Bridge fatigue life prediction method based on wavelet packet analysis

Abstract

The invention relates to the technical field of bridge fatigue life prediction, and provides a bridge fatigue life prediction method based on wavelet packet analysis. The method comprises the following steps: acquiring a deflection signal of a target bridge; adding white noise into the deflection signal to obtain a noise deflection signal, and performing empirical mode decomposition on the noise deflection signal to obtain a plurality of final mode components; denoising and reconstructing all the final modal components to obtain a reconstructed deflection signal, and performing wavelet packet analysis on the reconstructed deflection signal to obtain a deflection feature vector; encoding and decoding the deflection feature vector to obtain a final deflection feature vector; and predicting the bridge fatigue life of the target bridge according to the final deflection feature vector to obtain a bridge fatigue life prediction result of the target bridge. According to the method, the accuracy of bridge fatigue life prediction can be improved.

Description

Technical Field

[0001] The present application relates to the technical field of bridge fatigue life prediction, and in particular to a bridge fatigue life prediction method based on wavelet packet analysis. Background Art

[0002] When bridges are subjected to long-term traffic loads, fatigue cracks will occur. This fatigue damage poses a serious threat to the structural safety and service life of bridges. Therefore, accurate prediction of the fatigue life of bridges is crucial for formulating maintenance strategies and extending the service life of bridges. Traditional fatigue life prediction methods rely on a large number of structural or material fatigue tests, which are often computationally intensive, complex to operate, and require a large amount of test data support. At present, the methods for predicting bridge fatigue life mainly include methods based on statistical models and methods based on machine learning. Statistical model methods usually treat stress amplitudes as random variables and use gamma distribution, lognormal distribution, Weibull distribution, etc. for simulation. However, the simulation accuracy of these methods is low and they fail to fully utilize the information of measured data.

[0003] At present, the field of bridge fatigue life prediction widely uses technologies based on traditional statistical methods, such as the stress-life (SN) curve method. The SN curve method establishes the relationship between stress and fatigue life through experimental data, which can qualitatively describe the fatigue damage process of the bridge structure and quantitatively calculate its fatigue life. However, this method has some technical disadvantages in practical applications, such as strong dependence on experimental data, slow model update, and difficulty in adapting to complex and changing working conditions, resulting in limited prediction accuracy.

[0004] It can be seen that the current bridge fatigue life prediction method has the problem of low accuracy in bridge fatigue life prediction. Summary of the invention

[0005] The present application provides a bridge fatigue life prediction method based on wavelet packet analysis, which can solve the problem of low accuracy in bridge fatigue life prediction.

[0006] In a first aspect, an embodiment of the present application provides a bridge fatigue life prediction method based on wavelet packet analysis, and the bridge fatigue life prediction method includes: Obtaining the deflection signal of the target bridge; the deflection signal is used to describe the deflection of the target bridge at T moments under the joint action of multiple effects, where the Tth moment is the current moment; Adding white noise to the deflection signal to obtain a noisy deflection signal, and performing empirical mode decomposition on the noisy deflection signal to obtain multiple final modal components; Denoising and reconstructing all final modal components to obtain reconstructed deflection signals, and performing wavelet packet analysis on the reconstructed deflection signals to obtain deflection eigenvectors; Encode and decode the deflection feature vector to obtain the final deflection feature vector; The bridge fatigue life of the target bridge is predicted according to the final deflection eigenvector to obtain a bridge fatigue life prediction result of the target bridge; the bridge fatigue life prediction result is used to describe the fatigue degree of the target bridge.

[0007] Optionally, adding white noise to the deflection signal to obtain a noisy deflection signal includes: Calculate the standard deviation of the deflection signal in multiple time windows; An adaptive white noise is generated based on the standard deviation, and the adaptive white noise is added to the deflection signal to obtain a noisy deflection signal.

[0008] Optionally, calculate the standard deviation of the deflection signal in multiple time windows, including: By the formula: Solve for the standard deviation ; in, represents the local mean, represents the time window, , Indicates the deflection signal the deflection of a historical moment; Generates adaptive white noise based on standard deviation, including: By formula: Generate adaptive white noise ; in, represents the amplitude, represents the noise frequency, Indicates Gaussian white noise, represents the basic noise frequency, Indicates the amount of change, represents the change threshold, represents the deflection signal, Indicates the previous deflection signal, represents the basic noise amplitude; Adding adaptive white noise to the deflection signal, a noisy deflection signal is obtained, including: By formula: Calculate the noise deflection signal .

[0009] Optionally, empirical mode decomposition is performed on the noise deflection signal to obtain multiple final modal components, including: By formula: Calculate the Final modal components ; in, Indicates modal components, represents the empirical mode decomposition, represents the number of modal components.

[0010] Optionally, all final modal components are denoised and reconstructed to obtain a reconstructed deflection signal, including: Perform wavelet packet decomposition on all final modal components to obtain multiple denoised wavelet packet coefficients; Perform soft threshold processing on each denoising wavelet packet coefficient to obtain the final denoising wavelet packet coefficient corresponding to each denoising wavelet packet coefficient; All final denoised wavelet packet coefficients are subjected to inverse wavelet packet transform to obtain the reconstructed deflection signal.

[0011] Optionally, soft threshold processing is performed on each denoised wavelet packet coefficient to obtain a final denoised wavelet packet coefficient corresponding to each denoised wavelet packet coefficient, including: By formula: Calculate the final denoised wavelet packet coefficients ; in, represents the denoised wavelet packet coefficients The symbol function of represents the threshold value, represents the noise standard deviation, Indicates the length of the signal.

[0012] Optionally, wavelet packet analysis is performed on the reconstructed deflection signal to obtain a deflection feature vector, including: The number of iterations is increased by 1, and the reconstructed deflection signal is decomposed by multi-layer wavelet packets according to the step size factor to obtain multiple signal nodes; Determine the target signal node from all signal nodes, and calculate the standard deviation and average value of the wavelet packet coefficients of the target signal node; The standard deviation of wavelet packet coefficients and the average value of wavelet packet coefficients are combined to obtain the initial deflection eigenvector; An error signal is calculated based on the initial deflection eigenvector, and the step factor is updated to obtain an updated step factor; Performing power detection on the error signal to obtain the power of the error signal, and determining whether a difference between the power of the error signal and the power of the error signal of the previous iteration is greater than a power difference threshold; If so, the updated step size factor is used as the step size factor, the number of iterations plus 1 is returned, and the reconstructed deflection signal is subjected to multi-layer wavelet packet decomposition according to the step size factor to obtain the steps of multiple signal nodes; Otherwise, the initial deflection eigenvector is taken as the deflection eigenvector.

[0013] Optionally, an error signal is calculated based on the initial deflection eigenvector, including: By formula: Calculate the error signal ; in, Indicates the expected signal, represents the output signal after the filter processes the initial deflection eigenvector; The step size factor is updated to obtain an updated step size factor, including: By formula: Calculate the The step size factor for the iteration ; in, Indicates The step size factor for the iteration, and All represent parameters. represents the mask ratio, represents the standard deviation of the wavelet packet coefficients, represents the average value of the wavelet packet coefficients, Represents a positive number, , Indicates the maximum number of iterations.

[0014] Optionally, the deflection feature vector is encoded and decoded to obtain a final deflection feature vector, including: Generate a mask matrix of deflection eigenvectors; The deflection feature vector is encoded according to the mask matrix to obtain a coding vector, and the coding vector is decoded to obtain a final deflection feature vector.

[0015] Optionally, generate a mask matrix for the deflection eigenvector, including: By formula: Calculate the mask matrix Line Column mask ; in, represents a random number uniformly distributed in the range [0, 1]. represents the mask probability, and All represent hyperparameters, express The expected value of express The fluctuation range of represents the density threshold, represents the strong signal threshold, Indicates the weak signal threshold, represents the range constant, , represents the index of the last row in the mask matrix, , Represents the index of the last column in the mask matrix.

[0016] In a second aspect, an embodiment of the present application provides a bridge fatigue life prediction device based on wavelet packet analysis, comprising: An acquisition module is used to acquire a deflection signal of a target bridge; the deflection signal is used to describe the deflection of the target bridge at T moments under the joint action of multiple effects, and the Tth moment is the current moment; An empirical mode decomposition module is used to add white noise to the deflection signal to obtain a noisy deflection signal, and perform empirical mode decomposition on the noisy deflection signal to obtain multiple final modal components; The wavelet packet analysis module is used to denoise and reconstruct all final modal components to obtain a reconstructed deflection signal, and to perform wavelet packet analysis on the reconstructed deflection signal to obtain a deflection feature vector; The encoding and decoding module is used to encode and decode the deflection feature vector to obtain the final deflection feature vector; The prediction module is used to predict the bridge fatigue life of the target bridge according to the final deflection characteristic vector to obtain the bridge fatigue life prediction result of the target bridge; the bridge fatigue life prediction result is used to describe the fatigue degree of the target bridge.

[0017] In a third aspect, an embodiment of the present application provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned bridge fatigue life prediction method based on wavelet packet analysis when executing the above-mentioned computer program.

[0018] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned bridge fatigue life prediction method based on wavelet packet analysis.

[0019] The above solution of the present application has the following beneficial effects: In an embodiment of the present application, the deflection signal of the target bridge is obtained, and then white noise is added to the deflection signal to obtain a noise deflection signal, and the noise deflection signal is subjected to empirical mode decomposition to obtain a final modal component, and then the final modal component is denoised and reconstructed to obtain a reconstructed deflection signal, and the reconstructed deflection signal is subjected to wavelet packet analysis to obtain a deflection feature vector, and then the deflection feature vector is encoded and decoded to obtain a final deflection feature vector, and finally the fatigue life of the target bridge is predicted according to the final deflection feature vector to obtain a bridge fatigue life prediction result of the target bridge. Among them, adding white noise to the deflection signal and then denoising and reconstructing it can effectively filter the noise part in the deflection signal, retain the useful components containing the fatigue characteristics of the bridge as much as possible, improve the accuracy of the reconstructed deflection signal, and perform wavelet packet analysis on the accurate reconstructed deflection signal, giving full play to the advantages of wavelet packet analysis in signal processing, which is fine and comprehensive, effectively improving the accuracy of the deflection feature vector, and predicting the fatigue life of the bridge based on the accurate deflection feature vector can improve the accuracy of the bridge fatigue life prediction result.

[0020] Other beneficial effects of the present application will be described in detail in the subsequent specific implementation section. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0022] Figure 1 A flow chart of a bridge fatigue life prediction method based on wavelet packet analysis provided in one embodiment of the present application; Figure 2 A schematic diagram of a three-layer wavelet packet decomposition tree provided in an embodiment of the present application; Figure 3A schematic diagram of the structure of a bridge fatigue life prediction device based on wavelet packet analysis provided in one embodiment of the present application; Figure 4 A schematic diagram of the structure of a terminal device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0023] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application may also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present application.

[0024] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or combinations thereof.

[0025] It should also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0026] As used in the specification and appended claims of this application, the term "if" can be interpreted as "when" or "uponce" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" can be interpreted as meaning "uponce it is determined" or "in response to determining" or "uponce [described condition or event] is detected" or "in response to detecting [described condition or event]", depending on the context.

[0027] In addition, in the description of the present application specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0028] References to "one embodiment" or "some embodiments" etc. described in the specification of this application mean that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0029] In response to the problem of low accuracy in existing bridge fatigue life predictions, an embodiment of the present application provides a bridge fatigue life prediction method based on wavelet packet analysis. The bridge fatigue life prediction method obtains the deflection signal of the target bridge, then adds white noise to the deflection signal to obtain a noisy deflection signal, performs empirical mode decomposition on the noisy deflection signal to obtain a final modal component, then denoises and reconstructs the final modal component to obtain a reconstructed deflection signal, performs wavelet packet analysis on the reconstructed deflection signal to obtain a deflection eigenvector, then encodes and decodes the deflection eigenvector to obtain a final deflection eigenvector, and finally predicts the bridge fatigue life of the target bridge based on the final deflection eigenvector to obtain a bridge fatigue life prediction result of the target bridge. Among them, adding white noise to the deflection signal and then denoising and reconstructing it can effectively filter out the noise part in the deflection signal, retain the useful components containing the fatigue characteristics of the bridge as much as possible, and improve the accuracy of the reconstructed deflection signal. Wavelet packet analysis is performed on the precise reconstructed deflection signal, which gives full play to the advantages of wavelet packet analysis in signal processing, effectively improves the accuracy of the deflection feature vector, and predicts the fatigue life of the bridge based on the accurate deflection feature vector, which can improve the accuracy of the bridge fatigue life prediction results.

[0030] Next, an exemplary description is given of the bridge fatigue life prediction method based on wavelet packet analysis provided in this application.

[0031] like Figure 1 As shown, the bridge fatigue life prediction method based on wavelet packet analysis provided in this application includes the following steps: Step 11, obtaining the deflection signal of the target bridge.

[0032] The above deflection signal is used to describe the deflection of the target bridge at T moments under the joint action of multiple effects, and the Tth moment is the current moment.

[0033] In some embodiments of the present application, the deflection of the bridge during structural changes can be collected in real time by installing displacement sensors, inclination sensors, strain sensors and other equipment at key monitoring points of the bridge.

[0034] It should be noted that the fatigue damage degree of the bridge can be preliminarily determined based on the deflection-related data. Specifically, based on the knowledge of material mechanics and using the deflection curve equation of a simply supported beam under external load, the stiffness expression can be obtained as follows: Among them, B is the section bending stiffness, M is the section bending moment, L is the span of the specimen, α is the deflection coefficient related to the support conditions, load, etc., and f is the deflection of the specimen at the corresponding section.

[0035] The fatigue damage degree characterized by the stiffness degradation law is defined as: in, is the damage variable (defined using the specimen stiffness, ∈[0,1]), is the initial stiffness of the specimen, is the stiffness of the specimen after being subjected to n times of fatigue loading, It is the stiffness of the specimen at fatigue failure (N times of repeated loading).

[0036] Step 12, adding white noise to the deflection signal to obtain a noisy deflection signal, and performing empirical mode decomposition on the noisy deflection signal to obtain multiple final modal components.

[0037] The above-mentioned multiple final modal components are used to describe the signal components of the deflection signal at multiple different frequencies, and each final modal component corresponds to a signal component of a different frequency.

[0038] In some embodiments of the present application, the steps of adding white noise to the deflection signal to obtain a noisy deflection signal, and performing empirical mode decomposition on the noisy deflection signal to obtain a plurality of final modal components include: In the first step, the standard deviation of the deflection signal in multiple time windows is calculated.

[0039] Specifically, through the formula: Solve for the standard deviation ; in, represents the local mean, represents the time window, , Indicates the last historical moment, Indicates the deflection signal The deflection of a historical moment.

[0040] In the second step, adaptive white noise is generated based on the standard deviation and added to the deflection signal to obtain a noisy deflection signal.

[0041] Specifically, through the formula: Generate adaptive white noise ; in, represents the amplitude, represents the noise frequency, Indicates The Gaussian white noise has a mean of 0 and a variance of 1. represents the basic noise frequency, Indicates the amount of change, represents the change threshold, represents the deflection signal, represents the previous deflection signal, that is, a deflection signal with the same time span as the deflection signal obtained in step 11, but earlier than the deflection signal and not overlapping with it, Indicates the basic noise amplitude.

[0042] By formula: Calculate the noise deflection signal .

[0043] The third step is to perform empirical mode decomposition on the noise deflection signal to obtain multiple final modal components.

[0044] Specifically, through the formula: Calculate the Final modal components ; in, Indicates modal components, represents the empirical mode decomposition, represents the number of modal components.

[0045] Step 13, denoising and reconstructing all final modal components to obtain a reconstructed deflection signal, and performing wavelet packet analysis on the reconstructed deflection signal to obtain a deflection eigenvector.

[0046] In some embodiments of the present application, the above-mentioned steps of denoising and reconstructing all final modal components to obtain a reconstructed deflection signal, and performing wavelet packet analysis on the reconstructed deflection signal to obtain a deflection feature vector include: In the first step, all final modal components are decomposed by wavelet packets to obtain multiple denoised wavelet packet coefficients.

[0047] Specifically, appropriate wavelet basis functions and decomposition levels are selected, and wavelet packet decomposition is performed on all final modal components to obtain multiple denoising wavelet packet coefficients.

[0048] In the second step, each denoised wavelet packet coefficient is subjected to soft threshold processing to obtain the final denoised wavelet packet coefficient corresponding to each denoised wavelet packet coefficient.

[0049] Specifically, through the formula: Calculate the final denoised wavelet packet coefficients ; in, Represents the denoised wavelet packet coefficients The symbol function of represents the threshold value, represents the noise standard deviation, Indicates the length of the signal.

[0050] In the third step, all the final denoised wavelet packet coefficients are subjected to inverse wavelet packet transform to obtain the reconstructed deflection signal.

[0051] In the fourth step, the number of iterations is increased by 1, and the reconstructed deflection signal is decomposed by multi-layer wavelet packet according to the step factor to obtain multiple signal nodes.

[0052] It should be noted that the initial number of iterations is 0. The above signal nodes are the sub-band signals obtained after decomposition. Each decomposition will divide the signal into multiple frequency bands, and each frequency band corresponds to a signal node.

[0053] For example, the number of layers is 3, and three-layer wavelet packet decomposition is performed. The obtained three-layer wavelet packet decomposition tree is as follows: Figure 2 As shown, represents the reconstructed deflection signal, Represents the signal node of layer 0, represents the first signal node of the first layer, Represents the second signal node of the first layer, represents the first signal node of the second layer, represents the second signal node of the second layer, Represents the third signal node of the second layer, Represents the fourth signal node of the second layer, Represents the first signal node of the third layer, Represents the second signal node of the third layer, Represents the third signal node of the third layer, Represents the fourth signal node of the third layer, Represents the fifth signal node of the third layer, Represents the sixth signal node of the third layer, Represents the seventh signal node of the third layer, It represents the eighth signal node of the third layer, and the arrows represent the signal decomposition relationship between signal nodes.

[0054] The fifth step is to determine the target signal node from all signal nodes, and calculate the standard deviation and average value of the wavelet packet coefficients of the target signal node.

[0055] It should be noted that after performing multi-layer wavelet packet decomposition, multiple signal nodes are obtained, and the wavelet packet coefficients of each data in the reconstructed deflection signal at each signal node are also obtained. All wavelet packet coefficients of each signal node are analyzed, and the signal node with the most obvious fluctuation of the wavelet packet coefficient is identified, and the signal node is used as the target signal node. Then, the standard deviation and average value of the wavelet packet coefficients of the target signal node are calculated based on all the wavelet packet coefficients of the target signal node. Standard deviation of wavelet packet coefficients The calculation formula is: in, represents the number of data in the reconstructed deflection signal, represents the first The wavelet packet coefficients of the data at the target signal node, Represents the average value of the wavelet packet coefficients of the target signal node.

[0056] In the sixth step, the standard deviation of the wavelet packet coefficients and the average value of the wavelet packet coefficients are combined to obtain the initial deflection eigenvector.

[0057] In the seventh step, the error signal is calculated based on the initial deflection eigenvector, and the step factor is updated to obtain an updated step factor.

[0058] Specifically, through the formula: Calculate the error signal .

[0059] in, Indicates the expected signal, Represents the output signal after the filter processes the initial deflection eigenvector.

[0060] By formula: Calculate the The step size factor for the iteration .

[0061] in, Indicates The step size factor for the iteration, and All represent parameters. represents the mask ratio, represents the standard deviation of the wavelet packet coefficients, represents the average value of wavelet packet coefficients, Represents a positive number, , Indicates the maximum number of iterations.

[0062] In the eighth step, power detection is performed on the error signal to obtain the power of the error signal, and it is determined whether the difference between the power of the error signal and the power of the error signal of the previous iteration is greater than a power difference threshold.

[0063] If so, the updated step size factor is used as the step size factor, the number of iterations plus 1 is returned, and the reconstructed deflection signal is subjected to multi-layer wavelet packet decomposition according to the step size factor to obtain the steps of multiple signal nodes; Otherwise, the initial deflection eigenvector is taken as the deflection eigenvector.

[0064] It should be noted that the process of performing power detection on the error signal to obtain the power of the error signal may specifically be: 1. Calculate the power of the error signal: Assume the error signal is , its power can be approximated by calculating its energy: Where N is the length of the signal and e(n) is the nth sample of the error signal.

[0065] 2. Compare power difference: The power of the error signal of the current iteration is calculated by the above formula: and the power of the error signal of the previous iteration .

[0066] Determine whether the power difference is greater than the set threshold : If the condition is met, the iteration continues; otherwise, the iteration stops.

[0067] Step 14, encoding and decoding the deflection feature vector to obtain the final deflection feature vector.

[0068] In some embodiments of the present application, the steps of encoding and decoding the deflection feature vector to obtain the final deflection feature vector include: In the first step, a mask matrix of deflection eigenvectors is generated.

[0069] Specifically, through the formula: Calculate the mask matrix Line Column mask ; in, represents a random number uniformly distributed in the range [0, 1]. represents the mask probability, and All represent hyperparameters, express For example, if the mean is 0.5, then on average, about half of the elements in the mask matrix will be masked. express The larger the fluctuation range, the more dispersed the distribution of the mask probability, and the sparsity of the generated mask matrix will be more random. represents the density threshold, represents the strong signal threshold, Indicates the weak signal threshold, represents the range constant, , represents the index of the last row in the mask matrix, , Represents the index of the last column in the mask matrix.

[0070] In the second step, the deflection feature vector is encoded according to the mask matrix to obtain the encoding vector, and the encoding vector is decoded to obtain the final deflection feature vector.

[0071] It should be noted that an asymmetric encoder-decoder model (such as a masked autoencoder) can be used to perform the above steps to obtain the final deflection feature vector.

[0072] Exemplarily, the encoder encodes only the unmasked data, while the decoder attempts to reconstruct all the data, including the masked parts. The encoder processes the visible data and extracts features, and the decoder uses these features to try to reconstruct the complete data. The data features extracted by the encoder are fed into the decoder, which attempts to reconstruct the complete data, including those masked parts. In this way, the data representation learned by the model can capture the intrinsic structure and pattern of the bridge deflection signal. The difference between the reconstructed data and the original data is calculated as a loss (usually a mean squared error), and the model is trained by minimizing this loss. This loss function is only calculated on the masked data, and the unmasked data does not participate in the calculation of the loss.

[0073] After the training is completed, the trained model is used to perform the above steps to obtain the final deflection feature vector, which can improve the accuracy of the final deflection feature vector.

[0074] Step 15, predicting the fatigue life of the target bridge according to the final deflection eigenvector to obtain a prediction result of the fatigue life of the target bridge.

[0075] The above bridge fatigue life prediction results are used to describe the fatigue degree of the target bridge.

[0076] Specifically, the fatigue degree of the target bridge can be inferred by analyzing the reconstruction error between the final deflection eigenvector and the deflection eigenvector. The reconstruction error can be quantified by calculating the mean square error (MSE) or mean absolute error (MAE) between the final deflection eigenvector and the deflection eigenvector. The calculation formula is as follows: in is the number of samples (which can be understood as the data dimension of the deflection eigenvector), It is The true value of the samples (i.e. the value in the deflection eigenvector), It is The predicted value of samples (i.e., the value in the final deflection eigenvector).

[0077] in is the sample size, It is The true value of the samples, It is The predicted value of a sample.

[0078] According to the value of MSE or MAE, the fatigue degree of the target bridge can be analyzed. The smaller the value of MSE or MAE, the lighter the fatigue degree, and the larger the value of MSE or MAE, the heavier the fatigue degree.

[0079] If the bridge structure is healthy, the reconstruction error of its deflection eigenvector in the model should be small. When the reconstruction error is large, it may indicate fatigue or damage problems in the bridge structure. If the reconstruction error exceeds the preset threshold or is significantly different from the reconstruction error in the healthy state, it can be considered that there is a problem with the bridge.

[0080] Exemplarily, the corresponding relationship between the MSE or MAE value and the fatigue degree is shown in Table 1.

[0081] Table 1

[0082] It should be noted that after obtaining the bridge life prediction results, corresponding measures can be taken for the target bridge according to the bridge life prediction results, such as reinforcement, so as to avoid accidents such as collapse caused by excessive bridge fatigue.

[0083] It is worth mentioning that adding white noise to the deflection signal and then denoising and reconstructing it can effectively filter out the noise part in the deflection signal, retain the useful components containing the fatigue characteristics of the bridge as much as possible, and improve the accuracy of the reconstructed deflection signal. Wavelet packet analysis is performed on the precise reconstructed deflection signal, which gives full play to the advantages of wavelet packet analysis in signal processing. It effectively improves the accuracy of the deflection feature vector, and predicts the fatigue life of the bridge based on the accurate deflection feature vector, which can improve the accuracy of the bridge fatigue life prediction results.

[0084] In addition, the deflection signal of the bridge is denoised to effectively filter out the noise part of the signal and retain the useful components containing the fatigue characteristics of the bridge as much as possible, providing effective input for subsequent damage identification.

[0085] In the fatigue life prediction of the overall structure of the bridge under complex environment, a method combining wavelet packet analysis in machine learning with unsupervised learning of masked autoencoding is proposed to identify the fatigue damage of the bridge in real time. This method gives full play to the advantages of wavelet packet analysis in signal processing, which is precise and comprehensive. Moreover, this method only needs to calculate and analyze the deflection data automatically collected in the bridge monitoring system, which can effectively improve the monitoring efficiency.

[0086] The method of the present application also has the following advantages: (1) Parameters for measuring the degree of bridge fatigue damage and indicators for identifying bridge fatigue damage are proposed, providing a scientific basis for accurately quantifying bridge fatigue damage.

[0087] (2) From the perspective of digital signals, it is proposed to convert the mid-span deflection monitoring data of the bridge into digital signals, and extract characteristic parameters that are highly correlated with the fatigue damage status of the bridge from the fluctuation characteristics of the deflection signal.

[0088] (3) The Masked Autoencoders Are Scalable Vision Learners (MAE) method for scalable vision learning is applied to the learning of the standard deviation and the mean of wavelet packet coefficients, opening up an innovative signal processing approach.

[0089] (4) MAE method This method innovatively utilizes the powerful representation learning ability of the masked autoencoder. By reconstructing the partially masked wavelet packet coefficients, the model can automatically learn the essential characteristics of the signal without labeling.

[0090] (5) The MAE method minimizes the mean absolute error in the reconstruction process. The MAE method not only improves the efficiency and accuracy of feature extraction, but also significantly enhances the model's sensitivity to signal changes.

[0091] (6) Noise reduction method based on complementary ensemble empirical mode decomposition (CEEM) and wavelet packet. Compared with the traditional wavelet denoising method, the CEEM combined with wavelet denoising algorithm can improve the signal denoising effect, effectively improve the signal quality and enhance the signal characteristics.

[0092] (7) A noise reduction method based on CEEM and wavelet packets is proposed. The original signal is preprocessed using wavelet packet noise reduction. Then, the CEEM algorithm is used to perform modal decomposition to obtain a series of modal components (IMF, Intrinsic Mode Function). Soft threshold noise reduction is performed on the noise-dominated mode, and adaptive threshold noise reduction is performed on the signal-dominated mode. Finally, the signal is reconstructed.

[0093] (8) The CEEM combined with wavelet denoising algorithm can simultaneously consider the time domain and frequency domain characteristics of the signal, which enables it to better extract the effective features of the signal while removing the noise components in the signal.

[0094] (9) Combining the variable step size algorithm of the error signal and the unsupervised learning of MAE can significantly improve the performance, robustness, and convergence speed of the algorithm and reduce the consumption of computing resources.

[0095] (10) Wavelet packet analysis can provide a multi-resolution representation of the signal, which helps to extract features that are more sensitive to damage from the dynamic response signal of the bridge. Combined with the unsupervised learning ability of MAE, it can further identify and locate damage in the bridge structure and improve the accuracy of damage detection.

[0096] The following is an exemplary description of the bridge fatigue life prediction device based on wavelet packet analysis provided in the present application.

[0097] like Figure 3 As shown, the embodiment of the present application provides a bridge fatigue life prediction device based on wavelet packet analysis, and the bridge fatigue life prediction device 300 based on wavelet packet analysis includes: The acquisition module 301 is used to acquire the deflection signal of the target bridge; the deflection signal is used to describe the deflection of the target bridge at T moments under the joint action of multiple effects, and the Tth moment is the current moment; An empirical mode decomposition module 302 is used to add white noise to the deflection signal to obtain a noisy deflection signal, and perform empirical mode decomposition on the noisy deflection signal to obtain a plurality of final modal components; The wavelet packet analysis module 303 is used to denoise and reconstruct all final modal components to obtain a reconstructed deflection signal, and perform wavelet packet analysis on the reconstructed deflection signal to obtain a deflection feature vector; The encoding and decoding module 304 is used to encode and decode the deflection feature vector to obtain the final deflection feature vector; The prediction module 305 is used to predict the fatigue life of the target bridge according to the final deflection characteristic vector to obtain the fatigue life prediction result of the target bridge; the fatigue life prediction result of the target bridge is used to describe the fatigue degree of the target bridge.

[0098] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of the present application. Their specific functions and technical effects can be found in the method embodiment part and will not be repeated here.

[0099] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0100] like Figure 4 As shown, an embodiment of the present application provides a terminal device. The terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 4 Only one processor is shown in the figure), a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 implements the steps of any of the above-mentioned method embodiments when executing the computer program D102.

[0101] Specifically, when the processor D100 executes the computer program D102, it obtains the deflection signal of the target bridge, then adds white noise to the deflection signal to obtain a noisy deflection signal, performs empirical mode decomposition on the noisy deflection signal to obtain a final modal component, then denoises and reconstructs the final modal component to obtain a reconstructed deflection signal, performs wavelet packet analysis on the reconstructed deflection signal to obtain a deflection eigenvector, then encodes and decodes the deflection eigenvector to obtain a final deflection eigenvector, and finally predicts the fatigue life of the target bridge based on the final deflection eigenvector to obtain a bridge fatigue life prediction result of the target bridge. Among them, adding white noise to the deflection signal and then denoising and reconstructing it can effectively filter out the noise part in the deflection signal, retain the useful components containing the fatigue characteristics of the bridge as much as possible, and improve the accuracy of the reconstructed deflection signal. Wavelet packet analysis is performed on the precise reconstructed deflection signal, which gives full play to the advantages of wavelet packet analysis in signal processing, effectively improves the accuracy of the deflection feature vector, and predicts the fatigue life of the bridge based on the accurate deflection feature vector, which can improve the accuracy of the bridge fatigue life prediction results.

[0102] The processor D100 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0103] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may also be an external storage device of the terminal device D10, such as a plug-in hard disk, a smart memory card (SMC, SmartMedia Card), a secure digital (SD, Secure Digital) card, a flash card (Flash Card), etc. equipped on the terminal device D10. Further, the memory D101 may also include both an internal storage unit of the terminal device D10 and an external storage device. The memory D101 is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of the computer program, etc. The memory D101 may also be used to temporarily store data that has been output or is to be output.

[0104] An embodiment of the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the above-mentioned method embodiments can be implemented.

[0105] An embodiment of the present application provides a computer program product. When the computer program product runs on a terminal device, the terminal device can implement the steps in the above-mentioned method embodiments when executing the computer program product.

[0106] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment method, which can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may at least include: any entity or device, recording medium, computer memory, read-only memory (ROM, Read-OnlyMemory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium that can carry the computer program code to the bridge fatigue life prediction method device / terminal device based on wavelet packet analysis. For example, a USB flash drive, a mobile hard disk, a disk or an optical disk.

[0107] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0108] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0109] The above is a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles described in the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A bridge fatigue life prediction method based on wavelet packet analysis, characterized in that: include: Obtaining the deflection signal of the target bridge; The deflection signal is used to describe the deflection of the target bridge at T moments under the joint action of multiple effects, and the Tth moment is the current moment; Adding white noise to the deflection signal to obtain a noise deflection signal, and performing empirical mode decomposition on the noise deflection signal to obtain a plurality of final modal components; Denoising and reconstructing all final modal components to obtain a reconstructed deflection signal, and performing wavelet packet analysis on the reconstructed deflection signal to obtain a deflection eigenvector; Encoding and decoding the deflection feature vector to obtain a final deflection feature vector; The bridge fatigue life of the target bridge is predicted according to the final deflection eigenvector to obtain a bridge fatigue life prediction result of the target bridge; the bridge fatigue life prediction result is used to describe the fatigue degree of the target bridge.

2. The bridge fatigue life prediction method according to claim 1 is characterized in that: The step of adding white noise to the deflection signal to obtain a noise deflection signal comprises: Calculating the standard deviation of the deflection signal in multiple time windows; An adaptive white noise is generated based on the standard deviation, and the adaptive white noise is added to the deflection signal to obtain a noise deflection signal.

3. The bridge fatigue life prediction method according to claim 2 is characterized in that: The step of calculating the standard deviation of the deflection signal in multiple time windows comprises: By the formula: Solve for the standard deviation ; in, represents the local mean, represents the time window, , Indicates the deflection signal the deflection of a historical moment; The step of generating adaptive white noise based on the standard deviation comprises: By formula: Generate adaptive white noise ; in, represents the amplitude, represents the noise frequency, Indicates Gaussian white noise, represents the basic noise frequency, Indicates the amount of change, represents the change threshold, represents the deflection signal, Indicates the previous deflection signal, represents the basic noise amplitude; The step of adding the adaptive white noise to the deflection signal to obtain a noise deflection signal comprises: By formula: Calculate the noise deflection signal .

4. The bridge fatigue life prediction method according to claim 3 is characterized in that: The step of performing empirical mode decomposition on the noise deflection signal to obtain a plurality of final modal components includes: By formula: Calculate the Final modal components ; in, Indicates modal components, represents the empirical mode decomposition, represents the number of modal components.

5. The bridge fatigue life prediction method according to claim 1, characterized in that: The denoising and reconstruction of all final modal components to obtain a reconstructed deflection signal includes: Perform wavelet packet decomposition on all final modal components to obtain multiple denoised wavelet packet coefficients; Performing soft threshold processing on each denoising wavelet packet coefficient to obtain a final denoising wavelet packet coefficient corresponding to each denoising wavelet packet coefficient; All final denoised wavelet packet coefficients are subjected to inverse wavelet packet transform to obtain the reconstructed deflection signal.

6. The bridge fatigue life prediction method according to claim 5, characterized in that: The step of performing soft threshold processing on each denoising wavelet packet coefficient to obtain a final denoising wavelet packet coefficient corresponding to each denoising wavelet packet coefficient includes: By formula: Calculate the final denoised wavelet packet coefficients ; in, represents the denoised wavelet packet coefficients The symbol function of represents the threshold value, represents the noise standard deviation, Indicates the length of the signal.

7. The bridge fatigue life prediction method according to claim 1, characterized in that: The step of performing wavelet packet analysis on the reconstructed deflection signal to obtain a deflection feature vector includes: The number of iterations is increased by 1, and the reconstructed deflection signal is subjected to multi-layer wavelet packet decomposition according to the step size factor to obtain multiple signal nodes; Determine a target signal node from all signal nodes, and calculate the standard deviation and average value of the wavelet packet coefficients of the target signal node; Combining the standard deviation of the wavelet packet coefficients and the average value of the wavelet packet coefficients to obtain an initial deflection eigenvector; Calculating an error signal based on the initial deflection eigenvector, and updating the step factor to obtain an updated step factor; Performing power detection on the error signal to obtain the power of the error signal, and determining whether a difference between the power of the error signal and the power of the error signal of the previous iteration is greater than a power difference threshold; If yes, the updated step size factor is used as the step size factor, the number of iterations plus 1 is returned, and the reconstructed deflection signal is subjected to multi-layer wavelet packet decomposition according to the step size factor to obtain a plurality of signal nodes; Otherwise, the initial deflection feature vector is used as the deflection feature vector.

8. The bridge fatigue life prediction method according to claim 7 is characterized in that: The calculating of the error signal based on the initial deflection eigenvector comprises: By formula: Calculate the error signal ; in, Indicates the expected signal, represents the output signal after the filter processes the initial deflection eigenvector; The updating of the step size factor to obtain an updated step size factor includes: By formula: Calculate the The step size factor for the iteration ; in, Indicates The step size factor for the iteration, and All represent parameters. represents the mask ratio, represents the standard deviation of the wavelet packet coefficients, represents the average value of the wavelet packet coefficients, Represents a positive number, , Indicates the maximum number of iterations.

9. The bridge fatigue life prediction method according to claim 1, characterized in that: The encoding and decoding of the deflection feature vector to obtain a final deflection feature vector includes: generating a mask matrix of the deflection feature vector; The deflection feature vector is encoded according to the mask matrix to obtain a coding vector, and the coding vector is decoded to obtain a final deflection feature vector.

10. The bridge fatigue life prediction method according to claim 9, characterized in that: The mask matrix for generating the deflection feature vector comprises: By formula: Calculate the mask matrix Line Column mask ; in, represents a random number uniformly distributed in the range [0, 1]. represents the mask probability, and All represent hyperparameters, express The expected value of express The fluctuation range of represents the density threshold, represents the strong signal threshold, Indicates the weak signal threshold, represents the range constant, , represents the index of the last row in the mask matrix, , Represents the index of the last column in the mask matrix.

Images