Method, system and equipment for evaluating corrosion of overhead ground wire and medium

Abstract

The invention discloses an overhead ground wire corrosion assessment method, system and device and a medium. The method comprises the following steps: acquiring an original signal; performing back propagation on the original signal to obtain a dispersion compensation signal, extracting a phase to obtain a signal symbol coherence factor and a waveform correlation factor, and performing signal reconstruction; preprocessing the reconstructed signal to obtain an input signal; inputting an input signal into the trained SEResNet50 network in a segmented manner, and connecting an SE block to the tail of each residual block of the SEResNet50 network; and obtaining the corrosion degree state through the SE block output. According to the method, the accuracy and sensitivity of corrosion feature extraction are improved by introducing a nonlinear ultrasonic guided wave technology and utilizing a remarkable nonlinear effect generated when the nonlinear ultrasonic guided wave technology interacts with corrosion defects; and an SEResNet50 deep learning network is adopted to construct a corrosion evaluation model with high generalization ability, so that the accuracy and reliability of an evaluation result are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of ground wire status prediction and evaluation, and in particular to an overhead ground wire corrosion evaluation method, system, equipment and medium. Background Art

[0002] Overhead ground wires are exposed to complex environments for a long time, and their corrosion process is hidden, gradual, and uneven, resulting in corrosion features being weak in conventional detection signals and difficult to accurately extract.

[0003] Traditional detection methods often rely on linear ultrasonic technology, which lacks sensitivity for detecting micro-corrosion or early-stage corrosion and struggles to capture the nonlinear acoustic effects caused by corrosion, limiting the accuracy and timeliness of corrosion assessments. Furthermore, the performance of existing deep learning algorithms is highly dependent on the quality and quantity of training data. In practical applications, obtaining sufficient and representative labeled data can be challenging, limiting the generalization and practicality of the algorithms. While deep learning algorithms can automatically learn complex patterns in data, their ability to model physical phenomena remains limited. This is particularly true when dealing with complex physical phenomena such as the interaction between nonlinear ultrasonic guided waves and material defects, where deep learning algorithms may not fully capture all relevant physical mechanisms and characteristics. In practical applications, material defect detection often requires both real-time and accuracy. Improving accuracy may require increasing model complexity and computational cost, sacrificing real-time performance; or meeting real-time requirements may require simplifying the model structure or reducing computational accuracy. Therefore, balancing these two requirements in real-time and efficient development of overhead ground wire corrosion assessment models is crucial. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides an overhead ground wire corrosion assessment method and system to solve the problem that the current assessment model is difficult to capture the nonlinear acoustic effects caused by corrosion, which limits the accuracy and timeliness of corrosion assessment. The deep learning algorithm may not be able to fully capture all relevant physical mechanisms and characteristics, and cannot balance the model's streamlined structure and accuracy.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for evaluating corrosion of an overhead ground wire, comprising:

[0008] Get the original signal;

[0009] The dispersion compensation signal is obtained by backpropagating the original signal, and the phase is extracted to obtain the signal symbol coherence factor and waveform correlation factor for signal reconstruction;

[0010] Preprocessing the reconstructed signal to obtain an input signal;

[0011] The input signal is segmented and input into the trained SE ResNet50 network, and the SE block is connected to the end of each residual block of the SE ResNet50 network;

[0012] The corrosion degree status is obtained through the SE block output.

[0013] As a preferred embodiment of the overhead ground wire corrosion assessment method of the present invention, the dispersion compensation signal is obtained by backpropagation, and the phase is extracted to obtain the signal symbol coherence factor and waveform correlation factor, and the signal is reconstructed, including:

[0014] The full-focus imaging algorithm based on phase information reconstruction uses the numerical solution of the wavenumber domain dispersion curve to reversely propagate the time domain signal of the Lamb wave. For each imaging point, all original signals are dispersion compensated to obtain the dispersion compensated signal.

[0015] The symbol coherence factor and waveform correlation factor between the pulse echo and the excitation waveform are extracted from the dispersion compensation signal to reconstruct the scattered signal.

[0016] The beneficial effects of this preferred solution are: it can correct the waveform shape change caused by dispersion, and at the same time eliminate the flight time of the pulse echo so that the phase of the pulse echo after dispersion compensation should be consistent with the excitation signal.

[0017] As a preferred embodiment of the method for evaluating corrosion of an overhead ground wire according to the present invention, the SE block in the SEResNet50 network includes a conversion operation;

[0018] The conversion operation performs a general transformation on a first input X of a given first dimension to obtain a second output X′ of a second dimension.

[0019] As a preferred embodiment of the overhead ground wire corrosion assessment method of the present invention, the SE block in the SEResNet50 network further includes:

[0020] The SE block performs a squeeze operation, takes the second output of the second dimension as input, and compresses the spatial information of each channel into statistics through global average pooling to obtain a channel statistics vector;

[0021] The SE block performs the excitation operation and uses the s-type activation gating mechanism to excite the channel statistics vector to obtain the channel weight vector;

[0022] Based on the second output and the channel weight vector, a recalibrated feature map is obtained by rescaling.

[0023] The beneficial effects of this preferred solution are: it can automatically learn the complex nonlinear relationship between corrosion characteristics and detection signals, construct a corrosion assessment model with high generalization ability, and effectively improve the accuracy and reliability of the assessment results.

[0024] As a preferred solution of the overhead ground wire corrosion assessment method described in the present invention, the full-focus imaging algorithm based on phase information reconstruction of the signal utilizes the numerical solution of the wavenumber domain dispersion curve to reversely propagate the time domain signal of the Lamb wave. For each imaging point, all original signals are dispersion compensated to obtain a dispersion compensated signal, which is expressed as:

[0025]

[0026] Among them, S pq (r, t) is the Lamb wave signal after dispersion compensation, ω is the angular frequency, k(ω) is the wave number of the angular frequency, i is the imaginary unit, d pq (r) is the propagation distance of the Lamb wave from the transmitter p to the focus r and then to the receiver q.

[0027] As a preferred embodiment of the overhead ground wire corrosion assessment method described in the present invention, the method comprises extracting a symbol coherence factor and a waveform correlation factor between a pulse echo and an excitation waveform from a dispersion compensation signal to reconstruct a scattered signal, including:

[0028] Based on the dispersion compensation signal, the time period that matches the excitation signal is intercepted to extract the focused wave signal;

[0029] Based on the fusion of phase and waveform correlation of the focused wave signal and the original excitation signal, the symbol coherence factor and waveform correlation factor are extracted and the pulse echo is reconstructed;

[0030] For each imaging point in the reconstructed pulse echo, the defect imaging map is calculated using the full-focus imaging algorithm.

[0031] As a preferred solution of the overhead ground wire corrosion assessment method described in the present invention, the reconstructed signal is preprocessed to obtain an input signal, including:

[0032] De-noising the reconstructed signal;

[0033] Normalize the denoised signal and segment the normalized signal;

[0034] Each signal segment is labeled to determine its corresponding corrosion degree category as the input label.

[0035] In a second aspect, the present invention provides an overhead ground wire corrosion assessment system, comprising:

[0036] An acquisition module, used to obtain the original signal;

[0037] The signal reconstruction module is used to obtain the dispersion compensation signal by backpropagating the original signal, extract the phase to obtain the signal symbol coherence factor and waveform correlation factor, and perform signal reconstruction;

[0038] A preprocessing module, used for preprocessing the reconstructed signal to obtain an input signal;

[0039] The input and feature extraction module is used to input the input signal into the trained SE ResNet50 network in segments, and connect the SE block to the end of each residual block of the SE ResNet50 network;

[0040] The output module is used to obtain the corrosion degree status through the SE block output.

[0041] In a third aspect, the present invention provides a computer device, comprising:

[0042] memory and processor;

[0043] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the overhead ground wire corrosion assessment method are implemented.

[0044] In a fourth aspect, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the overhead ground wire corrosion assessment method.

[0045] Compared with the existing technology, the beneficial effects of the present invention are as follows: the present invention introduces nonlinear ultrasonic guided wave technology and utilizes its significant nonlinear effect when interacting with corrosion defects to enhance the performance of corrosion characteristics in the detection signal, thereby improving the accuracy and sensitivity of corrosion feature extraction; adopts the SEResNet50 deep learning network and utilizes its powerful feature learning and expression capabilities to automatically learn the complex nonlinear relationship between corrosion characteristics and detection signals, construct a corrosion assessment model with high generalization ability, and improve the accuracy and reliability of the assessment results; the combination of the two realizes online monitoring and automated evaluation of corrosion conditions, providing strong support for the intelligent operation and maintenance of power grid equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0047] Figure 1 The figure is a schematic diagram of the overall process of the overhead ground wire corrosion assessment method according to one embodiment of the present invention.

[0048] Figure 2 This is a structural diagram of the signal reconstruction process in the overhead ground wire corrosion assessment method described in the second embodiment of the present invention. DETAILED DESCRIPTION

[0049] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0050] Example 1, with reference to Figure 1 , as one embodiment of the present invention, provides a method for evaluating corrosion of an overhead ground wire, comprising:

[0051] S100: obtaining the original signal;

[0052] S200: Obtain a dispersion-compensated signal by backpropagating the original signal, extract the phase to obtain the signal symbol coherence factor and waveform correlation factor, and perform signal reconstruction;

[0053] S300: Preprocessing the reconstructed signal to obtain an input signal;

[0054] S400: Input the input signal into the trained SE ResNet50 network in segments, and connect the SE block at the end of each residual block of the SE ResNet50 network;

[0055] S500: Obtain the corrosion degree status through the SE block output.

[0056] It should be noted that traditional detection methods often rely on linear ultrasonic technology, which lacks sensitivity for detecting minor corrosion or early stages of corrosion. Corrosion assessment models are often based on empirical formulas or simple machine learning algorithms, which have poor generalization and adaptability when faced with complex and changing corrosion morphologies and operating conditions. These models often struggle to accurately reflect the complex nonlinear relationship between corrosion characteristics and detection signals, resulting in significant deviations in assessment results and failing to meet the high-precision requirements for corrosion assessment in practical engineering projects. These methods often involve complex model structures and extensive computing resources, leading to long model training times and high computational costs. They also require human input for data analysis and processing, which is not only inefficient but also susceptible to human influence, resulting in subjective and inconsistent assessment results. This makes them unable to meet the requirements of modern power grids for online equipment status monitoring and intelligent operation and maintenance. Therefore, by combining steps S100-S500 with nonlinear ultrasonic guided wave detection technology and the SEResNet50 network model, real-time and efficient corrosion assessment of overhead ground wires is achieved.

[0057] Example 2, reference Figure 1-Figure 2 , which is an embodiment of the present invention, provides an overhead ground wire corrosion assessment method based on the above embodiment.

[0058] In this embodiment of the present application, the raw signal acquired in step S100 can be detected using ultrasonic guided waves. A piezoelectric transducer array installed at one end of the overhead ground wire or at a specific monitoring point emits ultrasonic guided waves in a specific pattern. When encountering defects such as corrosion or broken strands, reflection, scattering, or mode conversion occurs. Typically, a piezoelectric transducer probe is connected to the ultrasonic guided wave excitation / acquisition equipment.

[0059] In an optional implementation, step S100 may also provide a time domain / frequency domain traveling wave signal through high frequency traveling wave detection (OPGW), but the resolution of such a signal may be insufficient and needs to be adjusted before proceeding to subsequent steps.

[0060] In the embodiment of this application, Figure 2 As shown, in step S200, a dispersion compensation signal is obtained by back propagation, and the phase is extracted to obtain the signal symbol coherence factor and the waveform correlation factor, and signal reconstruction is performed, including the following steps A1-A2:

[0061] A1: The all-focusing imaging algorithm based on phase information reconstruction uses the numerical solution of the wavenumber domain dispersion curve to reversely propagate the Lamb wave time domain signal. For each imaging point, all original signals are dispersion compensated to obtain the dispersion compensated signal.

[0062] It should be noted that when using ultrasonic guided waves for line survey imaging, a total focusing imaging algorithm that reconstructs the signal with phase information is preferred to address the inherent dispersion characteristics of ultrasonic guided waves and the grating and sidelobe problems existing in the classic total focusing imaging algorithm (TFM). This algorithm can correct the waveform shape changes caused by dispersion and eliminate the flight time of the pulse echo so that the phase of the pulse echo after dispersion compensation should be consistent with the excitation signal.

[0063] A2: Extract the symbol coherence factor and waveform correlation factor between the pulse echo and the excitation waveform from the dispersion compensation signal to reconstruct the scattered signal.

[0064] In the embodiment of the present application, in step A1 of step S200, the all-focusing imaging algorithm for reconstructing the signal based on phase information uses the numerical solution of the wavenumber domain dispersion curve to reversely propagate the time domain signal of the Lamb wave. Specifically, the phase and waveform of the Lamb wave will change during the propagation process. Therefore, for each imaging point, dispersion compensation is performed on all original signals to obtain a dispersion compensated signal, which is expressed as:

[0065]

[0066] Among them, S pq (r, t) is the Lamb wave signal after dispersion compensation, ω is the angular frequency, k(ω) is the wave number of the angular frequency, i is the imaginary unit, d pq (r) is the propagation distance of the Lamb wave from the transmitter p to the focus r and then to the receiver q.

[0067] In the embodiment of the present application, in step A2 of step S200, the symbol coherence factor (SCF) and waveform correlation factor (WCF) between the pulse echo and the excitation waveform are extracted from the dispersion compensation signal to reconstruct the scattered signal, including the following steps A2.1-A2.3:

[0068] A2.1: Based on the dispersion compensation signal, the time segment that matches the excitation signal is intercepted to extract the focused wave signal;

[0069] Specifically, since the wave packet after compensation in the previous step will be consistent with the phase and waveform of the excitation signal, the original pulse echo caused by the defect can be captured and the focused wave signal can be extracted, which is expressed as:

[0070] u pq (r,t)=S pq (r,t)| t∈[0,T] (2)

[0071] Among them, u pq (r, t) is the focused wave, and t is the length of the excitation signal in the time domain; choosing the appropriate focusing time point can realize the TFM imaging algorithm.

[0072] Furthermore, the amplitude-based imaging algorithm is susceptible to noise, while phase information has higher reliability. Therefore, the pulse echo for TFM imaging can be obtained by the following steps:

[0073] A2.2: Based on the phase and waveform correlation of the focused wave signal and the original excitation signal, the symbol coherence factor and waveform correlation factor are extracted and the pulse echo is reconstructed;

[0074] Specifically, it can be expressed as:

[0075]

[0076] Among them, U pq (r, t) is the reconstructed pulse echo, sign(), cov(), and σ() are the sign function, covariance function, and standard deviation, respectively.

[0077] It should be noted that, through the above formula (3), the pulse echo of the defect can be reconstructed based on the phase and waveform information of the signal.

[0078] A2.3: For each imaging point in the reconstructed pulse echo, a full-focus imaging algorithm is used to calculate the defect imaging image;

[0079] Specifically, the imaging algorithm can be expressed as:

[0080]

[0081] It should be noted that although formula (4) uses TFM, the imaging index is based on the phase and waveform information of the signal, so it has a strong anti-interference ability.

[0082] In an optional embodiment, the compensation mechanism in step S200 can also utilize multimodal dispersion compensation, processing both the symmetric mode (S0) and the antisymmetric mode (A0) of the Lamb wave simultaneously. This leverages the differential response of these two modes to corrosion to enhance feature recognition. Specifically, this can be achieved by performing modal separation on the original signal and independently performing backpropagation compensation. Signal reconstruction can then be performed using the symbol coherence factor and waveform correlation factor.

[0083] It should also be noted that the nonlinear ultrasonic guided wave technology introduced in step S200 produces significant nonlinear effects when interacting with corrosion defects, greatly enhancing the appearance of corrosion signatures in the detection signal. This significantly improves the accuracy and sensitivity of corrosion signature extraction, enabling more precise identification of overhead ground wire corrosion. This provides a more reliable data foundation for subsequent corrosion assessments, helping to promptly detect early-stage corrosion issues and ensure safe and stable power grid operation.

[0084] In the embodiment of the present application, the reconstructed signal is preprocessed in step S300 to obtain an input signal, which includes the following steps B1-B3:

[0085] B1: denoising the reconstructed signal;

[0086] It should be noted that the collected nonlinear ultrasonic guided wave signals are often affected by various factors such as environmental noise and equipment interference. These noises will mask the useful information in the signals and reduce the accuracy of corrosion assessment.

[0087] In an optional embodiment, step B1 can obtain a denoised signal through wavelet denoising. By selecting an appropriate wavelet basis function and number of decomposition levels, the signal is subjected to wavelet decomposition to obtain wavelet coefficients at different scales. Then, according to a threshold processing rule, the wavelet coefficients are thresholded to remove the wavelet coefficients corresponding to the noise. Finally, wavelet reconstruction is performed to obtain the denoised signal.

[0088] B2: Normalize the denoised signal and segment the normalized signal;

[0089] It should be noted that since the amplitude and range of the signals collected under different detection conditions may vary, the signals need to be normalized to facilitate subsequent feature extraction and model training. The denoised and normalized signals are segmented, and the length of each segment is determined based on the actual situation.

[0090] In another optional embodiment, the time-frequency feature extraction can be performed on the reconstructed signal to extract the reconstructed signal U pq (r, t) is subjected to synchronized compressed wavelet transform (SST), the time-frequency spectrum is divided into regions, and normalized to enhance the contrast between the corroded area and the non-corroded area.

[0091] B3: Label each signal segment and determine its corresponding corrosion severity category as the input label.

[0092] For example, based on the actual corrosion detection results, each signal segment is labeled to determine its corresponding corrosion degree category, such as no corrosion, mild corrosion, moderate corrosion, and severe corrosion; the above-mentioned labeling information will be used as a label for training the network.

[0093] Furthermore, step S400 segments the input signal and inputs it into the trained SEResNet50 network.

[0094] It should be noted that the SE ResNet50 network is constructed by integrating the Squeeze and Excitation (SE) block into the ResNet50 network. Its goal is to improve the quality of the representation generated by the network by explicitly modeling the interdependence between its convolutional feature channels.

[0095] In the embodiment of the present application, the SE block in the SEResNet50 network in step S400 includes a conversion operation F tr ;

[0096] The conversion operation is to obtain a second output X′ with a second dimension of H2×W2×C2 by a general transformation, such as a convolution operation, for a first input X with a given first dimension of H1×W1×C1.

[0097] It should be noted that the conversion operation F tr Strictly speaking it is not part of the SE block, but part of the original network.

[0098] Furthermore, unlike traditional CNNs, the next three operations of the SE block are designed to recalibrate the previously obtained features.

[0099] In the embodiment of the present application, the SE block in the SE ResNet50 network in step S400 further includes the following steps C1-C3:

[0100] C1: SE block performs a squeeze operation, taking the second output of the second dimension as input and compressing the spatial information of each channel into statistics through global average pooling to obtain a channel statistics vector;

[0101] Specifically, the squeeze operation is a process of generating channel statistics through global average pooling, and its formula can be described as:

[0102]

[0103] Among them, X′ c (i, j) represents the value at position (i, j) of the c-th channel of the input. The spatial dimension of the input X′ is H2×W2, and the channel number is c. Z represents the statistic of the c-th channel.

[0104] Therefore, the above equation converts the input of H2×W2×C2 into the output of 1×1×C2.

[0105] C2: SE block performs excitation operation and uses s-type activation gating mechanism to excite the channel statistics vector to obtain the channel weight vector;

[0106] Specifically, it can be expressed as:

[0107] s=Fex (Z,D)=σ(g(Z,D))=σ(D2δ(D1,Z)) (6)

[0108] Where σ is the sigmoid function, δ is the rectified linear unit (ReLU) function, and represents two fully connected (FC) layers, where the reduction ratio r is the number of hidden nodes in the middle layer.

[0109] It should be noted that the purpose of the excitation operation is to fully capture the channel correlation.

[0110] C3: Based on the second output and the channel weight vector, the recalibrated feature map is obtained by rescaling.

[0111] Specifically, the final output is obtained by rescaling X′ with the activation s after the excitation operation, which is expressed as:

[0112] X c =F scale (X′ c ,s c )=s c X′ c (7)

[0113] Where X=[X1,X2,…,X c ], F scale is the product on the channel, so it is equivalent to each X′ c Multiply it by its corresponding s c .

[0114] It should be noted that the above conversion operations and steps C1-C3 are detailed implementations of the SE block, which is embedded into the conventional ResNet50 network to form the SE ResNet50 model. Similar to the ResNet50 network, the SE ResNet50 network can be divided into six parts: Stem Block, stage 1, stage 2, stage 3, stage 4, and subsequent processes. The difference between the two network models is that the SE ResNet50 network connects the SE block after the convolution block (Conv block) and the identity block (Identity block) in stages 1-5, that is, the last small block of the residual block.

[0115] In the embodiment of the present application, in step S500 , the corrosion degree status is obtained through the SE block output.

[0116] Specifically, the SEResNet50 network first performs preliminary feature extraction on the input signal using a Stem Block. This Stem Block typically consists of a convolutional layer, a batch normalization layer, and an activation function layer, which captures the signal's essential features. The signal then passes through stages 1 through 5. In the convolutional layer, the signal is convolved with a convolution kernel to extract local features.

[0117] The signal is then processed by the SE block. In the SE block, a squeeze operation is first performed to generate channel statistics through global average pooling. The spatial dimensions of each channel are compressed to obtain the channel statistic Z. Next, an excitation operation is performed, using a sigmoid activation gating mechanism. Two fully connected layers capture the correlation between channels and generate activation values. Finally, the output of the convolutional layer is rescaled with the activation value to obtain the output of the SE block. As the signal passes through each layer of the network, the feature maps at different stages contain different levels of information. Through feature fusion and further convolution and SE block processing, the network can extract more advanced and abstract features that can better reflect the corrosion status of the power grid overhead ground wire.

[0118] For example, the specific content of the output can be a corrosion degree category, such as no corrosion, mild corrosion, moderate corrosion, and severe corrosion; this label is also used in pre-training.

[0119] It should be noted that the SEResNet50 deep learning network employed has powerful feature learning and representation capabilities. It can automatically learn the complex nonlinear relationships between corrosion characteristics and detection signals, constructing a corrosion assessment model with high generalization capabilities. This model can adapt to different corrosion scenarios and operating conditions, effectively improving the accuracy and reliability of assessment results, and providing a more scientific and effective method for corrosion assessment of power grid overhead ground wires.

[0120] Example 3. The above is a schematic scheme of a method for evaluating the corrosion of an overhead ground wire. It should be noted that the technical scheme of this system for evaluating the corrosion of an overhead ground wire is based on the same concept as the technical scheme of the method described above. For details not described in detail in the technical scheme of the system for evaluating the corrosion of an overhead ground wire in this example, please refer to the description of the technical scheme of the method described above.

[0121] This embodiment also provides another overhead ground wire corrosion assessment system, including:

[0122] An acquisition module, used to obtain the original signal;

[0123] The signal reconstruction module is used to obtain the dispersion compensation signal by backpropagating the original signal, extract the phase to obtain the signal symbol coherence factor and waveform correlation factor, and perform signal reconstruction;

[0124] A preprocessing module, used for preprocessing the reconstructed signal to obtain an input signal;

[0125] The input and feature extraction module is used to input the input signal into the trained SE ResNet50 network in segments and connect the SE block at the end of each residual block of the SE ResNet50 network;

[0126] Output module, used to obtain the corrosion degree status through SE block output.

[0127] This embodiment also provides a computer device suitable for overhead ground wire corrosion assessment, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the overhead ground wire corrosion assessment method proposed in the above embodiment.

[0128] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for evaluating the corrosion of an overhead ground wire as proposed in the above embodiment is implemented.

[0129] The storage medium proposed in this embodiment and the method for implementing overhead ground wire corrosion assessment proposed in the above embodiment belong to the same inventive concept. Technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0130] Through the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented with the help of software and necessary general hardware, and of course can also be implemented by hardware. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of the present invention.

[0131] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for evaluating corrosion of an overhead ground wire, characterized in that: include: Get the original signal; The dispersion compensation signal is obtained by backpropagating the original signal, and the phase is extracted to obtain the signal symbol coherence factor and waveform correlation factor for signal reconstruction; Preprocessing the reconstructed signal to obtain an input signal; The input signal is segmented and input into the trained SE ResNet50 network, and the SE block is connected to the end of each residual block of the SE ResNet50 network; The corrosion degree status is obtained through the SE block output.

2. The method for evaluating corrosion of an overhead ground wire according to claim 1, wherein: The dispersion compensation signal is obtained by back propagation, and the phase is extracted to obtain the signal symbol coherence factor and the waveform correlation factor, and the signal is reconstructed, including: The full-focus imaging algorithm based on phase information reconstruction uses the numerical solution of the wavenumber domain dispersion curve to reversely propagate the time domain signal of the Lamb wave. For each imaging point, all original signals are dispersion compensated to obtain the dispersion compensated signal. The symbol coherence factor and waveform correlation factor between the pulse echo and the excitation waveform are extracted from the dispersion compensation signal to reconstruct the scattered signal.

3. The method for evaluating corrosion of an overhead ground wire according to claim 2, wherein: The SE block in the SEResNet50 network includes a conversion operation; The conversion operation performs a general transformation on a first input X of a given first dimension to obtain a second output X′ of a second dimension.

4. The method for evaluating corrosion of an overhead ground wire according to claim 3, wherein: The SE block in the SEResNet50 network also includes: The SE block performs a squeeze operation, takes the second output of the second dimension as input, and compresses the spatial information of each channel into statistics through global average pooling to obtain a channel statistics vector; The SE block performs the excitation operation and uses the s-type activation gating mechanism to excite the channel statistics vector to obtain the channel weight vector; Based on the second output and the channel weight vector, a recalibrated feature map is obtained by rescaling.

5. The method for evaluating corrosion of an overhead ground wire according to claim 4, wherein: The full-focusing imaging algorithm based on phase information reconstruction uses the numerical solution of the wavenumber domain dispersion curve to reversely propagate the time domain signal of the Lamb wave. For each imaging point, all original signals are dispersion compensated to obtain the dispersion compensated signal, which is expressed as: Among them, S pq (r, t) is the Lamb wave signal after dispersion compensation, ω is the angular frequency, k(ω) is the wave number of the angular frequency, i is the imaginary unit, d pq (r) is the propagation distance of the Lamb wave from the transmitter p to the focus r and then to the receiver q.

6. The method for evaluating corrosion of an overhead ground wire according to claim 5, wherein: Extract the symbol coherence factor and waveform correlation factor between the pulse echo and the excitation waveform from the dispersion compensation signal to reconstruct the scattered signal, including: Based on the dispersion compensation signal, the time period that matches the excitation signal is intercepted to extract the focused wave signal; Based on the fusion of phase and waveform correlation of the focused wave signal and the original excitation signal, the symbol coherence factor and waveform correlation factor are extracted and the pulse echo is reconstructed; For each imaging point in the reconstructed pulse echo, the defect imaging map is calculated using the full-focus imaging algorithm.

7. The method for evaluating corrosion of an overhead ground wire according to claim 6, wherein: The reconstructed signal is preprocessed to obtain the input signal, including: De-noising the reconstructed signal; Normalize the denoised signal and segment the normalized signal; Each signal segment is labeled to determine its corresponding corrosion degree category as the input label.

8. An overhead ground wire corrosion assessment system, applying the method according to any one of claims 1 to 7, characterized in that: include: An acquisition module, used to obtain the original signal; The signal reconstruction module is used to obtain the dispersion compensation signal by backpropagating the original signal, extract the phase to obtain the signal symbol coherence factor and waveform correlation factor, and perform signal reconstruction; A preprocessing module, used for preprocessing the reconstructed signal to obtain an input signal; The input and feature extraction module is used to input the input signal into the trained SE ResNet50 network in segments, and connect the SE block to the end of each residual block of the SE ResNet50 network; The output module is used to obtain the corrosion degree status through the SE block output.

9. A computer device comprising: memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the overhead ground wire corrosion assessment method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions, when executed by a processor, implement the steps of the overhead ground wire corrosion assessment method according to any one of claims 1 to 7.

Images