Classification method and system for detecting anchoring quality grade of anchor rod through ultrasonic guided wave

Abstract

The invention discloses an anchor rod anchoring quality grade classification method and system through ultrasonic guided wave detection. The method comprises the steps that A, ultrasonic guided wave defect signals are obtained through an ultrasonic guided wave detection platform; b, performing modal decomposition on the ultrasonic guided wave defect signal by using an HO-VMD algorithm, and performing signal noise reduction processing by taking PCC-E as a fitness function to obtain a noise-reduced ultrasonic guided wave characteristic signal; c, constructing a data set by using the de-noised ultrasonic guided wave characteristic signals and grade labeling data; d, training the constructed anchor rod anchoring quality grade classification model by using the data set to obtain a trained anchor rod anchoring quality grade classification model; and E, judging the input to-be-classified data by using the trained anchor rod anchoring quality grade classification model, and finally obtaining an anchor rod anchoring quality classification grade corresponding to the to-be-classified data. According to the method, accurate and efficient classification of anchoring quality grades in a complex environment can be realized based on an intelligent classification method fusing adaptive signal decomposition and depth time sequence modeling.

Description

Technical Field

[0001] The present invention relates to the field of non-destructive testing of anchor bolt anchoring quality, and in particular to a method and system for classifying anchor bolt anchoring quality grades using ultrasonic guided wave testing based on deep learning. Background Art

[0002] In large-scale geotechnical engineering projects such as slopes, bridges, and tunnels, anchor bolting technology is an important flexible reinforcement and key support method, and its quality directly affects the stability of the structure. Traditional anchor bolting quality inspection methods are mainly divided into two categories: destructive inspection and non-destructive inspection. Destructive inspection evaluates the pull-out bearing capacity of anchor bolts by analyzing the load-displacement curve, but the operation is complicated and inefficient. It is only suitable for small-scale spot checks and may damage the structure. Non-destructive testing is represented by ultrasonic guided wave technology, which uses the propagation characteristics of high-frequency sound waves in anchor bolts to infer internal defects. It has the advantages of being non-invasive and capable of batch inspection. However, in complex engineering environments, it still faces problems such as strong signal noise interference, insufficient feature extraction, and insufficient classification accuracy. Intelligent improvements are urgently needed to improve detection reliability.

[0003] In summary, existing technologies fail to effectively address the high noise interference, feature aliasing, and insufficient robustness of classification models associated with ultrasonic guided wave signals, limiting their applicability in engineering applications for anchor bolt quality inspection. Therefore, an intelligent classification method combining adaptive signal decomposition and deep time series modeling is urgently needed to accurately and efficiently determine anchor bolt quality in complex environments. Summary of the Invention

[0004] The purpose of the present invention is to provide a method and system for classifying anchor bolt anchoring quality grades under ultrasonic guided wave detection, which can achieve accurate and efficient classification of anchoring quality grades in complex environments based on an intelligent classification method that integrates adaptive signal decomposition and deep time series modeling.

[0005] The present invention adopts the following technical solutions:

[0006] A method for classifying anchor bolt anchoring quality by ultrasonic guided wave detection comprises the following steps:

[0007] A: Use the ultrasonic guided wave testing platform to perform ultrasonic guided wave testing on the anchor body containing defects to obtain the ultrasonic guided wave defect signal;

[0008] B: Use the HO-VMD algorithm to perform modal decomposition on the ultrasonic guided wave defect signal, use PCC-E as the fitness function, perform signal noise reduction, and obtain the ultrasonic guided wave characteristic signal after noise reduction;

[0009] The HO-VMD algorithm is a VMD algorithm optimized based on the HO algorithm. The HO algorithm represents the Hippo optimization algorithm, and the VMD algorithm represents the variational mode decomposition algorithm. PCC represents the Pearson correlation coefficient, and E represents the power spectrum entropy.

[0010] C: Construct a dataset using the noise-reduced ultrasonic guided wave feature signals and graded labeled data;

[0011] D: Using the data set to train the constructed anchor bolt anchoring quality grade classification model to obtain the trained anchor bolt anchoring quality grade classification model;

[0012] E: Use the trained anchor bolt anchoring quality grade classification model to judge the input data to be classified, and finally obtain the anchor bolt anchoring quality classification grade corresponding to the data to be classified.

[0013] Step B includes the following steps:

[0014] B1: Set the number of iterations of the HO algorithm, initialize the ratio of predators to joiners, and the modal layer number K and penalty factor α of the VMD algorithm;

[0015] B2: Initialize the population size of the HO algorithm with the Logistic-Tent chaotic map to obtain the VMD algorithm optimized by the HO algorithm, namely the HO-VMD algorithm;

[0016] B3: Construct PCC-E as the fitness function of the HO algorithm;

[0017] B4: Use HO-VMD algorithm to reduce noise of ultrasonic guided wave defect signals;

[0018] B5: Using the HO-VMD algorithm and combining it with the fitness function, the optimal modal layer number K and penalty factor α are obtained through iterative updates, and the denoised ultrasonic guided wave characteristic signal is output.

[0019] In step B1, the VMD algorithm is used to decompose the ultrasonic guided wave defect signal into K decomposition layers through adaptive quasi-orthogonal transformation. The ultrasonic guided wave defect signal is composed of the sum of K natural mode components of different frequency bands.

[0020] In step B3, the optimization process of PCC-E is as follows:

[0021] a. Calculate the Pearson correlation coefficient between each intrinsic modal component and the original signal, record it as PCC, and calculate the Pearson correlation coefficient between the intrinsic modal component and the next extant intrinsic modal component, record it as PCC′;

[0022] b. Calculate the correlation coefficient ratio of each intrinsic modal component R = PCC / PCC′;

[0023] c. Set the correlation coefficient ratio threshold G to filter out the intrinsic modal components whose correlation coefficient ratio R is greater than the correlation coefficient ratio threshold G;

[0024] d. Calculate the power spectrum entropy E of the filtered intrinsic modal components P ;

[0025] e. Calculate the output value of the fitness function fitness = minE P .

[0026] In step D, the anchor bolt anchoring quality grade classification model consists of a CNN model and a BiLSTM model; the CNN model performs convolution and pooling operations on the denoised ultrasonic guided wave characteristic signal to extract the potential features of the data, and the BiLSTM model is used to extract the local features and long-term regularities of the denoised ultrasonic guided wave characteristic signal, and performs ultrasonic guided wave signal detection on anchor bolts with different anchoring qualities according to the anchoring density of the anchoring section, and then enhances the data by adding interference to the original signal; finally, the output of the BiLSTM is classified by a fully connected neural network classifier to obtain the prediction result of the anchor bolt anchor body quality grade type.

[0027] The ultrasonic guided wave detection platform includes an ultrasonic guided wave signal excitation device, a signal acquisition device and an anchor bolt anchor body; the ultrasonic guided wave signal excitation device is used to transmit an ultrasonic guided wave signal modulated by a sine wave to the anchor bolt anchor body; the signal acquisition device is used to receive the ultrasonic guided wave defect signal; the anchor bolt anchor body is used to simulate resin anchor bolts of different quality grades.

[0028] The ultrasonic guided wave signal excitation device is composed of a signal generator, a power amplifier and an ultrasonic generating probe which are electrically connected in sequence; the signal acquisition device is composed of an ultrasonic receiving probe and a digital oscilloscope which are electrically connected in sequence; the anchor bolt anchor body adopts a mining threaded steel resin anchor bolt.

[0029] Step A includes the following steps:

[0030] A1: Build the ultrasonic guided wave signal excitation device and signal acquisition device, and apply coupling agent between the ultrasonic probe and the anchor body to be tested;

[0031] A2: Turn on the signal generator and power amplifier, and set the parameters of the excitation signal;

[0032] A3: Turn on the digital oscilloscope and use the ultrasonic receiving probe to collect the excitation signal of the signal transmitter and the ultrasonic guided wave defect signal;

[0033] A4: Import the waveform signal data saved by the digital oscilloscope into the processing module.

[0034] Ultrasonic guided wave testing anchor bolt anchoring quality classification system, including ultrasonic guided wave testing platform and processing module;

[0035] Ultrasonic guided wave detection platform, used to obtain ultrasonic guided wave defect signals of different defect locations and sizes;

[0036] The processing module is used to perform modal decomposition on the ultrasonic guided wave defect signal using the HO-VMD algorithm, perform signal denoising processing using PCC-E as the fitness function, and obtain the denoised ultrasonic guided wave characteristic signal. Then, the trained anchor bolt anchoring quality grade classification model is used to judge the input data to be classified based on the mapping relationship between the ultrasonic guided wave signal characteristic data of different anchoring defects and the anchor bolt anchoring quality grade, and finally obtain the anchor bolt anchoring quality classification grade corresponding to the data to be classified.

[0037] The ultrasonic guided wave detection platform includes an ultrasonic guided wave signal excitation device, a signal acquisition device and an anchor bolt anchor body; the ultrasonic guided wave signal excitation device is used to transmit an ultrasonic guided wave signal modulated by a sine wave to the anchor bolt anchor body; the signal acquisition device is used to receive the ultrasonic guided wave defect signal; the anchor bolt anchor body is used to simulate resin anchor bolts of different quality grades.

[0038] The present invention performs ultrasonic guided wave detection on an anchor body containing defects through an ultrasonic guided wave detection platform to obtain an ultrasonic guided wave signal containing defect characteristics, retains key classification information for determining the anchor quality grade of the anchor, then uses a variational mode decomposition (VMD) algorithm optimized by a Hippo optimization algorithm (HO) to perform modal decomposition on the ultrasonic guided wave signal containing defect characteristics to obtain a noise-reduced ultrasonic guided wave characteristic signal, then combines the noise-reduced ultrasonic guided wave signal characteristic with the anchor quality grade, and finally learns the mapping relationship between the ultrasonic guided wave characteristic signal and the anchor quality grade through CNN-BiLSTM network training, thereby effectively improving the accuracy of anchor quality grade determination, which has important application value for the determination of anchor quality grade. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the process of the grade classification method of the present invention;

[0040] Figure 2 Schematic diagram of the time domain of the natural modal component in the present invention;

[0041] Figure 3 Schematic diagram of the frequency domain of the natural modal component in the present invention;

[0042] Figure 4 Schematic diagram of the ultrasonic guided wave characteristic signal after noise reduction in the present invention;

[0043] Figure 5 Schematic diagram of the composition structure of the hierarchical classification system in the present invention. DETAILED DESCRIPTION

[0044] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.

[0045] like Figure 1 As shown, the method for classifying anchor bolt anchoring quality levels by ultrasonic guided wave detection according to the present invention comprises the following steps:

[0046] A: Use the ultrasonic guided wave testing platform to perform ultrasonic guided wave testing on the anchor body containing defects to obtain ultrasonic guided wave defect signals containing defect characteristics;

[0047] In step A, an ultrasonic guided wave detection platform is built to collect ultrasonic guided wave defect signals of different defect positions and sizes, so that the differences in ultrasonic guided waves can be analyzed in subsequent steps to complete the classification of the stability state of the anchor rod and anchor body.

[0048] In this embodiment, the ultrasonic guided wave detection platform includes an ultrasonic guided wave signal excitation device, a signal acquisition device, and an anchor bolt anchor body, wherein:

[0049] The ultrasonic guided wave signal excitation device is used to transmit the ultrasonic guided wave signal modulated by the sine wave to the anchor body of the anchor rod; the ultrasonic guided wave signal excitation device is composed of a signal generator, a power amplifier and an ultrasonic generating probe which are electrically connected in sequence;

[0050] A signal acquisition device for receiving ultrasonic guided wave defect signals; the signal acquisition device is composed of an ultrasonic receiving probe and a digital oscilloscope electrically connected in sequence;

[0051] Anchor rod anchor body is used to simulate resin anchor rods of different quality grades by controlling different anchoring positions of resin anchoring agent.

[0052] In this embodiment, the anchor bolt anchor body can be a mining threaded steel resin anchor bolt sample, in which concrete is used to simulate the surrounding rock, and resin mastic is used to simulate the anchoring agent;

[0053] In this embodiment, the electrical connection between the components in the ultrasonic guided wave signal excitation device and the signal acquisition device can be achieved by connecting with dual BNC signal source lines.

[0054] In step A, the steps for obtaining an ultrasonic guided wave defect signal containing defect characteristics using an ultrasonic guided wave detection platform are as follows:

[0055] A1: Build an ultrasonic guided wave signal excitation device and a signal acquisition device, and apply coupling agent between the ultrasonic probe and the anchor body to be tested to improve the transmission efficiency of the ultrasonic wave and reduce interference from air or other media;

[0056] A2: Turn on the signal generator and power amplifier, and set the parameters of the required excitation signal;

[0057] A3: Turn on the digital oscilloscope and use the ultrasonic receiving probe to collect the excitation signal of the signal transmitter and the ultrasonic guided wave defect signal;

[0058] A4: Import the waveform signal data saved by the digital oscilloscope into the computer for further analysis in subsequent steps.

[0059] B: The variational mode decomposition (VMD) algorithm optimized by the Hippo optimization algorithm (HO) is used to perform modal decomposition on the ultrasonic guided wave defect signal containing defect characteristics. The Pearson correlation coefficient is used as the fitness function, and the signal components are selected for signal denoising to obtain the denoised ultrasonic guided wave characteristic signal.

[0060] In step B, the ultrasonic guided wave defect signal containing defect characteristics is modally decomposed by the HO-VMD algorithm optimized by the Hippo optimization algorithm (HO) to obtain the denoised ultrasonic guided wave feature signal, thereby increasing the accuracy of classification.

[0061] Since the ultrasonic guided wave signals collected on-site are subject to multiple interferences (dispersion effects, boundary reflections, and electromagnetic noise), the defect features and noise are highly aliased in the time-frequency domain; traditional methods such as threshold filtering or Fourier transform cannot distinguish between real defect echoes and false signals in the same frequency band, which directly reduces the reliability of the classification model. Although empirical mode decomposition (EMD) has been attempted for feature extraction, its heuristic recursive screening mechanism has fundamental limitations: when the instantaneous frequency of the signal suddenly changes (such as defect interface reflection), the high-frequency component is engulfed in the low-frequency IMF, generating false modes. Pulse interference will distort the envelope, causing an "error transfer effect", and the decomposition process has no orthogonality guarantee, resulting in characteristic energy leakage. Variational mode decomposition (VMD) ensures the orthogonality of the IMF frequency band by constraining the variational model, thereby avoiding modal aliasing. However, when it is applied to ultrasonic guided wave defect signals for anchor detection, the following problems exist:

[0062] (1) Slight deviations in the number of modal layers K and the penalty factor α can lead to decomposition failure.

[0063] (2) The combined search space of (K,α) reaches 10 4 The computational cost of traditional grid search is too high.

[0064] (3) Conventional indicators (such as envelope entropy) cannot distinguish between defect components and residual noise.

[0065] To solve the above problems, the present invention makes full use of the coupling relationship between the propagation mechanism of ultrasonic guided waves and optimization mathematics, creatively integrates chaos theory, information entropy and correlation analysis across disciplines, adopts an improved Hippo algorithm global search (K, α), introduces the Logistic-Tent chaos mapping to break the premature convergence, and designs a new fitness function to retain only the components that are strongly correlated with the original signal and have a sparse spectrum, thereby solving the above-mentioned defects in the prior art.

[0066] In the present invention, the step B comprises the following specific steps:

[0067] B1: Set the number of iterations of the HO algorithm, initialize the ratio of predators to joiners, and the parameter combination (K, α) of the VMD algorithm;

[0068] In this embodiment, the population size of the HO algorithm is set to 30 and the number of iterations is set to 40;

[0069] In the present invention, the VMD algorithm is first used to decompose the composite signal s(t), i.e., the ultrasonic guided wave defect signal, into K decomposition layers, i.e., the inherent modes (IMFs) of different frequency bands, through adaptive quasi-orthogonal transformation. The composite signal s(t) consists of K inherent mode components m of different frequency bands. k (t) is added together. Each natural modal component m k (t) represents an amplitude A that changes with time t k (t) and phase The modulated cosine wave has the following mathematical model:

[0070]

[0071] Where m k (t) is the k-th layer intrinsic modal component, k = (1,…,K), A k () is the amplitude modulation function, A k (t) represents m k The instantaneous amplitude of (t), is the phase function, Represents m k The instantaneous phase of (t), represents the instantaneous frequency;

[0072] Since each natural modal component m k The spectrum of (t) is concentrated at its center frequency ω k (t), the bandwidth can be estimated by using Gaussian smoothed signals to make the obtained components sparse. Therefore, VMD can express the signal decomposition problem as a constrained variation problem:

[0073] In order to make each natural mode component concentrate only on a specific frequency band, the spectrum of each natural mode component is made at its center frequency ω k As much as possible, the bandwidth of each natural mode component is as narrow as possible. Therefore, for each natural mode component m k (t), first through the Hilbert transform kernel Converted to an analytical signal and then multiplied by Implement the frequency shift operation to shift the center frequency of the modal signal to zero frequency, then calculate the time derivative of the result to approximately measure its bandwidth, and finally square and sum the bandwidth of all modes to minimize it as the objective function, that is,

[0074]

[0075] in, Represents minimizing the objective function, adjusting the modal function and center frequency; represents the Hilbert transform kernel; δ(t) represents the Dirac delta function; Indicates moving the signal to the center frequency; represents the L2 norm squared, represents the partial derivative with respect to time t, j represents the imaginary unit;

[0076] The above objective is a constrained variational problem, because in addition to minimizing the bandwidth of each intrinsic mode component, it is also necessary to ensure that all eigenmode components combined can reconstruct the original signal s(t). Therefore, the VMD method described in this invention introduces an augmented Lagrangian function, which transforms the original constrained problem into an unconstrained optimization problem. This function is composed of the sum of the bandwidths of the eigenmode components, the reconstruction error, and the constraint penalty term award, and is expressed as:

[0077]

[0078] Where L represents the augmented Lagrangian function; {m k} represents the set of intrinsic modal components {m1(t), m2(t), ..., m K (t)}; {ω k} represents the set of mode center frequencies {ω1(t), ω2(t), …, ω K (t)}; λ represents the Lagrange multiplier; α is the penalty coefficient; the superscript s represents the exponent of the analytical signal operator; λ(t) represents the time-varying Lagrange multiplier;

[0079] The alternating direction multiplier algorithm is then used to iteratively find the optimal solution to the constrained variational model. When the sum of the relative changes in all intrinsic modal components between two consecutive iterations is less than the set error precision threshold ε, the algorithm has converged and the decomposition is complete.

[0080] In summary, the effectiveness of signal decomposition is determined by the total number of decomposition layers, K, and the penalty coefficient, α, in variational mode decomposition (VMD). Setting K too small may result in insufficient signal decomposition, leading to modal aliasing. Conversely, setting K too large may lead to over-decomposition of the signal, causing modal loss. For the penalty coefficient, α, if its value is too large, the bandwidth of the eigenmodal components will be reduced, leading to information loss. However, if α is too small, the bandwidth of the modal components may be increased, resulting in overlapping frequency centers and modal aliasing. Therefore, the proper selection and optimization of these two parameters is crucial to the decomposition effect of VMD. Other parameters, such as the time-varying Lagrange multiplier λ(t) and the error precision threshold ε, have relatively little impact on the decomposition results and can therefore be set to fixed initial values.

[0081] B2: Initialize the population size of the HO algorithm using the Logistic-Tent chaotic map.

[0082] In this embodiment, the Hippopotamus Optimization (HO) algorithm is a swarm-based optimization algorithm that simulates the position updates, defense strategies, and avoidance methods of hippos in rivers or ponds. The specific process of the HO algorithm is as follows:

[0083] Step 1: Construct a randomly initialized population;

[0084] Step 2: Update the position of male hippos in the population;

[0085] Step 3: Update the position of the predator in the population;

[0086] Step 4: When the population position update encounters a local trap or boundary, the population will try to leave this area and enter a random location near the current position;

[0087] The specific process of the above HO algorithm is conventional technology in this field and will not be repeated here;

[0088] After each iteration of the HO algorithm, the positions of individual population members are updated according to Steps 2 to 4, continuing until the final iteration. During the algorithm's execution, the dominant solution is continuously searched for and stored. After the algorithm completes, the last dominant solution is considered the final solution. However, like all metaheuristic algorithms, the initial constraint of the HO algorithm is that it cannot guarantee a global optimum due to the random search process. Therefore, the present invention improves the HO algorithm to address the specific characteristics of the problem.

[0089] In this embodiment, the Logistic-Tent chaotic mapping is a method for improving the population position of the HO algorithm. In the HO algorithm, due to its random search process, individuals may easily fall into local optimality. The Logistic-Tent chaotic mapping system combines the complex dynamic characteristics of the Logistic mapping with the higher iteration speed and stronger autocorrelation of the Tent mapping. The sequence generated by this chaotic mapping exhibits excellent randomness and ergodicity. During the algorithm's iterative process, by using this chaotic sequence to update the individual position, nonlinear and irregular movement of individuals in the search space can be achieved, thereby enhancing the algorithm's exploration ability and effectively avoiding early convergence.

[0090] The Logistic-Tent chaotic map expression is as follows:

[0091]

[0092] Among them, χ i represents the chaotic state value at the current moment (i-th step); i+1 represents the chaotic state value at the next moment (i+1 step); r represents the control parameter of the chaotic system; mod 1 represents the modulo operation, and the modulus is 1;

[0093] B3: Construct the Pearson correlation coefficient-power spectrum entropy (PCC-E) and use it as the fitness function of the HO algorithm;

[0094] In this embodiment, the Pearson correlation coefficient-power spectrum entropy (PCC-E) is a new comprehensive objective function created in the present invention. The core of the VMD algorithm is to determine the number of decomposition layers K and the penalty factor α. The parameter combination K, α directly affects the decomposition effect of the ultrasonic signal, and thus determines the effect of signal denoising. The choice of fitness function plays a decisive role in the process of optimizing signal decomposition. Power spectral entropy (PSE) can effectively describe the spectral characteristics of the signal, especially the randomness of its frequency distribution. When the spectrum of the signal is more concentrated, the value of the power spectral entropy tends to be extremely small, and the calculation formula of its power spectral density estimate is as follows:

[0095]

[0096] Among them, f n represents the angular frequency, and the subscript n represents the index of the frequency component; e(f n ) represents the frequency point f n The power spectrum density estimate in the angular frequency domain at ; L is the signal data length; x(ω) is the discrete Fourier transform of the sample signal;

[0097] Normalize the power spectrum to obtain the power spectrum probability density function P n , the calculation formula is:

[0098]

[0099] Among them, P n Indicates the nth f n The normalized power of ; N represents the total number of frequency components.

[0100] Then the power spectrum entropy E P The calculation formula is:

[0101]

[0102] Among them, lnP n Represents the probability P n Take the natural logarithm.

[0103] In practical applications, simply using power spectrum entropy as a fitness function often results in the decomposed signal tending towards a simple harmonic form. To address this problem, the present invention also introduces the Pearson correlation coefficient (PCC) as a supplementary fitness function.

[0104] The Pearson correlation coefficient (PCC) is used to measure the degree of linear correlation between two variables, and its value range is [-1, +1]. When the PCC is negative, it means that as one variable increases, the other variable tends to decrease; when the PCC is positive, it means that both variables increase together.

[0105] Based on these two objective functions, power spectrum entropy and Pearson correlation coefficient, the present invention proposes a new comprehensive objective function as a fitness function, namely Pearson correlation coefficient-power spectrum entropy (PCC-E). The optimization process using the above fitness function is as follows:

[0106] a. Calculate the Pearson correlation coefficient PCC between each intrinsic modal component and the original signal, and calculate the Pearson correlation coefficient PCC′ between the intrinsic modal component and the next adjacent intrinsic modal component, such as the Pearson correlation coefficient between IMF1 and IMF2, and between IMF2 and IMF3;

[0107] b. Calculate the correlation coefficient ratio R = PCC / PCC′ for each intrinsic modal component. For example, when calculating intrinsic modal component IMF1, PCC is the Pearson correlation coefficient between intrinsic modal component IMF1 and the original signal, and PCC′ is the Pearson correlation coefficient between intrinsic modal component IMF1 and intrinsic modal component IMF2. When calculating intrinsic modal component IMF2, PCC is the Pearson correlation coefficient between intrinsic modal component IMF2 and the original signal, and PCC′ is the Pearson correlation coefficient between intrinsic modal component IMF2 and intrinsic modal component IMF3.

[0108] c. Set the correlation coefficient ratio threshold G = 10 to filter out the natural mode components whose correlation coefficient ratio R is greater than the correlation coefficient ratio threshold G;

[0109] By setting the correlation coefficient ratio threshold G = 10, if the correlation between a certain IMF and the original signal is significantly higher than the correlation between adjacent IMFs (that is, higher than the set correlation coefficient ratio threshold), then the component can be considered more "effective" and has a higher correlation with the original signal;

[0110] d. Calculate the power spectrum entropy E of the filtered intrinsic modal components P , to quantify the spectral characteristics of the components.

[0111] e. Calculate the output value of the fitness function fitness = minE P ,R>10.

[0112] B4: Use HO-VMD algorithm to reduce noise of ultrasonic guided wave defect signals;

[0113] Since ultrasonic guided wave signals are often affected by noise in actual detection, resulting in a decrease in signal quality and affecting the accuracy of anchor bolt anchoring quality grade identification, in the present invention, the HO-VMD algorithm is used to reduce the noise of ultrasonic guided wave defect signals and normalize the signal amplitude.

[0114] B5: Using the HO-VMD algorithm and the fitness function, the optimal parameter combination (K, α) is obtained through iterative updates, and the noise-reduced ultrasonic guided wave characteristic signal containing the defect characteristics is output;

[0115] In the present invention, by iteratively updating the fitness value and selecting the optimal position, the optimal parameter combination (K, α) is found after the algorithm termination condition is met, and the ultrasonic guided wave characteristic signal containing the defect characteristics after noise reduction is output.

[0116] In this embodiment, the Pearson correlation coefficient-power spectrum entropy is used as the fitness function, the population size is set to 30, the number of iterations is set to 40, and finally K=8 and α=963.24 are determined by the variational mode decomposition algorithm (VMD) optimized by the Hippo optimization algorithm (HO). Then, the decomposed intrinsic mode components are transformed by FFT to obtain the time domain and frequency domain diagrams of the intrinsic mode components, as shown in the figure. Figure 2 and Figure 3 shown.

[0117] In the process of signal decomposition, the Pearson correlation coefficient between each intrinsic modal component and the original signal is calculated, and the intrinsic modal component with a high correlation with the original signal is selected for signal reconstruction, and finally the ultrasonic guided wave characteristic signal after noise reduction is obtained, such as Figure 4 shown.

[0118] In the present invention, after the VMD denoising in step B, the noise signal in the original signal can be effectively removed, so that the detection signal and the normal signal can be accurately extracted, which can increase the accuracy of classification.

[0119] C: Combine the de-noised ultrasonic guided wave characteristic signal with the graded labeled data to form a dataset, and divide the dataset into a training set and a test set according to a set ratio. At the same time, divide each graded data into the two datasets in proportion to reduce the difference between the datasets.

[0120] In this example, anchor bolt anchoring quality is graded into Class I, Class II, Class III, and Class IV based on the density of the anchoring section. Ultrasonic guided wave signal detection is performed on anchor bolts of varying anchoring quality. Data enhancement is also performed by adding Gaussian white noise and impulse noise to the original signal to simulate random interference and electromagnetic interference that may occur under different experimental conditions. Data enhancement can effectively improve model robustness and generalization capabilities. In this example, a total of 1,107 data samples were collected, with the training and test sets divided into a ratio of 7:3.

[0121] D: Input the training set into the anchor bolt anchoring quality grade classification model constructed by the CNN-BiLSTM network for training. This allows the anchor bolt anchoring quality grade classification model to learn the weights of the influence of the ultrasonic guided wave signal feature data of different anchoring defects on the grade classification results. Finally, the trained anchor bolt anchoring quality grade classification model is obtained.

[0122] In the present invention, the signal obtained after signal decomposition and noise reduction by the HO-VMD algorithm is first fed into a one-dimensional convolutional neural network model (1D-CNN). After convolution and pooling operations, the potential features of the data (including local pulse amplitude, pulse width, oscillation decay pattern and other features) are further extracted.

[0123] The BiLSTM model is then used to mine the local features (including defect echo time delay, intervals between adjacent reflected pulses) and long-term patterns (including guided wave attenuation trends and dispersion curve offsets) of the de-noised ultrasonic signal data. Ultrasonic guided wave signal detection is performed on anchor bolts of varying anchoring qualities, based on the density of the anchoring section (in this example, anchor bolt anchoring quality is categorized as Class I, Class II, Class III, and Class IV). Data enhancement is performed by adding Gaussian white noise and pulse noise to the original signal to simulate random interference and electromagnetic interference that may occur under different experimental conditions. Data enhancement can effectively improve model robustness and generalization capabilities. Finally, the BiLSTM output is connected to a fully connected neural network classifier for classification, ultimately resulting in a prediction of the anchor bolt's anchor quality grade type.

[0124] In this embodiment, the one-dimensional convolutional neural network (1D-CNN) includes a convolution layer and a pooling layer. The convolution kernel size of the convolution layer is 1×3, the number of convolution kernels is 64, and relu is used as the activation function; the pooling layer adopts the maximum pooling method, and performs maximum pooling with a step size of 2 units.

[0125] The BiLSTM network includes a layer of BiLSTM, where the number of LSTM model units is 32 and the relu function is used for activation.

[0126] When performing multi-classification tasks, this paper uses four core metrics as model evaluation criteria: accuracy, precision, recall, and F1 score. These metrics can comprehensively and accurately understand and evaluate the performance of the classification model, providing a basis for overall model evaluation.

[0127]

[0128] Where Acc is the accuracy rate, Re is the recall rate, Pre is the precision rate, F1 is the F1 score, TP is the number of samples correctly predicted as positive, TN is the number of samples correctly predicted as negative, FP is the number of samples incorrectly predicted as positive, and FN is the number of samples incorrectly predicted as negative.

[0129] In order to further verify the classification performance of the HO-VMD-CNN-BiLSTM anchor body quality grade proposed in the present invention and to more comprehensively evaluate the classification performance of the improved model, the present invention carried out a comparative analysis of multiple evaluation indicators, including CNN, SVM and BP neural networks. To ensure the effectiveness of the comparative test, the CNN structure of each test model is the same, and the parameters such as the number of model training, learning rate, regularization parameter and learning rate adjustment factor are all set to the same standard.

[0130] Data shows that the proposed HO-VMD-CNN-BiLSTM classification model demonstrates excellent performance across various evaluation metrics. Specifically, the model achieves 96.78% accuracy, 94.79% precision, 93.47% recall, and an F1 score of 0.94.

[0131] Compared with the traditional deep learning CNN model, the accuracy, precision, recall and F1 score of the HO-VMD-CNN-BiLSTM model increased by 2.17%, 3.26%, 4.44% and 3.29% respectively;

[0132] Compared with the traditional machine learning SVM model, the accuracy, precision, recall and F1 score of the HO-VMD-CNN-BiLSTM model increased by 6.82%, 4.40%, 9.30% and 6.82% respectively;

[0133] Compared with the classic BP neural network model, the accuracy, precision, recall and F1 score of the HO-VMD-CNN-BiLSTM model increased by 9.30%, 7.80%, 11.90% and 10.59% respectively;

[0134] The HO-VMD-CNN-BiLSTM model employed in this paper combines the feature extraction capabilities of convolutional neural networks (CNNs) with the sequence modeling capabilities of bidirectional long short-term memory networks (BiLSTMs), enabling it to better capture the temporal characteristics and contextual information of time series data. This model significantly outperforms traditional CNN, SVM, and BP neural network models across all comparison metrics, particularly in recall and F1 scores.

[0135] E: Using the trained anchor bolt anchoring quality grade classification model, the input data to be classified is judged according to the mapping relationship between the ultrasonic guided wave signal feature data of different anchoring defects and the anchor bolt anchoring quality grade, and finally the anchor bolt anchoring quality classification grade corresponding to the data to be classified is obtained.

[0136] When using the ultrasonic guided wave detection anchor bolt anchoring quality grade classification method described in the present invention to perform ultrasonic guided wave detection, first ensure the safety of the on-site environment and conduct comprehensive tests on all equipment to verify their normal operation. Subsequently, the ultrasonic guided wave sensor is arranged according to the on-site conditions, and then the signal excitation device is started to transmit the signal, and the signal acquisition device is used to receive the reflected signal. Finally, the HO-VMD algorithm is used to reduce the noise of the signal, extract clearer signal features, and input these features into the trained anchor bolt anchoring quality grade classification model. The anchor bolt anchoring quality grade classification model will output the predicted results of the anchor bolt anchor body quality grade. These results will be compared with the actual on-site conditions or the known anchoring quality to verify the accuracy of the model, and the model parameters will be adjusted and optimized according to the comparison results to improve the classification accuracy. In addition, the predicted quality grade and the classification results of the relevant ultrasonic guided wave signal features will be recorded and compiled into a detailed inspection report to provide data support for subsequent structural safety management and facility maintenance. Users can input the ultrasonic guided wave detection signal to be classified and load the trained model. The signal after noise reduction processing by the HO-VMD algorithm will be imported into the model for quality grade classification prediction, thereby achieving accurate assessment of the quality grade of the anchor bolt and anchor body.

[0137] like Figure 5 As shown, the ultrasonic guided wave detection anchor bolt anchoring quality grade classification system based on the above classification method of the present invention includes an ultrasonic guided wave detection platform and a processing module;

[0138] The ultrasonic guided wave detection platform is used to collect ultrasonic guided wave defect signals of different defect locations and sizes, so as to analyze the differences in ultrasonic guided waves in subsequent steps and complete the classification of the stability state of the anchor bolt and anchor body.

[0139] The ultrasonic guided wave detection platform includes an ultrasonic guided wave signal excitation device, a signal acquisition device, and an anchor bolt, wherein:

[0140] The ultrasonic guided wave signal excitation device is used to transmit the ultrasonic guided wave signal modulated by the sine wave to the anchor body of the anchor rod; the ultrasonic guided wave signal excitation device is composed of a signal generator, a power amplifier and an ultrasonic generating probe which are electrically connected in sequence;

[0141] A signal acquisition device for receiving ultrasonic guided wave defect signals; the signal acquisition device is composed of an ultrasonic receiving probe and a digital oscilloscope electrically connected in sequence;

[0142] Anchor rod anchor body is used to simulate resin anchor rods of different quality grades by controlling different anchoring positions of resin anchoring agent.

[0143] The processing module is used to perform modal decomposition on ultrasonic guided wave signals containing defect characteristics based on the variational mode decomposition (VMD) algorithm optimized by the Hippo optimization algorithm (HO), select signal components for signal noise reduction using the Pearson correlation coefficient as the fitness function, and obtain the noise-reduced ultrasonic guided wave characteristic signals. The trained anchor bolt anchoring quality grade classification model constructed based on the CNN-BiLSTM network is then used to judge the input data to be classified based on the mapping relationship between the ultrasonic guided wave signal characteristic data of different anchoring defects and the anchor bolt anchoring quality grade, and finally obtain the anchor bolt anchoring quality classification grade corresponding to the data to be classified.

[0144] In this embodiment, the anchor bolt anchor body can be a mining threaded steel resin anchor bolt sample, in which concrete is used to simulate the surrounding rock, and resin mastic is used to simulate the anchoring agent;

[0145] In this embodiment, the electrical connection between the components in the ultrasonic guided wave signal excitation device and the signal acquisition device can be achieved by connecting with dual BNC signal source lines.

[0146] In this embodiment, the steps of obtaining an ultrasonic guided wave defect signal containing defect characteristics using an ultrasonic guided wave detection platform are as follows:

[0147] A1: Build an ultrasonic guided wave signal excitation device and a signal acquisition device, and apply coupling agent between the ultrasonic probe and the anchor body to be tested to improve the transmission efficiency of the ultrasonic wave and reduce interference from air or other media;

[0148] A2: Turn on the signal generator and power amplifier, and set the parameters of the required excitation signal;

[0149] A3: Turn on the digital oscilloscope and use the ultrasonic receiving probe to collect the excitation signal of the signal transmitter and the ultrasonic guided wave defect signal;

[0150] A4: Import the waveform signal data saved by the digital oscilloscope into the computer for further analysis in subsequent steps.

[0151] In the present invention, the detailed steps and methods for classifying the anchoring quality grades of anchor bolts using the processing module have been described in detail above and will not be repeated here.

[0152] It should be noted that those skilled in the art will also understand that the various illustrative logical blocks and steps listed in the embodiments of the present application can be implemented by electronic hardware, computer software, or a combination of both. Whether such functions are implemented by hardware or software depends on the specific application and the design requirements of the entire system. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of the present application.

[0153] Furthermore, the word "exemplary" is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as "exemplary" is not necessarily to be construed as advantageous over other aspects or designs. Rather, the use of the word exemplary is intended to present concepts in a concrete manner. As used herein, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified or clear from the context, "X applies to A or B" is intended to mean any of the natural inclusive permutations. That is, if X applies to A; X applies to B; or X applies to both A and B, then "X applies to A or B" satisfies any of the aforementioned instances. Furthermore, the articles "a" and "an," as used in this application and the appended claims, are generally understood to mean "one or more," unless otherwise specified or clear from the context to refer to the singular form.

[0154] Likewise, although the present disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art after reading and understanding the specification and drawings. The present disclosure includes all such modifications and variations and is limited only by the scope of the claims. In particular, with respect to the various functions performed by the components described above (e.g., elements, resources, etc.), unless otherwise indicated, the terms used to describe such components are intended to correspond to any component (functionally equivalent) that performs the specific functions of the described components, even if structurally not equivalent to the disclosed structures. In addition, although specific features of the present disclosure may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations as may be desired and beneficial for any given or specific application. In addition, with respect to the terms "including," "having," "having," "having," or variations thereof used in the specific embodiments or claims, such terms are intended to be inclusive in a manner similar to the term "comprising."

[0155] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0156] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A method for classifying anchor bolt anchoring quality using ultrasonic guided wave testing, characterized by: The following steps are involved: A: Use the ultrasonic guided wave testing platform to perform ultrasonic guided wave testing on the anchor body containing defects to obtain the ultrasonic guided wave defect signal; B: Use the HO-VMD algorithm to perform modal decomposition on the ultrasonic guided wave defect signal, use PCC-E as the fitness function, perform signal noise reduction, and obtain the ultrasonic guided wave characteristic signal after noise reduction; The HO-VMD algorithm is a VMD algorithm optimized based on the HO algorithm. The HO algorithm represents the Hippo optimization algorithm, and the VMD algorithm represents the variational mode decomposition algorithm. PCC represents the Pearson correlation coefficient, and E represents the power spectrum entropy. C: Construct a dataset using the noise-reduced ultrasonic guided wave feature signals and graded labeled data; D: Using the data set to train the constructed anchor bolt anchoring quality grade classification model to obtain the trained anchor bolt anchoring quality grade classification model; E: Use the trained anchor bolt anchoring quality grade classification model to judge the input data to be classified, and finally obtain the anchor bolt anchoring quality classification grade corresponding to the data to be classified.

2. The method for classifying anchor bolt anchoring quality by ultrasonic guided wave detection according to claim 1, characterized in that: Step B includes the following steps: B1: Set the number of iterations of the HO algorithm, initialize the ratio of predators to joiners, and the modal layer number K and penalty factor α of the VMD algorithm; B2: Initialize the population size of the HO algorithm with the Logistic-Tent chaotic map to obtain the VMD algorithm optimized by the HO algorithm, namely the HO-VMD algorithm; B3: Construct PCC-E as the fitness function of the HO algorithm; B4: Use HO-VMD algorithm to reduce noise of ultrasonic guided wave defect signals; B5: Using the HO-VMD algorithm and combining it with the fitness function, the optimal modal layer number K and penalty factor α are obtained through iterative updates, and the denoised ultrasonic guided wave characteristic signal is output.

3. The method for classifying anchor bolt anchoring quality by ultrasonic guided wave detection according to claim 2, characterized in that: In step B1, the VMD algorithm is used to decompose the ultrasonic guided wave defect signal into K decomposition layers through adaptive quasi-orthogonal transformation. The ultrasonic guided wave defect signal is composed of the sum of K natural mode components of different frequency bands.

4. The method for classifying anchor bolt anchoring quality by ultrasonic guided wave detection according to claim 3, characterized in that: In step B3, the optimization process of PCC-E is as follows: a. Calculate the Pearson correlation coefficient between each intrinsic modal component and the original signal, record it as PCC, and calculate the Pearson correlation coefficient between the intrinsic modal component and the next extant intrinsic modal component, record it as PCC′; b. Calculate the correlation coefficient ratio of each intrinsic modal component R = PCC / PCC′; c. Set the correlation coefficient ratio threshold G to filter out the intrinsic modal components whose correlation coefficient ratio R is greater than the correlation coefficient ratio threshold G; d. Calculate the power spectrum entropy E of the filtered intrinsic modal components P ; e. Calculate the output value of the fitness function fitness = minE P .

5. The method for classifying anchor bolt anchoring quality using ultrasonic guided wave testing according to claim 1, characterized in that: In step D, the anchor bolt anchoring quality grade classification model consists of a CNN model and a BiLSTM model; the CNN model performs convolution and pooling operations on the denoised ultrasonic guided wave characteristic signal to extract the potential features of the data, and the BiLSTM model is used to extract the local features and long-term regularities of the denoised ultrasonic guided wave characteristic signal, and performs ultrasonic guided wave signal detection on anchor bolts with different anchoring qualities according to the anchoring density of the anchoring section, and then enhances the data by adding interference to the original signal; finally, the output of the BiLSTM is classified by a fully connected neural network classifier to obtain the prediction result of the anchor bolt anchor body quality grade type.

6. The method for classifying anchor bolt anchoring quality using ultrasonic guided wave testing according to claim 1, characterized in that: The ultrasonic guided wave detection platform includes an ultrasonic guided wave signal excitation device, a signal acquisition device and an anchor bolt anchor body; the ultrasonic guided wave signal excitation device is used to transmit an ultrasonic guided wave signal modulated by a sine wave to the anchor bolt anchor body; the signal acquisition device is used to receive the ultrasonic guided wave defect signal; the anchor bolt anchor body is used to simulate resin anchor bolts of different quality grades.

7. The method for classifying anchor bolt anchoring quality using ultrasonic guided wave testing according to claim 6, characterized in that: The ultrasonic guided wave signal excitation device is composed of a signal generator, a power amplifier and an ultrasonic generating probe which are electrically connected in sequence; the signal acquisition device is composed of an ultrasonic receiving probe and a digital oscilloscope which are electrically connected in sequence; the anchor bolt anchor body adopts a mining threaded steel resin anchor bolt.

8. The method for classifying anchor bolt anchoring quality by ultrasonic guided wave detection according to claim 7, characterized in that: Step A includes the following steps: A1: Build the ultrasonic guided wave signal excitation device and signal acquisition device, and apply coupling agent between the ultrasonic probe and the anchor body to be tested; A2: Turn on the signal generator and power amplifier, and set the parameters of the excitation signal; A3: Turn on the digital oscilloscope and use the ultrasonic receiving probe to collect the excitation signal of the signal transmitter and the ultrasonic guided wave defect signal; A4: Import the waveform signal data saved by the digital oscilloscope into the processing module.

9. An ultrasonic guided wave inspection anchor bolt anchoring quality grade classification system based on the quality grade classification method according to any one of claims 1 to 8, characterized in that: Includes ultrasonic guided wave detection platform and processing module; Ultrasonic guided wave detection platform, used to obtain ultrasonic guided wave defect signals of different defect locations and sizes; A processing module is used to perform modal decomposition on the ultrasonic guided wave defect signal using the HO-VMD algorithm, and to perform signal noise reduction using PCC-E as a fitness function to obtain the ultrasonic guided wave characteristic signal after noise reduction; Then, the trained anchor bolt anchoring quality grade classification model is used to judge the input data to be classified according to the mapping relationship between the ultrasonic guided wave signal feature data of different anchoring defects and the anchor bolt anchoring quality grade, and finally the anchor bolt anchoring quality classification grade corresponding to the data to be classified is obtained.

10. The method for classifying anchor bolt anchoring quality using ultrasonic guided wave testing according to claim 9, characterized in that: The ultrasonic guided wave detection platform includes an ultrasonic guided wave signal excitation device, a signal acquisition device and an anchor bolt anchor body; the ultrasonic guided wave signal excitation device is used to transmit an ultrasonic guided wave signal modulated by a sine wave to the anchor bolt anchor body; the signal acquisition device is used to receive the ultrasonic guided wave defect signal; the anchor bolt anchor body is used to simulate resin anchor bolts of different quality grades.

Images