A method and system for evaluating the state of a power fitting

Abstract

The application discloses a kind of electric power fitting state evaluation method and system, belong to ultrasonic flaw detection technical field.The application emits sinusoidal pulse train to one end of electric power fitting, and receives echo signal at monitoring point;Wigner-Ville time-frequency analysis is carried out to echo signal, and time-frequency ridge line is extracted, the frequency interval and candidate time of guided wave and slow wave are divided, and the arrival time of guided wave and slow wave is obtained based on energy peak value;Guided wave signal and slow wave signal in echo signal are separated;Wavelet packet decomposition is carried out to guided wave and slow wave signal respectively, and respective wavelet packet relative distortion vector is constructed;The time-domain total energy of sinusoidal pulse train, guided wave signal and slow wave signal is calculated simultaneously, and damage dissipation coefficient is obtained;Finally, the above features are processed using dual-mode compensation network, and the fitting state evaluation result is output.The application combines the dual-mode information of guided wave and slow wave and compensation mechanism, and significantly improves the accuracy of electric power fitting state evaluation.

Description

Technical Field

[0001] This invention relates to the field of ultrasonic defect detection technology, and specifically to a method and system for assessing the condition of power fittings. Background Technology

[0002] In existing technologies, an ultrasonic transducer typically transmits an ultrasonic excitation signal to the power fitting and receives the echo signal at another location on the power fitting. Since the propagation of ultrasonic waves within the fitting is affected by the material state and structural integrity, existing technologies usually compare the total energy of the received echo signal with the energy of the transmitted signal, construct an energy loss index based on the degree of energy attenuation, and use this index to determine whether the power fitting is abnormal and the degree of abnormality.

[0003] However, when ultrasound propagates inside power fittings, it does not propagate in a single mode, but rather forms multiple wave components with different propagation speeds and characteristics. The guided wave signal mainly propagates along the fitting structure and is more sensitive to the overall structural state; the slow wave signal carries some response information generated by structural defects, interface scattering, and local energy conversion. The received echo signal is actually a composite signal formed by the superposition of multiple propagation modes.

[0004] Current technologies for calculating energy loss indices typically perform energy statistics directly on the overall echo signal without distinguishing and analyzing guided wave signals and slow wave signals within the echo signal. Instead, they uniformly mix the state information carried by different propagation modes into the total energy parameter for processing. The resulting energy loss index only reflects the overall energy change and fails to reflect the differences in the response of different propagation modes to changes in the hardware's state.

[0005] Furthermore, when power fittings are in the defect initiation or minor damage stage, the sensitivity of guided wave signals and slow wave signals to abnormal conditions often differs, while the change in the total echo energy may not be significant. In this case, relying solely on the overall echo energy attenuation for condition assessment can easily mask some abnormal information, making it difficult to accurately reflect the true operating state of the power fittings. Simultaneously, different damage forms may produce similar total energy attenuation results, further increasing the uncertainty in condition determination. Therefore, existing technologies suffer from low accuracy in power fitting condition assessment. Summary of the Invention

[0006] In view of the above-mentioned shortcomings in the prior art, the present invention provides a method and system for assessing the condition of power fittings, which solves the problem of low accuracy in assessing the condition of power fittings in the prior art.

[0007] To achieve the above-mentioned objectives, the technical solution adopted by this invention is: a method for assessing the condition of power fittings, comprising the following steps: S1. A sinusoidal pulse train modulated by the Hanning window is transmitted to one end of the power fitting, and the echo signal is received at the monitoring point. S2. Perform Wigner-Ville time-frequency analysis on the echo signal, extract the time-frequency ridge, divide the frequency range and candidate time corresponding to the guided wave and slow wave, and obtain the arrival time of the guided wave and the arrival time of the slow wave based on the energy peak. S3. Based on the arrival time of the guided wave and the arrival time of the slow wave, separate the guided wave signal and the slow wave signal from the echo signal; S4. Perform wavelet packet decomposition on the guided wave signal to obtain the guided wave energy of each node in each layer and construct the relative distortion vector of the guided wavelet packet. S5. Perform wavelet packet decomposition on the slow wave signal to obtain the slow wave energy of each node in each layer and construct the relative distortion vector of the slow wavelet packet. S6. Obtain the total time-domain energy of the sinusoidal pulse train, slow wave signal and guided wave signal respectively, and calculate the damage dissipation coefficient; S7. A dual-modal compensation network is used to process the relative distortion vector of the guided wavelet packet, the relative distortion vector of the slow wavelet packet, and the damage dissipation coefficient to obtain the hardware condition assessment results.

[0008] Furthermore, S2 includes the following sub-steps: S21. At the monitoring point, the Wigner-Ville distribution is used to obtain the energy distribution matrix of the echo signal in the time-frequency two-dimensional plane. S22. Extract time-frequency ridges from the energy distribution matrix; S23. Find the guided wave frequency range and the slow wave frequency range in the time-frequency ridge, extract the corresponding time of the guided wave frequency range to obtain the guided wave candidate time set, and extract the corresponding time of the slow wave frequency range to obtain the slow wave candidate time set. S24. In the candidate time set of guided waves, find the time corresponding to the maximum energy value in the energy distribution matrix to obtain the arrival time of the guided waves; S25. In the set of candidate times for slow waves, find the time corresponding to the maximum energy value in the energy distribution matrix to obtain the arrival time of the slow waves.

[0009] Furthermore, S3 includes the following sub-steps: S31. Taking the waveguide arrival time as the center, and based on the preset waveguide time width, obtain the waveguide arrival time range; S32. Using the slow wave arrival time as the center, and based on the preset slow wave time width, obtain the slow wave arrival time range; S33. Within the waveguide arrival time range, extract the corresponding segment from the echo signal to obtain the waveguide signal; S34. Within the arrival time range of the slow wave, extract the corresponding segment from the echo signal to obtain the slow wave signal.

[0010] Furthermore, S4 includes the following sub-steps: S41. Perform wavelet packet decomposition on the guided wave signal to obtain the guided wavelet packet coefficients of each node in each layer. S42. Calculate the guided wave energy of each node in each layer based on the guided wavelet packet coefficients of each node in each layer. S43. Calculate the relative rate of change of guided wave energy at each node in each layer based on the difference between the guided wave energy at each node and the stored health reference guided wave energy. S44. Take the average of the relative guided wave energy change rate of each node in each layer to obtain the relative distortion value of the guided wavelet packet of each layer. S45. Construct the wavelet packet relative distortion vector by using the relative distortion values ​​of the wavelet packets of each layer as elements.

[0011] Furthermore, S5 includes the following sub-steps: S51. Perform wavelet packet decomposition on the slow wave signal to obtain the slow wavelet packet coefficients of each node in each layer. S52. Calculate the slow wave energy of each node in each layer based on the slow wavelet packet coefficients of each node in each layer. S53. Calculate the relative slow wave energy change rate of each node in each layer based on the difference between the slow wave energy of each node and the stored health baseline slow wave energy. S54. Take the average of the relative slow wave energy change rate of each node in each layer to obtain the relative distortion value of the slow wave wavelet packet. S55. Construct a slow wavelet packet relative distortion vector by using the relative distortion values ​​of the slow wavelet packets of each layer as elements.

[0012] Furthermore, S6 includes the following sub-steps: S61. Add the energy of the guided wave signal at each moment to obtain the total energy of the guided wave in the time domain; S62. Add the energy of the slow wave signal at each moment to obtain the total energy of the slow wave in the time domain. S63. Add up the energy of the sinusoidal pulse train at each moment to obtain the total energy in the transmission time domain; S64. Calculate the waveguide loss anomaly based on the difference between the total energy in the transmission time domain and the total energy in the guided wave time domain; S65. Calculate the damage dissipation coefficient based on the waveguide loss anomaly and the total energy in the slow wave time domain.

[0013] Furthermore, the formula for calculating the damage dissipation coefficient in S65 is as follows: , in, The damage dissipation coefficient is... The total energy in the time domain at launch. The total time-domain energy of the guided wave. This represents the total energy in the slow-wave time domain.

[0014] Furthermore, S7 includes the following sub-steps: S71. The first fully connected layer is used to process the relative distortion vector of the guided wavelet packet to obtain the hidden layer feature vector of the guided wave distortion. S72. The second fully connected layer is used to process the relative distortion vector of the slow wavelet packet to obtain the feature vector of the slow wavelet distortion hidden layer. S73. Add the waveguide distortion hidden layer feature vector and the slow wave distortion hidden layer feature vector using adder A1 to obtain the distortion hidden layer feature fusion vector. S74. The distorted hidden layer feature fusion vector is processed using the third fully connected layer to obtain the initial state value; S75. The damage dissipation coefficient is processed using the fourth fully connected layer to obtain the damage dissipation hidden layer feature vector. S76. The damage dissipation hidden layer feature vector is processed using the fifth fully connected layer to obtain the damage coefficient; S77. The initial state value is compensated by the damage coefficient through the compensation output unit to obtain the hardware condition assessment result.

[0015] Furthermore, the expression for the compensation output unit in S77 is: , in, This is the result of the hardware condition assessment. The initial state value, The damage coefficient is... It is a natural constant.

[0016] A power fitting condition assessment system, comprising: The transmitting and receiving subsystem is used to transmit a sinusoidal pulse train modulated by the Hanning window to one end of the power fitting and to receive the echo signal at the monitoring point. The arrival time acquisition subsystem is used to perform Wigner-Ville time-frequency analysis on the echo signal, extract the time-frequency ridge, divide the frequency range and candidate time corresponding to the guided wave and slow wave, and obtain the arrival time of the guided wave and the arrival time of the slow wave based on the energy peak. The signal separation subsystem is used to separate the guided wave signal and the slow wave signal from the echo signal based on the guided wave arrival time and the slow wave arrival time. The guided wave distortion vector construction subsystem is used to perform wavelet packet decomposition on the guided wave signal, obtain the guided wave energy of each node in each layer, and construct the relative distortion vector of the guided wavelet packet. The slow wave distortion vector construction subsystem is used to perform wavelet packet decomposition on the slow wave signal, obtain the slow wave energy of each node in each layer, and construct the relative distortion vector of the slow wave wavelet packet. The damage dissipation coefficient calculation subsystem is used to obtain the total energy in the time domain for sinusoidal pulse trains, slow wave signals and guided wave signals respectively, and to calculate the damage dissipation coefficient. The evaluation subsystem is used to process the relative distortion vector of the guided wavelet packet, the relative distortion vector of the slow wavelet packet, and the damage dissipation coefficient using a dual-modal compensation network to obtain the hardware condition evaluation results.

[0017] The beneficial effects of this invention are as follows: 1. This invention achieves effective separation of the two propagation modes by performing Wigner-Ville time-frequency analysis on the echo signal and accurately dividing the frequency ranges and arrival times of guided and slow waves based on the time-frequency ridge. Compared with existing technologies that directly calculate the total energy of the overall echo, this invention can extract the state information from the guided wave signal and the slow wave signal separately, avoiding the overlap of response characteristics of different propagation modes, thereby significantly improving the ability to identify abnormal states of power fittings (especially the defect initiation or minor damage stages).

[0018] 2. In the early stages of defects or minor damage, the overall change in total echo energy is often insignificant. This invention, by constructing relative distortion vectors for guided wavelet packets and slow wavelet packets respectively, can capture subtle local energy changes and distortion features in each propagation mode, solving the problem in existing technologies where anomalous information is masked due to insignificant changes in total energy.

[0019] 3. Different damage types may lead to similar overall energy attenuation results, thus increasing the risk of misjudgment. This invention introduces a damage dissipation coefficient and combines it with the relative distortion vectors of wavelet packets of guided and slow waves to form a multi-dimensional feature description system. Finally, through fusion processing using a dual-modal compensation network, it can effectively distinguish the different effects of different damage types on the two propagation modes, improving the accuracy of power fitting condition assessment. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method for assessing the condition of electrical fittings.

[0021] Figure 2 This is a schematic diagram of the launch of the fittings; Figure 3 The waveform diagram is for a sinusoidal pulse train. Detailed Implementation

[0022] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0023] Example 1, such as Figure 1 As shown, a method for assessing the condition of electrical fittings includes the following steps: S1. A sinusoidal pulse train modulated by the Hanning window is transmitted to one end of the power fitting, and the echo signal is received at the monitoring point. S2. Perform Wigner-Ville time-frequency analysis on the echo signal, extract the time-frequency ridge, divide the frequency range and candidate time corresponding to the guided wave and slow wave, and obtain the arrival time of the guided wave and the arrival time of the slow wave based on the energy peak. S3. Based on the arrival time of the guided wave and the arrival time of the slow wave, separate the guided wave signal and the slow wave signal from the echo signal; S4. Perform wavelet packet decomposition on the guided wave signal to obtain the guided wave energy of each node in each layer and construct the relative distortion vector of the guided wavelet packet. S5. Perform wavelet packet decomposition on the slow wave signal to obtain the slow wave energy of each node in each layer and construct the relative distortion vector of the slow wavelet packet. S6. Obtain the total time-domain energy of the sinusoidal pulse train, slow wave signal and guided wave signal respectively, and calculate the damage dissipation coefficient; S7. A dual-modal compensation network is used to process the relative distortion vector of the guided wavelet packet, the relative distortion vector of the slow wavelet packet, and the damage dissipation coefficient to obtain the hardware condition assessment results.

[0024] In this embodiment, the expression for the sinusoidal pulse train is: , in, It is a sinusoidal pulse train. The center frequency of the ultrasound. The duration of a single pulse. The pulse repetition period, The number of pulses. The pulse number. It is a time variable.

[0025] exist Figure 2 In this process, a sinusoidal pulse train is emitted from one end of the fitting, and the echo signal is received at the monitoring point at the other end. Figure 3 The waveform of a sinusoidal pulse train is shown, with the ultrasound center frequency set to 100kHz. Figure 3 It can be seen that the sinusoidal pulse train modulated by the Hanning window has a smooth envelope and a finite time-domain duration, which is beneficial for generating high-energy ultrasonic pulses at the transmitting end and reducing waveform tailing and signal aliasing.

[0026] In this embodiment, S2 includes the following sub-steps: S21. At the monitoring point, the Wigner-Ville distribution is used to obtain the energy distribution matrix of the echo signal in the time-frequency two-dimensional plane. S22. Extract time-frequency ridges from the energy distribution matrix; S23. Find the guided wave frequency range and the slow wave frequency range in the time-frequency ridge, extract the corresponding time of the guided wave frequency range to obtain the guided wave candidate time set, and extract the corresponding time of the slow wave frequency range to obtain the slow wave candidate time set. S24. In the candidate time set of guided waves, find the time corresponding to the maximum energy value in the energy distribution matrix to obtain the arrival time of the guided waves; S25. In the set of candidate times for slow waves, find the time corresponding to the maximum energy value in the energy distribution matrix to obtain the arrival time of the slow waves.

[0027] In this embodiment, the formula for calculating the Wigner-Ville distribution in S21 is: , in, The energy distribution matrix is... For time variables, For frequency variables, To postpone echo signal, In advance The complex conjugate of the echo signal, For integration variables, It is a natural constant. It is the imaginary unit.

[0028] In this embodiment, the expression for the time-frequency ridge in S22 is: , in, For time-frequency ridges, To traverse all frequencies Find the frequency corresponding to the maximum energy value as The instantaneous dominant frequency.

[0029] In this embodiment, the guided wave frequency range is set to [90kHz, 110kHz], and the slow wave frequency range is set to [30kHz, 70kHz].

[0030] In S24, the energy values ​​belonging to the guided wave candidate time set range of the energy distribution matrix are extracted, and the maximum value is found to obtain the guided wave arrival time. In S25, the energy values ​​belonging to the slow wave candidate time set range of the energy distribution matrix are extracted, and the maximum value is found to obtain the slow wave arrival time.

[0031] In this embodiment, S3 includes the following sub-steps: S31. Taking the waveguide arrival time as the center, and based on the preset waveguide time width, obtain the waveguide arrival time range; S32. Using the slow wave arrival time as the center, and based on the preset slow wave time width, obtain the slow wave arrival time range; S33. Within the waveguide arrival time range, extract the corresponding segment from the echo signal to obtain the waveguide signal; S34. Within the arrival time range of the slow wave, extract the corresponding segment from the echo signal to obtain the slow wave signal.

[0032] In this embodiment, the waveguide arrival time range is: Slow wave arrival time range: ,in, For the guided wave arrival time, To preset the waveguide time width, For slow wave arrival time, To preset the slow wave time width, , .

[0033] This invention achieves accurate acquisition of candidate times for guided and slow waves by performing Wigner-Ville time-frequency analysis on the echo signal and extracting the time-frequency ridge. It determines the arrival time of the guided and slow waves by combining the energy peak value, and then extracts corresponding segments from the echo signal with the arrival time as the center of the effective time range, thereby obtaining the guided wave signal and the slow wave signal. This invention can effectively separate signals of different propagation modes, reduce mutual interference, and improve the accuracy of signal analysis.

[0034] In this embodiment, S4 includes the following sub-steps: S41. Perform wavelet packet decomposition on the guided wave signal to obtain the guided wavelet packet coefficients of each node in each layer. S42. Calculate the guided wave energy at each node in each layer based on the guided wavelet packet coefficients: ,in, For the first Layer The guided wave energy of the node, For the first Layer Node number Guided wavelet packet coefficients at sampling points, The number of sampling points, The sampling point number, For the node number, The layer number; S43. Calculate the relative rate of change of guided wave energy at each node in each layer based on the difference between the guided wave energy at each node and the stored health reference guided wave energy. S44. Take the average of the relative guided wave energy change rate of each node in each layer to obtain the relative distortion value of the guided wavelet packet of each layer. S45. Construct the wavelet packet relative distortion vector by using the relative distortion values ​​of the wavelet packets of each layer as elements.

[0035] The expression for the relative distortion value of the guided wavelet packet is: ,in, For the first Relative distortion of wavelet packets in guided layers. For the first Layer Nodal guided wave energy, For the first Layer The health reference guided wave energy stored in the node For the number of each node, For the layer number, This is used to avoid parameters that are 0.

[0036] This invention decomposes guided wave signals using wavelet packets, refining the guided wave energy to different decomposition levels and frequency band nodes. It then compares the current guided wave energy distribution with a baseline guided wave energy distribution under healthy conditions to construct a relative distortion vector of the guided wavelet packets. This allows it to reflect the redistribution characteristics of guided wave energy across frequency bands. Compared to analysis using only the total guided wave energy, the S4 step-by-step approach not only characterizes the overall energy attenuation during guided wave propagation but also reveals the impact of abnormal states such as defects, cracks, or loosening on the propagation characteristics of different frequency components. This enhances the sensitivity to minor damage and early anomalies, improving the characterization ability of local damage features in power fittings.

[0037] In this embodiment, S5 includes the following sub-steps: S51. Perform wavelet packet decomposition on the slow wave signal to obtain the slow wavelet packet coefficients of each node in each layer. S52. Calculate the slow wave energy of each node in each layer based on the slow wavelet packet coefficients of each node: ,in, For the first Layer Slow-wave energy of the node For the first Layer Node number Slow wavelet packet coefficients at the sampling points, The number of sampling points, The sampling point number, For the node number, The layer number; S53. Calculate the relative slow wave energy change rate of each node in each layer based on the difference between the slow wave energy of each node and the stored health baseline slow wave energy. S54. Take the average of the relative slow wave energy change rate of each node in each layer to obtain the relative distortion value of the slow wave wavelet packet. S55. Construct a slow wavelet packet relative distortion vector by using the relative distortion values ​​of the slow wavelet packets of each layer as elements.

[0038] The expression for the relative distortion value of the slow wavelet packet is: ,in, For the first Relative distortion value of slow wavelet packet in layered slow wavelet, For the first Layer Nodal slow wave energy, For the first Layer The node stores the health benchmark slow-wave energy. For the number of each node, For the layer number, This is used to avoid parameters that are 0.

[0039] This invention performs wavelet packet decomposition on slow-wave signals, refining the slow-wave energy to different decomposition levels and frequency band nodes. By comparing the slow-wave energy with a healthy baseline, the relative slow-wave energy change rate at each level node is calculated, thereby constructing a relative distortion vector for the slow-wave wavelet packet. This method can reveal the local energy change characteristics of slow waves during propagation. Compared to simple total energy analysis, it can more sensitively reflect low-frequency abnormal energy distribution changes caused by minor damage, structural loosening, or cracks inside power fittings, helping to detect abnormal signals early and improve the accuracy and reliability of condition assessment.

[0040] In this embodiment, S6 includes the following sub-steps: S61. Add the energy of the guided wave signal at each moment to obtain the total energy of the guided wave in the time domain; S62. Add the energy of the slow wave signal at each moment to obtain the total energy of the slow wave in the time domain. S63. Add up the energy of the sinusoidal pulse train at each moment to obtain the total energy in the transmission time domain; S64. Calculate the waveguide loss anomaly based on the difference between the total energy in the transmission time domain and the total energy in the guided wave time domain; S65. Calculate the damage dissipation coefficient based on the waveguide loss anomaly and the total energy in the slow wave time domain.

[0041] In S61, S62, and S63, the energy is the square of the amplitude at each moment.

[0042] In this embodiment, the formula for calculating the damage dissipation coefficient in S65 is: , in, The damage dissipation coefficient is... The total energy in the time domain at launch. The total time-domain energy of the guided wave. This represents the total energy in the slow-wave time domain.

[0043] By simultaneously considering waveguide loss and slow wave energy changes, this invention can not only quantify the energy attenuation caused by defects during waveguide propagation, but also capture the abnormal accumulation of slow wave signals, thereby achieving sensitive detection of internal damage to power fittings.

[0044] In this embodiment, the dual-modal compensation network includes: a first fully connected layer, a second fully connected layer, an adder A1, a third fully connected layer, a fourth fully connected layer, a fifth fully connected layer, and a compensation output unit. S7 includes the following sub-steps: S71. The first fully connected layer is used to process the relative distortion vector of the guided wavelet packet to obtain the hidden layer feature vector of the guided wave distortion. S72. The second fully connected layer is used to process the relative distortion vector of the slow wavelet packet to obtain the feature vector of the slow wavelet distortion hidden layer. S73. Add the waveguide distortion hidden layer feature vector and the slow wave distortion hidden layer feature vector using adder A1 to obtain the distortion hidden layer feature fusion vector. S74. The distorted hidden layer feature fusion vector is processed using the third fully connected layer to obtain the initial state value; S75. The damage dissipation coefficient is processed using the fourth fully connected layer to obtain the damage dissipation hidden layer feature vector. S76. The damage dissipation hidden layer feature vector is processed using the fifth fully connected layer to obtain the damage coefficient; S77. The initial state value is compensated by the damage coefficient through the compensation output unit to obtain the hardware condition assessment result.

[0045] In this embodiment, the expression for the compensation output unit in S77 is: , in, This is the result of the hardware condition assessment. The initial state value, The damage coefficient is... It is a natural constant.

[0046] In this embodiment, the number of output nodes of the first fully connected layer is set to 8, the number of output nodes of the second fully connected layer is set to 8, the number of output nodes of the third fully connected layer is set to 1, the number of output nodes of the fourth fully connected layer is set to 8, and the number of output nodes of the fifth fully connected layer is set to 1.

[0047] This invention utilizes a first fully connected layer and a second fully connected layer to extract high-dimensional feature information from the relative distortion vectors of guided wavelet packets and slow wavelet packets, respectively. A dual-mode distortion feature fusion is achieved through adder A1, thus simultaneously characterizing guided wave propagation anomalies and slow wave propagation anomalies. Subsequently, a third fully connected layer is used to establish a nonlinear mapping relationship between distortion features and fitting status, obtaining initial state values. Further, a fourth and fifth fully connected layer are used to extract features from the damage dissipation coefficient, obtaining a damage coefficient reflecting the degree of energy dissipation. A compensation output unit is then used to compensate for the initial state values. This invention not only utilizes the energy redistribution characteristics of guided and slow wave frequency bands but also integrates the damage dissipation information characterized by guided wave loss and slow wave energy anomalies. This enables the state assessment results to simultaneously possess frequency band distortion sensing and energy dissipation sensing capabilities, thereby improving the accuracy and reliability of identifying and assessing abnormal states such as early damage, local cracks, and loose connections in power fittings.

[0048] In this embodiment, At that time, the hardware was in good condition; when At that time, the hardware showed a slight abnormality; when At that time, the hardware showed obvious damage.

[0049] Example 2: A power fitting condition assessment system, comprising: The transmitting and receiving subsystem is used to transmit a sinusoidal pulse train modulated by the Hanning window to one end of the power fitting and to receive the echo signal at the monitoring point. The arrival time acquisition subsystem is used to perform Wigner-Ville time-frequency analysis on the echo signal, extract the time-frequency ridge, divide the frequency range and candidate time corresponding to the guided wave and slow wave, and obtain the arrival time of the guided wave and the arrival time of the slow wave based on the energy peak. The signal separation subsystem is used to separate the guided wave signal and the slow wave signal from the echo signal based on the guided wave arrival time and the slow wave arrival time. The guided wave distortion vector construction subsystem is used to perform wavelet packet decomposition on the guided wave signal, obtain the guided wave energy of each node in each layer, and construct the relative distortion vector of the guided wavelet packet. The slow wave distortion vector construction subsystem is used to perform wavelet packet decomposition on the slow wave signal, obtain the slow wave energy of each node in each layer, and construct the relative distortion vector of the slow wave wavelet packet. The damage dissipation coefficient calculation subsystem is used to obtain the total energy in the time domain for sinusoidal pulse trains, slow wave signals and guided wave signals respectively, and to calculate the damage dissipation coefficient. The evaluation subsystem is used to process the relative distortion vector of the guided wavelet packet, the relative distortion vector of the slow wavelet packet, and the damage dissipation coefficient using a dual-modal compensation network to obtain the hardware condition evaluation results.

[0050] Example 2 is implemented in the same way as Example 1.

[0051] This invention achieves effective separation of the two propagation modes by performing Wigner-Ville time-frequency analysis on the echo signal and accurately dividing the frequency ranges and arrival times of guided and slow waves based on the time-frequency ridge. Compared with existing technologies that directly calculate the total energy of the overall echo, this invention can extract the state information from the guided wave signal and the slow wave signal separately, avoiding the overlap of response characteristics of different propagation modes, thereby significantly improving the ability to identify abnormal states of power fittings (especially the defect initiation or minor damage stages).

[0052] In the early stages of defects or minor damage, the overall change in total echo energy is often insignificant. This invention, by constructing relative distortion vectors for guided wavelet packets and slow wavelet packets respectively, can capture subtle local energy changes and distortion features in each propagation mode, solving the problem in existing technologies where anomalous information is masked due to insignificant changes in total energy.

[0053] Different damage types can lead to similar overall energy attenuation results, increasing the risk of misjudgment. This invention introduces a damage dissipation coefficient and combines it with the relative distortion vectors of wavelet packets of guided and slow waves to form a multi-dimensional feature description system. Finally, a dual-modal compensation network is used for fusion processing, which can effectively distinguish the different effects of different damage types on the two propagation modes, improving the accuracy of power fitting condition assessment.

[0054] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing the condition of power fittings, characterized in that, Includes the following steps: S1. A sinusoidal pulse train modulated by the Hanning window is transmitted to one end of the power fitting, and the echo signal is received at the monitoring point. S2. Perform Wigner-Ville time-frequency analysis on the echo signal, extract the time-frequency ridge, divide the frequency intervals and candidate times corresponding to the guided wave and slow wave, and obtain the arrival time of the guided wave and the arrival time of the slow wave based on the energy peak. S3. Based on the arrival time of the guided wave and the arrival time of the slow wave, separate the guided wave signal and the slow wave signal from the echo signal; S4. Perform wavelet packet decomposition on the guided wave signal to obtain the guided wave energy of each node in each layer and construct the relative distortion vector of the guided wavelet packet. S5. Perform wavelet packet decomposition on the slow wave signal to obtain the slow wave energy of each node in each layer and construct the relative distortion vector of the slow wavelet packet. S6. Obtain the total time-domain energy for the sinusoidal pulse train, slow wave signal, and guided wave signal respectively, and calculate the damage dissipation coefficient: ,in, The damage dissipation coefficient is... The total energy in the time domain at launch. The total time-domain energy of the guided wave. This represents the total energy in the slow-wave time domain. S7. A dual-modal compensation network is used to process the relative distortion vector of the guided wavelet packet, the relative distortion vector of the slow wavelet packet, and the damage dissipation coefficient to obtain the hardware condition assessment results: The wavelet packet relative distortion vector is processed by the first fully connected layer to obtain the wavelet distortion hidden layer feature vector; The relative distortion vector of the slow wavelet packet is processed by the second fully connected layer to obtain the feature vector of the slow wavelet distortion hidden layer; Adder A1 is used to add the eigenvectors of the guided wave distortion hidden layer and the slow wave distortion hidden layer to obtain the fused eigenvector of the distortion hidden layer. The distorted hidden layer feature fusion vector is processed using a third fully connected layer to obtain the initial state value; The damage dissipation coefficient is processed using the fourth fully connected layer to obtain the damage dissipation hidden layer feature vector; The damage coefficient is obtained by processing the feature vector of the damage dissipation hidden layer using the fifth fully connected layer. The initial state value is compensated by the damage coefficient through the compensation output unit, and the hardware condition assessment result is obtained: ,in, This is the result of the hardware condition assessment. The initial state value, The damage coefficient is... It is a natural constant.

2. The method for assessing the condition of power fittings according to claim 1, characterized in that, S2 includes the following steps: S21. At the monitoring point, the Wigner-Ville distribution is used to obtain the energy distribution matrix of the echo signal in the time-frequency two-dimensional plane. S22. Extract time-frequency ridges from the energy distribution matrix; S23. Find the guided wave frequency range and the slow wave frequency range in the time-frequency ridge, extract the corresponding time of the guided wave frequency range to obtain the guided wave candidate time set, and extract the corresponding time of the slow wave frequency range to obtain the slow wave candidate time set. S24. In the candidate time set of guided waves, find the time corresponding to the maximum energy value in the energy distribution matrix to obtain the arrival time of the guided waves; S25. In the set of candidate times for slow waves, find the time corresponding to the maximum energy value in the energy distribution matrix to obtain the arrival time of the slow waves.

3. The method for assessing the condition of power fittings according to claim 1, characterized in that, S3 includes the following steps: S31. Taking the waveguide arrival time as the center, and based on the preset waveguide time width, obtain the waveguide arrival time range; S32. Using the slow wave arrival time as the center, and based on the preset slow wave time width, obtain the slow wave arrival time range; S33. Within the waveguide arrival time range, extract the corresponding segment from the echo signal to obtain the waveguide signal; S34. Within the arrival time range of the slow wave, extract the corresponding segment from the echo signal to obtain the slow wave signal.

4. The method for assessing the condition of power fittings according to claim 1, characterized in that, S4 includes the following steps: S41. Perform wavelet packet decomposition on the guided wave signal to obtain the guided wavelet packet coefficients of each node in each layer. S42. Calculate the guided wave energy of each node in each layer based on the guided wavelet packet coefficients of each node in each layer. S43. Calculate the relative rate of change of guided wave energy at each node in each layer based on the difference between the guided wave energy at each node and the stored health reference guided wave energy. S44. Take the average of the relative guided wave energy change rate of each node in each layer to obtain the relative distortion value of the guided wavelet packet of each layer. S45. Construct the wavelet packet relative distortion vector by using the relative distortion values ​​of the wavelet packets of each layer as elements.

5. The method for assessing the condition of power fittings according to claim 1, characterized in that, S5 includes the following steps: S51. Perform wavelet packet decomposition on the slow wave signal to obtain the slow wavelet packet coefficients of each node in each layer. S52. Calculate the slow wave energy of each node in each layer based on the slow wavelet packet coefficients of each node in each layer. S53. Calculate the relative slow wave energy change rate of each node in each layer based on the difference between the slow wave energy of each node and the stored health baseline slow wave energy. S54. Take the average of the relative slow wave energy change rate of each node in each layer to obtain the relative distortion value of the slow wave wavelet packet. S55. Construct a slow wavelet packet relative distortion vector by using the relative distortion values ​​of the slow wavelet packets of each layer as elements.

6. The method for assessing the condition of power fittings according to claim 1, characterized in that, S6 includes the following steps: S61. Add the energy of the guided wave signal at each moment to obtain the total energy of the guided wave in the time domain; S62. Add the energy of the slow wave signal at each moment to obtain the total energy of the slow wave in the time domain. S63. Add up the energy of the sinusoidal pulse train at each moment to obtain the total energy in the transmission time domain; S64. Calculate the waveguide loss anomaly based on the difference between the total energy in the transmission time domain and the total energy in the guided wave time domain; S65. Calculate the damage dissipation coefficient based on the waveguide loss anomaly and the total energy in the slow wave time domain.

7. A power fitting condition assessment system, implementing the power fitting condition assessment method according to any one of claims 1 to 6, characterized in that, include: The transmitting and receiving subsystem is used to transmit a sinusoidal pulse train modulated by the Hanning window to one end of the power fitting and to receive the echo signal at the monitoring point. The arrival time acquisition subsystem is used to perform Wigner-Ville time-frequency analysis on the echo signal, extract the time-frequency ridge, divide the frequency range and candidate time corresponding to the guided wave and slow wave, and obtain the arrival time of the guided wave and the arrival time of the slow wave based on the energy peak. The signal separation subsystem is used to separate the guided wave signal and the slow wave signal from the echo signal based on the guided wave arrival time and the slow wave arrival time. The guided wave distortion vector construction subsystem is used to perform wavelet packet decomposition on the guided wave signal, obtain the guided wave energy of each node in each layer, and construct the relative distortion vector of the guided wavelet packet. The slow wave distortion vector construction subsystem is used to perform wavelet packet decomposition on the slow wave signal, obtain the slow wave energy of each node in each layer, and construct the relative distortion vector of the slow wave wavelet packet. The damage dissipation coefficient calculation subsystem is used to obtain the total energy in the time domain for sinusoidal pulse trains, slow wave signals and guided wave signals respectively, and to calculate the damage dissipation coefficient. The evaluation subsystem is used to process the relative distortion vector of the guided wavelet packet, the relative distortion vector of the slow wavelet packet, and the damage dissipation coefficient using a dual-modal compensation network to obtain the hardware condition evaluation results.

Images