Corrosion positioning method for metal pressure hull of underwater equipment based on enhanced guided wave signal
Patent Information
- Application Number
- CN202511733946.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-24
- Publication Date
- 2026-10-27
AI Technical Summary
然而，由于水下装备在复杂环境中的使用条件（如高水压、低温、高湿等），实际损伤数据难以大规模收集
[0021]上述基于增强导波信号的水下装备金属耐压结构腐蚀定位方法，在每条路径的导波信号的瞬时幅度与相位中分别引入独立高斯噪声，并在相位中添加预设范围随机相位偏移，采用幅值和相位双重扰动的方式来模拟环境因素对导波信号的影响，从而增强模型训练数据的多样性，提高模型的泛化能力和鲁棒性。通过KPCA模型有效从复杂的非线性数据中提取出与损伤相关的低维特征，降低数据的维度，并保留信号中的重要损伤信息，从而有效减少计算复杂性，并保留导波信号中的损伤信息，为后续的神经网络模型提供高质量的输入特征。使用一维卷积神经网络来处理导波信号，进行损伤定位，并使用保留损伤敏感的非线性特征对一维卷积神经网络模型进行损伤定位训练，使一维卷积神经网络模型能够自动学习信号中的损伤特征并进行高精度的定位，提高了模型的训练效率和定位精度。
Smart Images
Figure CN122983199A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of underwater equipment corrosion damage detection technology, and in particular to a method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals. Background Technology
[0002] In existing technologies, corrosion damage detection of metallic pressure-resistant structures in underwater equipment primarily relies on guided wave technology. Guided wave technology is an important method in structural health monitoring (SHM), particularly suitable for damage detection of large-area, complex structures. Guided wave signals (such as Lamb waves) have advantages such as long propagation distance and sensitivity to early damage, thus they are widely used in the detection of underwater equipment and ship hulls. Furthermore, convolutional neural networks are commonly used for automated damage detection and can learn damage patterns from large amounts of data through training. However, existing deep learning methods typically rely on large amounts of real damage data for training. However, due to the complex operating conditions of underwater equipment (such as high water pressure, low temperature, and high humidity), it is difficult to collect actual damage data on a large scale. This results in a severe shortage of training datasets, failing to support efficient training of deep learning models, thus causing a "data hunger" problem. Existing technologies typically rely on a large number of real damage samples, a problem that cannot be solved in practical applications. On the other hand, multipath guided wave signals bring high-dimensional data input, resulting in a large computational load during training, making it difficult for models to process efficiently; moreover, the signals are affected by noise and environmental factors, leading to low positioning accuracy. Traditional feature extraction methods are ineffective in handling complex guided wave signals, especially when nonlinear and high-dimensional data are present, making it difficult to extract effective damage features from the original signal. Existing methods also suffer from decreased positioning accuracy in real-world environments, particularly when affected by various environmental factors, and lack sufficient adaptability.
[0003] Therefore, a technical solution is urgently needed to solve the above problems. Summary of the Invention
[0004] Therefore, it is necessary to provide a corrosion positioning method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals that can reduce computational complexity, has strong adaptability, and high positioning accuracy, in response to the above-mentioned technical problems.
[0005] A method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals includes the following steps: Collect guided wave signals from different paths of the metal pressure-resistant structure of underwater equipment, and standardize the guided wave signals. Extract the instantaneous amplitude and phase information of the guided wave signal for each path, introduce independent Gaussian noise into the amplitude and phase respectively, and add a preset range of random phase shift to the phase so that the processed amplitude and phase constitute an enhanced guided wave signal. The physical features of the enhanced guided wave signal for each path are extracted and input into the KPCA model for nonlinear dimensionality reduction, while retaining the damage-sensitive nonlinear features. Damage-sensitive nonlinear features are retained and input into a one-dimensional convolutional neural network model for damage localization training. Damage location was determined using a trained one-dimensional convolutional neural network model.
[0006] In one embodiment, guided wave signals from different paths along the metal pressure-resistant structure of underwater equipment are acquired, and the guided wave signals are standardized, including: PZT sensors are arranged according to the shape of the underwater equipment's metal pressure-resistant structure. Each PZT sensor excites a guided wave signal, and the remaining PZT sensors receive the reflected wave, forming different paths. The guided wave signals for each path are acquired, and the amplitude of the guided wave signals is normalized using a standardization method.
[0007] In one embodiment, the instantaneous amplitude and phase information of the guided wave signal for each path is extracted, independent Gaussian noise is introduced into the amplitude and phase respectively, and a random phase shift within a preset range is added to the phase, so that the processed amplitude and phase constitute an enhanced guided wave signal, including: Perform a Hilbert transform on the time-domain signal of the guided wave signal for each path:
[0008] in, It is a complex analytic signal; It is a time-domain signal; For Hilbert transform operators; The imaginary unit; Extracting instantaneous amplitude and phase information from complex analytic signals:
[0009] in, For amplitude; For phase; Introduce independent Gaussian noise into the amplitude:
[0010] in, The amplitude after processing; This is the amplitude adjustment factor; For amplitude disturbance; This represents the amplitude variance. Independent Gaussian noise is introduced into the phase, and a random phase offset within a preset range is added as a fixed offset:
[0011] in, The processed phase; For phase adjustment factor; For phase perturbation; For fixed offset; For phase variance; The processed amplitude and phase are used to construct an enhanced guided wave signal:
[0012] in, To enhance the guided wave signal; {·} represents the operation of extracting the real part.
[0013] In one embodiment, the amplitude variance is calculated from the amplitude signal power and a preset signal-to-noise ratio, and the phase variance is calculated from the phase signal power and a preset signal-to-noise ratio.
[0014] Wherein, SNR is the preset signal-to-noise ratio; This refers to the amplitude signal power. This represents the phase signal power.
[0015] In one embodiment, the physical characteristics include kurtosis, standard deviation, root mean square value, peak factor, and Hilbert envelope peak.
[0016] In one embodiment, the physical features of the enhanced guided wave signal for each path are extracted, and the extracted physical features are input into a KPCA model for nonlinear dimensionality reduction, retaining damage-sensitive nonlinear features, including: The physical features of the enhanced guided wave signal for each path are extracted and then input into the KPCA model. The KPCA model performs nonlinear mapping on the input physical features using a Gaussian radial basis kernel function. By using the Nyström approximation method, the first k principal components are selected, where k is a preset value, and the variance is retained to a value greater than 95, resulting in damage-sensitive nonlinear features.
[0017] In one embodiment, the one-dimensional convolutional neural network model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.
[0018] In one embodiment, damage-sensitive nonlinear features are input into a one-dimensional convolutional neural network model for damage localization training, including: Introduce damage-sensitive nonlinear features into the input layer; Multiple convolutional layers are used to extract local features from the nonlinear features in the input layer. The convolutional kernels learn to obtain the damage features in the guided wave signal. Each convolutional layer is followed by a pooling layer, which uses max pooling to reduce the dimensionality of the convolutional layer output. The features extracted by convolution are flattened and input into a fully connected layer for regression prediction of the damage location. The coordinates of the predicted damage location are then output through the output layer to complete the damage localization training.
[0019] In one embodiment, during damage localization training, the Adam optimizer is used, and a segmented decay strategy is employed for training, with the loss function being the mean squared error.
[0020] A corrosion locating device for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals, comprising: The signal acquisition module is used to acquire guided wave signals from different paths of the metal pressure-resistant structure of underwater equipment and to standardize the guided wave signals. The data augmentation module is used to extract the instantaneous amplitude and phase information of the guided wave signal for each path, introduce independent Gaussian noise into the amplitude and phase respectively, and add a preset range of random phase shift to the phase so that the processed amplitude and phase constitute the enhanced guided wave signal. The dimension reduction module is used to extract the physical features of the enhanced guided wave signal for each path, and input the extracted physical features into the KPCA model for nonlinear dimension reduction processing, while retaining the damage-sensitive nonlinear features. The model training module is used to input damage-sensitive nonlinear features into a one-dimensional convolutional neural network model for damage localization training. The testing module is used to locate damage points using a trained one-dimensional convolutional neural network model.
[0021] The aforementioned corrosion localization method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals introduces independent Gaussian noise into the instantaneous amplitude and phase of the guided wave signal for each path, and adds a preset range of random phase shift to the phase. This dual perturbation of amplitude and phase simulates the influence of environmental factors on the guided wave signal, thereby enhancing the diversity of model training data and improving the model's generalization ability and robustness. The KPCA model effectively extracts low-dimensional features related to damage from complex nonlinear data, reducing the data dimensionality while retaining important damage information in the signal. This effectively reduces computational complexity and preserves damage information in the guided wave signal, providing high-quality input features for subsequent neural network models. A one-dimensional convolutional neural network is used to process the guided wave signal for damage localization. The damage-sensitive nonlinear features are used to train the one-dimensional convolutional neural network model for damage localization, enabling the model to automatically learn damage features in the signal and perform high-precision localization, improving the model's training efficiency and localization accuracy.
[0022] This invention can effectively solve the problems of data scarcity, complexity of high-dimensional data processing, difficulty in feature extraction, and poor environmental adaptability in the detection of corrosion damage in underwater equipment, and significantly improve the accuracy and robustness of damage localization. Attached Figure Description
[0023] Figure 1 This is an illustration of an application scenario for a corrosion localization method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals, as shown in one embodiment. Figure 2 This is a flowchart illustrating a corrosion location method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals in one embodiment. Figure 3 This is a schematic diagram of the location and path of the PZT sensor in one embodiment; Figure 4 This is a dispersion curve of an 8mm aluminum alloy in one embodiment. Figure 4 (a) is the group velocity curve of an 8mm aluminum alloy. Figure 4 (b) is the phase velocity curve of an 8mm aluminum alloy; Figure 5 This is a schematic diagram of the excitation signal in one embodiment; Figure 6 This is a schematic diagram of enhancing the guided wave signal in one embodiment, wherein, Figure 6 (a) is a diagram of the guided wave signal. Figure 6 (b) is a comparison diagram of the guided wave signal and the enhanced guided wave signal; Figure 7 This is a schematic diagram of a one-dimensional convolutional neural network model structure in one embodiment; Figure 8 This is a structural block diagram of a corrosion positioning device for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals, as shown in one embodiment. Detailed Implementation
[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0025] The corrosion location method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals provided in this application can be applied to, for example... Figure 1 In the application environment shown, the signal acquisition device 102 acquires the guided wave signal from the underwater equipment 104. The signal acquisition device 102 can be a piezoelectric sensor.
[0026] In one embodiment, such as Figure 2As shown, a method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals is provided, including the following steps: Step 201: Collect guided wave signals from different paths of the underwater equipment's metal pressure-resistant structure and standardize the guided wave signals.
[0027] Step 202: Extract the instantaneous amplitude and phase information of the guided wave signal for each path, introduce independent Gaussian noise into the amplitude and phase respectively, and add a preset range of random phase offset to the phase so that the processed amplitude and phase constitute an enhanced guided wave signal.
[0028] Step 203: Extract the physical features of the enhanced guided wave signal for each path, and input the extracted physical features into the KPCA model for nonlinear dimensionality reduction processing to retain the damage-sensitive nonlinear features.
[0029] KPCA, or Kernel Principal Component Analysis, is a technique for dimensionality reduction that maps data to a high-dimensional feature space and uses kernel functions to capture the nonlinear characteristics of the data.
[0030] Step 204: Input the retained damage-sensitive nonlinear features into the one-dimensional convolutional neural network model for damage localization training.
[0031] Step 205: Use the trained one-dimensional convolutional neural network model to locate the damage location.
[0032] In the aforementioned corrosion localization method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals, independent Gaussian noise is introduced into the instantaneous amplitude and phase of the guided wave signal for each path, and a preset range of random phase shift is added to the phase. This dual perturbation of amplitude and phase simulates the influence of environmental factors on the guided wave signal, thereby enhancing the diversity of model training data and improving the model's generalization ability and robustness. The KPCA model effectively extracts low-dimensional features related to damage from complex nonlinear data, reducing the dimensionality of the data while retaining important damage information in the signal. This effectively reduces computational complexity and preserves damage information in the guided wave signal, providing high-quality input features for subsequent neural network models. A one-dimensional convolutional neural network is used to process the guided wave signal for damage localization. The one-dimensional convolutional neural network model is trained for damage localization using damage-sensitive nonlinear features, enabling it to automatically learn damage features in the signal and perform high-precision localization, thus improving the model's training efficiency and localization accuracy.
[0033] In one embodiment, step 201 involves acquiring guided wave signals from different paths of the underwater equipment's metal pressure-resistant structure and standardizing the guided wave signals, including: PZT sensors are arranged according to the shape of the underwater equipment's metal pressure-resistant structure. Each PZT sensor excites a guided wave signal, and the remaining PZT sensors receive the reflected wave, forming different paths.
[0034] The guided wave signals for each path are acquired, and the amplitude of the guided wave signals is normalized using a standardization method.
[0035] Specifically, refer to Figure 3 In one embodiment, the underwater equipment's metal pressure-resistant structure is a cylindrical shell made of aluminum alloy 6061. Four PZT (piezoelectric) sensors are arranged on the surface of the cylindrical shell, 10mm from the edge of the plate. Each PZT sensor excites a guided wave signal, and the remaining PZT sensors receive the reflected waves, forming six paths: S1 excites S2, S3, and S4 receive; S2 excites S3 and S4 receive; and S3 excites S4 receive. 3. When using guided waves for detection, an excitation signal of appropriate frequency is required. The group velocity curve and phase velocity curve of the guided wave propagation in 8mm of aluminum alloy 6061 material are calculated using the numerical software Dispersion Calculator, as shown below. Figure 4 As shown, when the excitation frequency is 100-200kHz, the guided wave modes only have A0 and S0 modes, and the guided wave signal is relatively pure, making it easy to identify damage information. Due to limitations of the experimental equipment, 160kHz was chosen as the excitation frequency for the guided wave. A 5-cycle narrowband sinusoidal signal modulated by a Hanning window was used as the excitation source. Figure 5 Its expression is:
[0036] in, The center frequency; The period is [period], and the total duration is [duration]. .
[0037] Signals for each path were acquired using a guided wave acquisition system at a sampling frequency of 12 MHz, with 4000 data points collected for each path. A normalization method was used to normalize the signal amplitude, ensuring that amplitude variations did not affect subsequent analysis.
[0038] In one embodiment, step 202 involves extracting the instantaneous amplitude and phase information of the guided wave signal for each path, introducing independent Gaussian noise into both the amplitude and phase, and adding a preset range of random phase shift to the phase, so that the processed amplitude and phase constitute an enhanced guided wave signal, including: Perform a Hilbert transform on the time-domain signal of the guided wave signal for each path:
[0039] in, It is a complex analytic signal; It is a time-domain signal; For Hilbert transform operators; The imaginary unit; The Hilbert transform is a mathematical tool used for signal analysis that extracts the instantaneous amplitude and phase of a signal by generating an analytical form of the signal.
[0040] Extracting instantaneous amplitude and phase information from complex analytic signals:
[0041] in, For amplitude; For phase; Introduce independent Gaussian noise into the amplitude:
[0042] in, The amplitude after processing; This is the amplitude adjustment factor; For amplitude disturbance; This represents the amplitude variance. Independent Gaussian noise is introduced into the phase, and a random phase offset within a preset range is added as a fixed offset:
[0043] in, The processed phase; For phase adjustment factor; For phase perturbation; For a fixed offset, ; For phase variance; ; The processed amplitude and phase are used to construct an enhanced guided wave signal:
[0044] in, To enhance the guided wave signal; The {·} operator is used to extract the real part of a complex analytic signal, which is then used to restore the complex analytic signal to an observable real signal.
[0045] In this embodiment, independent Gaussian noise is introduced into the amplitude and phase to simulate measurement disturbances, thereby simulating measurement errors or environmental interference. A preset range of random phase offset is added as a fixed offset, and based on this, a disturbance adjustment factor ( and Adjusting the perturbation amplitude increases the randomness of the generated signal. Considering the actual magnitude of the impact of environmental factors such as temperature and humidity on the guided wave signal, the amplitude and phase changes typically occur by about 3%, with phase shifts generally within ±30°. This range can effectively simulate signal changes under typical environmental perturbations, ensuring the realism and physical interpretability of the perturbation, while enhancing data diversity and improving model robustness. The enhanced guided wave signal constructed in this embodiment is as follows: Figure 6 As shown in (b).
[0046] In one embodiment, the amplitude variance is calculated from the amplitude signal power and a preset signal-to-noise ratio, and the phase variance is calculated from the phase signal power and a preset signal-to-noise ratio.
[0047] Wherein, SNR is the preset signal-to-noise ratio; This refers to the amplitude signal power. This represents the phase signal power.
[0048] In one embodiment, the physical characteristics include kurtosis, standard deviation, root mean square value, peak factor, and Hilbert envelope peak.
[0049] Kurtosis: Describes the transient characteristics of a signal; damage causes significant changes in kurtosis. Standard Deviation: Reflects the dispersion of the signal amplitude; damage may cause changes in the signal's fluctuation range. Root Mean Square (RMS): Quantifies the energy distribution of the signal; damage may cause signal energy attenuation or concentration. Peak Factor: Measures the impulse characteristics of a signal; reflected waves caused by damage may increase the peak factor. Peak of Hilbert Envelope Spectrum: Extracts the maximum amplitude of the envelope spectrum, reflecting the energy concentration at the characteristic frequencies of damage.
[0050] In one embodiment, step 203 involves extracting the physical features of the enhanced guided wave signal for each path and inputting the extracted physical features into a KPCA model for nonlinear dimensionality reduction, retaining damage-sensitive nonlinear features, including: The physical features of the enhanced guided wave signal for each path are extracted and then input into the KPCA model. The KPCA model performs nonlinear mapping on the input physical features using a Gaussian radial basis kernel function. By using the Nyström approximation method, the first k principal components are selected, where k is a preset value, and the variance is retained to a value greater than 95, resulting in damage-sensitive nonlinear features.
[0051] In this embodiment, the KPCA model effectively reduces the dimensionality of the signal, thereby enabling the subsequent one-dimensional convolutional neural network model to process the data more efficiently and improve the accuracy of damage localization.
[0052] In one embodiment, reference Figure 7 A one-dimensional convolutional neural network model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.
[0053] In this embodiment, the one-dimensional convolutional neural network model can automatically extract local features of the signal through convolutional layers and reduce the dimensionality of the features through pooling layers, ultimately outputting the accurate location of the damage.
[0054] In one embodiment, step 204 involves inputting damage-sensitive nonlinear features into a one-dimensional convolutional neural network model for damage localization training, including: Introduce damage-sensitive nonlinear features into the input layer; Multiple convolutional layers are used to extract local features from the nonlinear features in the input layer. The convolutional kernels learn to obtain the damage features in the guided wave signal. Each convolutional layer is followed by a pooling layer, which uses max pooling to reduce the dimensionality of the convolutional layer output. The features extracted by convolution are flattened and input into a fully connected layer for regression prediction of the damage location. The coordinates of the predicted damage location are then output through the output layer to complete the damage localization training.
[0055] In one embodiment, during damage localization training, the Adam optimizer is used, and a piecewise decay strategy is employed for training. The loss function is mean squared error, and the model is optimized by minimizing the difference between the predicted coordinates and the true coordinates.
[0056] It should be understood that, although Figure 2 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 2 At least some of the steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least some of the sub-steps or stages of other steps.
[0057] In one embodiment, such as Figure 8 As shown, a corrosion locating device for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals is provided, comprising: The signal acquisition module 901 is used to acquire guided wave signals from different paths of the metal pressure-resistant structure of underwater equipment and to standardize the guided wave signals. The data enhancement module 902 is used to extract the instantaneous amplitude and phase information of the guided wave signal of each path, introduce independent Gaussian noise into the amplitude and phase respectively, and add a preset range of random phase shift to the phase so that the processed amplitude and phase constitute the enhanced guided wave signal. The dimension reduction module 903 is used to extract the physical features of the enhanced guided wave signal of each path and input the extracted physical features into the KPCA model for nonlinear dimension reduction processing, while retaining the damage-sensitive nonlinear features. Model training module 904 is used to input damage-sensitive nonlinear features into a one-dimensional convolutional neural network model for damage localization training. Test module 905 is used to locate damage locations using a trained one-dimensional convolutional neural network model.
[0058] Specific limitations regarding the corrosion location device for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals can be found in the limitations of the corrosion location method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals mentioned above, and will not be repeated here. Each module in the aforementioned corrosion location device for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.
[0059] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0060] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals, characterized in that, Includes the following steps: Collect guided wave signals from different paths of the metal pressure-resistant structure of underwater equipment, and standardize the guided wave signals. Extract the instantaneous amplitude and phase information of the guided wave signal for each path, introduce independent Gaussian noise into the amplitude and phase respectively, and add a random phase shift within a preset range into the phase so that the processed amplitude and phase constitute an enhanced guided wave signal. The physical features of the enhanced guided wave signal for each path are extracted and input into the KPCA model for nonlinear dimensionality reduction, while retaining the damage-sensitive nonlinear features. Damage-sensitive nonlinear features are retained and input into a one-dimensional convolutional neural network model for damage localization training. Damage location was determined using a trained one-dimensional convolutional neural network model.
2. The method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 1, characterized in that, Collect guided wave signals from different paths along the metal pressure-resistant structure of underwater equipment, and standardize the guided wave signals, including: PZT sensors are arranged according to the shape of the underwater equipment's metal pressure-resistant structure. Each PZT sensor excites a guided wave signal, and the remaining PZT sensors receive the reflected wave, forming different paths. The guided wave signals for each path are acquired, and the amplitude of the guided wave signals is normalized using a standardization method.
3. The method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 1, characterized in that, The instantaneous amplitude and phase information of the guided wave signal for each path is extracted. Independent Gaussian noise is introduced into both the amplitude and phase, and a random phase shift within a preset range is added to the phase. The processed amplitude and phase constitute an enhanced guided wave signal, including: Perform a Hilbert transform on the time-domain signal of the guided wave signal for each path: in, It is a complex analytic signal; It is a time-domain signal; For Hilbert transform operators; The imaginary unit; Extracting instantaneous amplitude and phase information from complex analytic signals: in, For amplitude; For phase; Introduce independent Gaussian noise into the amplitude: in, The amplitude after processing; This is the amplitude adjustment factor; For amplitude disturbance; This represents the amplitude variance. Independent Gaussian noise is introduced into the phase, and a random phase offset within a preset range is added as a fixed offset: in, The processed phase; For phase adjustment factor; For phase perturbation; For fixed offset; For phase variance; The processed amplitude and phase are used to construct an enhanced guided wave signal: in, To enhance the guided wave signal; {·} represents the operation of extracting the real part.
4. The method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 3, characterized in that, The amplitude variance is calculated from the amplitude signal power and the preset signal-to-noise ratio, and the phase variance is calculated from the phase signal power and the preset signal-to-noise ratio. Wherein, SNR is the preset signal-to-noise ratio; This refers to the amplitude signal power. This represents the phase signal power.
5. The method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 1, characterized in that, The physical characteristics include kurtosis, standard deviation, root mean square value, peak factor, and Hilbert envelope peak value.
6. The method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 5, characterized in that, The physical features of the enhanced guided wave signal for each path are extracted, and these features are then input into the KPCA model for nonlinear dimensionality reduction, preserving damage-sensitive nonlinear features, including: The physical features of the enhanced guided wave signal for each path are extracted and then input into the KPCA model. The KPCA model performs nonlinear mapping on the input physical features using a Gaussian radial basis kernel function. By using the Nyström approximation method, the first k principal components are selected, where k is a preset value, and the variance is retained to a value greater than 95, resulting in damage-sensitive nonlinear features.
7. The method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 6, characterized in that, The one-dimensional convolutional neural network model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer.
8. The method for locating corrosion of underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 7, characterized in that, Damage-sensitive nonlinear features are input into a one-dimensional convolutional neural network model for damage localization training, including: Introduce damage-sensitive nonlinear features into the input layer; Multiple convolutional layers are used to extract local features from the nonlinear features in the input layer. The convolutional kernels learn to obtain the damage features in the guided wave signal. Each convolutional layer is followed by a pooling layer, which uses max pooling to reduce the dimensionality of the convolutional layer output. The features extracted by convolution are flattened and input into a fully connected layer for regression prediction of the damage location. The coordinates of the predicted damage location are then output through the output layer to complete the damage localization training.
9. The corrosion location method for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals according to claim 8, characterized in that, In the damage localization training, the Adam optimizer is used, and a piecewise decay strategy is employed for training. The loss function is the mean squared error.
10. A corrosion positioning device for underwater equipment metal pressure-resistant structures based on enhanced guided wave signals, characterized in that, include: The signal acquisition module is used to acquire guided wave signals from different paths of the metal pressure-resistant structure of underwater equipment and to standardize the guided wave signals. The data augmentation module is used to extract the instantaneous amplitude and phase information of the guided wave signal for each path, introduce independent Gaussian noise into the amplitude and phase respectively, and add a preset range of random phase shift to the phase so that the processed amplitude and phase constitute the enhanced guided wave signal. The dimension reduction module is used to extract the physical features of the enhanced guided wave signal for each path, and input the extracted physical features into the KPCA model for nonlinear dimension reduction processing, while retaining the damage-sensitive nonlinear features. The model training module is used to input damage-sensitive nonlinear features into a one-dimensional convolutional neural network model for damage localization training. The testing module is used to locate damage points using a trained one-dimensional convolutional neural network model.