Cable aging on-line monitoring method and device based on spread spectrum time domain reflection method

Abstract

The application provides a cable aging online monitoring method and device based on a spread spectrum time domain reflection method, which comprises the following steps: establishing a frequency-variable distribution parameter model based on the physical structure of a cable, performing background noise spectrum scanning on the cable to identify a spectrum hole, and taking the spectrum hole as a working frequency band of a detection signal; generating a detection signal matched with the working frequency band, and injecting the detection signal into a live cable; collecting echo signals of the cable, performing adaptive signal decomposition on the echo signals to obtain purified echoes, performing cross-correlation operation on the purified echoes and the detection signal, and extracting reflection response features; performing phase dispersion compensation on the reflection response features to obtain dispersion-corrected reflection responses; extracting multi-dimensional feature vectors of the dispersion-corrected reflection responses, inputting the multi-dimensional feature vectors into a physical information neural network, and inversely obtaining insulation aging quantitative indexes and cable aging grades of each position of the cable.

Description

Technical Field

[0001] This application mainly relates to the field of cable condition monitoring technology, and in particular to a method and device for online monitoring of cable aging based on spread spectrum time-domain reflectometry. Background Technology

[0002] Cables, as the core carriers of energy transmission and signal control in nuclear power plants, are the "blood vessels" and "nerves" of these plants. Because cables in nuclear power plants are exposed to harsh environments such as high temperatures and high radiation for extended periods, their insulation materials undergo slow but irreversible aging, leading to decreased dielectric strength and increased leakage current. In severe cases, this can cause short circuits or even fires, endangering nuclear safety. Therefore, online monitoring and accurate assessment of cable aging conditions have become a rigid requirement in the industry.

[0003] Some solutions employ Spread Spectrum Time Domain Reflectometry (SSTDR) to monitor the aging status of cables. The core principle of SSTDR is to inject a pseudo-random spread spectrum probe signal into the cable. When the signal encounters impedance discontinuities during its transmission along the cable (insulation aging causes local dielectric parameter changes, leading to impedance discontinuities), it generates reflected echoes. By analyzing the time difference, amplitude, and phase characteristics between the echo and the original probe signal, the aging location and status can be determined. However, existing SSTDR systems face three major bottlenecks: first, strong electromagnetic noise at nuclear power plant sites easily drowns out weak aging reflection signals; second, the dispersion effect caused by insulation aging broadens the reflection peak and significantly reduces positioning accuracy; and third, there is a lack of a quantitative mapping model from reflection signal characteristics to physical aging indicators, making it difficult to support accurate operation and maintenance throughout the cable's entire lifecycle.

[0004] Therefore, there is an urgent need for a new method that can locate and quantitatively assess the aging status of cables online with high precision. Summary of the Invention

[0005] The technical problem to be solved by this application is to provide a method and device for online monitoring of cable aging based on spread spectrum time-domain reflectometry, so as to solve the problem of low accuracy in aging location positioning in existing cable aging monitoring technologies.

[0006] To address the aforementioned technical problems, this application provides an online monitoring method for cable aging based on spread spectrum time-domain reflectometry, comprising: establishing a frequency-varying distributed parameter model based on the physical structure of the cable; performing background noise spectrum scanning on the cable to identify spectral holes, and using the spectral holes as the operating frequency band of the detection signal; generating a detection signal matching the operating frequency band, and injecting the detection signal into the energized cable; acquiring the echo signal of the cable; performing adaptive signal decomposition on the echo signal to remove non-stationary background noise to obtain a purified echo; performing cross-correlation operation on the purified echo and the detection signal to extract reflection response features; performing phase dispersion compensation on the reflection response features to obtain a dispersion-corrected reflection response; extracting the multidimensional feature vector of the dispersion-corrected reflection response, inputting it into a physical information neural network, and inverting to obtain the insulation aging quantification index and cable aging level at each location of the cable, wherein the loss function of the physical information neural network includes the physical constraint loss of the transmission line telegraph equations.

[0007] Optionally, generating a detection signal matching the operating frequency band includes: generating an original spread spectrum sequence with sharp autocorrelation characteristics, modulating the original spread spectrum sequence, and obtaining a detection signal matching the operating frequency band.

[0008] Optionally, the original spread spectrum sequence is subjected to binary phase shift keying modulation to obtain the detection signal as follows:

[0009] Where t is the time variable, The detection signal, The original spreading sequence is denoted as n, where n is the chip number of the original spreading sequence. For pulse shaping function, For chip cycle, The carrier frequency, the carrier frequency Select within the operating frequency band.

[0010] Optionally, the center frequency of the spectral hole is used as the carrier frequency. The selected value ensures that the detected signal and background noise are orthogonal and avoid each other in the frequency domain.

[0011] Optionally, the adaptive signal decomposition is at least one of variational mode decomposition, adaptive noise complete set empirical mode decomposition, wavelet packet transform, or nonlinear filtering operator based on autoencoder network.

[0012] Optionally, the adaptive signal decomposition is performed as variational mode decomposition. Performing variational mode decomposition on the echo signal to remove non-stationary background noise and obtain a purified echo includes: adaptively determining the number of decomposition levels of the variational mode decomposition based on the center frequency of the echo signal; decomposing the echo signal into a number of narrowband components equal to the number of decomposition levels by solving a constrained variational problem, and minimizing the sum of the bandwidths of each narrowband component; removing narrowband components that do not overlap with the operating frequency band of the detection signal, and reconstructing the remaining narrowband components to obtain the purified echo.

[0013] Optionally, the phase dispersion compensation is at least one of the following: fractional Fourier transform, time-reversal mirror compensation, frequency domain all-pass filter phase correction, or compensation scheme based on time-frequency distribution energy ridge extraction.

[0014] Optionally, the phase dispersion compensation is a fractional Fourier transform. A fractional Fourier transform is performed on the reflection response characteristics, and the focusing order is determined by gradient search to cancel the phase dispersion, thereby obtaining the dispersion-corrected reflection response.

[0015] Optionally, the method for obtaining the dispersion-corrected reflection response includes: transforming the reflection response features to a fractional-order domain and calculating the kurtosis of the relevant peaks in the fractional-order domain; adjusting the rotation angle of the fractional-order Fourier transform through gradient search to maximize the kurtosis, and determining the focusing order using the rotation angle corresponding to the maximum value; performing a fractional-order Fourier transform on the reflection response features again using the focusing order to cancel the phase quadratic nonlinear distortion caused by insulation aging, refocus the energy of the relevant peaks, and obtain the dispersion-corrected reflection response.

[0016] Optionally, the loss function of the physical information neural network includes the sum of data fitting loss and physical constraint loss. The data fitting loss is the mean square error between the inverted insulation aging quantification index and the pre-calibrated true value. The physical constraint loss is the sum of the squares of the voltage equation residuals and the squares of the current equation residuals in the transmission line telegraph equation.

[0017] Optionally, the expression for the physical constraint loss is:

[0018] in, For the physical constraint loss, The complex amplitude of the voltage on the transmission line. The complex amplitude of the current on the transmission line, These are the position coordinates along the length of the cable. Resistance per unit length Inductance per unit length Conductivity per unit length Capacitance per unit length Angular frequency, It is the imaginary unit.

[0019] Optionally, establishing a frequency-varying distributed parameter model based on the physical structure of the cable includes: modeling shielded cables and unshielded power cables separately; establishing a distributed capacitance model for shielded cables, in which the complex permittivity follows the Cole-Cole polarization model to characterize the relaxation time shift caused by aging; and introducing an equivalent structural operator for unshielded power cables to correct electromagnetic field spillover effects.

[0020] Optionally, the quantitative index of insulation aging is the dielectric loss tangent. Its inversion formula is:

[0021] in, Angular frequency, and These represent the imaginary and real parts of a complex signal, respectively. The reflection coefficient spectrum is obtained by Fourier transforming the dispersion-corrected reflection response. For shielded cables, the structural compensation factor is... The value is 1 for unshielded power cables. The value is determined by the equivalent structural operator.

[0022] Optionally, injecting the detection signal into the energized cable includes: using a coupling module to couple the detection signal to the cable, wherein the coupling module is a high-voltage DC blocking capacitor architecture, forming a high-pass filter with a cutoff frequency of less than 100kHz at the physical level to block power frequency energy and suppress common-mode electromagnetic interference inside the nuclear power plant; or, the coupling module is replaced with a non-contact inductive coupling based on a high-frequency magnetic ring sensor, a flexible Rogowski coil, or a distributed capacitor clamp to achieve electromagnetic induction non-contact injection.

[0023] Optionally, it also includes an early warning output step: mapping the cable health status obtained by inversion to a preset full life cycle asset management coordinate system, automatically triggering a graded early warning mechanism when the insulation aging quantitative index exceeds the preset early deterioration threshold, and generating a monitoring report containing the aging location and expected remaining life.

[0024] To address the aforementioned technical problems, this application provides an online monitoring device for cable aging based on spread spectrum time-domain reflectometry, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor performs the method described in this application.

[0025] Compared with the prior art, this application has the following advantages: This application presents a cable aging online monitoring method and device based on spread spectrum time-domain reflectometry (SSTDR). Through frequency selection using spectral holes and adaptive signal decomposition, it actively avoids strong electromagnetic noise and eliminates non-stationary interference, significantly improving the signal-to-noise ratio. Utilizing phase dispersion compensation and its optimized search, it counteracts phase distortion caused by aging, achieving centimeter-level positioning. By combining physical information neural networks and transmission line equation constraints, it inversely derives quantitative indicators of insulation aging from multidimensional reflection characteristics, overcoming the limitation of existing SSTDR methods that only classify normal / abnormal conditions. This application achieves integrated monitoring of cable aging online, strong noise resistance, high-precision positioning, and quantitative assessment. Attached Figure Description

[0026] The accompanying drawings are included to provide a further understanding of this application. They are incorporated into and constitute a part of this application. The drawings illustrate embodiments of this application and, together with this specification, serve to explain the principles of this application.

[0027] Figure 1 This is a flowchart of an embodiment of the cable aging online monitoring method based on spread spectrum time-domain reflectometry according to this application.

[0028] Figure 2 This is a flowchart of an embodiment of the present application of performing variational mode decomposition on an echo signal to remove non-stationary background noise.

[0029] Figure 3 This is a schematic diagram of obtaining purified echo according to an embodiment of this application.

[0030] Figure 4 This is a flowchart of performing a fractional Fourier transform on the reflection response characteristics according to an embodiment of this application.

[0031] Figure 5 This is a schematic diagram of the reflection response after dispersion correction according to an embodiment of the application.

[0032] Figure 6 A schematic diagram of a physical information neural network according to an embodiment of this application.

[0033] Figure 7 This application presents an embodiment of an online monitoring device for cable aging based on spread spectrum time-domain reflectometry. Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this application.

[0035] This application provides a method and device for online monitoring of cable aging based on spread spectrum time-domain reflectometry, which is particularly suitable for harsh environments with extremely high cable reliability requirements, such as nuclear power plants, aerospace, and rail transportation. The following explanation uses online monitoring of cable aging in a nuclear power plant as an example.

[0036] The unique characteristics of nuclear power plant sites impose stringent requirements on cable aging monitoring: safety-grade cables in nuclear power plants must be continuously monitored during normal unit operation, and any testing method requiring power outage is unacceptable. Simultaneously, nuclear power plant sites present a series of objective physical factors that hinder reflection-based measurements: strong electromagnetic noise near cables, such as 50Hz power frequency currents, inverter pulses, and solenoid valve actuation, easily drowns out weak aging reflection signals. Insulation aging causes nonlinear changes in the dielectric constant with frequency, leading to electromagnetic wave propagation dispersion and significantly reducing the positioning accuracy of traditional reflection methods, with errors reaching several meters or even tens of meters. Furthermore, existing online monitoring methods cannot directly provide the degree of insulation aging; they typically only distinguish between normal and abnormal states, failing to provide continuous quantitative indicators of aging or predictions of remaining life.

[0037] To address the aforementioned pain points, this embodiment will elaborate on the core concept of this application: by actively sensing and avoiding strong noise frequency bands, adaptively decomposing echoes to remove non-stationary interference, performing phase dispersion compensation on reflected signals to restore positioning resolution, and introducing a physical constraint inversion model that conforms to the laws of electromagnetic propagation, a complete technical link from signal acquisition to quantitative evaluation is formed, thereby realizing online, strong interference resistant, centimeter-level positioning and quantitative inversion of insulation aging indicators for cable aging monitoring.

[0038] Figure 1 This is a flowchart of an embodiment of the online monitoring method for cable aging based on spread spectrum time-domain reflectometry according to this application. Figure 1 As shown, the online monitoring method for cable aging based on spread spectrum time-domain reflectometry includes: Step S1: Establish a frequency-varying distributed parameter model based on the physical structure of the cable, perform background noise spectrum scanning on the cable to identify spectral holes, and use the spectral holes as the working frequency band of the detection signal.

[0039] Optionally, establishing a frequency-varying distributed parameter model based on the physical structure of the cable includes: modeling shielded cables and unshielded power cables separately.

[0040] For shielded cables, the electromagnetic field is confined within the insulation layer due to the presence of the metallic shielding layer, eliminating outward radiation losses. Therefore, the classical coaxial cable capacitance formula is used as a basis, and a frequency-dependent capacitance is introduced. Complex permittivity as a function of aging time t Establish a distributed capacitance model:

[0041] in, For distributed capacitance model, The complex permittivity follows the Cole-Cole polarization model to characterize the relaxation time shift caused by aging. Let the radius be the outer conductor. Let be the radius of the inner conductor.

[0042] For unshielded power cables, since these cables do not have a metallic shielding layer, the electromagnetic field has an outward spillover effect, which will cause deviations between the actual distributed capacitance and conductivity parameters and the ideal coaxial model. Therefore, an equivalent structure operator η is introduced to correct the calculation results of the ideal model. The corrected parameters can truly reflect the electromagnetic transmission characteristics of unshielded cables and avoid mismatch between the reference library data and the actual cable characteristics.

[0043] Optionally, a hardware control unit is used to perform a background noise spectrum scan on the live cable, identify the energy distribution of power frequency harmonics and their overtone interference, automatically find spectral holes, and select the center frequency of the spectral hole as the carrier frequency of the probe signal. This application achieves orthogonal avoidance between the probe signal and environmental noise in the frequency domain by concentrating the energy of the probe signal on the spectral hole, thereby suppressing strong interference such as power frequency harmonics and inverter pulses at the source. Traditional SSTDR relies on spread spectrum processing gain to suppress noise, but when the interference power is extremely high, the receiving front end is prone to entering the nonlinear region or even saturation, leading to correlation peak distortion or failure. This application ensures that the probe signal and the working echo avoid strong interference frequency bands by pre-sensing the noise distribution and actively avoiding it, so that the received signal is always within the linear dynamic range, fundamentally solving the problem of strong noise drowning out weak aging reflection signals.

[0044] Step S2: Generate a detection signal that matches the operating frequency band and inject the detection signal into the energized cable.

[0045] Optionally, generating a detection signal matching the operating frequency band includes: generating an original spread spectrum sequence with sharp autocorrelation characteristics; modulating the original spread spectrum sequence to extend its signal energy to a wide bandwidth of 5 MHz to 24 MHz. The generated detection signal retains the inherent characteristics of the original spread spectrum sequence, such as strong anti-interference capability and high range resolution, while also matching the frequency band with spectral holes identified in the pre-processing stage, further reducing the probability of frequency band overlap with electromagnetic noise in the field. The sharp autocorrelation characteristic ensures that even weak aging reflection signals can be effectively identified during subsequent cross-correlation calculations, preventing correlation peaks from being submerged in strong noise.

[0046] Optionally, the original spread spectrum sequence includes, but is not limited to, m-sequences, Gold sequences, Kasami sequences, complementary sequences, or chaotic sequences with ideal autocorrelation properties.

[0047] Optionally, the methods for modulating the original spread spectrum sequence include, but are not limited to, binary phase shift keying (BPSK), quadrature amplitude modulation (QAM), multiple phase shift keying (MPSK), or linear frequency modulation (LFM) pulse compression modulation.

[0048] When the original spread spectrum sequence is modulated using binary phase shift keying, the generated probe signal is:

[0049] Where t is the time variable, To detect signals, Let n be the original spreading sequence, and n be the chip number of the original spreading sequence. For pulse shaping function, For chip cycle, Carrier frequency, carrier frequency Select within the operating frequency band.

[0050] Preferably, during modulation, the center frequency of the spectral hole is used as the carrier frequency. The selected value ensures that the detected signal and background noise are orthogonal and avoid each other in the frequency domain.

[0051] Optionally, injecting the detection signal into the energized cable includes: using a coupling module to couple the detection signal to the cable. The coupling module is a high-voltage DC blocking capacitor architecture, forming a high-pass filter with a cutoff frequency below 100kHz at the physical level to block power frequency energy and suppress common-mode electromagnetic interference within the nuclear power plant, while maintaining a 5MHz frequency. A 24MHz spread spectrum signal provides a low-loss injection channel; alternatively, the coupling module can be replaced with a non-contact inductive coupling based on a high-frequency magnetic ring sensor, a flexible Rogowski coil, or a distributed capacitive clamp to achieve electromagnetic induction non-contact injection.

[0052] Step S3: Acquire the echo signal from the cable, perform adaptive signal decomposition on the echo signal to remove non-stationary background noise, obtain purified echo, perform cross-correlation operation on the purified echo and the detection signal, and extract reflection response features.

[0053] The acquired echo signals typically contain strong interference, such as power frequency harmonics and inverter pulses, which are non-stationary noises. Therefore, it is necessary to remove non-stationary background noise to obtain purified echoes. This application performs adaptive signal decomposition on the echo signals to remove non-stationary background noise. Optionally, the adaptive signal decomposition can be Variational Mode Decomposition (VMD), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), wavelet packet transform, or a nonlinear filtering operator based on an autoencoder network.

[0054] Taking Variational Mode Decomposition (VMD) as an example, such as Figure 2 As shown, variational mode decomposition is performed on the echo signal to remove non-stationary background noise, resulting in a purified echo, including: Step S31: Adaptively determine the number of decomposition layers for variational mode decomposition based on the center frequency of the echo signal.

[0055] In traditional variational mode decomposition (VMD) applications, the number of decomposition levels, K, is typically preset by the user based on experience or selected through repeated trials. However, in the specific scenario of online monitoring of nuclear power cable aging, a fixed or empirical K value can lead to the following problems: The complexity of the echo signal dynamically changes with the degree of cable aging. If a constant K is used, an excessively small K will result in under-decomposition, causing the effective signal to be mixed with noise; an excessively large K will result in over-decomposition, generating spurious narrowband components and introducing additional computational overhead and errors. Secondly, the intensity and frequency distribution of interference such as power frequency harmonics and inverter pulses at nuclear power plants change with equipment start-up and shutdown and load variations. A fixed K cannot dynamically match the non-stationary characteristics of noise, causing VMD to be unable to effectively separate interference components and affecting the quality of the purified echo.

[0056] Therefore, this application proposes to adaptively determine the number of decomposition layers based on the center frequency of the echo signal. In one embodiment, a Fast Fourier Transform (FFT) is performed on the acquired echo signal f(t) to obtain its power spectral density. The number of peaks, noise floor, and signal energy distribution within the operating frequency band of the probe signal are analyzed, and the optimal K value is dynamically calculated. This maximizes the removal of non-stationary interference while preserving effective reflection characteristics, providing a high-quality purified echo for subsequent cross-correlation demodulation and dispersion compensation. For example, for a slightly aged cable, when the inverter is not running, the echo signal power spectrum only shows two significant peaks within the operating frequency band (corresponding to reflections at the cable's beginning and end), and the system automatically sets K=4. When the inverter starts and generates strong pulse interference, two significant broadband bulges are added to the power spectrum, and the system automatically adjusts K to 6. In this way, VMD can always separate the effective reflected signal from non-stationary noise with the most appropriate number of decomposition layers, avoiding noise residue caused by under-decomposition and preventing computational burden and spurious components caused by over-decomposition.

[0057] Step S32: Decompose the echo signal into a number of narrowband components equal to the number of decomposition layers by solving a constrained variational problem, and minimize the sum of the bandwidths of each narrowband component.

[0058] The constrained variational problem is to decompose a strongly disturbed echo signal f(t) into K narrowband components. And minimize the sum of the bandwidths of each narrowband component.

[0059] The expression for the constrained variational problem is:

[0060] The constraints are: .

[0061] in, For echo signal, For the Kth narrowband component, The center frequency of the Kth narrowband component is... For time variables, Let be the Dirac impulse function. The partial derivative with respect to time t is used to measure the bandwidth of the signal; K is the number of narrowband components, which is equal to the number of mode decomposition layers.

[0062] Step S33: Remove narrowband components that do not overlap with the operating frequency band of the detection signal, and reconstruct the remaining narrowband components to obtain the purified echo.

[0063] like Figure 3 As shown, the raw echo signal f(t) is first received. This signal contains not only weak reflections from various locations on the cable (aging interfaces, connectors, etc.), but also strong non-stationary background noise specific to nuclear power plants, such as power frequency harmonics, inverter pulses, and solenoid valve actions. Based on the center frequency of the echo signal and the spectral complexity within the operating frequency band, the number of decomposition layers K is adaptively determined. In this example, it is shown that K narrowband components IMF1, IMF2…IMFK are obtained. The operating frequency band of the probe signal is used as the selection criterion. If the Kth narrowband component… center frequency With the carrier frequency of the detection signal If the distance to the target signal exceeds a preset threshold, the component is considered non-stationary background noise and is removed; otherwise, it is retained as a valid component related to aging reflections. The filtered narrowband components are then linearly superimposed to reconstruct the purified echo. The purified echo significantly reduces the energy of non-stationary noise while preserving the integrity of the weak aging reflection signal.

[0064] Purify the echo With detection signal The data is fed into the cross-correlation unit for cross-correlation demodulation.

[0065] in, Let be the time delay variable, representing the time delay between the probe signal s(t) and the purified echo. The relative time delay between the two signals. R(τ) is the cross-correlation function and also the reflection response characteristic of this application. The larger its value, the better matched the two signals are under that time delay. In this application, the peak position of R(τ) corresponds to the reflection delay at the impedance discontinuity point in the cable, the peak amplitude reflects the degree of impedance change, and the peak width can be used to assess the severity of the dispersion effect. By finding the maximum point of R(τ), the aging location of the cable can be determined. L = vp τ peak / 2, where τpeak is the time delay variable corresponding to the maximum point of R(τ). vp For the propagation speed of the wave.

[0066] Step S4: Perform phase dispersion compensation on the reflection response characteristics to obtain the dispersion-corrected reflection response.

[0067] In online monitoring of cable aging, the aging of insulation materials causes a nonlinear change in their dielectric constant with frequency (i.e., dispersion effect). When the spread spectrum detection signal propagates in the cable, different frequency components experience different phase velocities, resulting in a second-order nonlinear phase distortion in the received reflected signal. The macroscopic manifestation of this distortion is that the originally sharp correlation peak is broadened, the amplitude decreases, and even multiple false peaks appear, directly leading to two serious consequences: first, the positioning accuracy is greatly reduced, with errors reaching several meters or even tens of meters; second, the amplitude attenuation of the reflected peak distorts the quantitative assessment of the aging degree.

[0068] Traditional SSTDR can only perform coarse positioning using a fixed wave velocity, and cannot fundamentally eliminate waveform distortion caused by dispersion. Therefore, after extracting the reflection response characteristics, it is necessary to implement specialized phase dispersion compensation to restore the energy refocusing of the relevant peaks, thereby obtaining accurate aging locations and high-fidelity reflection characteristics.

[0069] In this application, "phase dispersion compensation" is a broader concept. Any technical means that can offset or reduce the phase second-order nonlinear distortion caused by insulation aging and refocus the relevant peak energy can be used as an equivalent replacement for this application. Specifically, it includes, but is not limited to, at least one of the following: fractional Fourier transform (FrFT), time-reversed mirror (TRM) compensation, phase correction of frequency domain all-pass filter, and compensation scheme based on time-frequency distribution energy ridge extraction.

[0070] The principle of the fractional Fourier transform (FrFT) is to rotate the signal to the optimal fractional domain and then use a linear frequency-modulated kernel function to cancel out the conjugate of the quadratic phase distortion, thus achieving energy refocusing. The following uses the fractional Fourier transform as an example to explain in detail how gradient search can be used to determine the focusing order and achieve phase dispersion compensation. Figure 4 As shown, it includes: Step S41: Transform the reflection response features to the fractional domain and calculate the kurtosis of the relevant peaks in the fractional domain.

[0071] Let the reflection response features extracted from the cross-correlation operation be... Its related peaks broaden due to the influence of aging dispersion. Define the rotation angle. The fractional Fourier transform of the following is:

[0072] in, Let be the fractional Fourier transform kernel function, and its expression is:

[0073] in, The rotation angle is... For time variables, These are fractional-order field coordinates.

[0074] Kurtosis is a high-order statistic used in statistics and signal processing to describe the shape of a probability distribution; it measures the thickness or peak of the tail of the data distribution. In this embodiment, for each candidate rotation angle... Calculate the transformed spectrum The kurtosis. The greater the kurtosis, the more concentrated the energy is at a few points, that is, the better the focusing.

[0075] Step S42: Adjust the rotation angle of the fractional Fourier transform by gradient search to maximize the kurtosis, and determine the focusing order by the rotation angle corresponding to the maximum value.

[0076] Set rotation angle The search interval is determined using the golden section search method. In each iteration, the kurtosis corresponding to the two trial points is calculated, and the interval is continuously narrowed until the interval length is less than a preset threshold. The rotation angle corresponding to the maximum kurtosis is taken as the optimal rotation angle, and the focusing order is determined.

[0077] Step S43: Perform a fractional Fourier transform on the reflection response characteristics again using the focusing order, so that the phase second nonlinear distortion caused by insulation aging is canceled, the energy of the relevant peak is refocused, and the reflection response after dispersion correction is obtained.

[0078] Using the optimal rotation angle again A fractional Fourier transform is performed. In this transform domain, the quadratic phase term caused by aging cancels out the linear frequency modulation term in the FrFT kernel, and the correlation peak energy is compressed to an extremely narrow range. Subsequently, the signal is transformed back to the time domain by an inverse fractional Fourier transform to obtain the dispersion-corrected reflection response.

[0079] like Figure 5 As shown, the reflection response features extracted from the cross-correlation calculation are obtained without phase dispersion compensation. Due to the dispersion effect caused by insulation aging, the resulting phenomena are: severe broadening of the correlation peak, decrease in amplitude, and asymmetrical waveform, as illustrated in the figure as a flat and wide bulge. If the distance is calculated directly based on the peak position of this correlation peak, the positioning error can reach several meters or even tens of meters because the peak point may drift due to the broadening effect; at the same time, the amplitude after broadening will also be distorted, and it cannot accurately reflect the degree of aging.

[0080] To address the aforementioned issues, this application employs FrFT transform for adaptive dispersion compensation, compressing the originally dispersed energy into an extremely narrow range to obtain the dispersion-corrected reflection response. . The correlation peaks were restored to sharp, symmetrical pulses with widths close to the theoretical resolution. The peak position corresponds to the round-trip time delay of the aging point. Combined with the propagation wave velocity, the aging position can be calculated. Through the above compensation, the positioning accuracy is improved from meters to centimeters.

[0081] Step S5: Extract the multidimensional feature vector of the reflection response after dispersion correction, input it into the physical information neural network, and invert to obtain the insulation aging quantification index and cable aging level at each location of the cable. The loss function of the physical information neural network includes the physical constraint loss of the transmission line telegraph equations.

[0082] like Figure 6 As shown, the multidimensional feature vector X is input into the Physics-Informed Neural Network (PINN). In this embodiment, the multidimensional feature vector X includes at least: the time-domain peak envelope entropy, the frequency-domain energy spectrum centroid shift, and the fractional-order energy concentration.

[0083] The multidimensional feature vector is fed into multiple fully connected hidden layers after entering the input layer (the diagram shows three hidden layers; the specific number of layers and neurons can be adjusted according to the amount of training data). Each layer performs linear transformations and nonlinear activations, abstracting the input features layer by layer into higher-level representations, ultimately outputting a quantitative indicator of insulation aging—the dielectric loss tangent. Both domestic and international cable aging management standards will... These are listed as recommended or mandatory indicators for cable insulation condition assessment. Therefore, this application uses online inversion... It can be directly integrated with existing standards and management systems without additional conversion, making it convenient for engineering applications.

[0084] Unlike purely data-driven neural networks, the PINN in this application, during training and inference, The loss function of the neural network includes not only data fitting loss but also physical constraint loss. The data fitting loss is the mean square error between the inverted insulation aging quantification index and the pre-calibrated true value. In this embodiment, it is the predicted value. With the real The mean square error.

[0085] The physical constraint loss is the sum of the squares of the residuals of the voltage equation and the current equation in the transmission line telegraph equations. The design principle of the physical constraint loss is to force the output of the neural network to satisfy the transmission line telegraph equations: Voltage equation:

[0086] Current equation:

[0087] The expression for physical constraint loss is:

[0088] in, For physical constraint loss, The complex amplitude of the voltage on the transmission line. The complex amplitude of the current on the transmission line, These are the position coordinates along the length of the cable. Resistance per unit length Inductance per unit length Conductivity per unit length Capacitance per unit length Angular frequency, The unit is the imaginary unit. Among these parameters, the resistance per unit length R, the inductance per unit length L, the conductance per unit length G, and the capacitance per unit length C can be calculated using a frequency-varying distributed parameter model.

[0089] By minimizing the loss function of the neural network, a trained physical information neural network is obtained. The trained physical information neural network directly outputs the dielectric loss tangent from the multi-dimensional feature vector. Cable aging rating. Cable aging ratings include, but are not limited to, healthy, slightly aged, moderately aged, and severely aged. The cable aging rating is determined according to... The value is determined by comparing it with the pre-review threshold, which can be determined based on the frequency-varying distribution parameter model.

[0090] Optionally, it also includes an early warning output step: mapping the cable health status obtained by inversion to a preset full life cycle asset management coordinate system, automatically triggering a graded early warning mechanism when the insulation aging quantitative index exceeds the preset early deterioration threshold, and generating a monitoring report containing the aging location and expected remaining life.

[0091] Optionally, the dielectric loss tangent The inversion formula is:

[0092] in, Angular frequency, and These represent the imaginary and real parts of a complex signal, respectively. The reflection coefficient spectrum is obtained by Fourier transform of the dispersion-corrected reflection response. For shielded cables, the structural compensation factor is... The value is 1 for unshielded power cables. The value is determined by the equivalent structure operator. The structure compensation coefficient corrects for errors caused by differences in cable geometry and shielding / unshielding, making the online inversion... It closely approximates the true value of offline measurements.

[0093] like Figure 7 As shown, this application provides an online monitoring device for cable aging based on spread spectrum time-domain reflectometry, which mainly includes: a host computer 71, a signal generation and acquisition unit 72, and a broadband isolation coupling module 73. The entire system operates in the strong electromagnetic noise environment of a nuclear power plant (the diagram shows strong electromagnetic interference injected by noise sources such as frequency converters / motors).

[0094] The host computer 71 runs core algorithms such as VMD adaptive signal decomposition, FrFT dispersion compensation, and PINN inversion. It communicates with the signal generation and acquisition unit 72 via USB or Ethernet interface, sends control parameters (such as the carrier frequency of the probe signal, chip rate, number of decomposition layers, etc.), and receives echo data for real-time processing, display, and storage.

[0095] The signal generation and acquisition unit 72 undertakes the dual tasks of signal generation and acquisition. On the one hand, it generates an original spread spectrum sequence with sharp autocorrelation characteristics according to the instructions sent by the host computer 71, and performs BPSK modulation to convert the probe signal into an analog output. On the other hand, it acquires the echo signal reflected back through the cable at high speed, performs necessary digital down-conversion and downsampling, and then uploads it to the host computer 71.

[0096] The broadband isolation coupling module 73 enables the injection of the detection signal and the extraction of the echo signal, while simultaneously blocking the 50Hz power frequency high voltage and common-mode interference in the cable from entering the signal generation and acquisition unit 72. The broadband isolation coupling module 73 in the figure adopts a high-voltage DC blocking capacitor architecture (which can also be replaced by a high-frequency magnetic ring, Rogowski coil, or other non-contact coupling method), forming a high-pass filter with a cutoff frequency below 100kHz. This allows the high-frequency detection signal (5~24MHz) to pass through with low loss, while the power frequency and its harmonics are significantly attenuated.

[0097] The workflow of the cable aging online monitoring device based on spread spectrum time-domain reflectometry is as follows: The host computer 71 performs a background noise spectrum scan on the live cable through the signal generation and acquisition unit 72, identifies the distribution of power frequency harmonics and overtone interference, automatically finds spectral holes, and determines the optimal carrier frequency. The frequency and chip rate parameters are then sent to the signal generation and acquisition unit 72.

[0098] The signal generation and acquisition unit 72 generates a detection signal according to the instructions of the host computer 71. This signal is then injected into the cable under test via the broadband isolation coupling module 73. As the detection signal propagates along the cable, it encounters impedance discontinuities caused by aging, resulting in reflected echoes. Simultaneously, strong electromagnetic interference generated by noise sources such as frequency converters is also coupled to the cable.

[0099] The signal generation and acquisition unit 72 acquires the echo signal superimposed with noise through the broadband isolation coupling module 73, and uploads it to the host computer 71 in real time via USB / Ethernet.

[0100] The host computer 71 first performs adaptive signal decomposition (such as VMD) to remove non-stationary background noise and obtain purified echo; then it performs cross-correlation operation with the probe signal to extract reflection response features; next, it uses fractional Fourier transform (or other compensation schemes such as time-reversal mirror) to perform phase dispersion compensation to achieve centimeter-level positioning; finally, it uses a physical information neural network to inversely derive the dielectric loss tangent from the compensated reflection response. It also outputs health status and early warning information.

[0101] This device can achieve online, uninterrupted monitoring of cable aging conditions and output... Quantitative indicators and centimeter-level positioning information provide reliable hardware support for the full life-cycle management of cables in nuclear power plants.

[0102] The cable aging online monitoring method and device based on spread spectrum time-domain reflectometry of this application have the following beneficial effects: (1) This application actively identifies spectral holes by scanning the background noise spectrum and dynamically sets the carrier frequency of the probe signal to the center frequency of the spectral hole, thereby achieving orthogonal avoidance between the probe signal and environmental noise in the frequency domain and suppressing strong interference such as power frequency harmonics and inverter pulses at the nuclear power plant site from the source. On this basis, adaptive signal decomposition (such as variational mode decomposition) is used to decompose the echo signal and remove interference components that do not overlap with the operating frequency band of the probe signal, further stripping away non-stationary background noise. The above-mentioned dual-layer anti-interference architecture of "active avoidance + adaptive decomposition" significantly improves the signal-to-noise ratio compared with the traditional SSTDR which only relies on spread spectrum processing gain, ensuring that weak aging reflection signals can still be reliably extracted in a strong noise environment.

[0103] (2) This application introduces fractional Fourier transform into SSTDR cable aging monitoring to compensate for the phase dispersion effect caused by insulation aging. By maximizing the kurtosis of the fractional domain correlation peak through gradient search, the optimal rotation angle (focusing order) is adaptively determined. The linear frequency modulation term of the FrFT kernel function is used to conjugate and cancel the secondary phase nonlinear distortion caused by aging, thereby achieving refocusing of the correlation peak energy. The positioning accuracy is improved from the meter level to the centimeter level, overcoming the positioning deviation problem.

[0104] (3) This application constructs a physical information neural network with physical constraints embedded in the transmission line telegraph equations. It takes multidimensional reflection features as input and outputs insulation aging quantitative index. According to the specific value of the insulation aging quantitative index, it can be further subdivided into multiple levels such as healthy, mild aging, moderate aging, and severe aging. It can also combine historical monitoring data to fit the aging trend curve and predict the remaining life.

[0105] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously. Furthermore, other operations may be added to these processes, or one or more steps may be removed from these processes.

[0106] This application also provides a computer program product containing instructions. The computer program product may be software or program products containing instructions, capable of running on a network device or stored on any available medium. When the computer program product is run on at least one network device, it causes the at least one network device to perform an online cable aging monitoring method.

[0107] This application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that a network device can store, or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state drive). The computer-readable storage medium includes instructions that instruct the network device to perform an online cable aging monitoring method.

[0108] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, these terms have no special meaning and therefore should not be construed as limiting the scope of protection of this application. In addition, although the terminology used in this application is selected from commonly known and used terms, some terms mentioned in this application's specification may have been chosen by the applicant according to his or her judgment, and their detailed meanings are explained in the relevant sections of this description. Moreover, this application should be understood not only through the actual terms used, but also through the meaning implied by each term.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the protection scope of the technical solutions of the embodiments of the present invention.

[0110] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0111] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for online monitoring of cable aging based on spread spectrum time-domain reflectometry, characterized in that, include: A frequency-varying distributed parameter model is established based on the physical structure of the cable. The background noise spectrum of the cable is scanned to identify spectral holes, and the spectral holes are used as the working frequency band of the detection signal. Generate a detection signal that matches the operating frequency band, and inject the detection signal into the energized cable; The echo signal of the cable is acquired, and adaptive signal decomposition is performed on the echo signal to remove non-stationary background noise and obtain purified echo. The purified echo and the detection signal are cross-correlated to extract the reflection response features. Phase dispersion compensation is performed on the reflection response characteristics to obtain the dispersion-corrected reflection response; The multidimensional feature vector of the reflection response after dispersion correction is extracted and input into the physical information neural network to invert the insulation aging quantification index and cable aging level at each location of the cable. The loss function of the physical information neural network includes the physical constraint loss of the transmission line telegraph equations.

2. The method as described in claim 1, characterized in that, Generating a detection signal that matches the operating frequency band includes: generating an original spread spectrum sequence with sharp autocorrelation characteristics, modulating the original spread spectrum sequence, and obtaining a detection signal that matches the operating frequency band.

3. The method as described in claim 2, characterized in that, The original spread spectrum sequence is subjected to binary phase shift keying modulation to obtain the detection signal: Where t is the time variable, The detection signal, The original spreading sequence is denoted as n, where n is the chip number of the original spreading sequence. For pulse shaping function, For chip cycle, The carrier frequency, the carrier frequency Select within the operating frequency band.

4. The method as described in claim 3, characterized in that, The center frequency of the spectral hole is used as the carrier frequency. The selected value ensures that the detected signal and background noise are orthogonal and avoid each other in the frequency domain.

5. The method as described in claim 1, characterized in that, The adaptive signal decomposition is at least one of variational mode decomposition, adaptive noise complete set empirical mode decomposition, wavelet packet transform, or nonlinear filtering operator based on autoencoder network.

6. The method as described in claim 1, characterized in that, The adaptive signal decomposition is performed as variational mode decomposition. Variational mode decomposition is then applied to the echo signal to remove non-stationary background noise, resulting in a purified echo, including: The number of decomposition layers for variational mode decomposition is adaptively determined based on the center frequency of the echo signal. The echo signal is decomposed into a number of narrowband components equal to the number of decomposition layers by solving a constrained variational problem, and the sum of the bandwidths of each narrowband component is minimized. The narrowband components that do not overlap with the operating frequency band of the detection signal are removed, and the remaining narrowband components are reconstructed to obtain the purified echo.

7. The method as described in claim 1, characterized in that, The phase dispersion compensation is at least one of the following: fractional Fourier transform, time-reversal mirror compensation, frequency domain all-pass filter phase correction, or compensation scheme based on time-frequency distribution energy ridge extraction.

8. The method as described in claim 1, characterized in that, The phase dispersion compensation is a fractional Fourier transform. A fractional Fourier transform is performed on the reflection response characteristics, and the focusing order is determined by gradient search to cancel the phase dispersion, thus obtaining the dispersion-corrected reflection response.

9. The method as described in claim 8, characterized in that, Methods for obtaining the dispersion-corrected reflection response include: The reflection response characteristics are transformed to the fractional domain, and the kurtosis of the relevant peaks in the fractional domain is calculated. The rotation angle of the fractional Fourier transform is adjusted by gradient search to maximize the kurtosis, and the focusing order is determined by the rotation angle corresponding to the maximum value. The fractional Fourier transform of the reflection response characteristics is performed again using the focusing order to cancel out the phase second nonlinear distortion caused by insulation aging, and the energy of the relevant peak is refocused to obtain the dispersion-corrected reflection response.

10. The method as described in claim 1, characterized in that, The loss function of the physical information neural network includes the sum of data fitting loss and physical constraint loss. The data fitting loss is the mean square error between the inverted insulation aging quantification index and the pre-calibrated true value. The physical constraint loss is the sum of the squares of the voltage equation residuals and the squares of the current equation residuals in the transmission line telegraph equation.

11. The method as described in claim 10, characterized in that, The expression for the physical constraint loss is: in, For the physical constraint loss, The complex amplitude of the voltage on the transmission line. The complex amplitude of the current on the transmission line, These are the position coordinates along the length of the cable. Resistance per unit length Inductance per unit length Conductivity per unit length Capacitance per unit length Angular frequency, It is the imaginary unit.

12. The method as described in claim 1, characterized in that, The establishment of a frequency-varying distributed parameter model based on the physical structure of the cable includes: Shielded cables and unshielded power cables are modeled separately. For shielded cables, a distributed capacitance model is established, in which the complex permittivity follows the Cole-Cole polarization model to characterize the relaxation time shift caused by aging. For unshielded power cables, an equivalent structure operator is introduced to correct the electromagnetic field spillover effect.

13. The method as described in claim 12, characterized in that, The quantitative index for insulation aging is the dielectric loss tangent. Its inversion formula is: in, Angular frequency, and These represent the imaginary and real parts of a complex signal, respectively. The reflection coefficient spectrum is obtained by Fourier transforming the dispersion-corrected reflection response. For shielded cables, the structural compensation factor is... The value is 1 for unshielded power cables. The value is determined by the equivalent structural operator.

14. The method as described in claim 1, characterized in that, Injecting detection signals into a energized cable includes: A coupling module is used to couple the detection signal to the cable. The coupling module is a high-voltage DC blocking capacitor architecture, which forms a high-pass filter with a cutoff frequency of less than 100kHz at the physical level to block power frequency energy and suppress common-mode electromagnetic interference inside the nuclear power plant. Alternatively, the coupling module can be replaced with a non-contact inductive coupling based on a high-frequency magnetic ring sensor, a flexible Rogowski coil, or a distributed capacitor clamp to achieve electromagnetic induction non-contact injection.

15. The method as described in claim 1, characterized in that, It also includes an early warning output step: mapping the cable health status obtained by inversion to a preset full life cycle asset management coordinate system. When the insulation aging quantitative index exceeds the preset early deterioration threshold, a graded early warning mechanism is automatically triggered, and a monitoring report containing the aging location and expected remaining life is generated.

16. A cable aging online monitoring device based on spread spectrum time-domain reflectometry, characterized in that, include: At least one processor; as well as A memory that is communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, and when the computer program is executed by the at least one processor, it causes the at least one processor to perform the method according to any one of claims 1 to 15.

Images