing the acceleration sequence to define the impact contact time and calculate the initial energy absorption ratio is as follows: By performing zero-crossing analysis on the acceleration sequence, the contact trigger moment and separation moment crossing the zero baseline are identified, and the time difference between the two is the impact contact time. This time reflects the plastic deformation characteristics and stiffness of the concrete surface layer. Within the time interval from contact to separation, two time-dimensional integration operations are performed on the acceleration sequence to obtain velocity parameters, extracting the initial impact velocity and the initial rebound separation velocity. The specific logic for calculating the initial energy absorption ratio is as follows: calculate the difference between the square of the initial impact velocity and the square of the initial rebound separation velocity, and then divide this difference by the square of the initial impact velocity. This ratio is physically equivalent to the proportion of kinetic energy converted into plastic deformation energy and acoustic energy in concrete during a single impact, thus characterizing the local material dissipation characteristics.

[0007] Furthermore, the process of mapping the impact contact time to the cutoff frequency is as follows: To avoid division by zero or frequency abrupt changes caused by minute noise, the impact contact time is verified against a clock noise floor tolerance threshold. When the contact time exceeds this threshold, an inverse proportional function mapping is performed, that is, a preset constant mapping coefficient is divided by the impact contact time to obtain an initial cutoff limit. Subsequently, the system hardware's upper and lower frequency limits are used to perform boundary clamping processing on the initial cutoff limit; that is, when the limit value is higher than the upper limit, the upper limit value is used, and when it is lower than the lower limit, the lower limit value is used. The clamped value is used as the cutoff frequency for ultrasonic frequency domain analysis. This mechanism utilizes the physical correlation that a shorter mechanical impact contact time corresponds to stronger high-frequency excitation.

[0008] Furthermore, the technical principle for obtaining the dispersion attenuation coefficient is as follows: A Fast Fourier Transform (FFT) was performed on the captured full-wavelength ultrasonic signal with a Hanning window applied to obtain the discrete frequency domain amplitude spectrum. Using the calculated cutoff frequency as the boundary, the discrete spectrum was divided into low-frequency and high-frequency bands. The frequency band integral was calculated as follows: within the low-frequency band, the sum of the squared amplitude values ​​at each discrete frequency point was calculated to obtain the low-frequency band energy integral value; within the high-frequency band, the sum of the squared amplitude values ​​at each discrete frequency point was calculated to obtain the high-frequency band energy integral value. After confirming that the high-frequency band energy integral value did not fall into the noise floor region, the low-frequency band energy integral value was divided by the high-frequency band energy integral value to obtain the dispersion attenuation coefficient. This coefficient reflects the relatively severe attenuation of high-frequency short-wave components of ultrasound waves during propagation through the medium due to scattering at the microcracks and aggregate interface.

[0009] Furthermore, the adaptive feature selection logic for the impact of environmental moisture is as follows: Determine whether the calculated dispersion attenuation coefficient is greater than or equal to the set scattering threshold. If it is, it indicates that the medium has high internal moisture or strong micro-scattering. In this case, the test area is in a high-scattering humid state, and the surface impact characteristics are more representative. Therefore, the initial energy absorption ratio is directly assigned to the mechanical feature set. If it is less than the threshold, it indicates that the test area is in a low-scattering dry state. A single impact is easily affected by the randomness of aggregate distribution. In this case, the spatial array impact command is triggered to extract the pure mortar phase characteristics.

[0010] Furthermore, the specific process for extracting the characteristics of pure mortar phase is as follows: Multiple discrete impacts are performed within the minimum safe distance of the limiting plate constraint to obtain the corresponding energy absorption ratios and construct a sequence set. This set is then sorted in ascending order and converted into a one-dimensional monotonically increasing ordered sequence. A first-order relative difference operation is performed on the ordered sequence, i.e., the difference between the next element and the previous element is calculated. When the difference exceeds the material impedance transition threshold, the upper and lower boundary indices are recorded to eliminate outliers caused by extreme aggregates or voids at both ends of the sequence, retaining the valid core sub-sequence in the middle of the sequence.

[0011] The specific mathematical processing for calculating the characteristics of pure mortar phase is as follows: The arithmetic mean of all elements in the effective core subsequence is calculated as the expected value of the mortar phase energy. Simultaneously, the ratio of the standard deviation to the mean of this subsequence is calculated as the coefficient of variation. Subsequently, an exponentially decaying nonlinear weighted fusion is performed, which involves taking the base of the natural logarithm to the negative power of the coefficient of variation to obtain a penalty factor. This penalty factor is then multiplied by the expected value of the mortar phase energy. The final result is established as the characteristics of pure mortar phase, thereby suppressing the interference of excessively discrete data on the comprehensive evaluation.

[0012] Furthermore, the technical principle of the feature calculation module is as follows: An initial matrix is ​​constructed by aligning the propagation speed of sound, rebound value, impact contact time, dispersion attenuation coefficient, and mechanical feature set. The variance of each column is calculated, and redundant columns with variances below the minimum floating-point lower limit are removed. The remaining matrix is ​​then standardized by zero mean and unit variance for each column; that is, each element is subtracted from the mean of its column and then divided by the standard deviation of its column, thus unifying the data scale of features with different dimensions.

[0013] The extraction of latent variables employs partial least squares logic: the standardized matrix is ​​multiplied and accumulated along the first orthogonal weight vector to obtain the first orthogonal principal component latent variables. Next, an information stripping operation is performed, which involves subtracting the outer product of the first principal component and its corresponding loading vector from the original matrix to generate a residual matrix. This multiplication and stripping operation is repeated on the residual matrix until mutually independent orthogonal principal component latent variables of a predetermined dimension are obtained.

[0014] Finally, the latent variables of each orthogonal principal component are multiplied by their corresponding regression mapping coefficients, and the sum is calculated to obtain the comprehensive variable value. The comprehensive variable value is then added to the pre-calibrated strength benchmark bias constant to reconstruct the output concrete compressive strength.

[0015] This invention provides a comprehensive method for testing concrete strength based on the rebound method and ultrasonic method. It has the following beneficial effects: 1. This invention maps the impact contact time to the cutoff frequency of the ultrasonic frequency band integral, establishing a physical correlation between mechanical impact response and ultrasonic frequency domain analysis. Since the impact contact time directly reflects the surface stiffness of concrete, using it as a boundary line to calculate the dispersion attenuation coefficient allows the ultrasonic frequency cutoff to be dynamically adjusted according to the actual physical state of the test area. This mechanism avoids information loss or noise interference caused by the use of fixed filter bands in existing technologies, improving the accuracy of acoustic feature extraction under different ages and stiffness conditions.

[0016] 2. This invention achieves dynamic selection of detection features by comparing the dispersion attenuation coefficient with the scattering threshold. The dispersion attenuation coefficient is calculated from the energy integral values ​​of high and low frequency bands and can characterize the degree of scattering and moisture state of ultrasonic waves propagating inside the material. Based on the determination result of this coefficient, the system automatically switches between the energy absorption ratio of a single impact and the pure mortar phase characteristics extracted by multiple array impacts, reducing the interference of random distribution of local coarse aggregates and abnormal moisture content on a single detection parameter, and enhancing the adaptability of the detection method to complex working conditions.

[0017] 3. This invention constructs an initial comprehensive state matrix by combining the propagation speed of sound, rebound value, impact contact time, dispersion attenuation coefficient, and mechanical feature set, and extracts orthogonal principal component latent variables for the final strength calculation. Since the multi-source parameters obtained from ultrasonic and rebound detection overlap in physical mechanisms, direct regression calculations are prone to multicollinearity problems. This scheme extracts independent latent variables through orthogonal feature projection and information stripping operations, eliminating linear correlation interference between parameters and ensuring the stability of the compressive strength mapping model and the reliability of the output results. Attached Figure Description

[0018] Figure 1 This is a flowchart of the cross-modal adaptive comprehensive detection algorithm of the present invention; Figure 2 This is a diagram showing the overall hardware architecture and microarray interaction of the system according to the present invention; Figure 3 The diagram below illustrates the principle of multimodal adaptive decoupling and intensity reconstruction in a specific application embodiment of the present invention. Figure A shows the frequency domain segmentation diagram of the full waveform of ultrasound; Figure B shows the time history curve of rebound acceleration; and Figure C shows the spatial impact sequence and statistical filtering diagram. Detailed Implementation

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See attached document Figure 2 This invention provides a comprehensive concrete strength testing system based on rebound and ultrasonic methods, comprising: an ultrasonic acquisition module, a rebound acquisition module, a main control module, and a limiting plate.

[0021] The ultrasonic acquisition module is used to transmit and receive full-wavelength ultrasonic signals on the surface of the component under test, and to acquire full-wavelength ultrasonic signals.

[0022] The rebound acquisition module integrates an acceleration sensor to perform physical impact operations and simultaneously acquire acceleration sequences.

[0023] The main control module establishes communication connections with both the ultrasonic acquisition module and the rebound acquisition module. The main control module receives full-waveform time-domain signals and acceleration sequences, and performs data synchronization, signal truncation, statistical filtering, and regression calculation.

[0024] The limiting plate is a mechanical structural component with arrayed positioning guide holes. It is used in conjunction with the rebound acquisition module to constrain the physical impact coordinates of the rebound acquisition module.

[0025] See attached document Figure 1 This invention provides a comprehensive method for testing the strength of concrete based on the rebound method and the ultrasonic method, comprising the following steps: S10, the ultrasonic acquisition module acquires the full waveform signal of the ultrasonic wave in the preset test area of ​​the component under test and transmits it to the main control module. The main control module extracts the arrival time of the first ultrasonic wave and calculates the propagation speed of sound in combination with the transducer spacing. S20, the rebound acquisition module performs the first single impact at the center of the test area, and the main control module receives the rebound value and acceleration sequence recorded by the rebound acquisition module; S30, the main control module analyzes the acceleration sequence, locates the zero-crossing moment of the acceleration, defines the impact contact time, extracts the initial impact velocity and the initial rebound separation velocity through time-domain quadratic integration, and calculates the initial energy absorption ratio; S40, the main control module extracts the impact contact time and constructs a cutoff frequency for the ultrasonic signal. The cutoff frequency and the impact contact time are set to an inverse proportional function relationship. S50, the main control module uses the cutoff frequency as the frequency band division boundary, performs low-frequency band energy integration and high-frequency band energy integration on the amplitude spectrum of the full ultrasonic waveform signal, and calculates the dispersion attenuation coefficient of the ultrasonic first wave. S60, the main control module compares the real-time calculated dispersion attenuation coefficient with the preset scattering threshold. S70, when the dispersion attenuation coefficient is greater than or equal to the scattering threshold, the main control module confirms the mechanical feature set as the initial energy absorption ratio; when the dispersion attenuation coefficient is less than the scattering threshold, the main control module triggers the spatial array impact command, and the operator uses the limit plate to constrain the rebound acquisition module to perform multiple spatial discrete impacts in the test area. The main control module receives the spatial energy absorption ratio sequence set of multiple discrete impacts in sequence, performs the sequential statistical filtering algorithm on the spatial energy absorption ratio sequence set, and removes the extreme values ​​at both ends of the sequence by ascending order and truncated mean filtering, calculates the expected value of mortar phase energy, obtains the pure mortar phase characteristics and establishes them as the mechanical feature set; S80: The main control module aggregates the propagation speed of sound, dispersion attenuation coefficient, rebound value, impact contact time, and mechanical feature set to construct an initial comprehensive state matrix. The main control module inputs the initial comprehensive state matrix into the partial least squares regression solution model and extracts mutually independent orthogonal principal component latent variables through orthogonal weight matrix calculation. The main control module performs linear weight quantization and superposition on the extracted principal component latent variables, reconstructs and outputs the concrete compressive strength of the measuring point.

[0026] See attached document Figure 1 The main control module acquires basic multimodal physical quantities and performs time history analysis. The main control module then controls the ultrasonic acquisition module and the rebound acquisition module to perform the following sub-steps: In this embodiment, the propagation time of ultrasound in a heterogeneous multiphase composite medium can macroscopically reflect the density and mechanical reference state of the internal structure of the medium. Based on the above physical mechanism, in S101, the ultrasonic acquisition module arranges a transmitting transducer and a receiving transducer in a preset test area on the surface of the component under test, and calibrates the relative geometric straight-line distance between the transmitting and receiving transducers, which is recorded as the distance measurement. The main control module sends a pulse excitation signal to the ultrasonic acquisition module, driving the transmitting transducer to emit ultrasound into the component under test. The receiving transducer synchronously captures the full waveform signal of the ultrasound transmitted through the concrete medium, and the main control module caches the full waveform signal of the ultrasound in its internal storage unit.

[0027] S102, the main control module extracts the arrival time of the first ultrasonic wave from the full waveform signal. As a preferred method, considering the extreme error situation in actual engineering testing where poor coupling or signal abnormalities may cause the arrival time of the first ultrasonic wave to approach zero, the main control module pre-sets a minimum positive threshold allowed by the system clock accuracy before performing the division operation. When the extracted arrival time of the first ultrasonic wave is less than or equal to this minimum positive threshold, the system determines that the current ultrasonic acquisition is invalid and triggers an audible and visual alarm to prompt a re-acquisition; when the extracted arrival time of the first ultrasonic wave is greater than this minimum positive threshold, the main control module divides the distance measurement by the arrival time of the first ultrasonic wave to directly calculate and obtain the macroscopic ultrasonic propagation velocity. The physical dimension of this macroscopic ultrasonic propagation velocity is m / s, which is used to characterize the initial acoustic expectation value of the macroscopic elastic modulus of the test area. The method of applying the coupling agent to the ultrasonic transducer and the specific hardware topology of the pulse excitation circuit can be conventionally selected and configured by those skilled in the art according to the requirements of the actual testing environment; these are well-known technologies in the field and will not be elaborated here.

[0028] After acquiring macroscopic acoustic characteristics, in order to establish a cross-modal coupling relationship from the surface to the core, the system further introduces a micro-area dynamic rebound acquisition mechanism.

[0029] S201, the main control module determines the geometric midpoint of the ranging line as the center of the test area. The physical basis for selecting the center of the test area for impact is that this central location is least affected by the boundary stress field of the ultrasonic transducer, and can provide the purest initial mechanical state reference. The rebound acquisition module performs the first single impact at the center of the test area and acquires the traditional rebound value based on the displacement measurement of the impact hammer.

[0030] S202: Simultaneously with the physical impact, the accelerometer integrated within the rebound acquisition module acquires a continuous acceleration sequence at a preset sampling frequency. To ensure absolute alignment of operating conditions during subsequent multi-source cross-analysis, the main control module uses a global hardware clock source to allocate global timestamps for the sampling tasks of the ultrasonic acquisition module and the rebound acquisition module, eliminating asynchronous communication delays between heterogeneous sensors. The main control module synchronously receives the rebound value and acceleration sequence.

[0031] Typically, the collision process between the rebound acquisition module and the concrete surface is a very short-duration Hertzian contact and elastoplastic deformation coupling process. This requires precise capture of the mechanical response limits.

[0032] S301, the main control module performs zero-crossing analysis on the acceleration sequence to define the impact contact time limit. The main control module determines the contact trigger moment as the moment when the gradient in the acceleration sequence suddenly changes and first crosses the zero baseline. The main control module identifies the separation moment as the moment when the acceleration curve, after reaching its peak, first falls back and crosses the zero baseline again.

[0033] S302, the main control module calculates the impact contact time by subtracting the contact trigger time from the separation time. Based on the stiffness differences of conventional concrete grades, the normal range for this impact contact time is between several hundred microseconds and several milliseconds. The impact contact time objectively characterizes, from a dynamic perspective, the actual mechanical impedance duration between the mechanical impact hammer of the rebound acquisition module and the surface of the two-phase composite medium from contact to separation. Its value shows a strong negative correlation with the dynamic compressive stiffness of the mortar on the component surface.

[0034] After accurately obtaining the contact time interval, the system needs to further quantify the kinetic energy transferred from the projectile system to the inelastic medium, which depends on a deep deconstruction of the kinematic time sequence.

[0035] S303: Within the time interval defined by the contact trigger moment and the separation moment, the main control module performs a time-domain quadratic integral calculation on the acceleration sequence. Specifically, the main control module first reconstructs the acceleration into a velocity sequence through a single integral, and then performs a second integral to reconstruct the displacement sequence. During the integration initialization phase, the main control module forcibly calibrates the velocity of the impact hammer to zero at the moment of release. It obtains the true initial impact velocity by integrating the acceleration from release to the contact trigger moment, while simultaneously forcibly calibrating the displacement at the contact trigger moment to zero to eliminate low-frequency integration accumulation errors caused by sensor zero-point drift.

[0036] S304, based on the above reconstructed sequence, the main control module uses the maximum point of the displacement time program sequence to help confirm the stationary point of energy conversion, and then accurately extracts the initial impact velocity corresponding to the contact trigger moment and the initial rebound separation velocity corresponding to the separation moment from the velocity time program sequence. Considering that system abnormalities or human operation errors may lead to unsuccessful triggering of effective impact, causing the initial impact velocity to approach zero and triggering division overflow, the main control module pre-determines whether the absolute value of the initial impact velocity is greater than a preset impact failure threshold. This threshold is physically calibrated based on the rated energy release parameters of the rebound spring. If the absolute value of the initial impact velocity is not greater than the threshold, it is determined that the single impact kinetic energy is insufficient, the system discards it and prompts for retesting; if the absolute value of the initial impact velocity is greater than the threshold, the main control module calculates the initial energy absorption ratio using the initial impact velocity and the initial rebound separation velocity. The formula is as follows: ; In the formula, The initial energy absorption ratio is based on the principle of kinetic energy conservation and dissipation. Its theoretical value ranges from 0 to 1. This parameter is used to quantitatively characterize the proportion of kinetic energy conversion and consumption. The absolute value of the initial velocity of the rebound separation, extracted with the aid of time-domain quadratic integration, represents the residual kinetic energy velocity state of the impact hammer at the instant it leaves the medium surface. The absolute value of the initial impact velocity, extracted with the aid of time-domain quadratic integration, characterizes the initial energy release velocity state of the hammer at the moment of contact with the medium surface. This core algorithm effectively eliminates the physical errors caused by internal mechanical friction and spring fatigue in traditional rebound hammers from the underlying physical principles, achieving an unbiased estimation of the plastic energy dissipation characteristics of the component surface. The main control module stores the initial energy absorption ratio in the system register for subsequent state machine calls.

[0037] See attached document Figure 1 In conventional nondestructive testing theory for heterogeneous multiphase media, traditional ultrasonic testing often employs a fixed frequency domain analysis window, which is difficult to adapt to the spatial variability of concrete surface hardness and internal pore structure. To address the limitations caused by independent analysis of heterogeneous physical quantities, this embodiment establishes a mapping mechanism from surface mechanical impedance to the frequency domain truncation boundary. The main control module controls the execution of the following sub-steps: S401, the main control module calls the impact contact time stored in the system register. The physical reason for choosing this physical quantity as the cross-modal input parameter is that the impact contact time essentially reflects the macroscopic mechanical impedance characteristics of the surface mortar of the test area. When the test area surface is dense and has high stiffness, the momentum exchange during the impact process is rapid, and the impact contact time is short. Under this condition, high-frequency sound waves can propagate effectively in the medium. Conversely, when there are microcracks or a loose layer with high water content on the surface, the plastic buffering effect leads to a longer impact contact time. Under this condition, the high-frequency components of the ultrasonic wave will undergo strong scattering and attenuation. Based on the above physical causal relationship between mechanics and acoustics, the main control module uses the impact contact time as a cross-modal constraint to construct the cutoff frequency for the ultrasonic signal.

[0038] S402, considering that hardware acquisition anomalies or partial sensor detachment may lead to extreme distortion in the extracted time history data, the main control module performs a validity check on the impact contact time before performing frequency domain mapping calculation. The main control module determines whether the impact contact time is less than a preset clock noise floor tolerance threshold. Here, the specific value of this clock noise floor tolerance threshold is determined based on the highest sampling period of the system's analog-to-digital converter, and is usually set to 3 to 5 times the sampling period. The technical purpose is to avoid division overflow faults caused by the denominator approaching zero in subsequent inverse proportional calculations. To avoid one-sided judgment based solely on a single extreme value, the main control module also performs a weighted comparison with the historical average contact time of adjacent measurement points. If the currently extracted contact time is not only less than the above threshold, but also deviates from the historical average by more than a preset coefficient of dispersion, the main control module determines that the dynamic rebound feature is invalid and directly calls the preset nominal center frequency of the ultrasonic transducer as the default cutoff frequency; when the impact contact time is greater than the clock noise floor tolerance threshold and is within a reasonable fluctuation range, the main control module executes the determined inverse proportional function mapping calculation.

[0039] S403, as a preferred method, the main control module performs an inverse proportional function mapping calculation, dividing the preset mapping coefficient by the impact contact time to obtain the initial cutoff limit. The necessary technical formulas used in this invention are as follows: ; In the formula, The initial cutoff limit for ultrasonic signals is constructed, and its physical meaning is the basic value of the dynamic frequency band boundary that distinguishes whether ultrasonic waves mainly undergo viscous absorption attenuation or geometric scattering attenuation in the medium of the current test area. This refers to the impact contact time calculated in the preceding steps; These are the mapping coefficients. This is an empirical constant with normalized physical dimensions. Its specific value depends on the stiffness constant of the rebound spring and the resonance characteristics of the ultrasonic transducer. In practical engineering applications, those skilled in the art can determine its specific value through fitting and calibration using standard homogeneous test blocks. Its conventional value range is between 0.5 and 2.0.

[0040] S404, after completing the inverse proportional mapping calculation, to prevent the calculated initial cutoff limit from exceeding the physical response limit of the actual hardware system, the main control module further performs boundary clamping processing on the initial cutoff limit in conjunction with the upper and lower frequency limits of the system hardware. The main control module presets the upper and lower frequency limits. The upper frequency limit is constrained by the Nyquist sampling theorem and the low-pass filter cutoff frequency of the ultrasonic receiving circuit, and in this embodiment, it is typically set to 2.5 times the nominal center frequency of the ultrasonic transducer; the lower frequency limit is constrained by the low-frequency physical boundary of the transmitting transducer, and is typically set to 0.3 times the nominal center frequency of the ultrasonic transducer. When the calculated initial cutoff limit is greater than the upper frequency limit, the main control module forcibly assigns it to the upper frequency limit; when the calculated initial cutoff limit is less than the lower frequency limit, the main control module forcibly assigns it to the lower frequency limit. Through the above boundary clamping operation, the main control module uses the clamped initial cutoff limit as the final cutoff frequency for subsequent acoustic dispersion analysis. For the general logic of boundary clamping in digital signal processing, those skilled in the art can implement it using conventional conditional statements. The code execution method is a well-known technology in this field and will not be described in detail here.

[0041] See attached document Figure 1 In this embodiment, according to the acoustic theory of porous media, the high-frequency components of ultrasound are more sensitive to microcracks and moisture content changes within concrete, and are prone to strong Rayleigh scattering and absorption attenuation at heterogeneous interfaces. Based on the above physical phenomena, the main control module controls the execution of the following sub-steps to calculate the dispersion attenuation coefficient: In step S501, the main control module extracts the full-wavelength ultrasonic signal from its internal storage and truncates a preset time window. To preserve core physical characteristics, the width of this window is constrained to be two to three times the nominal period of the ultrasonic transducer. As a preferred method, to suppress the spectral leakage effect caused by time-domain truncation, the main control module applies a Hanning window to the truncated preset time window for smooth transition processing. The Hanning window is chosen because it can effectively suppress sidelobe interference, and its main lobe width is moderate, which is conducive to the accurate segmentation of the acoustic frequency band. After windowing, the main control module performs a Fast Fourier Transform on the time-domain signal sequence to obtain the corresponding discrete frequency domain amplitude spectrum. The underlying butterfly operation logic of the Fast Fourier Transform can be directly called by those skilled in the art using standard digital signal processing algorithm libraries; it is a well-known technology in the field and will not be elaborated upon here.

[0042] In step S502, the main control module uses the cutoff frequency output from the previous step as the frequency band division boundary to divide the discrete frequency domain amplitude spectrum into low-frequency and high-frequency bands. Specifically, within the low-frequency band formed by the set lower frequency limit and the cutoff frequency, the main control module performs discrete integration on the squared amplitude value of the discrete frequency domain amplitude spectrum to obtain the low-frequency band energy integral value. Similarly, within the high-frequency band formed by the cutoff frequency and the set upper frequency limit, the main control module performs the same discrete integration calculation to obtain the high-frequency band energy integral value. This energy integration operation macroscopically quantifies the absolute distribution of sound wave energy in different frequency bands from the frequency domain dimension, so as to truly reflect the energy transfer characteristics in the sound dispersion process.

[0043] S503, to eliminate non-target interference caused by power fluctuations in the transmitting transducer and differences in coupling layer thickness in the absolute energy value, the main control module needs to calculate the relative dispersion attenuation coefficient using a ratio method. Considering that in extremely deteriorated or severely internally vacuous concrete areas, high-frequency ultrasonic waves may be completely absorbed or geometrically scattered, causing the high-frequency energy integral value to approach zero, the main control module pre-determines whether the high-frequency energy integral value is less than a set spectral noise floor integration threshold before performing the division operation. This spectral noise floor integration threshold is physically calibrated based on the inherent thermal noise power spectral density of the system's receiving circuit. To avoid misjudgment caused by a single extreme value, the main control module simultaneously introduces the total energy signal-to-noise ratio across the entire frequency band as a two-dimensional verification indicator. When the high-frequency band energy integral value is less than or equal to the spectral noise floor integral threshold, and the total energy signal-to-noise ratio across the entire frequency band is lower than the preset distortion limit, the main control module determines that the high-frequency signal is completely attenuated and directly assigns a preset maximum constant penalty value to the dispersion attenuation coefficient. This constant penalty value is typically set to 99.9. When the high-frequency band energy integral value is greater than the spectral noise floor integral threshold, or the total energy signal-to-noise ratio across the entire frequency band is within the normal communication range, the main control module divides the low-frequency band energy integral value by the high-frequency band energy integral value to calculate the dispersion attenuation coefficient of the first ultrasonic wave, as shown in the following formula: ; In the formula, is the dispersion attenuation coefficient of the first ultrasonic wave, which is dimensionless and used to characterize the ability of microscopic defects and water content in the concrete of the test area to scatter and strip away high-frequency sound waves. This is the low-frequency band energy integral value, representing the total energy carried by ultrasonic waves during transmission within the low-frequency band. This is the high-frequency energy integral value, representing the total energy carried by ultrasonic waves during transmission within the high-frequency band. A larger value for this dispersion attenuation coefficient indicates a stronger dissipation effect of scatterers within the medium on high-frequency energy.

[0044] In S504, the main control module stores the calculated dispersion attenuation coefficient into its internal register queue. This characteristic parameter will serve as the core criterion for determining whether the acoustic attenuation mechanism in the current test area is dominated by internal water saturation in the subsequent state machine.

[0045] See attached document Figure 1 In this embodiment, the water saturation of the pores inside the concrete has a nonlinear dissipation effect on the high-frequency ultrasonic energy. When the component is in a high moisture content state, the ultrasonic scattering is amplified and attenuated. At this time, the surface mortar is in a generally softened state, and the mechanical feedback of a single measuring point is sufficient to characterize the macroscopic compressive strength benchmark. However, when the component is in a dry or low moisture content state, the high-frequency ultrasonic energy attenuation is weaker. At this time, the hardness difference between the aggregate and mortar phase on the concrete surface will cause spatial variability, and a single physical impact is prone to mechanical characteristic deviation due to accidental impact on coarse aggregate. Based on the above cross-modal physical constraint mechanism, the main control module controls and executes the following state machine determination and feature set update sub-steps: S601, the main control module extracts the ultrasonic first-wave dispersion attenuation coefficient calculated in real time in the previous step and calls the system's preset scattering threshold for numerical comparison. The physical relevance of choosing the dispersion attenuation coefficient as the input parameter for threshold determination lies in the fact that this coefficient macroscopically quantifies the viscous absorption effect of the liquid medium on high-frequency mechanical waves. The specific value of this scattering threshold is a priori calibrated based on the reference acoustic test data of concrete specimens of the same grade under a fully immersed and saturated state. To ensure the robustness of the comparison logic and avoid one-sided judgments based solely on a single extreme value, the main control module introduces a two-dimensional logic check using the time-domain sound velocity attenuation rate before performing the comparison. If the coefficient is given a maximum constant penalty value due to an anomaly, or if the spatial gradient difference between the coefficient and adjacent measurement points exceeds the preset robustness boundary, the high scattering state flag is directly triggered.

[0046] S602, the main control module executes branch state machine logic based on the numerical comparison results. When the dispersion attenuation coefficient is greater than or equal to the scattering threshold, the system determines that the current measurement area is in a high-scattering, humid, or water-saturated state. Under this physical state, the main control module directly confirms the current mechanical feature set as the initial energy absorption ratio and terminates the micro-area dynamic rebound acquisition action of the current measurement area.

[0047] S603, when the dispersion attenuation coefficient is less than the scattering threshold, the system determines that the current test area is in a low-scattering dry state, and further elimination of local impedance interference caused by the random distribution of surface aggregate is required. As a preferred method, the main control module triggers a spatial array impact command and prompts the operator through an audio-visual interactive interface. The operator uses a limiting plate to constrain the rebound acquisition module to perform multiple spatial discrete impacts within the test area. The fixed-space array of holes inside the limiting plate ensures, mechanically, that the spatial coordinates of each impact are independent, avoiding falsely low secondary impact measurements caused by overlapping plastic damage zones.

[0048] S604, with the continuous execution of multiple discrete impacts, the main control module, based on the system's strict internal clock synchronization mechanism, sequentially receives the rebound dynamics time sequence corresponding to each physical action, ensuring precise alignment of each spatial displacement with the timestamp. The main control module cyclically calls the time-domain quadratic integral and energy conversion analysis logic to calculate the kinetic energy conversion consumption ratio corresponding to each independent impact, thereby constructing a set of spatial energy absorption ratio sequences containing measurements from multiple discrete impacts.

[0049] S605, to eliminate extreme interference from local defects and high-hardness aggregates and extract the mechanical characteristics that best represent the strength of the true concrete matrix, the main control module performs an order statistic filtering algorithm on the spatial energy absorption ratio sequence set. Specifically, the main control module sorts all elements in the sequence set in ascending order, generating a one-dimensional ordered sequence. Considering that the minimum values ​​at the beginning of the sequence usually correspond to surface microcracks or pore voids, while the maximum values ​​at the end of the sequence usually correspond to high-hardness coarse aggregate regions, the main control module uses a truncated mean filtering mechanism to remove extreme values ​​at both ends of the sequence, calculating the arithmetic mean of the retained sequence as the expected value of the mortar phase energy. Before performing this mean calculation, the main control module pre-checks the remaining effective sample size to determine... Is it less than or equal to zero? When When the value is less than or equal to zero, to avoid division overflow caused by the denominator approaching zero in division operations, the main control module interrupts the calculation and triggers a shift retest instruction; when When the value is greater than zero, the main control module performs a calculation of the predetermined expected value. The necessary technical formulas used in this invention are as follows: ; In the formula, The expected value of the mortar phase energy obtained for calculation is, in physical terms, the statistical convergence value of the energy dissipated by the actual plastic deformation of the surface mortar phase after removing heterogeneous boundary disturbances. The total number of spatial discrete impacts is physically determined by the total number of array holes in the limiting plate; The first in the set of spatial energy absorption ratio sequences arranged in ascending order. One element; This is the parameter for the number of bilaterally truncated tails. Equal to the total number of times The value obtained by multiplying the result by the preset truncation coefficient and then rounding it down is typically between 0.15 and 0.25, ensuring that the filtering algorithm can effectively isolate outliers while retaining a sufficient number of valid samples.

[0050] In S606, the main control module fuses the calculated expected energy values ​​of the mortar phase to generate pure mortar phase features and establishes them as a mechanical feature set, replacing the original initial single-impact features. This mechanism uses acoustic scattering parameters as a priori gating switches to dynamically and adaptively adjust the spatial density of mechanical sampling, effectively balancing the execution efficiency of on-site detection with the extraction accuracy of the underlying multiphase material physical features.

[0051] See attached document Figure 1 In this embodiment, when the component is in a low-scattering dry state, the main control module triggers multiple discrete impact commands. To ensure the spatial independence of the rebound dynamics, the plastic damage zone induced by a single impact must be geometrically isolated. Based on the above stress field superposition constraint requirements, the system coordinates and executes the following microarray spatial impact and synchronization constraint sub-steps: S701, the operator attaches and fixes the limiting plate to the preset test area on the surface of the component to be tested. As a preferred method, the limiting plate is made of high-rigidity alloy material, and its interior has an array of positioning guide holes. The inner diameter of the array of positioning guide holes is clearance-fitted with the outer diameter of the probe of the rebound acquisition module. To further improve the validity of the test data, this mechanical adaptation design ensures that the probe's impact direction is strictly perpendicular to the component surface, eliminating the tangential kinetic energy loss component introduced by the tilting impact; on the other hand, it reduces the frictional resistance interference when the probe slides within the hole.

[0052] S702, to avoid stress unloading and microcrack penetration between adjacent impact points, the center distance between adjacent positioning guide holes of the limiting plate must be greater than the effective plastic influence radius formed by a single physical impact on the concrete surface. During the physical process of the projectile impacting the concrete surface, the impact kinetic energy is converted into local plastic deformation energy on the surface and elastic strain energy radiating inwards, thus forming a hemispherical damage influence zone below the contact point. Based on the above general mechanical principles, the main control module calculates this minimum safe distance using Hertzian contact theory, based on the system's nominal initial impact kinetic energy and estimated compressive strength. Considering that the estimated lower limit of the strength of extremely deteriorated components may approach zero, to avoid distance divergence or even overflow errors caused by minimizing the denominator in division operations, the main control module presets a material mechanics failure threshold value before calculation. When the input estimated strength is lower than this threshold value, the failure threshold value is forcibly used as the calculation benchmark, as shown in the following formula: ; In the formula, The minimum center distance constraint value between adjacent positioning guide holes is usually set in millimeters. Its physical meaning is to define the diffusion and attenuation boundary of a single impact energy in the non-homogeneous concrete body, so as to ensure that the mechanical response of each test point in the space is independent of each other. The nominal initial mechanical kinetic energy of a single physical impact by the rebound acquisition module is determined by the physical stiffness and extension stroke of the spring inside the rebounder. To estimate the compressive strength of the component under test, the selection of this parameter is based on the fact that the yield strength of the target is directly inversely proportional to the expansion scale of the plastic pit. This is the safety diffusion coefficient in the plastic zone, which is dimensionless. This coefficient is usually obtained through regression analysis of numerous destructive indentation experiments, and its typical value ranges from 2.5 to 3.5. It is designed to reserve sufficient undisturbed matrix around the periphery of the principal stress influence zone to cope with local variations in material gradation.

[0053] S703, based on the aforementioned geometric constraints, the operator sequentially inserts the rebound acquisition module into each positioning guide hole of the limiting plate to perform physical impacts. Accompanying the impact action, the main control module synchronously records the spatial coordinate index of the current data frame. To ensure strict alignment of multi-source data in both spatial and operational dimensions, the main control module generates a composite label containing a timestamp and two-dimensional matrix coordinates for each valid impact. In specific implementation, the main control module allocates a unique timestamp identifier based on its internal global hardware timer. For data locking and timestamp synchronization mechanisms in multi-threaded concurrent environments, those skilled in the art can use standard real-time operating system event scheduling components, which are well-known technologies in the field and will not be elaborated upon here.

[0054] In S704, after completing a limited number of spatial discrete impacts, the main control module performs an integrity check on the dynamic time sequence set with composite tags. By verifying whether there are any missing nodes in the two-dimensional matrix coordinates, the main control module determines whether the current round of microarray impact actions completely covers the survey area grid. To avoid misjudgment due to relying solely on spatial coordinates, the main control module simultaneously extracts the timestamp information from each composite tag and calculates the time interval between two adjacent impacts. If this time interval exceeds the preset reasonable operation time window (e.g., conventionally set to 1 to 5 seconds), it is determined that the operation has been delayed by human intervention or the probe has been misplaced. Based on the above multi-dimensional verification results, if there are missing or abnormal data frames, the main control module provides a point-to-point prompt for retesting through the audio-visual interactive interface; if the verification passes, the cleaned time series set is submitted to the statistical filtering module for the calculation and extraction of the expected value of the mortar phase energy. This mechanism effectively prevents the risk of insufficient spatial sampling rate caused by human error and non-standard operation, ensuring that the final output mechanical parameters are statistically representative.

[0055] See attached document Figure 1In this embodiment, concrete, as a typical multiphase composite material, has a macroscopic mechanical response determined by the combined effects of the coupling of the mortar matrix, coarse aggregate, and the interfacial transition zone. Based on the aforementioned principle of multiphase mechanical heterogeneity, the local contact response of a single physical impact often involves random aggregate rebound or microcrack dissipation, making it impossible to decouple the independent contributions of each phase. To achieve the mapping from discrete spatial sampling to a single stable parameter, the main control module performs multiphase decoupling calculations, aiming to separate the pure mortar phase characteristic parameters that best characterize the overall material degradation from the discrete spatial sampling sequence. The system executes the following sub-steps: In S705, after confirming that the microarray spatial impact action completely covers the test area, the main control module retrieves a set of effective spatial energy absorption ratio sequences with multi-dimensional spatiotemporal labels. As a preferred method, the main control module performs sorting and reconstruction on this sequence set, calling its internal fast sorting component to transform the disordered sequence containing multiple independent measurements into a one-dimensional monotonically increasing ordered sequence. By introducing this numerical analysis method, the mechanical response, originally randomly distributed in geometric space, is re-aggregated into a one-dimensional statistical space. In this statistical space, the low-value interval at the beginning of the sequence represents pores or microcracks with extremely low mechanical impedance; the high-value interval at the end of the sequence represents the coarse aggregate embedding zone with extremely high mechanical stiffness; and the stable distribution area in the middle of the sequence precisely corresponds to the matrix mortar phase to be detected.

[0056] To avoid overfitting or undersampling issues caused by using fixed empirical truncation coefficients, the main control module adaptively determines the two-sided truncation boundary of the sequence based on the principle of numerical differentiation. In actual physical detection, due to the large differences in the elastic impedance of the medium between different phases (such as defects and mortar, mortar and aggregate), when the impact potential energy crosses the internal phase boundary layer, the sequence will inevitably experience a sudden jump in the differential gradient in the boundary region. Accordingly, the main control module performs a first-order relative difference operation on the generated one-dimensional ordered sequence to quantify the severity of the transition in local contact stiffness between adjacent statistical sequences. When performing this relative difference operation to obtain the relative energy gradient, its underlying logic is the difference between the next sequence element and the current element divided by the current element. Considering that there may be surface penetrating holes in the test area that result in extremely low local energy absorption ratios, the main control module pre-determines whether the extracted value of the current element, which serves as the denominator, approaches zero. If the value falls into a preset hardware-aware dead zone (e.g., below the minimum quantization unit of the underlying analog-to-digital converter), the main control module forcibly sets the current relative energy gradient to a preset maximum saturation constant or skips the calculation of this set of indices to avoid overflow errors caused by division by zero. Based on the boundary position where the relative energy gradient crosses the preset material impedance transition threshold, the main control module dynamically optimizes and locks the lower and upper boundary truncation indices. The specific value of this material impedance transition threshold is calibrated based on the standard dispersion range of concrete with the same mix proportion under stress-free damage conditions, and the typical value range is between 5% and 10%. For the specific iterative optimization process of sequence differential scanning and mutation point capture, those skilled in the art can use a conventional sliding window differential detection algorithm, which is a well-known technique in the field and will not be elaborated upon here.

[0057] In S707, the main control module extracts the effective core sub-sequence located in the middle of the sequence based on the determined upper and lower boundary truncation indices, and reconstructs the comprehensive mechanical feature set of the test area through multi-dimensional evaluation logic. The main control module calculates the arithmetic mean of the effective core sub-sequence to obtain the expected value of mortar phase energy representing the pure mortar stiffness. To avoid insensitivity to extreme values ​​of internal material degradation caused by relying solely on a single mean, the main control module further introduces the coefficient of variation of the sub-sequence as a spatial dispersion penalty factor to construct a two-dimensional judgment vector. When using the standard deviation of the core sub-sequence divided by the arithmetic mean to obtain the coefficient of variation, the main control module pre-checks whether the arithmetic mean of the effective core sub-sequence is close to zero. If the mean is lower than the preset minimum tolerance (such as the floating-point lower limit set by the system's underlying layer), the main control module directly assigns the coefficient of variation to the upper limit saturation constant set by the system, thereby avoiding divergence anomalies and program crashes caused by minimizing the denominator in the division operation. The physical meaning of this coefficient of variation is to characterize the fluctuation range of mechanical properties inside the pure mortar matrix caused by uneven micro-hydration or local drying shrinkage after removing the extreme interference of aggregate and pores.

[0058] In S708, the main control module performs exponential decay nonlinear weighted fusion of the expected value of the mortar phase energy and the calculated coefficient of variation to generate pure mortar phase features and establish them as a mechanical feature set, which is then output to the subsequent data aggregation module. To specifically implement the above nonlinear weighted logic and avoid biased evaluation caused by relying solely on the arithmetic mean, this embodiment constructs a comprehensive feature calculation formula that incorporates a dispersion penalty mechanism: ; In the formula, The final aggregated pure mortar phase characteristics are established as the aforementioned mechanical characteristic set, which is dimensionless and serves as the core input variable for subsequent cross-modal regression solutions. The expected value of mortar phase energy obtained from the previous steps is dimensionless and is used to characterize the average mechanical impedance level of the pure mortar matrix in the test area. Spatial Dispersion Penalty Coefficient and coefficient of variation Together they constitute the negative exponential factor. ,because and All are positive numbers, so the value of this negative exponential factor is always between (0, 1). The more heterogeneous the component is internally (i.e., the coefficient of variation), the more likely it is to change. As the value of the index increases, its value decreases rapidly in a non-linear manner, thus affecting the expected value of the mortar phase energy. Implementing more stringent feature reduction perfectly aligns with the physical mechanism of conservatively approximating the true mechanical compressive lower limit of the component; The coefficient of variation of the effective core subsequence is dimensionless and represents the statistical dispersion of the effective data interval. The spatial dispersion penalty coefficient is dimensionless. The physical reason for choosing this exponential decay function as the fusion scheme is that: when the coefficient of variation of the mortar phase within the survey area... As the value increases, it indicates a hidden tendency for uneven degradation within the matrix of the indicator material. In this case, the negative exponent term is used to adjust the expected value. Applying nonlinear reduction allows for a more conservative approximation of the true lower limit of the structural member's compressive strength. This penalty coefficient... The value range is typically calibrated between 0.5 and 1.5, specifically determined based on the allowable structural risk tolerance in engineering testing standards. This adaptive filtering decoupling mechanism effectively isolates high-impedance hard spots and low-impedance voids in heterogeneous composite materials through the mapping transformation between physical space sampling and statistical space. Simultaneously, it utilizes the coefficient of variation to supplement the assessment of local homogeneity, ensuring that the extracted cross-modal input features can accurately and robustly reflect the effective compressive strength benchmark of the surface mortar of the tested component.

[0059] See attached document Figure 1In this embodiment, after completing the ultrasonic dispersion attenuation analysis and microarray impact feature extraction within the test area, the system obtains multidimensional state parameters characterizing different physical levels of the material. To address the magnitude differences and collinear coupling interference among the multiple physical quantities, the main control module executes the following feature convergence and orthogonal solution operator steps: S801, the main control module retrieves various independent test parameters cached in the underlying storage medium. These parameters specifically include the propagation velocity and dispersion attenuation coefficient, characterizing internal density and microcrack distribution, as well as the rebound value, impact contact time, and mechanical feature set, reflecting the stiffness of local plastic damage on the surface. To ensure absolute alignment of the multi-source heterogeneous data with the operating conditions before inputting it into the model, the main control module relies on a global hardware clock source and spatial coordinate index to perform rigorous label matching and frame synchronization on the aforementioned cross-modal parameters, thereby constructing an initial comprehensive state matrix containing the state data of all valid detection nodes. This matrix mathematically and physically binds the ultrasonic acoustic dimension and the mechanical dynamics dimension, serving as the original input base for subsequent high-dimensional space dimensionality reduction mapping.

[0060] S802. Considering that the speed of ultrasonic propagation is typically on the order of thousands, while the dispersion attenuation coefficient or mechanical characteristic parameters may be in the single digit or decimal range, directly measuring distance in the original coordinate system would cause the physical weights of small-scale features to be severely overwhelmed by high-value features. As a preferred approach, the main control module forces the initial integrated state matrix to undergo zero-mean and unit-variance standardization before performing orthogonal decomposition. During the standardization operation, the main control module pre-calculates the empirical variance of each feature column. If the variance of a feature column approaches the floating-point lower limit set by the system's underlying layer, it indicates that the detection feature is in a dead state and has lost its information-expressing ability in the current measurement area. The main control module will dynamically remove the dead feature column, thereby avoiding the denominator minimization overflow error in the standardization division operation and constructing the standardized initial integrated state matrix.

[0061] S803. Given that both ultrasonic velocity and rebound value are macroscopically constrained by the water-cement ratio and porosity of concrete, multicollinearity exists among these state parameters. Based on this physical correlation, if a conventional multiple linear regression algorithm is directly applied to calculate compressive strength, the covariance matrix generated during the calculation process will inevitably tend to be singular, leading to highly unstable regression coefficients and loss of generalization ability. To overcome this matrix collinearity defect, the main control module calls the partial least squares regression algorithm, using nonlinear projection to find the orthogonal projection direction with the largest variance in the original feature space and the strongest correlation with compressive strength. The main control module deconstructs the input data by iteratively extracting mutually independent orthogonal principal component latent variables. In the specific underlying mathematical projection, the main control module no longer performs the matrix inner product iteration during the model training phase, but directly calls the preset orthogonal weight vector in the internal memory. The main control module projects the standardized feature vector of the current test area along the first orthogonal weight vector using a linear dot product, generating the first-dimensional orthogonal principal component latent variable.

[0062] In step S804, after successfully extracting the orthogonal principal component latent variables of the current dimension, the main control module performs a dimensionality reduction stripping operation on the feature residual matrix and intensity residual vector. The main control module uses loading coefficients to strip the residual information explained by the principal component from the current matrix features, generating an updated residual matrix, thus constructing an orthogonal and non-overlapping mathematical background for the next round of feature extraction. For the specific algebraic operation logic of matrix stripping and updating and inner partial least squares iteration, those skilled in the art can use standard singular value decomposition (SVD) or nonlinear iterative partial least squares (NIPALS) basic algorithm components, which are well-known technologies in the field and will not be elaborated upon here.

[0063] S805, after the number of extracted principal components reaches the preset optimal principal component truncation dimension, the main control module terminates the iterative stripping of latent variables. The main control module then performs linear weighted quantization and superposition of the latent variables of each dimension of orthogonal principal components with preset regression mapping coefficients to reconstruct the final concrete compressive strength of this test area. The necessary technical formulas used in this invention are as follows: ; In the formula, The concrete compressive strength is reconstructed and output by the main control module. Its unit is megapascal (MPa), and its physical meaning is to characterize the true mechanical compressive performance limit of the component under the current service state. The strength reference bias constant, with its unit being megapascals (MPa), is determined by the overall average of historical samples during the model's factory calibration phase and is used to establish the baseline level of the compressive strength of the test area. The optimal principal component truncation dimension is selected for the system. It is dimensionless and its physical purpose is to cut off the interference of tail noise components. The conventional value is set between 2 and 4 to balance the generalization ability and prediction accuracy of the model. For the first The regression mapping coefficients corresponding to the latent variables of each orthogonal principal component are dimensionless and represent the physical contribution weights of the orthogonal dimension features determined by a large number of samples during the factory calibration stage to the final intensity change. The independent latent variable parameters extracted in the previous steps are dimensionless and serve as pure cross-modal feature inputs after eliminating collinearity interference. This reconstruction mechanism completely eliminates the surface state sensitivity blind spot present in single physical detection methods. Relying on feature fusion in a multivariable orthogonal space, it achieves a comprehensive dimensionality reduction solution for both microscopic pore defects and macroscopic mechanical strength within the material.

[0064] To better understand the technical solution of the present invention, the following example of a non-destructive evaluation of the compressive strength of a reinforced concrete bridge pier that has been in service for 15 years is used to illustrate the working process of the system in detail.

[0065] Specific application examples: Step 1: Initial State Acquisition and Ultrasonic Determination. The operator selects a flat area on the side of the bridge pier as the test area, with the ultrasonic transducer distance preset to 200mm. The system emits ultrasonic waves towards the surface to be tested, extracting the arrival time of the first wave as 52.6 microseconds. From this, the main control module calculates the macroscopic propagation speed of sound to be approximately 3802 m / s. Next, the operator performs the first physical single impact at the center point of the test area using the rebound acquisition module. By analyzing the zero-crossing point of the acceleration time history curve, the main control module precisely locks the contact trigger time at 0.50ms and the separation time at 1.95ms, thus extracting the impact contact time as 1.45ms. Since this contact time is within the normal range, the main control module, based on the mapping formula, calculates the cutoff frequency for this test area to be 65kHz.

[0066] Step 2: Dispersion Analysis and Multiphase Branch Triggering. The main control module performs a Fourier transform on the entire ultrasonic waveform and calculates the integral of the low-frequency band energy and the high-frequency band energy with 65kHz as the boundary. By calculating the ratio of these two frequency band energies, the dispersion attenuation coefficient of the ultrasonic first wave is found to be 1.25. The system's preset scattering threshold is 2.0. Since the currently measured attenuation coefficient (1.25) is less than the threshold (2.0), the main control module automatically determines that the current bridge pier surface is in a low-scattering dry / semi-dry state, which is highly susceptible to interference from local aggregate distribution. Therefore, the system illuminates the indicator light and triggers the spatial array impact command.

[0067] Step 3: Microarray Impact and Decoupling Reconstruction. The operator places the limiting plate with 3×3 array holes (9 holes in total) firmly against the surface of the test area. The main control module verifies in the background that the center distance between adjacent positioning guide holes is 30mm (greater than the calculated single plastic influence radius of 22mm, meeting the independence requirement). The operator performs physical impacts on the 9 holes sequentially. The main control module obtains 9 sets of independent spatial single energy absorption ratio sequences, which are arranged in ascending order as follows: 0.28, 0.50, 0.52, 0.53, 0.54, 0.55, 0.57, 0.81, 0.85. After first-order differential gradient optimization and truncated mean filtering (removing two extremely high values ​​of 0.81 and 0.85 that hit the coarse aggregate, and one extremely low value of 0.28 that hit the surface micropores), the main control module finally extracted the expected energy value of 0.535, which is the mean of the remaining 6 effective core subsequences between 0.50 and 0.57. Combined with the coefficient of variation of this subsequence, the pure mortar phase characteristics of the test area are generated and established as the mechanical feature set.

[0068] Step 4: Orthogonal Dimensionality Reduction and Strength Output Finally, the main control module aggregates the propagation speed of sound (3802 m / s), dispersion attenuation coefficient (1.25), and mechanical feature set into an initial comprehensive state matrix. After zero-mean standardization, this matrix is ​​input into a partial least squares regression (PLS) model. The system extracts three orthogonal principal component latent variables along the preset weight vector, performs linear weight quantization and superposition, and finally outputs on the screen the target concrete reconstructed compressive strength of the bridge pier test area as 34.6 MPa. Subsequent core drilling experiments confirmed that the actual core compressive strength of this area was 35.1 MPa, with an error of only 1.4%.

[0069] Experimental verification and effect comparison: To verify the technological advancements and beneficial effects of this invention, the inventors' team conducted large-scale comparative experiments in a laboratory environment.

[0070] Experimental Sample Preparation: Fifty standard concrete cube specimens of 150mm×150mm×150mm with different grades (design strength covering C20 to C60), different moisture contents (completely dry to completely water saturated), and different degrees of deterioration (different carbonization depths and microcrack densities) were selected for the experiment.

[0071] For each group of test blocks, the following three testing methods were used to predict the compressive strength. After the non-destructive testing was completed, a destructive compressive strength test was conducted on a universal testing machine to obtain the true compressive strength limit value as the evaluation criterion: Method A (Prior Art 1): Traditional single rebound method (conducted in accordance with industry standard JGJ / T23).

[0072] Method B (Prior Art 2): Traditional ultrasonic-rebound integrated method (using a fixed intensity measurement curve, without decoupling the mortar phase and frequency domain adaptive characteristics).

[0073] Method C (this invention): A comprehensive solution system based on multimodal frequency division scattering adaptive gating and sequential statistics filtering.

[0074] Comparison table of experimental data and results

[0075] The experimental conclusions can be seen from the table above: The prediction error was greatly reduced: the average relative error dropped sharply from 10.2% in the traditional comprehensive method to 2.8%. This is mainly due to the successive statistical filtering technology (S705-S708) of this invention, which successfully removed the hard point interference and the soft point interference of voids in the aggregate and extracted the mortar phase reference characteristics that best represent the macroscopic strength of the material.

[0076] Overcoming interference from environmental and age variations: For complex, variable samples with high water content or abnormally degraded surfaces, traditional methods are prone to errors exceeding 20%. This invention introduces an ultrasonic frequency domain adaptive truncation (S403) and a water saturation scattering threshold gating mechanism (S601-S602), eliminating collinearity interference between different phase states. It achieved zero misclassification in 50 samples, demonstrating strong engineering robustness and generalization ability.

[0077] Reference Appendix Figure 3 This paper demonstrates the core working mechanism of the present invention in achieving time-frequency cross-modal analysis and multiphase data decoupling when performing non-destructive evaluation on reinforced concrete bridge piers that have been in service for 15 years. The mechanism is specifically divided into three sub-figures: Figure B (top right): Analysis of zero crossing of rebound acceleration time history (corresponding to step 1 of the embodiment) Image Explanation: This image shows the acceleration time-history damped oscillation curve collected and processed by the main control module after the operator performs the first physical single impact at the center of the test area.

[0078] Corresponding Implementation Example: By accurately analyzing the zero-crossing point of the signal, the system pinpointed the contact time (0.50ms) and separation time (1.95ms) on the surface of the measuring point, thereby precisely extracting the core physical parameter: the impact contact time is 1.45ms (as shown by the dashed scale in the figure). This contact time is within the normal range, reflecting the transient elastic impedance of the surface concrete, and serves as the physical benchmark for subsequent triggering and calculation of adaptive segmentation in the ultrasonic frequency domain.

[0079] Figure A (top left): Ultrasound full waveform frequency domain segmentation (corresponding to steps 1 and 2 of the embodiment) Image Explanation: This image shows the amplitude spectrum distribution of the first ultrasonic wave after Fourier transform. The thick black dashed line in the image represents the cutoff frequency (65kHz) dynamically calculated by the system using a mapping formula based on the contact time (1.45ms) extracted from Figure B.

[0080] Corresponding Implementation Example: Using 65kHz as the boundary, the system strictly divides the ultrasonic spectrum area into two integration intervals: a light gray region for low-frequency energy and a dark gray region for high-frequency energy. By calculating the integration ratio of these two frequency bands, the dispersion attenuation coefficient is determined to be 1.25. Since the current attenuation coefficient (1.25) is less than the system's preset scattering threshold (2.0), the main control module successfully determines that the surface of the bridge pier is in a low-scattering, dry state, making it highly susceptible to interference from local aggregate distribution. This triggers the indicator light and the subsequent microarray impact command.

[0081] Figure C (bottom): Spatial array polymorphic sequence truncation filtering mechanism (corresponding to steps 3 and 4 of the embodiment) Image Explanation: This bar chart depicts nine spatially independent single-shot energy absorption ratio sequences obtained after an operator performs nine physical impacts using a limiting plate with a 3×3 array of holes, and is arranged in ascending order.

[0082] Corresponding Implementation Example: After processing with the unique truncated mean filtering algorithm of this invention, the system successfully identified and removed extreme interference terms, perfectly decoupling the multiphase data. Pure white bar (ascending index 1, value 0.28): represents an extremely low value that was rejected, corresponding to one invalid data point (soft spot defect interference) that hit the surface micropores or microcracks in the embodiment.

[0083] Dark gray bars (ascending indexes 8 and 9, values ​​0.81 and 0.85 respectively): represent extremely high values ​​that were rejected, corresponding to two invalid data points (hard point aggregate interference) that were directly hit on the high-hardness coarse aggregate below in the embodiment.

[0084] Light gray bars (ascending index 2-7, values ​​between 0.50 and 0.57): represent the remaining 6 valid core subsequences, characterizing the true pure mortar matrix phase.

[0085] The horizontal black dashed line represents the expected mean of the six valid light gray column data points obtained by the main control module, yielding an expected value of 0.535 for the mortar phase energy. This expected value completely eliminates random interference from aggregates and pores, and is established as the mechanical characteristic set representing the macroscopic strength of the test area. Finally, this characteristic set, along with the sound velocity (3802 m / s), attenuation coefficient (1.25), etc., is zero-mean standardized and input into the PLS model, accurately outputting a reconstructed compressive strength of 34.6 MPa (with a core drilling error of only 1.4%).

[0086] Light gray columns (index 2-7): represent the remaining 6 effective mortar phase data.

[0087] The thick black dashed line represents the expected value of mortar phase energy extracted by the system after averaging these 6 valid data points. This expected value completely eliminates random interference from aggregates and pores, becoming a pure mortar phase characteristic representing the macroscopic strength of the test area and establishing it as a mechanical feature set. Finally, it is input into the PLS model and outputs a high-precision reconstructed strength of 34.6 MPa.

Claims

1. A comprehensive method for testing concrete strength based on rebound and ultrasonic methods, characterized in that, The method is executed by a detection system and includes the following steps: The detection system acquires the full waveform signal of the ultrasonic field in the test area and calculates the propagation speed of sound by combining the transducer spacing. The rebound value and acceleration sequence obtained after the projectile impact in the test area are acquired. The acceleration sequence is analyzed to determine the projectile contact time and the initial energy absorption ratio is calculated. The impact contact time is mapped to a cutoff frequency, and the dispersion attenuation coefficient is calculated by performing a frequency band integral on the full ultrasonic waveform signal based on the cutoff frequency. The dispersion attenuation coefficient is compared with the scattering threshold, and the initial energy absorption ratio or pure mortar phase characteristics are established as the mechanical feature set based on the comparison results. The propagation speed of sound, the rebound value, the impact contact time, the dispersion attenuation coefficient, and the mechanical feature set are combined to construct a standardized initial integrated state matrix; The orthogonal principal component latent variables are extracted from the initial comprehensive state matrix and quantified and superimposed to output the concrete compressive strength.

2. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 1, characterized in that, The initial energy absorption ratio is calculated by the following steps: Zero-crossing point analysis is performed on the acceleration sequence to identify the contact trigger moment and separation moment that cross the zero baseline. The impact contact time is calculated by the time difference between the separation moment and the contact trigger moment. Within the time interval formed by the contact trigger moment and the separation moment, perform a second time-domain integral on the acceleration sequence to extract the initial impact velocity and the initial rebound separation velocity; Based on the extracted initial impact velocity and the initial rebound separation velocity, the kinetic energy conversion consumption ratio is calculated to obtain the initial energy absorption ratio.

3. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 1, characterized in that, Mapping the impact contact time to a cutoff frequency specifically includes the following steps: The impact contact time is verified and determined against a preset clock noise floor tolerance threshold. When the impact contact time is greater than the clock noise floor tolerance threshold, an inverse proportional function mapping calculation is performed, and the preset mapping coefficient is divided by the impact contact time to obtain the initial cutoff limit. Based on the upper and lower frequency limits of the system hardware, boundary clamping processing is performed on the initial truncation limit, and the clamped initial truncation limit is used as the truncation frequency.

4. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 1, characterized in that, The dispersion attenuation coefficient is calculated by the following steps: A preset time window is extracted from the full waveform of the ultrasound signal and a Hanning window is applied. The discrete frequency domain amplitude spectrum is then obtained by fast Fourier transform. Using the cutoff frequency as the dividing line, the amplitude square value of the discrete frequency domain amplitude spectrum is integrated within the set low-frequency band and high-frequency band intervals respectively to obtain the low-frequency band energy integral value and the high-frequency band energy integral value. After performing a noise floor check on the high-frequency band energy integral value, the low-frequency band energy integral value is divided by the high-frequency band energy integral value to obtain the dispersion attenuation coefficient.

5. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 1, characterized in that, The initial energy absorption ratio or pure mortar phase characteristics are established as a set of mechanical characteristics, specifically including the following steps: Determine whether the dispersion attenuation coefficient calculated in real time is greater than or equal to the scattering threshold; When the dispersion attenuation coefficient is greater than or equal to the scattering threshold, the test area is determined to be in a high scattering humid state, and the initial energy absorption ratio is directly confirmed and assigned to the mechanical feature set. When the dispersion attenuation coefficient is less than the scattering threshold, the test area is determined to be in a low scattering dry state, triggering a space array impact command to obtain the pure mortar phase characteristics and establish them as the mechanical feature set.

6. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 5, characterized in that, Triggering the spatial array impact command, acquiring the pure mortar phase characteristics and establishing them as the mechanical feature set, specifically includes the following steps: A limiting plate with arrayed positioning guide holes is attached to the surface of the test area. The minimum safe distance between the guide holes is calculated based on the estimated compressive strength to constrain adjacent impact points from overlapping local plastic damage areas. Multiple spatial discrete impacts are sequentially performed within the array positioning guide hole. The timestamps are recorded synchronously, and the corresponding energy absorption ratio is calculated by cyclically calling the time-domain quadratic integral logic to construct a set of spatial energy absorption ratio sequences containing multiple impact data. The spatial energy absorption ratio sequence set is reconstructed by ascending order, converted into a one-dimensional monotonically increasing ordered sequence, and input into the statistical filtering module. The pure mortar phase characteristics are obtained through the statistical filtering module and established as the mechanical feature set.

7. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 6, characterized in that, The process of obtaining the pure mortar phase characteristics through the statistical filtering module and establishing them as the mechanical feature set includes the following steps: Perform a first-order relative difference operation on the ordered sequence to calculate the relative energy gradient between adjacent elements; Based on the boundary position where the relative energy gradient crosses the preset material impedance transition threshold, the lower boundary truncation index and the upper boundary truncation index are dynamically optimized and locked to extract the effective core subsequence located in the middle of the sequence. The arithmetic mean of the effective core subsequences is calculated to generate the expected value of the mortar phase energy. The coefficient of variation of the effective core subsequences is extracted. The expected value of the mortar phase energy and the coefficient of variation are then subjected to exponential decay nonlinear weighted fusion to generate the pure mortar phase features and establish them as the mechanical feature set.

8. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 1, characterized in that, Constructing the standardized initial integrated state matrix includes the following steps: Based on the global hardware clock source, the propagation speed of sound, the rebound value, the impact contact time, the dispersion attenuation coefficient, and the execution time and spatial label of the mechanical feature set are aligned to construct an initial comprehensive state matrix; The empirical variance of each feature column in the initial integrated state matrix is ​​checked, and dead feature columns below the set floating-point lower limit are dynamically removed. The remaining matrix is ​​then standardized with zero mean and unit variance to construct the standardized initial integrated state matrix.

9. A comprehensive method for testing concrete strength based on rebound and ultrasonic methods according to claim 8, characterized in that, Extracting orthogonal principal component latent variables from the initial integrated state matrix specifically includes the following steps: The features of the initial integrated state matrix are projected onto a preset orthogonal weight vector using a linear dot product to extract the orthogonal principal component latent variables of the first dimension. Using loading coefficients, information residuals that have been explained by the orthogonal principal component latent variables of the first dimension are extracted from the current matrix features to generate an updated residual matrix; The orthogonal feature projection and information stripping logic is iteratively executed on the updated residual matrix until the number of extracted principal components reaches the preset optimal principal component truncation dimension, and a series of mutually independent orthogonal principal component latent variables are extracted.

10. The method for comprehensive testing of concrete strength based on rebound and ultrasonic methods according to claim 1, characterized in that, The latent variables of the orthogonal principal components are quantified and superimposed to output the concrete compressive strength, which specifically includes the following steps: Obtain the intensity benchmark bias constant set during the model calibration phase, as well as the regression mapping coefficients corresponding to the latent variables of each of the orthogonal principal components; The weights are assigned by multiplying each of the orthogonal principal component latent variables by its corresponding regression mapping coefficient. The weighted latent variables of all the orthogonal principal components are linearly superimposed with the strength benchmark bias constant to reconstruct and output the concrete compressive strength.

Images