Self-aware performance detection method and system for carbon fiber concrete

Abstract

The application relates to the technical field of civil engineering materials and structural health monitoring, and discloses a self-perception performance detection method and system for carbon fiber concrete. The method comprises the following steps: obtaining original resistance signals, synchronously correcting temperature and current data, and obtaining corrected resistance signals; obtaining a purified stress resistance component through wavelet transform decomposition and inverse transform; obtaining a local resistance change rate based on frequency domain analysis, and constructing a stress resistance characteristic distribution; separating a damage sub-path and performing multi-scale decomposition when the resistance rises, and obtaining a damage candidate signal; performing grid division, threshold comparison and boundary marking according to the signal, and obtaining a damage positioning coordinate sequence; calculating a cumulative damage index, performing stress distribution fitting, and constructing a damage continuous distribution model; extracting a real-time resistance path, mapping the real-time resistance path to the model for feature matching, and obtaining a crack recognition result. The method can realize accurate positioning and recognition of damage under a complex environment.

Description

Technical Field

[0001] This invention relates to the field of civil engineering materials and structural health monitoring technology, and in particular to a self-sensing performance testing method and system for carbon fiber concrete. Background Technology

[0002] Currently, in the field of modern construction and infrastructure, material performance testing and structural health monitoring (SHM) are crucial for ensuring engineering safety and the safety of life and property. With the widespread application of new smart materials such as carbon fiber concrete, utilizing their own piezoresistive effect to achieve real-time and accurate damage identification and early warning has become a research focus in the fields of intelligent transportation and large bridge maintenance. This self-sensing technology, through the deployment of intelligent sensors or acquisition systems, converts the microscopic changes inside the material into electrical signals, and is widely used in scenarios such as stress monitoring of long-term loaded components and early diagnosis of microcracks.

[0003] In a common existing technology, external patch sensors or offline sampling and detection methods are typically used. Stress data is acquired by installing devices such as strain gauges on the material surface, or the structure is periodically ultrasonically scanned to assess internal damage. This approach involves installing external measuring equipment at pre-selected monitoring points, collecting surface deformation or wave velocity data, and calculating the internal stress state using empirical formulas. However, this approach suffers from drawbacks. External devices struggle to capture the subtle evolutionary patterns within the material and are highly susceptible to nonlinear interference from external factors such as temperature fluctuations and stray currents in actual operating environments. This results in an extremely low signal-to-noise ratio for the detected signal, making it difficult to accurately identify the true damage characteristics.

[0004] In summary, existing technologies suffer from low damage identification accuracy and insufficient reliability of real-time sensing under complex multi-source interference. Summary of the Invention

[0005] This invention provides a self-sensing performance detection method and system for carbon fiber concrete to solve the problems of low damage identification accuracy and insufficient real-time sensing reliability under complex multi-source interference.

[0006] In a first aspect, to address the aforementioned technical problems, the present invention provides a method for detecting the self-sensing performance of carbon fiber concrete, comprising:

[0007] The original resistance signal, synchronous temperature data, and current data are acquired, and interference correction is performed to obtain the corrected resistance signal.

[0008] The correction resistor signal is decomposed into low-frequency stress components by wavelet transform at multiple levels, and the low-frequency stress components are then inversely transformed to obtain purified stress resistance components.

[0009] The local resistance change rate is obtained by frequency domain feature analysis based on the purified stress resistance components, and the correspondence between stress and resistance is constructed based on the local resistance change rate to obtain the stress resistance characteristic distribution.

[0010] If the stress resistance characteristic distribution shows that the resistance value continues to rise, then the damage sub-path is separated from the stress resistance characteristic distribution, and the damage sub-path is decomposed into multiple scales to obtain the damage candidate signal.

[0011] Based on the candidate damage signal, the region is located to obtain the location result, and the boundary is marked based on the location result to obtain the damage location coordinate sequence;

[0012] The cumulative damage index is calculated based on the damage location coordinate sequence. The cumulative damage index that exceeds the preset damage risk threshold is fitted with stress distribution to construct a continuous damage distribution model.

[0013] The real-time resistance change path is extracted from the damage candidate signal and mapped to the damage continuous distribution model for feature matching to obtain the crack identification result.

[0014] Secondly, the present invention provides a self-sensing performance testing system for carbon fiber concrete, comprising:

[0015] The signal correction module is used to acquire the original resistance signal, synchronize temperature data and current data, and perform interference correction to obtain the corrected resistance signal;

[0016] The signal purification module is used to perform wavelet transform multi-level decomposition on the correction resistor signal to obtain low-frequency stress components, and to perform inverse transform on the low-frequency stress components to obtain purified stress resistance components.

[0017] The feature association module is used to perform frequency domain feature analysis based on the purified stress resistance components to obtain the local resistance change rate, and to construct the correspondence between stress and resistance based on the local resistance change rate to obtain the stress resistance feature distribution.

[0018] An anomaly capture module is used to separate the damage sub-path from the stress resistance feature distribution if the resistance value shows a continuous increase, and to perform multi-scale decomposition on the damage sub-path to obtain damage candidate signals.

[0019] The damage localization module is used to perform area localization based on the damage candidate signal to obtain a localization result, and to perform boundary marking based on the localization result to obtain a damage localization coordinate sequence.

[0020] The continuous modeling module is used to calculate the cumulative damage index based on the damage location coordinate sequence, and to fit the stress distribution of the cumulative damage index that exceeds the preset damage risk threshold to construct a continuous damage distribution model.

[0021] The crack identification module is used to extract the real-time resistance change path from the damage candidate signal and map it to the damage continuous distribution model for feature matching to obtain the crack identification result.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) This invention calculates the interference compensation factor by analyzing the fluctuation characteristics of temperature and current data, and uses this factor to perform regression adjustment on the original resistance signal to determine the interference removal result. This scheme quantifies the influence of environmental factors on the piezoresistive characteristics of materials, and removes the nonlinear noise generated by temperature fluctuations and current drift from the underlying signal, thereby achieving accurate correction of the original monitoring data and reducing the interference distortion of multi-source environmental factors on the structural detection results.

[0024] (2) This invention employs wavelet transform to perform multi-level decomposition of the correction signal, utilizes the decomposition level selection to separate low-frequency stress components from high-frequency noise transient interference, and combines inverse transform for signal reconstruction. This scheme effectively solves the problem of difficult identification of internal stress changes and chaotic electrical signals in carbon fiber concrete through in-depth deconstruction of frequency domain features, thereby obtaining high-purity stress resistance components and improving the signal-to-noise ratio and stress correlation in the signal processing process.

[0025] (3) This invention extracts the spatial distribution features of the damaged area and marks the location coordinate sequence through spatial grid division, preset threshold comparison and path model association technology, and then verifies the output through cumulative damage index and continuous distribution model. This scheme maps one-dimensional resistance change into a three-dimensional spatial damage field, realizing the whole process tracking of micro-cracks from initial location to boundary determination, thereby not only improving the sensitivity of damage identification, but also ensuring the timeliness and accuracy of bridge beam damage evaluation. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the self-sensing performance testing method for carbon fiber concrete provided in the first embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of the structure of the self-sensing performance testing system for carbon fiber concrete provided in the second embodiment of the present invention. Detailed Implementation

[0028] The technical solutions of 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.

[0029] Reference Figure 1 The first embodiment of the present invention provides a method for testing the self-sensing performance of carbon fiber concrete, comprising the following steps:

[0030] S11: Acquire the original resistance signal, synchronous temperature data and current data, and perform interference correction to obtain the corrected resistance signal;

[0031] S12, perform wavelet transform multi-level decomposition on the correction resistor signal to obtain low-frequency stress components, and perform inverse transform on the low-frequency stress components to obtain purified stress resistance components.

[0032] S13, perform frequency domain feature analysis on the purified stress resistance components to obtain the local resistance change rate, and construct the correspondence between stress and resistance based on the local resistance change rate to obtain the stress resistance characteristic distribution;

[0033] S14, if the stress resistance characteristic distribution shows that the resistance value continues to rise, then the damage sub-path is separated from the stress resistance characteristic distribution, and the damage sub-path is decomposed into multiple scales to obtain the damage candidate signal.

[0034] S15, perform region localization based on the damage candidate signal to obtain the localization result, and perform boundary labeling based on the localization result to obtain the damage localization coordinate sequence;

[0035] S16, calculate the cumulative damage index based on the damage location coordinate sequence, and fit the stress distribution of the cumulative damage index that exceeds the preset damage risk threshold to construct a continuous damage distribution model.

[0036] S17, extract the real-time resistance change path from the damage candidate signal and map it to the damage continuous distribution model for feature matching to obtain the crack identification result.

[0037] In step S11, the original resistance signal, synchronous temperature data, and current data are acquired, and interference correction is performed to obtain the corrected resistance signal, including:

[0038] Time-series analysis is performed on the synchronized temperature data and the current data, and the fluctuation frequency of the synchronized temperature data and the current data is extracted by Fourier transform;

[0039] An interference model is constructed based on the fluctuation frequency, and the interference model is used to solve the original resistance signal to obtain the interference compensation factor.

[0040] The original resistance signal is linearly adjusted using the interference compensation factor to obtain the corrected resistance signal.

[0041] In this embodiment, a sensor array pre-embedded inside the carbon fiber concrete structure synchronously acquires multi-source physical data. The original resistance signal refers to the sequence of resistance values ​​that characterize the initial electrical response of the material, captured in real time through embedded electrodes; the synchronous temperature data refers to the ambient temperature values ​​that correspond one-to-one with the original resistance signal on the time axis, obtained using a temperature sensor spatially adjacent to the electrodes; and the current data refers to the real-time current intensity value fed back by the excitation source in the monitoring circuit.

[0042] The original resistance signal, synchronous temperature data, and current data naturally contain resistance offset characteristics, signal noise characteristics, and signal distortion characteristics during the acquisition process. The resistance offset characteristic refers to the baseline shift caused by the zero-point drift of the sensor. The signal noise characteristic refers to the high-frequency random fluctuations caused by environmental electromagnetic interference. The signal distortion characteristic refers to the signal clipping or abrupt changes caused by poor electrode contact.

[0043] This embodiment first performs time-series analysis on the synchronized temperature data and the current data, and extracts the fluctuation frequencies of the synchronized temperature data and the current data through Fourier transform. Specifically, the discrete sequences of the synchronized temperature data and the current data that change over time are mapped from the time dimension to the frequency dimension. In the generated spectrum distribution, the frequency components with the highest energy proportion are extracted by searching for the extreme points of the amplitude envelope, and these are determined as the temperature fluctuation frequency and the current fluctuation frequency, respectively. The two together constitute the fluctuation frequency.

[0044] It should be noted that, in order to isolate the resistance increase caused by all non-damaging factors (such as temperature decrease), this embodiment constructs an interference model based on the fluctuation frequency. The interference model is an environmental sensitivity mapping model established based on multiple linear regression, used to quantify the contribution of non-stress factors to the resistance change. The construction process of the interference model involves, under a controlled environment without external mechanical loads, gradually adjusting the rate of change of ambient temperature and the disturbance frequency of the excitation current, recording the resistance shift of carbon fiber concrete under different frequency combinations; using a least squares fitting algorithm to establish the functional relationship between the fluctuation frequency and the resistance shift, determining the temperature influence coefficient and the current influence coefficient, thereby completing the construction of the interference model.

[0045] It is worth noting that during the training of the disturbance model, the temperature change rate was gradually adjusted from 0.02 degrees Celsius per minute to 0.2 degrees Celsius per minute in a no-load constant temperature chamber, and the current disturbance frequency was adjusted from 0.01 Hz to 0.1 Hz. A total of 25 sets of resistance offset data under different frequency combinations were collected, and multiple linear regression was used for fitting.

[0046] Subsequently, this embodiment utilizes the interference model to solve the original resistance signal and obtain the interference compensation factor. Specifically, the real-time extracted fluctuation frequency is input into the interference model, which calculates the pseudo-bias in resistance caused by the current environmental disturbance based on the temperature influence coefficient and the current influence coefficient. The pseudo-bias in resistance is divided by a preset reference resistance value to obtain the offset ratio, which is then determined as the interference compensation factor. The preset reference resistance value is obtained by measuring carbon fiber concrete specimens under no-load constant temperature conditions using the four-electrode method, and the arithmetic mean of 10 consecutive minutes of sampling data is taken as the reference resistance value.

[0047] Finally, this embodiment uses the interference compensation factor to linearly adjust the original resistance signal to obtain a corrected resistance signal. Specifically, the value of the original resistance signal at each moment is multiplied by a correction coefficient obtained by subtracting the interference compensation factor from 1. Through this linear adjustment process, this embodiment eliminates thermal drift caused by temperature fluctuations and electrical noise caused by current instability in the original signal, thereby outputting a corrected resistance signal that objectively reflects the stress changes in the internal structure of the material. It should be noted that the interference correction specifically uses a carbon fiber concrete "temperature-resistance" correlation dataset pre-collected in a controlled constant temperature chamber, and obtains the temperature compensation coefficient by fitting it using the least squares method. During correction, the compensation increment is calculated based on the synchronous temperature data and subtracted from the original resistance. In addition, using the voltage divider circuit principle, the collected voltage is divided by the synchronous current data to offset the influence of power supply ripple on the resistance measurement. Furthermore, for small fluctuations in the power supply current, the system monitors the current intensity at the input terminal in real time, and uses a modified description of Ohm's law to divide the collected voltage signal by the real-time current intensity to obtain a normalized resistance value, thereby eliminating the interference of power supply instability on the self-sensing performance.

[0048] For example, the sensor array acquires an original resistance signal with an average value of 130.5 ohms, simultaneously monitoring a temperature fluctuation frequency of 0.05 Hz and a current fluctuation frequency of 0.02 Hz. In this embodiment, the above fluctuation frequencies are substituted into the interference model, which calculates that the spurious resistance increment caused by environmental interference is 0.7 ohms. This embodiment divides 0.7 ohms by the reference value of 130 ohms to obtain an interference compensation factor of approximately 0.0054. Through linear adjustment, the incremental component corresponding to this ratio is removed from the original signal, resulting in a final output corrected resistance signal of 129.8 ohms.

[0049] It should be noted that the reference resistance value needs to be set based on the mix design document and laboratory calibration data of the specific carbon fiber concrete component. For example, for a standard specimen with a carbon fiber volume content of 0.5%, the initial resistance value under standard curing conditions at 20 degrees Celsius can be set as the reference resistance value. Those skilled in the art can determine the corresponding reference value based on the test report of the specific engineering material. This calibration process complies with the technical specifications for multi-source environmental noise compensation in the health monitoring of civil engineering structures, ensuring the reference stability of subsequent damage identification data.

[0050] In step S12, the correction resistance signal is decomposed into low-frequency stress components using wavelet transform at multiple levels, and the low-frequency stress components are then subjected to inverse transform to obtain purified stress resistance components, including:

[0051] The correction resistor signal is subjected to five-level wavelet decomposition to obtain the high-frequency coefficient matrix and the low-frequency coefficient matrix;

[0052] The high-frequency coefficient matrix is ​​set to zero according to a preset frequency threshold to obtain the reconstructed coefficient matrix;

[0053] The low-frequency coefficient matrix is ​​superimposed with the reconstructed coefficient matrix to obtain the low-frequency stress components;

[0054] The low-frequency stress component is subjected to inverse discrete wavelet transform to obtain the purified stress resistance component.

[0055] First, the system performs a five-level wavelet decomposition on the correction resistor signal to obtain a high-frequency coefficient matrix and a low-frequency coefficient matrix. It should be noted that the correction resistor signal, as the direct input source for step S12, undergoes a data integrity check before being transmitted to the wavelet decomposition module. This ensures that the corrected resistance value is within a preset physical reasonable range, for example, avoiding negative values ​​or values ​​exceeding the range of infinity, thus guaranteeing the numerical stability of subsequent multi-level decompositions. The preset physical reasonable range has a lower limit of 0.5 times the reference resistance value and an upper limit of 2.0 times the reference resistance value. Values ​​exceeding this range are considered sensor malfunctions or signal anomalies, suspending subsequent processing and issuing an alarm. Specifically, this embodiment uses a multi-scale analysis wavelet basis function with good compact support and orthogonality, and employs a multi-resolution analysis algorithm to project the correction resistor signal in the time domain onto the wavelet space. By continuously performing five iterative decompositions, the original signal is decomposed into a fifth-level low-frequency coefficient matrix representing the overall signal profile, and first to fifth-level high-frequency coefficient matrices representing different levels of local detail features. The five-layer decomposition depth is based on the acoustic emission signal characteristics generated by the propagation of microcracks inside carbon fiber concrete. Its effective energy frequency band has been verified by offline spectrum analysis and is mainly distributed in the 1 / 32 range of the sampling frequency. This characteristic frequency band can be accurately separated through five-layer decomposition.

[0056] Subsequently, the system sets the high-frequency coefficient matrix to zero according to a preset frequency threshold, obtaining the reconstructed coefficient matrix. In this embodiment, the preset frequency threshold is a physical cutoff boundary determined based on the sampling frequency and the number of decomposition layers. The system clears all values ​​in the high-frequency coefficient matrices from the first to the fifth layer to zero, aiming to filter out random noise generated by external electromagnetic induction or the data acquisition equipment itself in the original resistance signal through coefficient shielding in the spatial domain, retaining only trend information that has a direct causal relationship with structural stress and strain. The setting of the number of decomposition layers is determined based on the matching relationship between the sampling frequency and the target signal frequency band. Since the stress change of carbon fiber concrete is a low-frequency signal (usually below 5Hz), while the frequency of environmental electromagnetic interference is high, under the condition of a sampling frequency of 400Hz, the signal can be divided into multiple frequency bands after 5 layers of decomposition. The frequency range of the low-frequency component in the fifth layer exactly covers 0 to 6.25Hz, which can completely wrap the stress characteristics and effectively eliminate high-frequency noise.

[0057] Next, the system superimposes the low-frequency coefficient matrix with the reconstructed coefficient matrix to obtain the low-frequency stress components. The technical essence of this step is to perform frequency-domain selective reconstruction preparation of the signal, that is, to logically merge the retained fifth-layer low-frequency coefficients with the high-frequency coefficients that have already been zeroed, thereby providing a complete feature source containing pure low-frequency information for the subsequent inverse transform.

[0058] Finally, this embodiment performs inverse discrete wavelet transform on the low-frequency stress components to obtain purified stress-resistance components. Specifically, the system follows the reverse logic of the decomposition, performing upsampling and mirror filtering reconstruction layer by layer starting from the highest decomposition level, ultimately mapping the wavelet space coefficients back to the time domain. The obtained purified stress-resistance components eliminate high-frequency spike interference and accurately preserve the resistance trend changes caused by the compression of the concrete internal structure, providing high signal-to-noise ratio foundational data for the quantitative assessment of stress state in subsequent steps.

[0059] It should be noted that the order of the basis wavelet function selected in this embodiment is determined based on the dynamic balance point between waveform smoothness and computational overhead. Those skilled in the art can select a decomposition depth that can cover the main frequency band of stress variation, for example, four to six layers, based on the actual sampling rate of the sensing system and by analyzing the power distribution characteristics of the signal. This does not depart from the core idea of ​​the present invention.

[0060] It is worth noting that, during the inverse discrete wavelet transform process, in order to ensure the integrity of the signal energy and the accuracy of the reconstruction result, the coefficients of the reconstruction filter must strictly satisfy the orthogonal mirror condition. In one implementation, if the reconstructed signal exhibits numerical fluctuations at the edges of the time axis, the system will automatically employ a symmetrical physical extension method to preprocess the beginning and end of the signal to ensure the numerical continuity of the purified stress resistance component at the beginning and end of the observation period.

[0061] It is worth noting that in this embodiment, the preset frequency threshold is equal to the lowest frequency value corresponding to the high-frequency coefficients of the first layer after the five-layer wavelet decomposition; setting all coefficients in the high-frequency coefficient matrix to zero is equivalent to filtering out all components with frequencies higher than the threshold, thus achieving low-pass filtering; the threshold is calculated based on the sampling frequency and the number of decomposition layers, specifically the sampling frequency divided by two and the number of decomposition layers plus one.

[0062] For example, during the compressive strength calibration test of a carbon fiber concrete beam, the original collected resistance data contained high-frequency vibration signals caused by large construction machinery in the vicinity. In step S12, the system forcibly cleared the interference components with frequencies higher than a specific cutoff point through a five-level decomposition. After inverse transformation reconstruction, the purified stress-resistance components eliminated most of the background noise, making the physical characteristic of the resistance value decreasing linearly with increasing load more prominent, thus ensuring the scientific validity and reproducibility of the subsequent strain sensitivity calculation results.

[0063] In step S13, frequency domain feature analysis is performed on the purified stress-resistivity components to obtain the local resistance change rate, and the correspondence between stress and resistance is constructed based on the local resistance change rate to obtain the stress-resistivity characteristic distribution, including:

[0064] A fast Fourier transform is performed on the purified stress resistance component to obtain the characteristic component frequency. The amplitude ratio is calculated based on the characteristic component frequency to obtain the local resistance change rate.

[0065] A short-time Fourier transform is performed on the local resistance change rate to generate a time-frequency spectrum, and a time-frequency mapping matrix is ​​constructed based on the time-frequency spectrum;

[0066] Principal component analysis is performed on the time-frequency mapping matrix to obtain characteristic principal components, and linear combination projection is performed on the characteristic principal components to obtain the stress resistance characteristic distribution.

[0067] First, the system performs a Fast Fourier Transform on the purified stress-resistance component to obtain the characteristic component frequency. Based on the characteristic component frequency, an amplitude ratio is calculated to obtain the local resistance change rate. Specifically, in this embodiment, the signal purified in step S12 is transformed from the time domain to the frequency domain, and the characteristic component frequency is determined by identifying the frequency point with the most concentrated energy in the spectrum. Subsequently, the ratio of the square of the amplitude at this characteristic frequency to the square of the root mean square value of the total energy of the original signal is calculated, thereby quantifying the relative drastic change in resistance with stress in the energy domain. It should be noted that, to ensure the scientific validity of the data distribution, a base-10 logarithmic operation is performed on the ratio, and the result is normalized to the decibel level to obtain the local resistance change rate. The local resistance change rate reflects the intensity of resistance fluctuation caused by the closure of microcracks or the reorganization of the conductive network inside the carbon fiber concrete within a specific observation window.

[0068] In one implementation, the local resistance change rate is obtained by performing differential calculation on the purified stress resistance component in the time domain to obtain the instantaneous change in resistance value. This instantaneous change is then divided by the resistance reference value at the current sampling point, thereby directly quantizing the dimensionless local resistance change rate in the time domain. Subsequently, this time-domain change rate is used as the input source for the subsequent short-time Fourier transform to generate a time-frequency spectrum.

[0069] Subsequently, the system performs a short-time Fourier transform on the local resistance change rate to generate a time-frequency spectrum, and constructs a time-frequency mapping matrix based on the spectrum. In this embodiment, a fixed-length sliding window function is used to segment the local resistance change rate sequence, and frequency analysis is performed on each segment to obtain dynamic information on the evolution of resistance characteristics over time. The system arranges the frequency energy distributions corresponding to different times in columns to construct a two-dimensional numerical matrix, namely the time-frequency mapping matrix. Each element in this matrix represents the contribution of a specific frequency component at a specific time, which can characterize the non-stationary characteristics of resistance change during stress loading.

[0070] Next, the system performs principal component analysis on the time-frequency mapping matrix to obtain eigenvalues, and then performs linear combination projection based on these eigenvalues ​​to obtain the stress-resistance characteristic distribution. Specifically, the time-frequency mapping matrix is ​​first centered, its covariance matrix is ​​calculated, and the top few eigenvectors with the highest contribution rates are extracted through eigenvalue decomposition, defined as the eigenvalues. The extraction criterion for the eigenvalues ​​is that the cumulative variance contribution rate reaches 90%. By performing singular value decomposition on the time-frequency mapping matrix, the first five eigenvectors are retained. These five eigenvectors carry most of the nonlinear information of resistance evolution with stress, thus achieving data dimensionality reduction without losing key damage features. These principal components concentrate most of the resistance evolution law in the stress response and eliminate redundant secondary interference. Finally, the system projects the original time-frequency data onto a low-dimensional subspace composed of these eigenvalues, and generates a spatial distribution feature reflecting the correlation between stress intensity and resistance change, i.e., the stress-resistance characteristic distribution, through linear combination.

[0071] It should be noted that the selection of principal components for feature extraction in this embodiment is based on the cumulative contribution rate reaching a preset threshold. In this example, the cumulative contribution rate threshold is set to 85% to 95%, which is obtained through statistical analysis of all time-frequency characteristics generated by 30 groups of standard carbon fiber concrete specimens in loading failure tests. Experimental data shows that the first 3 to 5 principal components can carry more than 90% of the variance information in the original resistance evolution law. Selecting this range can achieve data dimensionality reduction and improve the real-time performance of calculations, while ensuring that key microcrack evolution characteristics are not missed. Those skilled in the art can set the cumulative contribution rate between 85% and 95% according to the complexity of the experimental data to ensure that key stress characteristic information is not lost while achieving data dimensionality reduction. This setting method aims to extract the core parameters that best represent the compressive state of the material.

[0072] It is worth noting that the overlap rate of the sliding window function is set based on a balance between time resolution and frequency resolution during the construction of the time-frequency mapping matrix. In one implementation, the system uses a 50% overlap rate to ensure that transient resistance jumps caused by sudden stress are not missed on the time axis. If the detected signal has strong non-stationarity, the system will automatically reduce the window function length to improve the tracking accuracy of the rapid loading process.

[0073] For example, when monitoring the load on a carbon fiber concrete bridge pier in service, step S13 receives the purified resistance sequence. A fast Fourier transform is used to identify characteristic frequencies corresponding to the vehicle passage cycle, and the local resistance change rate is calculated. Subsequently, a short-time transform is used to observe the frequency shift of the resistance value at the instant the vehicle passes, and principal component analysis is used to extract the core projection features representing the pier's compressive state. The resulting stress-resistance feature distribution clearly reveals the real-time correspondence between resistance changes and the pier's load-bearing capacity, providing a structured data matrix for accurate stress state classification in subsequent step S14.

[0074] In step S14, if the stress-resistance characteristic distribution shows a continuous increase in resistance value, then a damage sub-path is separated from the stress-resistance characteristic distribution, and the damage sub-path is decomposed into multiple scales to obtain damage candidate signals, including:

[0075] The real-time slope of the stress resistance characteristic distribution is obtained by taking the first derivative.

[0076] The damage sub-path is obtained by tracking the real-time slope change path that exceeds the preset slope threshold using a dynamic programming algorithm;

[0077] Five-level wavelet decomposition is performed on the damaged sub-path to obtain a set of high-frequency components, and the instantaneous energy distribution is obtained by calculating the energy concentration based on the set of high-frequency components.

[0078] Based on the instantaneous energy distribution, peak frequency is determined to obtain damage candidate signals.

[0079] First, the system performs a first-order derivative on the stress-resistance characteristic distribution to obtain the real-time change slope. Specifically, in this embodiment, the system performs a difference operation on the projected characteristic distribution sequence to calculate the rate of change of values ​​between adjacent characteristic points. The real-time change slope reflects the rate of evolution of the carbon fiber concrete resistance value with time or stress loading step. Since the resistance value of carbon fiber concrete usually decreases with pressure when the structure is intact, when the resistance value in the stress-resistance characteristic distribution shows a continuous upward trend, it indicates that microcracks may have occurred inside the material, leading to damage to the conductive path. Specifically, the system monitors the sign of the derivative of the stress-resistance characteristic distribution in the time dimension in real time. If the change in characteristic value is greater than zero for three consecutive sampling periods, it is determined that the resistance value is continuously rising, thereby triggering the damage sub-path separation process in step S14. In another case, if the stress-resistance characteristic distribution shows that the resistance value is in a downward trend or fluctuating equilibrium state, it is determined that the current concrete structure is in the normal piezoresistive induction stage and no microcrack damage has occurred. At this time, the system automatically skips steps S14 to S17, directly outputs the real-time stress monitoring results, and returns to step S11 to execute the signal acquisition for the next cycle.

[0080] It should be noted that, in order to avoid false triggering caused by signal noise, the system introduces a hysteresis comparison mechanism; the specific implementation of the determination that the resistance value is continuously rising is as follows: if the net increase of the feature value reaches four or more times within five consecutive sampling periods, and the maximum pullback amplitude during the period does not exceed five percent of the current peak value, then it is determined that the resistance value has entered the continuous rising stage and the damage sub-path separation is triggered.

[0081] Subsequently, the system uses a dynamic programming algorithm to track the path of change where the real-time slope exceeds a preset slope threshold, thus obtaining the damage sub-path. In this embodiment, the preset slope threshold is determined based on the baseline fluctuation range of the material during the elastic deformation stage. In this example, the slope threshold is set to 1.5 times the absolute value of the normal piezoresistive coefficient, and this threshold is determined by a benchmark calibration method. At the initial stage of system operation, the average slope of the resistance change of the material during the elastic deformation stage (when the stress is less than 30% of the rated strength) is first collected. Since the resistance increase caused by damage is a nonlinear abrupt change, setting a redundancy coefficient of 1.5 times can effectively avoid normal slope fluctuations caused by small temperature fluctuations or load oscillations, thereby accurately capturing irreversible damage signals inside the structure. If the real-time slope does not exceed the preset slope threshold, the material is determined to be in the elastic deformation stage or a stable piezoresistive response range. The system does not perform damage sub-path separation and directly skips the subsequent damage location and identification processes from S15 to S17, returning to S11 to continue real-time monitoring. This branching mechanism ensures that the system does not generate false damage warning signals under normal load conditions.

[0082] It should be noted that the slope residual term is the absolute value of the difference between the real-time change slope of the current sampling point and the preset slope threshold; the smoothing term is the absolute value of the difference between the slope values ​​of two adjacent sampling points when they are selected into the damaged sub-path; the cost is equal to the sum of the slope residual term and the smoothing term, wherein the weight coefficient of the smoothing term is 0.5; the dynamic programming calculates the path with the minimum cumulative cost point by point starting from the first sampling point, and when the local slope of three consecutive sampling points exceeds the slope threshold, the path segment is marked as a damaged sub-path.

[0083] Subsequently, the system uses a dynamic programming algorithm to search for a path with the minimum cumulative cost across the entire feature space. This path connects all points that satisfy the abnormal slope growth characteristic, thereby extracting the nonlinear signal segment representing material structural damage from the overall resistance distribution, which is defined as the damage sub-path. The cost function of the dynamic programming algorithm consists of a slope residual term and a smoothing term. Its path search rule is to start from the beginning of the time series and find a trajectory that minimizes the rate of change of the cumulative slope of the entire path. When the local slope exceeds the slope threshold for three consecutive sampling points, sub-path marking is automatically triggered to ensure that the physical process of instantaneous crack initiation leading to a jump in resistance is captured.

[0084] Next, the system performs a five-level wavelet decomposition on the damaged sub-path to obtain a set of high-frequency components, and calculates the instantaneous energy distribution based on the energy concentration of the high-frequency component set. Specifically, a basis wavelet function with high-resolution characteristics is selected to perform multi-scale fine decomposition on the damaged sub-path, extracting high-frequency detail features from the first to the fifth level to form the high-frequency component set. Subsequently, the system calculates the sum of squares of the high-frequency components at each sampling time and performs normalization processing to obtain the instantaneous energy distribution reflecting the distribution of signal abrupt change energy on the time axis.

[0085] Finally, the system determines the peak frequency based on the instantaneous energy distribution to obtain damage candidate signals. In this embodiment, the system identifies the time point in the instantaneous energy distribution where energy is significantly concentrated and extracts the frequency component corresponding to that point. If the frequency is within a preset crack propagation characteristic frequency band, the signal segment is determined to be an acoustic emission or resistance jump signal caused by internal structural damage, and it is encapsulated as the damage candidate signal. The obtained damage candidate signal eliminates the influence of steady-state load and focuses on revealing the transient characteristics of microscopic damage inside the concrete; wherein, the preset crack propagation characteristic frequency band is obtained by synchronously acquiring cracking signals during the three-point bending loading process of the carbon fiber concrete specimen through an acoustic emission sensor, and the dominant frequency range is determined to be 80 Hz to 800 Hz through spectral analysis.

[0086] It should be noted that the step size of the first-order derivative method used in this embodiment is selected based on the signal sampling frequency of the sensing system. Those skilled in the art can employ a multi-point smoothing derivative method to eliminate the interference of local minor fluctuations on the calculation of the real-time slope, based on the signal-to-noise ratio of the experimental environment; this does not depart from the core idea of ​​the present invention.

[0087] It is worth noting that the dynamic programming algorithm introduces a path smoothing penalty term when searching for damaged sub-paths. In one implementation, if the slope of adjacent feature points changes too abruptly and does not conform to physical logic, the algorithm will automatically adjust the path weights to ensure the continuity of the extracted sub-paths, thereby accurately pinpointing the occurrence process of internal material damage.

[0088] For example, when monitoring a carbon fiber reinforced concrete member subjected to fatigue loads, step S14 identifies an abnormal upward fluctuation in the resistance value at a certain stage. First-order differentiation reveals that the slope exceeds the critical point of the normal piezoresistive effect. The system uses a dynamic programming algorithm to accurately locate and separate this abnormal fluctuation path. After five-level wavelet decomposition, a peak energy value is found in the high-frequency component at a certain instant. Frequency determination shows that this peak value matches the characteristics of microcrack propagation, thus confirming it as a candidate damage signal. This signal is then passed to step S15 for subsequent precise quantitative analysis of the damage degree.

[0089] In step S15, region localization is performed based on the candidate damage signal to obtain a localization result, and boundary labeling is performed based on the localization result to obtain a damage localization coordinate sequence, including:

[0090] Obtain surface data of the bridge beam and divide the bridge beam surface into a set of grid cells based on the surface data;

[0091] The damage candidate signal is mapped to the set of grid cells to obtain cell resistance distribution data containing cell resistance values;

[0092] The unit resistance distribution data is scanned and compared with a preset resistance variation threshold to obtain an abnormal grid set in which the unit resistance value exceeds the resistance variation threshold.

[0093] Boundary fitting is performed on the abnormal mesh set to obtain the damage location coordinate sequence.

[0094] First, the system acquires surface data of the bridge beam and divides it into a set of grid cells based on this data. Specifically, in this embodiment, the geometric dimensions and surface morphology features of the beam are acquired through a pre-established bridge structural information model or on-site laser scanning. The system divides the monitored area of ​​the beam into equally spaced geometric grids based on the arrangement array of carbon fiber concrete sensors. Each grid represents an independent stress monitoring element, thus constructing the set of grid cells. The grid density is selected based on the spacing of the sensor electrodes and the expected minimum damage size, aiming to ensure that the positioning accuracy can cover the microscopic areas where cracks develop.

[0095] Subsequently, the system maps the candidate damage signals to the set of grid cells, obtaining unit resistance distribution data containing unit resistance values. In this embodiment, an interpolation algorithm or an electrical tomography algorithm is used to restore the candidate damage signals extracted in step S14 to their corresponding spatial locations. Specifically, based on the resistance jump information collected by the electrodes, the resistance distribution of each grid cell at the instant of damage occurrence is calculated through sensitivity matrix solving, thereby generating a spatial matrix reflecting the damaged state of the conductive network inside the beam, i.e., the unit resistance distribution data. Each value in the matrix corresponds to the real-time resistance level of a specific grid cell.

[0096] It should be noted that in this embodiment, sixteen electrodes are arranged at equal intervals on the surface or inside of the carbon fiber concrete structure. A constant current is applied to each pair of adjacent electrodes in sequence using an adjacent excitation method, and the potential difference between the remaining electrode pairs is measured. The sensitivity matrix is ​​pre-calculated based on the finite element method, and a finite element model completely identical to the set of mesh elements is established. The unit resistance change is set in each mesh element in sequence, and the potential difference change under each electrode combination is calculated, thereby constructing the Jacobian matrix between the electrode response and the unit resistance change. During mapping, the resistance jump amplitude in the damage candidate signal is used as the measurement vector, and the resistance change of each mesh element is solved by regularized least squares method to obtain the unit resistance distribution data.

[0097] Next, the system scans and compares the unit resistance distribution data with a preset resistance variation threshold, filtering out abnormal grid sets where the unit resistance value exceeds the resistance variation threshold. In this embodiment, the preset resistance variation threshold is determined based on the material's resistance value benchmark under healthy conditions and the range of environmental noise fluctuations. In this example, the resistance variation threshold is set to three times the standard deviation of the local resistance change rate, and this threshold is preset using statistical distribution. By collecting resistance data of the beam under static load equilibrium for at least 10 minutes, the standard deviation of the resistance value of each grid unit is calculated. According to the characteristics of normal distribution, values ​​exceeding three times the standard deviation are extremely rare abnormal events, usually corresponding to the physical interruption of the internal conductive path of the material. This setting method ensures that the positioning results have a very high degree of confidence. The system traverses every element in the unit resistance distribution data through a global scanning algorithm, comparing the resistance value of each unit with the threshold one by one. If the resistance value of a certain grid unit increases significantly and exceeds the threshold, it is determined that the unit is on the damage path through which the crack passes, and it is recorded in the abnormal grid set. If the resistance values ​​of all grid cells do not exceed the resistance variation threshold, the current signal fluctuation is determined to be within the allowable noise range of the sensor system. The system outputs a "structure healthy" status indicator and returns to step S11 to execute the next round of acquisition.

[0098] Finally, the system performs boundary fitting on the abnormal mesh set to obtain a damage location coordinate sequence. Specifically, the system extracts the geometric center coordinates of all meshes in the abnormal mesh set and uses the least squares method or spline curve fitting method to extract edges and fit trajectories to these coordinate points. By identifying the contour boundaries of the abnormal region, a series of coordinate points representing the geometry and extension direction of the damage are generated, i.e., the damage location coordinate sequence. This sequence not only pinpoints the specific spatial location of the damage but also quantifies the length and expansion trend of the damage, providing intuitive data support for subsequent structural safety assessments.

[0099] It should be noted that the size setting of the grid cells in this embodiment is determined based on a dynamic balance between spatial resolution and computational load. Those skilled in the art can set the grid side length in the range of centimeters to decimeters, depending on the bridge span and its importance. If suspected large-area damage is detected, the system will automatically activate the "refine grid" strategy to perform secondary densification of the abnormal grid set and its neighborhood, thereby improving the geometric accuracy of the positioning results.

[0100] It is worth noting that during the boundary fitting process, the system incorporates connected component analysis logic to eliminate isolated noise points caused by poor sensor contact. In one implementation, if no other abnormal grids exist around a given abnormal grid, it is identified as a false alarm and removed from the set. Boundary fitting is only performed on abnormal grids that form physically continuous clusters. This discrimination method aims to ensure that the damage location coordinate sequence accurately reflects the structural deterioration within the concrete, rather than being caused by random errors during signal acquisition.

[0101] For example, during health monitoring of the concrete box girder of an overpass, step S15 receives a candidate damage signal triggered by the passage of heavy-load vehicles. The system maps the signal to a preset mesh model and finds that the resistance of six consecutive mesh cells near the support exceeds a set resistance variation threshold. Through a boundary fitting algorithm, the system identifies a diagonally extending trajectory, and the resulting damage location coordinate sequence precisely points to the location of the shear crack on the side of the support. This result is then output to a visualization platform to guide maintenance personnel in performing targeted adhesive injection repair or reinforcement.

[0102] In step S16, the cumulative damage index is calculated based on the damage location coordinate sequence. The cumulative damage index exceeding a preset damage risk threshold is then fitted with a stress distribution model to construct a continuous damage distribution model, including:

[0103] Force calculations are performed on the damage location coordinate sequence to obtain the cumulative damage index and stress data of the corresponding region;

[0104] If the cumulative damage index exceeds the preset damage risk threshold, the stress data is smoothed to obtain smoothed stress data.

[0105] Gaussian fitting is performed on the smoothed stress data to construct a continuous damage distribution model.

[0106] First, the system performs stress calculations on the damage location coordinate sequence to obtain the cumulative damage index and stress data of the corresponding region. Specifically, in this embodiment, a fracture mechanics model is used to calculate the local stress state of the abnormal mesh set containing the damage location coordinate sequence. The stress calculation is achieved by analyzing the stress concentration and crack tip opening displacement within the carbon fiber concrete. In one implementation, a local stress field calculation relationship is established based on the stress intensity factor and material impedance characteristics, as shown in Formula 1 below:

[0107] ;

[0108] In the formula, This indicates that the geometric centroid corresponding to the damage location coordinate sequence is taken as the origin, and the polar radius at a distance r from the origin in the polar coordinate system is r with a polar angle of θ. Local stress tensor components at the location; This represents the stress intensity factor for a type of open crack. Indicates the preset polar angle The dimensionless angular distribution function. In this embodiment, the extracted resistance variation threshold deviation parameter is converted into a type-one open crack stress intensity factor. By combining the polar coordinate position parameters of the current mesh element, the stress data of each discrete abnormal mesh element is calculated.

[0109] It should be noted that the cumulative damage index refers to a dimensionless parameter characterizing the degree of degradation of the internal microstructure of carbon fiber concrete and the proportion of remaining service life consumed. This embodiment establishes a quantitative correlation between the cumulative damage index and stress data, strain data, and the rate of change of local resistance based on a nonlinear continuous damage mechanics model. In this embodiment, the quantitative relationship of the cumulative damage index is shown in Formula 2 below:

[0110] ;

[0111] In the formula, D represents the cumulative damage index, which is defined as a continuous real number between 0 and 1, where 0 represents a perfect state without damage and 1 represents a complete structural fracture failure, and is used to represent the percentage of local fatigue life consumed by the propagation of microcracks in the current monitoring area. This indicates the real-time change in resistance of the current grid cell; This represents the initial interference-free resistance value of the grid cell; This represents the local maximum principal stress value calculated using Formula 1; This represents the ultimate compressive strength constant of carbon fiber concrete, as determined by materials mechanics tests. This is a constant corresponding to the calibrated carbon fiber concrete material, used to represent the nonlinear sensitivity of the rate of change of resistance to damage; This is a constant corresponding to the calibration of carbon fiber reinforced concrete, used to represent the driving index of stress level on damage; for concrete with a carbon fiber volume content of approximately 0.5%, the calibration result is as follows: The value is 1.2 ± 0.2. The value is taken as 2.1 ± 0.3. In this embodiment, the cumulative damage index of the corresponding region is obtained by collecting the resistance deviation value and instantaneous mechanical state of the abnormal grid set on the time axis and substituting them into Formula 2.

[0112] If the cumulative damage index exceeds a preset damage risk threshold, the stress data is smoothed to obtain smoothed stress data. In this embodiment, the preset damage risk threshold is a physical critical value determined based on the ultimate strength and safety reserve factor of carbon fiber concrete. Specifically, it is determined through axial compression failure tests on standard carbon fiber concrete specimens. At least five groups of standard carbon fiber concrete specimens are subjected to axial compression failure tests, and the cumulative damage index at the time of failure of each specimen is recorded. The average value of these failure indices is taken and multiplied by a safety factor of 0.6 as the preset damage risk threshold. When the calculated index exceeds this threshold, the structure is determined to have a failure risk. The system uses a moving average filtering algorithm or a median filtering algorithm to smooth the extracted stress data, aiming to eliminate numerical spikes caused by fluctuations in the local contact resistance of the sensor, thereby extracting the smoothed stress data that reflects the essential trend of the structure's stress. If the cumulative damage index does not exceed the preset damage risk threshold, the system determines that the structure is in a fatigue safety period, does not perform Gaussian fitting, directly updates the historical damage database, and returns to step S11.

[0113] Finally, the system performs Gaussian fitting on the smoothed stress data to construct a continuous damage distribution model. It should be noted that, due to the quasi-brittle mechanical properties of carbon fiber concrete, the initiation and aggregation of microcracks within it macroscopically manifests as stress redistribution. According to the theory of nonlocal damage, the stress concentration at the crack tip and the energy release rate induced by damage exhibit isotropic or anisotropic characteristics in three-dimensional space, with the peak at the damage core and an exponential decay outwards. Since carbon fibers are randomly dispersed within the matrix, the physical evolution of stress wave propagation and strain gradient localization at the crack front edge statistically closely matches a normal distribution. Therefore, this embodiment uses a Gaussian probability distribution function to continuously fit the discrete spatial stress field. In this embodiment, since the bridge beam surface or monitoring area is a two-dimensional planar structure, the objective function of the continuous damage distribution model uses a planar bivariate Gaussian density function as the basic fitting unit, and its mathematical expression is shown in Formula 3:

[0114] ;

[0115] In the formula, S(x, y) represents the predicted value of continuous stress at spatial coordinates (x, y); A represents the amplitude of the Gaussian distribution, used to characterize the severity of stress concentration in the damage core region, and the unit is megapascals; and These represent the mean values ​​of the Gaussian distribution along the x-axis and y-axis, respectively, and are used to indicate the coordinates of the damage geometric center inside the carbon fiber concrete member. and These represent the variances along the x-axis and y-axis, respectively, and are used to represent the spatial influence range of the damage stress concentration zone and the gradient diffusion boundary. This embodiment utilizes a least-squares iterative algorithm, using the spatial coordinates of each discrete anomaly grid point... Using the corresponding smoothed stress data as fitting samples, the sum of squared residuals is minimized to solve for the parameter combination in the objective function, namely the amplitude A, mean parameter, and variance parameter. Finally, a continuous damage distribution model reflecting the continuous evolution state of the stress tensor of microcracks inside the structure is constructed.

[0116] It should be noted that the weighting coefficients for calculating the cumulative damage index in this embodiment are determined based on statistical data from material fatigue tests. Those skilled in the art can dynamically adjust the damage accumulation rate based on the bridge's service life and the environmental corrosion level to ensure the accuracy of the assessment results. This method aims to solve the technical challenge of accurately quantifying the damage level of carbon fiber concrete under long-term loads.

[0117] It is worth noting that, during the Gaussian fitting process, spatial correlation constraints are introduced to enhance the model's interpretability. In one implementation, if the fitted model exhibits abnormally high stress peaks in non-damaged regions, the system automatically adjusts the kernel bandwidth of the Gaussian function, reducing the fitting residuals through iterative optimization until the model's confidence level across the entire domain reaches a preset standard. This adaptive fitting method ensures that the continuously distributed damage model can realistically and continuously reflect the physical performance degradation state within the concrete beam.

[0118] For example, in the precise quantitative analysis of damage to a simply supported beam of a high-speed railway, step S16 receives the damage coordinates located at the bottom of the mid-span. The system calculation shows that the cumulative damage index at this location reaches 0.75, exceeding the preset safety threshold of 0.65. By smoothing the stress characteristics of this area and performing Gaussian fitting, the system constructs a funnel-shaped continuous damage distribution model, clearly showing that the stress concentration point is located at the tip of the main crack and decreases towards both flanges. The final generated model provides a quantitative basis for assessing the degradation of the bridge's load-bearing capacity, ensuring the safety of railway operations.

[0119] In step S17, the real-time resistance change path is extracted from the damage candidate signal and mapped to the damage continuous distribution model for feature matching to obtain the crack identification result, including:

[0120] The root mean square error analysis was performed on the continuous damage distribution model to obtain the self-sensing performance verification coefficients.

[0121] If the self-sensing performance verification coefficient is lower than the preset deviation threshold, then the real-time resistance change path is extracted from the damage candidate signal, and the real-time resistance change path is mapped to the damage continuous distribution model to obtain the mapping result.

[0122] Based on the mapping results, dynamic change capture and positioning are performed to obtain crack identification results.

[0123] First, the system performs root mean square error analysis on the continuous damage distribution model to obtain the self-sensing performance verification coefficient. This self-sensing performance verification coefficient is a scalar value calculated by taking the square root of the sum of squares of the predicted residuals of the entire grid cells. It is used to quantify the sensing accuracy and reliability of carbon fiber reinforced concrete material as a sensor under the current damage state.

[0124] It should be noted that in the root mean square error analysis, the actual observed values ​​used for comparison are strain values ​​collected by conventional strain gauges or force sensors that have been calibrated concurrently. For example, the strain value measured by an external sensor placed within the grid cell is multiplied by the known elastic modulus of the material to obtain the true stress value at that point; then, the residual between the stress value predicted by the damage continuous distribution model and the above true stress value is calculated, and the square root of the sum of the squares of the residuals across the entire grid cell is taken to obtain the self-sensing performance verification coefficient.

[0125] Subsequently, if the self-perceived performance verification coefficient is lower than a preset deviation threshold, the real-time resistance change path is extracted from the damage candidate signal, and the real-time resistance change path is mapped to the damage continuous distribution model to obtain the mapping result. The mapping process uses the grid cell set divided in S15 as a common spatial coordinate reference, forcibly aligning the spatial index of the real-time extracted resistance change path with the stress gradient coordinates in the damage continuous distribution model to eliminate spatial displacement deviations between different algorithm modules. In this embodiment, the preset deviation threshold is determined according to the accuracy level specifications for bridge structural health monitoring systems in the transportation industry. It is used to determine whether the model can truly represent the internal physical evolution logic of the material. By calculating the root mean square error between the fitted model and the measured values, the error is strictly controlled within 5% to ensure that the final output crack identification result has effective data support. When the verification coefficient is lower than this threshold (i.e., the model accuracy meets the standard), the system extracts the dynamic trajectory of the resistance value evolving over time from the damage candidate signal, defining it as the real-time resistance change path.

[0126] Next, the system uses a spatial transformation matrix to project the resistance change trajectory on the time axis into the geometric space of the continuous damage distribution model. The mapping result is obtained by calculating the overlap between path points and stress concentration areas in the model. This overlap is achieved by extracting regions with stress values ​​greater than a preset limit strength from the continuous damage distribution model as high-stress masks. The number of path points falling within this high-stress mask range in the real-time resistance change path is then counted. Dividing this number by the total number of path points yields the spatial overlap value. If this value exceeds a preset 90% threshold, the mapping result is considered a successful match. The preset limit strength used to generate the high-stress mask is 1.2 times the standard value of the tensile strength of carbon fiber concrete, which is obtained through a splitting test. If the self-perceived performance verification coefficient is higher than a preset deviation threshold, the system determines that the currently constructed model cannot accurately reflect the true stress field of the structure. At this point, the system triggers the "perception closed-loop failure" command, automatically retrieves historical data on ambient temperature and current fluctuations from the previous monitoring period, and re-executes the interference correction parameter calibration in step S11 until the verification coefficient returns to within the safe threshold, thereby ensuring the rigor of the identification results.

[0127] It should be noted that the spatial transformation matrix is ​​constructed by matching the two-dimensional plane coordinates of the sixteen electrodes with the center coordinates of each grid in the grid cell set, and using bilinear interpolation to calculate the electrode response weight corresponding to each grid, thereby constructing a transformation matrix from the electrode measurement space to the grid space.

[0128] Finally, the system dynamically captures and locates the crack based on the mapping results, obtaining the crack identification result. Specifically, the system calculates the second derivative of the real-time resistance change path on the time axis, identifying the inflection point where the slope change is most drastic. If this inflection point coincides with the maximum stress point in the continuous damage distribution model by more than 90% in spatial location, then this location is identified as the real-time crack tip, thus achieving dynamic capture of the crack propagation trajectory. The system ultimately outputs structured information including the crack initiation location, propagation length, opening width, and development stage classification, which is the crack identification result. This result achieves a precise transformation from abstract resistance fluctuations to concrete physical crack morphology.

[0129] It should be noted that the method for extracting the real-time resistance change path in this embodiment selects its time window based on the duration of the pulse width of the damage candidate signal. Those skilled in the art can set the sampling time window within the millisecond range based on the acoustic emission characteristic frequency at the moment of concrete cracking to capture the key characteristics of instantaneous crack initiation. This setting method aims to solve the technical problem of insufficient sensitivity in identifying micro-cracks under dynamic interference.

[0130] It is worth noting that, in order to eliminate the influence of ambient temperature and humidity fluctuations on the resistance path during the mapping process, the system introduces a dynamic reference compensation mechanism. In one implementation, if a slow synchronous drift in the global background resistance is detected, the system automatically subtracts the temperature drift component from the real-time resistance change path, thereby ensuring that the mapping result purely reflects the local resistance jump caused by structural damage. This data cleaning method significantly improves the rigor and technical reproducibility of the crack identification results.

[0131] Specifically, the dynamic baseline compensation mechanism calculates the median of the resistance values ​​of all grid cells at the current moment as the background baseline, updates the baseline with a 10-second sliding window, and subtracts the background baseline value point by point in the real-time resistance change path to obtain the path after temperature drift correction.

[0132] For example, during crack source analysis of a prestressed concrete continuous beam under a large temperature difference environment, step S17 received an abnormal resistance change sequence. The system calculation found that the root mean square error of the damage continuous distribution model was 0.03, lower than the preset deviation threshold of 0.05, indicating that the model was highly reliable. By mapping the real-time resistance change path to the stress distribution model, the system observed that the resistance decrease path completely coincided with the direction of maximum shear stress in the model. Through dynamic capture, the system accurately identified this as an oblique shear crack located in the beam web and determined its length to be 12.5 centimeters. The final crack identification result provided a core basis for the safety classification of the bridge, ensuring the closed-loop integrity of the monitoring scheme.

[0133] In summary, this invention achieves real-time and accurate detection of microcracks in carbon fiber concrete structures, efficient delineation of damage area boundaries, and closed-loop verification of self-sensing performance through the logical association and systematic integration of core technologies such as multi-source data interference elimination, multi-level signal decomposition and purification, stress resistance path tracing, and spatial grid damage localization.

[0134] Reference Figure 2 The second embodiment of the present invention provides a self-sensing performance testing system for carbon fiber concrete, comprising:

[0135] The signal correction module is used to acquire the original resistance signal, synchronize temperature data and current data, and perform interference correction to obtain the corrected resistance signal;

[0136] The signal purification module is used to perform wavelet transform multi-level decomposition on the correction resistor signal to obtain low-frequency stress components, and to perform inverse transform on the low-frequency stress components to obtain purified stress resistance components.

[0137] The feature association module is used to perform frequency domain feature analysis based on the purified stress resistance components to obtain the local resistance change rate, and to construct the correspondence between stress and resistance based on the local resistance change rate to obtain the stress resistance feature distribution.

[0138] An anomaly capture module is used to separate the damage sub-path from the stress resistance feature distribution if the resistance value shows a continuous increase, and to perform multi-scale decomposition on the damage sub-path to obtain damage candidate signals.

[0139] The damage localization module is used to perform area localization based on the damage candidate signal to obtain a localization result, and to perform boundary marking based on the localization result to obtain a damage localization coordinate sequence.

[0140] The continuous modeling module is used to calculate the cumulative damage index based on the damage location coordinate sequence, and to fit the stress distribution of the cumulative damage index that exceeds the preset damage risk threshold to construct a continuous damage distribution model.

[0141] The crack identification module is used to extract the real-time resistance change path from the damage candidate signal and map it to the damage continuous distribution model for feature matching to obtain the crack identification result.

[0142] It should be noted that the self-sensing performance testing system for carbon fiber concrete provided in this embodiment of the invention is used to perform all the process steps of the self-sensing performance testing method for carbon fiber concrete described in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0143] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0144] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for testing the self-sensing properties of carbon fiber concrete, characterized in that, include: The original resistance signal, synchronous temperature data, and current data are acquired, and interference correction is performed to obtain the corrected resistance signal. The correction resistor signal is decomposed into low-frequency stress components by wavelet transform at multiple levels, and the low-frequency stress components are then inversely transformed to obtain purified stress resistance components. The local resistance change rate is obtained by frequency domain feature analysis based on the purified stress resistance components, and the correspondence between stress and resistance is constructed based on the local resistance change rate to obtain the stress resistance characteristic distribution. If the stress resistance characteristic distribution shows that the resistance value continues to rise, then the damage sub-path is separated from the stress resistance characteristic distribution, and the damage sub-path is decomposed into multiple scales to obtain the damage candidate signal. Based on the damage candidate signal, the region is located to obtain the location result, and the boundary is marked based on the location result to obtain the damage location coordinate sequence; The cumulative damage index is calculated based on the damage location coordinate sequence. The cumulative damage index that exceeds the preset damage risk threshold is fitted with stress distribution to construct a continuous damage distribution model. The real-time resistance change path is extracted from the damage candidate signal and mapped to the damage continuous distribution model for feature matching to obtain the crack identification result.

2. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, The process of acquiring the original resistance signal, synchronous temperature data, and current data, and performing interference correction to obtain the corrected resistance signal includes: Time-series analysis is performed on the synchronized temperature data and the current data, and the fluctuation frequency of the synchronized temperature data and the current data is extracted by Fourier transform; An interference model is constructed based on the fluctuation frequency, and the interference model is used to solve the original resistance signal to obtain the interference compensation factor. The original resistance signal is linearly adjusted using the interference compensation factor to obtain the corrected resistance signal.

3. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, The step of performing wavelet transform multi-level decomposition on the corrected resistance signal to obtain low-frequency stress components, and then performing inverse transform on the low-frequency stress components to obtain purified stress-resistance components, includes: The correction resistor signal is subjected to five-level wavelet decomposition to obtain the high-frequency coefficient matrix and the low-frequency coefficient matrix; The high-frequency coefficient matrix is ​​set to zero according to a preset frequency threshold to obtain the reconstructed coefficient matrix; The low-frequency coefficient matrix is ​​superimposed with the reconstructed coefficient matrix to obtain the low-frequency stress components; The low-frequency stress component is subjected to inverse discrete wavelet transform to obtain the purified stress resistance component.

4. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, The step of performing frequency domain feature analysis based on the purified stress-resistivity components to obtain the local resistance change rate, and constructing the correspondence between stress and resistance based on the local resistance change rate to obtain the stress-resistivity characteristic distribution, includes: A fast Fourier transform is performed on the purified stress resistance component to obtain the characteristic component frequency. The amplitude ratio is calculated based on the characteristic component frequency to obtain the local resistance change rate. A short-time Fourier transform is performed on the local resistance change rate to generate a time-frequency spectrum, and a time-frequency mapping matrix is ​​constructed based on the time-frequency spectrum; Principal component analysis is performed on the time-frequency mapping matrix to obtain characteristic principal components, and linear combination projection is performed on the characteristic principal components to obtain the stress resistance characteristic distribution.

5. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, If the stress-resistance characteristic distribution shows a continuous increase in resistance value, then a damage sub-path is separated from the stress-resistance characteristic distribution, and the damage sub-path is decomposed into multiple scales to obtain damage candidate signals, including: The real-time slope of the stress resistance characteristic distribution is obtained by taking the first derivative. The damage sub-path is obtained by tracking the real-time slope change path that exceeds the preset slope threshold using a dynamic programming algorithm; Five-level wavelet decomposition is performed on the damaged sub-path to obtain a set of high-frequency components, and the instantaneous energy distribution is obtained by calculating the energy concentration based on the set of high-frequency components. Based on the instantaneous energy distribution, peak frequency is determined to obtain damage candidate signals.

6. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, The step of obtaining a location result by performing region localization based on the candidate damage signal, and then performing boundary labeling based on the location result to obtain a damage location coordinate sequence, includes: Obtain surface data of the bridge beam and divide the bridge beam surface into a set of grid cells based on the surface data; The damage candidate signal is mapped to the set of grid cells to obtain cell resistance distribution data containing cell resistance values; The unit resistance distribution data is scanned and compared with a preset resistance variation threshold to obtain an abnormal grid set in which the unit resistance value exceeds the resistance variation threshold. Boundary fitting is performed on the abnormal mesh set to obtain the damage location coordinate sequence.

7. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, The step of calculating the cumulative damage index based on the damage location coordinate sequence, and fitting the stress distribution of the cumulative damage index exceeding a preset damage risk threshold to construct a continuous damage distribution model includes: Force calculations are performed on the damage location coordinate sequence to obtain the cumulative damage index and stress data of the corresponding region; If the cumulative damage index exceeds the preset damage risk threshold, the stress data is smoothed to obtain smoothed stress data. Gaussian fitting is performed on the smoothed stress data to construct a continuous damage distribution model.

8. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, The step of extracting the real-time resistance change path from the candidate damage signal and mapping it to the continuous damage distribution model for feature matching to obtain the crack identification result includes: The root mean square error analysis was performed on the continuous damage distribution model to obtain the self-sensing performance verification coefficients. If the self-sensing performance verification coefficient is lower than the preset deviation threshold, then the real-time resistance change path is extracted from the damage candidate signal, and the real-time resistance change path is mapped to the damage continuous distribution model to obtain the mapping result. Based on the mapping results, dynamic change capture and positioning are performed to obtain crack identification results.

9. The self-sensing performance testing method for carbon fiber concrete according to claim 1, characterized in that, The original resistance signal, synchronous temperature data, and current data include resistance offset characteristics, signal noise characteristics, or signal distortion characteristics.

10. A self-sensing performance testing system for carbon fiber concrete, characterized in that, include: The signal correction module is used to acquire the original resistance signal, synchronize temperature data and current data, and perform interference correction to obtain the corrected resistance signal; The signal purification module is used to perform wavelet transform multi-level decomposition on the correction resistor signal to obtain low-frequency stress components, and to perform inverse transform on the low-frequency stress components to obtain purified stress resistance components. The feature association module is used to perform frequency domain feature analysis based on the purified stress resistance components to obtain the local resistance change rate, and to construct the correspondence between stress and resistance based on the local resistance change rate to obtain the stress resistance feature distribution. An anomaly capture module is used to separate the damage sub-path from the stress resistance feature distribution if the resistance value shows a continuous increase, and to perform multi-scale decomposition on the damage sub-path to obtain damage candidate signals. The damage localization module is used to perform area localization based on the damage candidate signal to obtain a localization result, and to perform boundary marking based on the localization result to obtain a damage localization coordinate sequence. The continuous modeling module is used to calculate the cumulative damage index based on the damage location coordinate sequence, and to fit the stress distribution of the cumulative damage index that exceeds the preset damage risk threshold to construct a continuous damage distribution model. The crack identification module is used to extract the real-time resistance change path from the damage candidate signal and map it to the damage continuous distribution model for feature matching to obtain the crack identification result.

Images