Bridge crack propagation evaluation method and system based on acoustic emission signals

Abstract

The invention belongs to the technical field of bridge body monitoring, and discloses a bridge crack propagation evaluation method and system based on acoustic emission signals. The method comprises the following steps: arranging an acoustic emission sensor array at a preset part of a bridge to collect original acoustic emission signals in a preset time period; processing by adopting a band-pass filter to obtain a filtered acoustic emission signal, and separating the filtered acoustic emission signal by using an empirical mode decomposition technology to obtain an effective acoustic emission signal; converting the effective acoustic emission signal by using time-frequency transformation to obtain a signal time-frequency diagram; constructing a crack propagation evaluation model, and obtaining a crack propagation rate and a serious state grade; obtaining a crack space positioning result through crack space positioning; respectively carrying out weighted fusion on the crack propagation rate and the serious state grade of the J sensors, combining with a spatial positioning result and a bridge model, and presenting in a three-dimensional form; according to the invention, more accurate crack positions can be obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge beam monitoring, and more specifically, to a method and system for evaluating the crack propagation of a bridge based on acoustic emission signals. Background Art

[0002] Traditional bridge detection methods, such as visual inspection, ultrasonic testing, strain gauge measurement, etc., have certain limitations in crack detection. Visual inspection relies on the experience and subjective judgment of inspectors and is difficult to detect tiny cracks and cracks hidden inside the structure; ultrasonic testing requires manual point-by-point scanning, with low detection efficiency and limited detection effect on complex structures; strain gauge measurement is mainly used to detect the stress changes of the structure and is not direct enough for the early detection and propagation evaluation of cracks.

[0003] The patent application with the publication number CN117607266A discloses a method and system for monitoring the cracking of a bridge beam based on acoustic emission: Step S1: Install an acoustic emission signal acquisition device. Specifically, the acoustic emission signal acquisition device includes an acoustic emission sensor and an acquisition control unit. On the bridge beam, at least two acoustic emission monitoring lines are arranged along the length direction of the bridge beam. A plurality of acoustic emission sensors are provided on the acoustic emission monitoring lines. All acoustic emission sensors are located on the plane below the bridge beam and are connected in series with the acquisition control unit; Step S2: Locate the acoustic emission sensors. Specifically, a plane rectangular coordinate system is constructed and each acoustic emission sensor is numbered, and the coordinates of each acoustic emission sensor are measured based on the plane rectangular coordinate system; Step S3: Monitor the bridge beam. Specifically, the acoustic emission ring count, acoustic emission energy, and signal acquisition time points are collected by the acoustic emission sensors to determine whether there is a cracking risk in the bridge beam. If there is a cracking risk in the bridge beam, then enter Step S4; Step S4: Locate the acoustic emission source. Specifically, a plurality of acoustic emission sensors close to the acoustic emission source are found according to the acoustic emission ring count and acoustic emission energy. The time difference for the acoustic emission elastic wave to reach the above-mentioned plurality of acoustic emission sensors to monitor the acoustic emission source is calculated according to the signal acquisition time points, and the horizontal and vertical coordinates of the acoustic emission source are solved according to the acoustic emission source calculation formula. This technical solution realizes the full-section monitoring and accurate positioning of the cracking of the bridge beam, and has an early warning function. Staff can view the online data in real time to prevent the occurrence of structural safety accidents.

[0004] Although the above method can meet most scenarios, through research and practical application of the above method and the existing technology, it is found that the above method and the existing technology have at least the following partial defects:

[0005] The quantitative evaluation of the crack propagation rate and the division of the severe state level are not perfect, and it is impossible to further clarify the crack propagation speed in the future period of time and the specific impact degree of the crack on the safety of the beam structure; only presenting the crack monitoring results in the form of data reports or simple two-dimensional graphs lacks intuitiveness and integrity, which is not conducive to users quickly understanding and grasping the overall situation of bridge cracks.

[0006] In view of this, the present invention proposes a method and system for evaluating the crack propagation of a bridge based on acoustic emission signals to solve the above problems. Summary of the Invention

[0007] In order to overcome the above defects of the prior art, the present invention provides the following technical solutions: A method for evaluating the crack propagation of a bridge based on acoustic emission signals, comprising the following steps:

[0008] Arrange an acoustic emission sensor array at a preset part of the bridge to collect the original acoustic emission signals within a preset time period;

[0009] Use a band-pass filter to remove the low-frequency ambient noise and high-frequency sensor interference of the original acoustic emission signals, obtain the filtered acoustic emission signals, and use the empirical mode decomposition technology to separate the filtered acoustic emission signals to obtain the effective acoustic emission signals;

[0010] Use time-frequency transformation to transform the effective acoustic emission signals to obtain a signal time-frequency diagram;

[0011] Construct a crack propagation evaluation model, use the signal time-frequency diagram as the input of the crack propagation evaluation model, and obtain the crack propagation rate and the corresponding severe state level; the crack propagation evaluation model includes a feature extraction network and an improved Paris crack propagation model;

[0012] Weightedly fuse the crack propagation rates and severe state levels of J sensors respectively to obtain the weighted crack propagation rate and severe state level, and combine the weighted crack propagation rate and severe state level with the bridge model and present it in a three-dimensional form.

[0013] Furthermore, the method further includes: performing crack spatial positioning by monitoring the original acoustic emission signals through a multi-channel sensor to obtain a crack spatial positioning result; combining the crack spatial positioning result with the bridge model and presenting it in a three-dimensional form;

[0014] The method for obtaining the crack spatial positioning result includes:

[0015] Step A: Calculate the time difference Δt between the original acoustic emission signals reaching any two sensors u and v uv , and use the time difference positioning formula to locate the crack source:

[0016] Step B: Based on the time difference Δtuv Construct a probability distribution model of crack positions to obtain the crack spatial positioning result;

[0017] The method for obtaining the crack spatial positioning result includes:

[0018] Step B1: Obtain the crack source located in step A to get the crack position (x, y);

[0019] Step B2: Obtain the corresponding value range [x min , y min ×[x max , y max , where x min is the minimum value of the lateral position of the crack on the bridge; y min is the minimum value of the longitudinal position of the crack on the bridge; x max is the maximum value of the lateral position of the crack on the bridge; y max is the maximum value of the longitudinal position of the crack on the bridge; obtain the corresponding prior probability distribution

[0020] Step B3: Calculate all values of the crack position (x, y) according to the time difference positioning formula; for the crack position (x, y), the time difference Δt uv obeys a normal distribution, calculate the likelihood function of the corresponding time difference Δt uv where σ is the standard deviation of the measurement error of the time difference Δt uv ; e is a mathematical constant; uv is the actually observed time difference; is the theoretical time difference calculated according to the crack position (x, y); based on L groups of time differences Δt (u, v = 1, 2, …, L, u ≠ v), calculate to obtain the total likelihood function P(Δt|x, y) = Π uv P u≠v Δt ( |x, y); uv

[0021] Step B4: Integrate all crack positions (x, y) to obtain the crack position evidence P(Δt) = ∫∫P(Δt|x, y)P(x, y)dxdy; where P(Δt) is the crack position evidence, that is, the integral value obtained by integrating all crack positions (x, y);

[0022] Step B5: Calculate the posterior probability distribution corresponding to the crack position (x, y) according to Bayes' formula, obtain the probability distribution model of crack positions, and take the position corresponding to the maximum value of the posterior probability distribution in the probability distribution model of crack positions as the crack spatial positioning result.​

[0023] Further, the method for obtaining the crack propagation rate and the corresponding severity level includes:

[0024] Step 1.1: Build a feature extraction network based on a convolutional neural network. The feature extraction network includes an input layer, n 1 convolutional layers, n 2 pooling layers, a self-attention module, a fully connected layer, and an output layer. Use the signal time-frequency diagram as the input of the feature extraction network to obtain the output feature vector. Among them, the convolutional layer and the pooling layer are used to extract the local time-frequency features of the signal time-frequency diagram to obtain the feature map R. In the self-attention module, perform a linear transformation on the feature map R obtained after the convolutional layer and the pooling layer to obtain the query matrix Q, the key matrix K, and the value matrix V, and calculate the attention score Score based on the query matrix Q, the key matrix K, and the value matrix V. Multiply the attention score by the value matrix to obtain the attention output.

[0025] Step 1.2: Obtain the stress intensity factor range based on the feature vector.

[0026] Step 1.3: Calculate the crack propagation rate based on the improved Paris crack propagation model, compare the crack propagation rate with the crack propagation rate interval corresponding to the preset expansion severity level, and obtain the corresponding severity level.

[0027] Further, the method for obtaining the query matrix Q, the key matrix K, and the value matrix V includes:

[0028] Q = R·W Q ;

[0029] K = R·W K ;

[0030] V = R·W V ;

[0031] where W Q , W K and W V are the weight matrices corresponding to the query matrix Q, the key matrix K, and the value matrix V, respectively;

[0032] The method for obtaining the attention score Score includes:

[0033]

[0034] where μ k is the dimension of the key matrix K; T is the transpose of the vector; softmax is the normalization mapping function;

[0035] The method for obtaining the attention output includes:

[0036] A = Score × V;

[0037] Wherein, A is the attention output.

[0038] Furthermore, the method for obtaining the stress intensity factor range includes:

[0039] Step 2.1: Obtain historical feature vectors and corresponding stress intensity factors;

[0040] Step 2.2: Establish a non - linear model of historical feature vectors and corresponding stress intensity factors;

[0041] Step 2.3: Fit the non - linear model based on historical feature vectors and corresponding historical stress intensity factor ranges, and optimize with the goal of minimizing the sum of squared residuals to obtain corresponding non - linear model parameter values. Substitute the corresponding non - linear model parameter values into the non - linear model to obtain the fitted non - linear model;

[0042] Step 2.4: Use the feature vector as the input of the fitted non - linear model to obtain the corresponding minimum stress intensity factor and maximum stress intensity factor. Take the difference between the maximum stress intensity factor and the minimum stress intensity factor as the stress intensity factor range.

[0043] Furthermore, the method for obtaining the improved Paris crack propagation model includes:

[0044] Obtain the correction coefficient corresponding to the Paris crack propagation model according to the bridge concrete compressive strength, elastic modulus and crack width. Obtain a fatigue test data set composed of stress intensity factor range test values. Obtain the fatigue crack propagation rate corresponding to each stress intensity factor test value in the fatigue test data set. Fit the first material constant C, the second material constant m and the correction coefficient in the improved Paris crack propagation model by the least - squares method; correct the Paris crack propagation model based on the fitted first material constant C, the second material constant m and the correction coefficient to obtain the improved Paris crack propagation model.

[0045] Furthermore, the method for obtaining the effective acoustic emission signal includes:

[0046] Step 3.1: Determine local extreme points by comparing the magnitude relationship between each data point in the filtered acoustic emission signal and the two adjacent data points. If the data point is greater than the two adjacent data points, the data point is a local maximum point; conversely, if the data point is less than the two adjacent data points, the data point is a local minimum point;

[0047] Step 3.2: Connect all local maximum points to form an upper envelope bl max(τ), where τ is time; connect all local minimum points to form the lower envelope bl min (τ);

[0048] Step 3.3: Calculate the average value of the upper envelope and the lower envelope Calculate the difference between the filtered acoustic emission signal g(τ) and the average value m(τ) to obtain the difference signal h(τ) = g(τ) - m(τ); where g(τ) is the filtered acoustic emission signal;

[0049] Step 3.4: Check whether the difference signal h(τ) satisfies the IMF condition; if not, use the difference signal h(τ) as the new filtered first acoustic emission signal, and repeat Steps 3.1 - 3.3 until the IMF condition is satisfied to obtain the first IMF; if satisfied, directly obtain the first IMF, and mark the first IMF as c 1 (τ);

[0050] Step 3.5: Subtract the first IMF c 1 (τ) from the filtered acoustic emission signal g(τ) to obtain the first residual signal r 1 (τ) = g(τ) - c 1 (τ);

[0051] Step 3.6: Use the first residual signal r 1 (τ) as the new filtered second acoustic emission signal, repeat Step 3.4 to obtain the second IMF, and mark the second IMF as c 2 (τ), repeat Step 3.5 to calculate and obtain the second residual signal r 2 (τ) = r 1 (τ) - c 2 (τ);

[0052] Step 3.7: Repeat Step 3.6 until the residual component becomes a monotonic function or the number of extreme points does not exceed two, then end the decomposition process; represent the filtered acoustic emission signal g(τ) as where c i (τ) is the IMF; is the residual component; n g is the number of IMFs into which the filtered acoustic emission signal g(τ) is divided;

[0053] Step 3.8: Perform a fast Fourier transform on each IMF to obtain the corresponding frequency spectrum range, extract the IMF as the effective component according to the preset frequency range of the effective signal, and reconstruct the effective component to obtain the useful signal g eff (τ) = ∑ i∈S c i(τ), where S is the index set of IMFs containing the active ingredient; the useful signal g eff (τ) = ∑ i∈S c i (τ) is used as the effective acoustic emission signal.

[0054] Furthermore, the IMF conditions include: within the entire data segment of the acoustic emission signal, the number of extreme points and the number of zero-crossing points must be equal or differ by at most one; at any data point, the average value of the upper envelope line and the lower envelope line defined by the local maximum points and local minimum points is zero.

[0055] Furthermore, the method for obtaining the time-frequency diagram of the signal includes:

[0056] Step 4.1: Divide the effective acoustic emission signal g eff (τ) into d small frames, each frame having a length of n fft , and the frame shift is hop length .

[0057] Step 4.2: Apply a preset window function w(τ) to each frame of the effective acoustic emission signal;

[0058] Step 4.3: Apply the fast Fourier transform to each frame of the effective acoustic emission signal after applying the preset window function w(τ), convert the time-domain signal of each frame of the effective acoustic emission signal after applying the preset window function w(τ) into a frequency-domain signal, and combine the frequency-domain signals to obtain a spectrum matrix;

[0059] The method for obtaining the spectrum matrix includes:

[0060] Construct a spectrum matrix framework with the number of signal frames corresponding to the rows and the number of frequency points of the effective acoustic emission signal corresponding to the columns;

[0061] Fill the fast Fourier transform results of each frame of the effective acoustic emission signal into the spectrum matrix framework to obtain the spectrum matrix;

[0062] Step 4.4: Based on the amplitude values of the elements in the spectrum matrix representing the intensity of the signal time-frequency diagram, that is, the pixel values corresponding to the signal time-frequency diagram, map the time in the spectrum matrix to the horizontal axis of the signal time-frequency diagram, map the frequency in the spectrum matrix to the vertical axis of the signal time-frequency diagram, and map each element value in the spectrum matrix to the corresponding position on the signal time-frequency diagram according to the intensity of the signal time-frequency diagram to obtain the signal time-frequency diagram.

[0063] Furthermore, the method for obtaining the weighted crack growth rate and the severity level includes:

[0064] Weight the crack growth rates of J sensors to obtain the weighted crack growth rate where α k is the crack propagation rate of the k-th sensor; w 1,k is the weight corresponding to α k The severity level of J sensors is weighted to obtain the weighted severity level where β k is the severity level of the k-th sensor; w 2,k is the weight corresponding to α k corresponding weight.

[0065] A bridge crack propagation evaluation system based on acoustic emission signals, used to implement the above-mentioned bridge crack propagation evaluation method based on acoustic emission signals, includes:

[0066] Data acquisition module: An acoustic emission sensor array is arranged at a preset part of the bridge to collect the original acoustic emission signals within a preset time period;

[0067] Pretreatment module: A band-pass filter is used to remove the low-frequency environmental noise and high-frequency sensor interference of the original acoustic emission signals to obtain the filtered acoustic emission signals, and the empirical mode decomposition technology is used to separate the filtered acoustic emission signals to obtain the effective acoustic emission signals;

[0068] Feature extraction module: The effective acoustic emission signals are transformed by time-frequency transformation to obtain the signal time-frequency diagram;

[0069] Expansion evaluation module: A crack propagation evaluation model is constructed, and the signal time-frequency diagram is used as the input of the crack propagation evaluation model to obtain the crack propagation rate and the corresponding severity level; the crack propagation evaluation model includes a feature extraction network and an improved Paris crack propagation model;

[0070] Crack location module: The original acoustic emission signals are monitored through multi-channel sensors for crack spatial location to obtain the crack spatial location result;

[0071] Visualization module: The crack propagation rate and severity level of J sensors are respectively weighted and fused to obtain the weighted crack propagation rate and severity level, and the weighted crack propagation rate, weighted severity level and spatial location result are combined with the bridge model and presented in a three-dimensional form.

[0072] The technical effects and advantages of the bridge crack propagation evaluation method and system based on acoustic emission signals of the present invention:

[0073] The present invention comprehensively considers the factors related to crack propagation. By deeply analyzing the acoustic emission signals and combining with physical models, it comprehensively evaluates information such as the crack propagation rate and the severity level, so as to more accurately grasp the development dynamics of the cracks and the impact on the bridge structure. It also obtains accurate spatial positioning results based on the time difference positioning formula combined with the crack position probability distribution model, and improves the positioning accuracy and robustness through the weighted fusion mechanism of multi-channel sensors to obtain a more accurate crack position. At the same time, the probability distribution positioning method can consider the uncertainty in the positioning process and give the probability distribution of the crack appearing at different positions, further improving the reliability of the positioning. It also combines the crack propagation degree, the severity level and the spatial positioning results with the bridge model to visually display the crack distribution and trend in three dimensions, making the evaluation results more intuitive and easy to understand, facilitating bridge management and maintenance personnel to quickly obtain key information and make decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is a schematic flow chart of a method for evaluating bridge crack propagation based on acoustic emission signals according to the present invention;

[0075] Figure 2 It is a schematic flow chart of a method for obtaining the crack propagation rate and the severity level according to the present invention;

[0076] Figure 3 It is a schematic flow chart of a method for obtaining the stress intensity factor range according to the present invention;

[0077] Figure 4 It is a block diagram of a system for evaluating bridge crack propagation based on acoustic emission signals according to the present invention;

[0078] Figure 5 It is a schematic diagram of the interface of a system for evaluating bridge crack propagation based on acoustic emission signals according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0079] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0080] Embodiment 1

[0081] Please refer to Figure 1 As shown, a method for evaluating bridge crack propagation based on acoustic emission signals in this embodiment includes the following steps:

[0082] Arrange an acoustic emission sensor array at the preset part of the bridge to collect the original acoustic emission signals within the preset time period; when cracks occur in the bridge, tiny acoustic emission signals will be generated. By arranging the sensor array at the preset part, these weak signals can be captured to achieve early detection of cracks, facilitating further analysis of the crack propagation rate, severity level, and spatial positioning results of the bridge in the later stage, helping to timely discover potential safety hazards and effectively avoid structural damage caused by further crack expansion.

[0083] Use a band-pass filter to remove the low-frequency environmental noise and high-frequency sensor interference of the original acoustic emission signal, obtain the filtered acoustic emission signal, and use the empirical mode decomposition technique to separate the filtered acoustic emission signal to obtain the effective acoustic emission signal.

[0084] When using a band-pass filter to process the original signal, the optimal passband frequency range of the band-pass filter can be selected according to the frequency characteristics of the original acoustic emission signal, and the noise and interference components outside the passband frequency range can be removed. It can quickly and accurately filter the signal without introducing additional distortion or delay, which helps to improve the analysis accuracy of the crack propagation rate, severity level, and spatial positioning results of the bridge in the later stage.

[0085] The methods for obtaining the effective acoustic emission signal include:

[0086] Determine the local extreme points by comparing the magnitude relationship between each data point in the filtered acoustic emission signal and the two adjacent data points corresponding to it. If the data point is greater than the two adjacent data points corresponding to it, the data point is a local maximum point; conversely, if the data point is less than the two adjacent data points corresponding to it, the data point is a local minimum point.

[0087] Connect all the local maximum points to form the upper envelope bl max (τ), where τ is time; connect all the local minimum points to form the lower envelope bl min (τ).

[0088] Calculate the average value of the upper envelope and the lower envelope Calculate the difference between the filtered acoustic emission signal g(τ) and the average value m(τ) to obtain the difference signal h(τ) = g(τ) - m(τ); where g(τ) is the filtered acoustic emission signal.

[0089] Check whether the difference signal h(τ) satisfies the IMF conditions (within the entire data segment of the acoustic emission signal, the number of extreme points and the number of zero-crossing points must be equal or differ by at most one; at any data point, the average value of the upper envelope and the lower envelope defined by the local maximum points and local minimum points is zero); if not, use the difference signal h(τ) as the new filtered first acoustic emission signal until the IMF conditions are satisfied to obtain the first IMF; if satisfied, directly obtain the first IMF and label the first IMF as c 1 (τ).

[0090] Subtract the first IMF c 1 (τ) from the filtered acoustic emission signal g(τ) to obtain the first residual signal r 1 (τ) = g(τ) - c 1 (τ).

[0091] Use the first residual signal r 1 (τ) as the new filtered second acoustic emission signal, repeat step 4 to obtain the second IMF, label the second IMF as c 2 (τ), and calculate to obtain the second residual signal r 2 (τ) = r 1 (τ) - c 2 (τ).

[0092] Until the residual component becomes a monotonic function or the number of extreme points does not exceed two, end the decomposition process; represent the filtered acoustic emission signal g(τ) as where c i (τ) is the IMF; is the residual component; n g is the number of IMFs into which the filtered acoustic emission signal g(τ) is divided.

[0093] Perform a fast Fourier transform on each IMF to obtain the corresponding frequency spectrum range, extract the IMFs as effective components according to the preset frequency range of the effective signal, and reconstruct the effective components to obtain the useful signal g eff (τ) = ∑ i∈S c i (τ), where S is the index set of the IMFs containing the effective components; use the useful signal g eff (τ) = ∑ i∈S c i (τ) as the effective acoustic emission signal.

[0094] Use time-frequency transformation to transform the effective acoustic emission signal to obtain the signal time-frequency diagram.

[0095] The methods for obtaining the signal time-frequency diagram include:

[0096] Divide the effective acoustic emission signal g eff (τ) into d small frames, each frame having a length of n fft , with a frame shift of hop length ,

[0097] Apply a preset window function w(τ) to each frame of the effective acoustic emission signal.

[0098] Apply the fast Fourier transform to each frame of the effective acoustic emission signal after applying the preset window function w(τ), convert the time-domain signal of each frame of the effective acoustic emission signal after applying the preset window function w(τ) into a frequency-domain signal, and combine the frequency-domain signals to obtain a spectrum matrix, where the spectrum matrix represents the intensity and phase of the signal in time and frequency; wherein, the rows represent time and the columns represent frequency.

[0099] Construct a spectrum matrix framework with the number of signal frames corresponding to the rows and the number of frequency points of the effective acoustic emission signal corresponding to the columns;

[0100] Fill the fast Fourier transform results of each frame of the effective acoustic emission signal into the spectrum matrix framework to obtain the spectrum matrix.

[0101] Based on the amplitude value of the elements in the spectrum matrix representing the intensity of the signal time-frequency diagram, i.e., the pixel value corresponding to the signal time-frequency diagram, map the time in the spectrum matrix to the horizontal axis of the signal time-frequency diagram, map the frequency in the spectrum matrix to the vertical axis of the signal time-frequency diagram, and map each element value in the spectrum matrix to the corresponding position on the signal time-frequency diagram according to the intensity of the signal time-frequency diagram to obtain the signal time-frequency diagram.

[0102] Construct a crack propagation evaluation model, use the signal time-frequency diagram as the input of the crack propagation evaluation model, and obtain the crack propagation rate and the severity status level; the crack propagation evaluation model includes a feature extraction network and a Paris crack propagation model.

[0103] Refer to Figure 2 , the method for obtaining the crack propagation rate and the corresponding severity status level includes:

[0104] Build a feature extraction network based on a convolutional neural network, and the feature extraction network includes an input layer, n 1 convolutional layers, n 2A pooling layer, a self-attention module, a fully connected layer, and an output layer; using the signal time-frequency diagram as the input of the feature extraction network to obtain the output feature vector; among them, the convolutional layer and the pooling layer are used to extract local time-frequency features to obtain the feature map R; in the self-attention module, the feature map R obtained after the convolutional layer and the pooling layer is linearly transformed to obtain the query matrix Q, the key matrix K, and the value matrix V, and the attention score Score is calculated according to the query matrix Q, the key matrix K, and the value matrix V; multiplying the attention score by the value matrix to obtain the attention output; the self-attention module can capture the global features of the signal and enhance the model's ability to identify crack propagation features.

[0105] The methods for obtaining the query matrix Q, the key matrix K, and the value matrix V include:

[0106] Q = R·W Q ;

[0107] K = R·W K ;

[0108] V = R·W V ;

[0109] where W Q , W K and W V are the weight matrices corresponding to the query matrix Q, the key matrix K, and the value matrix V respectively.

[0110] The methods for obtaining the attention score Score include:

[0111]

[0112] where μ k is the dimension of the key matrix K; T is the transpose of the vector; softmax is the normalization mapping function.

[0113] The methods for obtaining the attention output include:

[0114] A = Score×V;

[0115] where A is the attention output.

[0116] Based on the feature vector, obtain the stress intensity factor range.

[0117] Referring to Figure 3 , the methods for obtaining the stress intensity factor range include:

[0118] Step 2.1, obtain the historical feature vector and the corresponding stress intensity factor.

[0119] Step 2.2, establish a non-linear model between the historical feature vector and the corresponding stress intensity factor.

[0120] The methods for establishing the nonlinear model include:

[0121]

[0122] where D is the stress intensity factor; is the model parameter; n a is the highest order of the eigenvector; X is the eigenvector.

[0123] Step 2.3: Fit the above nonlinear model based on the historical eigenvectors and the corresponding historical stress intensity factors, and optimize it with the goal of minimizing the sum of squared residuals to obtain the corresponding model parameter values. Substitute the corresponding model parameter values into the nonlinear model to obtain the fitted nonlinear model.

[0124] Step 2.4: Use the eigenvector as the input of the fitted nonlinear model to obtain the corresponding minimum stress intensity factor and maximum stress intensity factor, and take the difference between the maximum stress intensity factor and the minimum stress intensity factor as the stress intensity factor range.

[0125] Obtain the corresponding correction coefficient according to the compressive strength, elastic modulus and crack width of the bridge concrete.

[0126]

[0127] where is the compressive strength correction coefficient; kc is the compressive coefficient, reflecting the degree of influence of the compressive strength on crack propagation; f c0 is the reference compressive strength; nc is the compressive index; f c is the compressive strength;

[0128]

[0129] where β E is the elastic modulus correction coefficient; E 0 is the reference elastic modulus; E is the elastic modulus;

[0130] β cw = 1 + k cw × cw;

[0131] where β cw is the crack width correction coefficient; cw is the crack width; k cw is the crack width coefficient, reflecting the degree of influence of the crack width on the acceleration of crack propagation;

[0132] Obtain the improved Paris crack propagation model based on the correction coefficient;

[0133] The improved Paris crack propagation model is as follows:

[0134]

[0135] Among them, a is the crack length; Y is the number of cycles; α is the crack growth rate, that is, the length of crack growth per cycle; ΔK is the stress intensity factor range, and both C and m are material constants. The material constant C is used to reflect the ability of bridge materials to resist fatigue crack growth, and the material constant m is used to characterize the influence degree of the stress intensity factor range on the fatigue crack growth rate.

[0136] Obtain a fatigue test data set composed of test values of the stress intensity factor range, obtain the fatigue crack growth rate corresponding to each stress intensity factor test value in the fatigue test data set, and fit the first material constant C, the second material constant m, and the correction coefficient in the improved Paris crack growth model by the least squares method; correct the Paris crack growth model based on the fitted first material constant C, the second material constant m, and the correction coefficient to obtain an improved Paris crack growth model;

[0137] Calculate the crack growth rate in the bridge based on the stress intensity factor range, the fitted first material constant C, and the fitted second material constant m. Calculate the crack growth rate based on the improved Paris crack growth model, and compare the crack growth rate with the crack growth rate interval corresponding to the preset expansion severity level (such as mild, moderate, severe) to obtain the corresponding severity level.

[0138] Monitor the original acoustic emission signal through a multi-channel sensor for crack spatial positioning to obtain the crack spatial positioning result.

[0139] The method for obtaining the crack spatial positioning result includes:

[0140] Calculate the time difference Δt between the original acoustic emission signal arriving at any two sensors u and v uv , and use the time difference positioning formula to locate the crack source:

[0141]

[0142] Among them, (x, y) is the crack position, and x and y respectively correspond to the lateral position and longitudinal position of the crack in the bridge; is the acoustic wave propagation speed; (x 1 , y 1 ) and (x 2 , y 2 ) are the coordinates of the first sensor and the second sensor respectively, x 1 and y 1 respectively correspond to the lateral position and longitudinal position of the first sensor in the bridge, x 2 and y 2They respectively correspond to the lateral position and longitudinal position of the second sensor on the bridge.

[0143] Based on the time difference Δt between the arrival times of the original acoustic emission signals at different sensors, a crack position probability distribution model is constructed to obtain the crack spatial positioning result.

[0144] The method for constructing the crack position probability distribution model includes:

[0145] The crack position (x, y) is obtained based on the located crack source.

[0146] According to the crack position (x, y), the corresponding value range [x min , y min × [x max , y max is obtained, where x min is the minimum value of the lateral position of the crack on the bridge; y min is the minimum value of the longitudinal position of the crack on the bridge; x max is the maximum value of the lateral position of the crack on the bridge; y max is the maximum value of the longitudinal position of the crack on the bridge; and the corresponding prior probability distribution P(x, y) is obtained.

[0147] The method for obtaining the corresponding prior probability distribution P(x, y) includes:

[0148]

[0149] All values of the crack position (x, y) are calculated according to the time difference positioning formula; for the crack position (x, y), the time difference Δt uv obeys a normal distribution, and the likelihood function P(Δt uv |x, y) of the corresponding time difference Δt uv is:

[0150]

[0151] where σ uv is the standard deviation of the measurement error of the time difference Δt uv ; e is a mathematical constant; is the actually observed time difference; is the theoretical time difference calculated according to the crack position (x, y).

[0152] Based on L groups of time differences Δt uv (u, v = 1, 2, …, L, u ≠ v), the total likelihood function P(Δt|x, y) is calculated as follows:

[0153] P(Δt|x, y) = Π u≠v P ( Δtuv |x,y);

[0154] Integrate over all crack positions (x,y) to obtain crack position evidence.

[0155] The method for obtaining crack position evidence includes:

[0156] P(Δt) = ∫∫P(Δt|x,y)P(x,y)dxdy;

[0157] where P(Δt) is the crack position evidence, i.e., the integral value obtained by integrating over all crack positions (x,y).

[0158] Calculate the posterior probability distribution corresponding to the crack position (x,y) according to Bayes' formula, obtain the crack position probability distribution model, and take the position corresponding to the maximum value of the posterior probability distribution in the crack position probability distribution model as the crack spatial positioning result.

[0159] Bayes' formula is as follows:

[0160]

[0161] where P(x,y|Δt) is the posterior probability distribution corresponding to the crack position (x,y).

[0162] Perform weighted fusion on the crack propagation rates and severity level grades of J sensors respectively to obtain the weighted crack propagation rate and the weighted severity level grade. Combine the weighted crack propagation rate, the weighted severity level grade, and the spatial positioning result with the bridge model and present it in three - dimensional form, which can intuitively display the crack distribution and trend.

[0163] The method for obtaining the weighted crack propagation rate and severity level grade includes:

[0164] Perform weighting on the crack propagation rates of J sensors to obtain the weighted crack propagation rate where α k is the crack propagation rate of the k - th sensor; w 1,k is the weight corresponding to α k Perform weighting on the severity level grades of J sensors to obtain the weighted severity level grade where β k is the severity level grade of the k - th sensor; w 2,k is the weight corresponding to α k The weights w 1,k 、w 2,k and w 3,k can be determined according to the signal - to - noise ratio and reliability of the sound propagation signal.

[0165] Example 2

[0166] Please refer to Figure 4 - Figure 5 As shown, a bridge crack propagation evaluation system based on acoustic emission signals in this embodiment includes:

[0167] Data acquisition module: An acoustic emission sensor array is arranged at a preset part of the bridge to collect the original acoustic emission signals within a preset time period;

[0168] Preprocessing module: A band-pass filter is used to remove the low-frequency ambient noise and high-frequency sensor interference of the original acoustic emission signals to obtain the filtered acoustic emission signals. The empirical mode decomposition technique is used to separate the filtered acoustic emission signals to obtain the effective acoustic emission signals;

[0169] Feature extraction module: The time-frequency transformation is used to transform the effective acoustic emission signals to obtain the signal time-frequency diagram;

[0170] Expansion evaluation module: A crack propagation evaluation model is constructed. The signal time-frequency diagram is used as the input of the crack propagation evaluation model to obtain the crack propagation rate and the corresponding severity level; The crack propagation evaluation model includes a feature extraction network and a Paris crack propagation model;

[0171] Crack location module: The multi-channel sensors are used to monitor the original acoustic emission signals for crack spatial location to obtain the crack spatial location results;

[0172] Visualization module: The crack propagation rates and severity levels of J sensors are respectively weighted and fused to obtain the weighted crack propagation rates and severity levels. The weighted crack propagation rates, weighted severity levels and spatial location results are combined with the bridge model and presented in a three-dimensional form.

[0173] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all of them should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0174] Finally: The above is only the preferred embodiment of the present invention and is not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A bridge crack extension assessment method based on acoustic emission signals, characterized in that: The steps include: An acoustic emission sensor array is arranged at a preset position of the bridge to collect original acoustic emission signals within a preset period of time; A bandpass filter is used to remove the low-frequency environmental noise and high-frequency sensor interference of the original acoustic emission signal to obtain a filtered acoustic emission signal. The filtered acoustic emission signal is separated using the empirical mode decomposition technique to obtain an effective acoustic emission signal. The effective acoustic emission signal is transformed by time-frequency transformation to obtain the signal time-frequency diagram; A crack extension assessment model is constructed, and the signal time-frequency diagram is used as the input of the crack extension assessment model to obtain the crack extension rate and the corresponding severity level; the crack extension assessment model includes a feature extraction network and an improved Paris crack extension model; The crack growth rates and severity levels of J sensors are weighted and fused respectively to obtain weighted crack growth rates and severity levels, which are then combined with the bridge model and presented in three-dimensional form.

2. A bridge crack extension assessment method based on acoustic emission signals according to claim 1, characterized in that: The method further includes: monitoring the original acoustic emission signal through a multi-channel sensor to perform spatial positioning of the crack, obtaining a spatial positioning result of the crack, and combining the spatial positioning result of the crack with a bridge model to present it in a three-dimensional form; The method for obtaining the crack spatial positioning result comprises: Step A: Calculate the time difference Δt between the original acoustic emission signal and any two sensors u and v uv , use the time difference positioning formula to locate the crack source: Step B: Based on the time difference Δt uv Construct a crack location probability distribution model to obtain crack spatial positioning results; Methods for obtaining fracture spatial positioning results include: Step B1, obtaining the crack source located in step A to obtain the crack position (x, y); Step B2: Obtain the corresponding value range [x min ,y min ]×[x max ,y max ], where x min is the minimum lateral position of the crack on the bridge; y min is the minimum longitudinal position of the crack in the bridge; x max is the maximum lateral position of the crack on the bridge; max is the maximum longitudinal position of the crack position on the bridge; obtain the corresponding prior probability distribution; Step B3: Calculate all values ​​of the crack position (x, y) according to the time difference positioning formula; for the crack position (x, y), the time difference Δt uv Obey the normal distribution and calculate the corresponding time difference Δt uv Likelihood function based on L groups of time differences Δt uv (u,v=1,2,…,L,u≠v) calculate and obtain the total likelihood function; Step B4, integrating all crack positions (x, y) to obtain crack position evidence; Step B5: Calculate the posterior probability distribution corresponding to the crack position (x, y) according to the Bayesian formula to obtain the crack position probability distribution model, and take the position corresponding to the maximum value of the posterior probability distribution in the crack position probability distribution model as the crack space positioning result.

3. The bridge crack extension assessment method based on acoustic emission signals according to claim 1 is characterized in that: The method for obtaining the crack growth rate and the corresponding severity level comprises: Step 1.1, build a feature extraction network based on a convolutional neural network, the feature extraction network includes an input layer, n1 convolutional layers, n2 pooling layers, a self-attention module, a fully connected layer and an output layer; use the signal time-frequency diagram as the input of the feature extraction network to obtain the output feature vector; wherein, the convolutional layer and the pooling layer are used to extract the local time-frequency features of the signal time-frequency diagram to obtain the feature map R; in the self-attention module, perform a linear transformation on the feature map R obtained after the convolutional layer and the pooling layer to obtain the query matrix Q, the key matrix K and the value matrix V, and calculate the attention score Score according to the query matrix Q, the key matrix K and the value matrix V; multiply the attention score by the value matrix to obtain the attention output; Step 1.2, obtaining the range of stress intensity factor based on the eigenvector; Step 1.3: Calculate the crack growth rate based on the improved Paris crack growth model, compare the crack growth rate with the crack growth rate interval corresponding to the preset growth severity level, and obtain the corresponding severity level.

4. A bridge crack extension assessment method based on acoustic emission signals according to claim 3, characterized in that: The method for obtaining the range of stress intensity factors comprises: Step 2.1, obtain the historical eigenvector and the corresponding stress intensity factor; Step 2.2: Establish a nonlinear model of historical eigenvectors and corresponding stress intensity factors: Step 2.3, fitting the nonlinear model based on the historical characteristic vector and the corresponding historical stress intensity factor range, and optimizing with the goal of minimizing the residual sum of squares to obtain the corresponding nonlinear model parameter values, and substituting the corresponding nonlinear model parameter values ​​into the nonlinear model to obtain the fitted nonlinear model; Step 2.4: Use the characteristic vector as the input of the fitted nonlinear model to obtain the corresponding minimum stress intensity factor and maximum stress intensity factor, and use the difference between the maximum stress intensity factor and the minimum stress intensity factor as the stress intensity factor range.

5. The bridge crack extension assessment method based on acoustic emission signals according to claim 2 is characterized in that: The method for obtaining the improved Paris crack propagation model includes: The correction coefficient corresponding to the Paris crack propagation model is obtained according to the compressive strength, elastic modulus and crack width of the bridge concrete, a fatigue test data set consisting of stress intensity factor range test values ​​is obtained, and the fatigue crack propagation rate corresponding to each stress intensity factor test value in the fatigue test data set is obtained. The first material constant C, the second material constant m and the correction coefficient in the improved Paris crack propagation model are fitted by the least squares method; the Paris crack propagation model is corrected based on the fitted first material constant C, the second material constant m and the correction coefficient to obtain the improved Paris crack propagation model.

6. The bridge crack extension assessment method based on acoustic emission signals according to claim 1 is characterized in that: The method for obtaining an effective acoustic emission signal comprises: Step 3.1, determining the local extreme point by comparing the size relationship between each data point in the filtered acoustic emission signal and the corresponding two adjacent data points, if the data point is larger than the corresponding two adjacent data points, then the data point is a local maximum point; conversely, if the data point is smaller than the corresponding two adjacent data points, then the data point is a local minimum point; Step 3.2: Connect all local maximum points to form the upper envelope bl max (τ), τ is time; connect all local minimum points to form the lower envelope Bl min (τ); Step 3.3, calculate the average value m(τ) of the upper envelope and the lower envelope, and calculate the difference between the filtered acoustic emission signal g(τ) and the average value m(τ) to obtain the difference signal h(τ); Step 3.4, check whether the difference signal h(τ) satisfies the IMF condition; if not, use the difference signal h(τ) as the new filtered first acoustic emission signal, repeat steps 3.1 to 3.3 until the IMF condition is met, and obtain the first IMF; if it is met, directly obtain the first IMF, and mark the first IMF as c1(τ); Step 3.5, subtract c1(τ) from the filtered acoustic emission signal g(τ) to obtain a first residual signal r1(τ); Step 3.6: Use the first residual signal r1(τ) as the new filtered second acoustic emission signal, repeat step 3.4 to obtain the second IMF, mark the second IMF as c2(τ), and repeat step 3.5 to calculate and obtain the second residual signal r2(τ); Step 3.7, repeat step 3.6 until the residual component When it becomes a monotonic function or the number of extreme points does not exceed two, the decomposition process ends; the filtered acoustic emission signal g(τ) is represented as n g The form of the IMF and residual components; Step 3.8: Perform fast Fourier transform on each IMF to obtain the corresponding spectrum range, extract the IMF as the effective component according to the preset effective signal frequency range, and reconstruct the effective component to obtain the useful signal g eff (τ), the useful signal g eff (τ) is taken as the effective acoustic emission signal.

7. A bridge crack extension assessment method based on acoustic emission signals according to claim 6, characterized in that: The IMF conditions include: within the entire data segment of the acoustic emission signal, the number of extreme points and the number of zero-crossing points must be equal or differ by at most one; at any data point, the average values ​​of the upper envelope and the lower envelope defined by the local maximum point and the local minimum point are zero.

8. The bridge crack propagation assessment method based on acoustic emission signals according to claim 6 is characterized in that: The method for obtaining a signal time-frequency diagram comprises: Step 4.1, dividing the effective acoustic emission signal into d small frames, and the frame shift is half of the preset length of each frame; Step 4.2, applying a preset window function w(τ) to each frame of valid acoustic emission signal; Step 4.3, applying fast Fourier transform to each frame of effective acoustic emission signal after applying the preset window function w(τ), converting the time domain signal of each frame of effective acoustic emission signal after applying the preset window function w(τ) into a frequency domain signal, and combining the frequency domain signals to obtain a spectrum matrix; The method for obtaining the frequency spectrum matrix comprises: Construct a spectrum matrix framework in which the rows correspond to the number of signal frames of the effective acoustic emission signal and the columns correspond to the number of frequency points of the effective acoustic emission signal; Fill the fast Fourier transform result of each frame of effective acoustic emission signal into the spectrum matrix frame to obtain the spectrum matrix; Step 4.4, based on the amplitude value of the elements in the spectrum matrix representing the intensity of the signal time-frequency diagram, that is, the pixel value corresponding to the signal time-frequency diagram, map the time in the spectrum matrix to the horizontal axis of the signal time-frequency diagram, map the frequency in the spectrum matrix to the vertical axis of the signal time-frequency diagram, and map each element value in the spectrum matrix to the corresponding position of the signal time-frequency diagram according to the intensity of the signal time-frequency diagram to obtain the signal time-frequency diagram.

9. The bridge crack propagation assessment method based on acoustic emission signals according to claim 1, characterized in that: The method for obtaining the weighted crack growth rate and severity level includes: Weighting the crack growth rates of J sensors to obtain a weighted crack growth rate; The severity levels of the J sensors are weighted to obtain a weighted severity level.

10. A bridge crack extension assessment system based on acoustic emission signals, used to implement a bridge crack extension assessment method based on acoustic emission signals as claimed in any one of claims 1 to 9, characterized in that: include: Data acquisition module: an acoustic emission sensor array is arranged at a preset position of the bridge to collect the original acoustic emission signal within a preset period of time; Preprocessing module: Use a bandpass filter to remove the low-frequency environmental noise and high-frequency sensor interference of the original acoustic emission signal to obtain the filtered acoustic emission signal, and use the empirical mode decomposition technology to separate the filtered acoustic emission signal to obtain the effective acoustic emission signal; Feature extraction module: Use time-frequency transformation to transform the effective acoustic emission signal and obtain the signal time-frequency diagram; Extension assessment module: construct a crack extension assessment model, use the signal time-frequency diagram as the input of the crack extension assessment model, and obtain the crack extension rate and the corresponding severity level; the crack extension assessment model includes a feature extraction network and an improved Paris crack extension model; Crack positioning module: monitors the original acoustic emission signal through a multi-channel sensor to locate the crack space and obtain the crack space positioning result; Visualization module: The crack growth rate and severity level of J sensors are weighted and fused to obtain the weighted crack growth rate and severity level. The weighted crack growth rate, weighted severity level and spatial positioning results are combined with the bridge model and presented in three-dimensional form.

Images