Apparatus and Method for Diagnosing Cable Faults Based on Hierarchical Instantaneous Frequency Estimation

Abstract

The present invention relates to a cable defect diagnosis device and method based on hierarchical instantaneous frequency estimation that can quantitatively diagnose the overall condition of a cable by utilizing hierarchical instantaneous frequency ridge estimation and signal-to-noise ratio gating. The device comprises: a signal preprocessing unit that receives a reflected signal of a Gaussian envelope linear chirp (GELC) signal applied to a cable and performs preprocessing; an instantaneous frequency estimation unit that calculates a raw instantaneous frequency (IF) based on the phase difference value of the preprocessed signal; a hierarchical IF ridge estimation unit that estimates a ridge on the raw IF through local window-based slope rate voting filtering and global window-based ridge completion processing; a parameter optimization unit that calculates a local window length, a slope rate threshold, and a voting filter gain based on the hierarchical IF ridge estimation result and the signal-to-noise ratio (SNR) of the applied signal; an SNR gating unit that applies a distance-based SNR model to the estimated IF ridge to allow only sections with a possibility of reflection to pass through; and a detection index calculation unit that calculates a signal detection index (SD) that quantifies the defect location and degree of reflection based on the SNR gating pass rate. It includes.

Description

Technology Field

[0001] The present invention relates to cable defect diagnosis, and more specifically, to a cable defect diagnosis apparatus and method based on hierarchical instantaneous frequency estimation that enables quantitative diagnosis of the overall condition of a cable by utilizing hierarchical instantaneous frequency ridge estimation and signal-to-noise non-gating. Background Technology

[0002] Various power cables used in power facilities may experience impedance mismatches due to deterioration, mechanical damage, degradation of insulation materials, and external shocks in long-term operating environments, and such defects have a significant impact on the stability and safety of the power supply.

[0003] Therefore, technology capable of early detection of various failure types occurring within cables, such as partial discharge, insulation degradation, short circuits, and grounding failures, and precisely locating their positions, is essential for the preventive maintenance and asset management of power facilities.

[0004] In response to this need, cable defect diagnosis technology using reflectometry is being widely studied as a non-destructive diagnostic tool. In particular, time-frequency domain reflectometry (TFDR) has the advantage of being able to detect defects at high resolution by applying a Gaussian Envelope Linear Chirp (GELC) signal with broadband frequency characteristics to a cable and simultaneously analyzing the reflected signal in the time and frequency domains.

[0005] TFDR can estimate the location of defects based on the arrival time of reflections in the time domain and can analyze frequency-dependent defects or dispersion characteristics of cables through the shape of the reflection signal in the frequency domain, thereby providing higher resolution and diagnostic information compared to conventional time domain reflection measurement (TDR).

[0006] However, conventional TFDR-based techniques have the following practical limitations.

[0007] First, there is a threshold ambiguity problem in which the threshold setting varies sensitively depending on the analysis environment and measurement conditions, which reduces the reliability of determining whether reflection is present.

[0008] Second, the time-frequency conversion process used in TFDR analysis is prone to cross-term artifacts, which degrades the accuracy of reflected wave analysis.

[0009] Third, since existing reflection analysis primarily utilizes amplitude-based indicators, there is a problem in that the actual reflection intensity or the degree of impedance mismatch is distorted or not accurately reflected due to amplitude insensitivity.

[0010] Fourth, analysis using instantaneous frequency (IF) originally assumes smooth phase change estimation, but in low SNR environments, IF estimation becomes unstable and contains a large amount of noise-based jitter or impulse spikes.

[0011] Due to these effects, it is difficult to ensure the continuity of the IF ridge, and consequently, it is difficult to reliably isolate the ridge caused by the actual defect.

[0012] In addition, cables are dispersion media with propagation constants and characteristic impedances defined by RLGC (resistance, inductance, capacitance, conductance) parameters, and their propagation speed and attenuation vary nonlinearly with frequency. Consequently, when using broadband signals in the form of chirps, not all frequency components have the same group delay, which complicates the cross-correlation results for fault reflections and makes accurate analysis difficult.

[0013] In particular, for long-distance cables, signal attenuation accumulates, causing the SNR of reflected signals to drop sharply, making it difficult to separate effective reflections using existing analysis methods. Due to these technical limitations, existing commercial cable diagnostic equipment has difficulty performing accurate diagnoses in long cable sections of approximately 1.5 km or more, and there has been a problem where analysis results become unstable depending on changes in field conditions.

[0014] Therefore, there is a continuously growing need for an automated reflection analysis technique that can reliably estimate IF ridges even in low SNR environments, accurately interpret reflection characteristics in dispersed media, and eliminate threshold ambiguity.

[0015] Furthermore, there is an increasing industrial demand for universal and highly reliable cable fault diagnosis technology that can operate consistently with the same algorithm even in facility environments with various cable types, lengths, and propagation characteristics. Prior art literature

[0016] Republic of Korea Published Patent No. 10-2025-0026522 Republic of Korea Published Patent No. 10-2025-0070405 Republic of Korea Registered Patent No. 10-2749554 The problem to be solved

[0017] The present invention aims to solve the problems of conventional cable defect diagnosis technology by providing a cable defect diagnosis device and method based on hierarchical instantaneous frequency estimation that can quantitatively diagnose the overall condition of a cable by utilizing hierarchical instantaneous frequency ridge estimation and signal-to-noise non-gating.

[0018] The purpose of the present invention is to provide a cable fault diagnosis device and method based on hierarchical instantaneous frequency estimation, which applies a hierarchical instantaneous frequency (IF) ridge estimation technique to eliminate noisy steep ridges in local regions and stably form ridges that maintain global continuity, thereby enabling clear distinction of reflected signals even in low-quality signals or long-distance cables.

[0019] The present invention aims to provide a cable fault diagnosis device and method based on hierarchical instantaneous frequency estimation that can significantly reduce false positives by suppressing noise-induced ridges that do not correspond to actual reflections, by introducing an SNR-based gating mechanism to physically verify the possibility of the existence of reflected signals and enabling probabilistic determination considering attenuation and cable dispersion characteristics by distance.

[0020] The purpose of the present invention is to provide a cable defect diagnosis device and method based on hierarchical instantaneous frequency estimation, which provides an automatic parameter optimization function so that key parameters of the analysis algorithm, such as local window length, slope ratio threshold, and voting filter gain, are automatically calculated, thereby ensuring consistent analysis performance even under various cable types, lengths, and environmental conditions.

[0021] The present invention aims to provide a cable defect diagnosis device and method based on hierarchical instantaneous frequency estimation, which spatially reflects the validity ratio of the instantaneous frequency ridge in the signal detection index (SD) to quantitatively evaluate not only the defect location but also the reflection intensity (magnitude of impedance mismatch).

[0022] The present invention aims to provide a cable fault diagnosis device and method based on hierarchical instantaneous frequency estimation that enables stable and accurate fault detection even for long-distance cables of 1.5 km or more, which have been difficult to diagnose, thereby providing high-reliability diagnosis in long-distance power grids, underground cables, industrial plants, and energy facilities.

[0023] Other objectives of the present invention are not limited to those mentioned above, and other unmentioned objectives will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0024] A cable defect diagnosis device based on hierarchical instantaneous frequency estimation according to the present invention for achieving the above-mentioned purpose comprises: a signal preprocessing unit that receives a reflected signal of a Gaussian envelope linear chirp (GELC) signal applied to a cable and performs preprocessing; an instantaneous frequency estimation unit that calculates a raw instantaneous frequency (IF) based on the phase difference value of the preprocessed signal; a hierarchical IF ridge estimation unit that estimates a ridge through local window-based slope rate voting filtering and global window-based ridge completion processing on the raw IF; a parameter optimization unit that calculates a local window length, a slope rate threshold, and a voting filter gain based on the hierarchical IF ridge estimation result and the signal-to-noise ratio (SNR) of the applied signal; an SNR gating unit that applies a distance-based SNR model to the estimated IF ridge to allow only sections with a possibility of reflection to pass through; and a detection index calculation unit that calculates a signal detection index (SD) that quantifies the defect location and the degree of reflection based on the SNR gating pass rate.

[0025] Here, the signal preprocessing unit is characterized by including a Hankel locus matrix generation unit that generates a Hankel locus matrix, a PCA-based dimensionality reduction unit that reduces signal dimensions using principal component analysis (PCA), a baseline ripple removal unit that removes low-frequency ripples present in the baseline of the reflected signal, an ICA-based independent component separation unit that separates noise components through independent component analysis (ICA), and an effective SNR estimation unit that calculates the minimum SNR within a global window.

[0026] In addition, the instantaneous frequency estimation unit is characterized by calculating the raw IF using the Boash phase difference estimation formula.

[0027] The hierarchical IF ridge estimation unit is characterized by including a slope rate voting filter unit that compares IF slopes within a local window and retains or removes samples through a slope ratio test, a noisy ridge removal unit that prunes the ridge by evaluating the continuity of the ridge's duration based on the local window length, and a final IF ridge generation unit that generates a final refined IF ridge by evaluating continuity in a global window through global window voting filtering, correcting the remaining ridge segment using linear interpolation, and applying a median filter.

[0028] And the parameter optimization unit is characterized by including an initial slope deviation calculation unit that calculates an IF slope standard deviation based on an effective SNR, a local window length calculation unit that calculates a local analysis window length using a Lambert-W function, a voting filter gain calculation unit that calculates a slope ratio threshold and a voting filter gain, and a reflection detection criterion optimization unit that calculates a global window length and a reflection detection criterion.

[0029] And the SNR gating unit is characterized by including a distance-based SNR model calculation unit that calculates distance-specific SNR based on a cable attenuation model and a reflection coefficient, a minimum required SNR calculation unit that calculates a minimum required SNR satisfying detection probability conditions, an SNR gate mask generation unit that generates an SNR gate mask that passes and blocks IF ridge samples based on SNR, and an SNR-IF combined filtering unit that combines the refined IF ridge mask and the SNR gate mask.

[0030] The detection indicator calculation unit is characterized by including: an effective ridge integration mask generation unit that generates an effective ridge mask indicating whether each sample on the IF ridge satisfies all criteria for statistical validity and SNR condition satisfaction; a global window sliding unit that calculates the number of effective ridge samples included at each location by sliding the ridge mask one sample at a time based on the global window length; a ridge pass rate calculation unit that calculates the ratio of ridge samples satisfying the SNR condition within the global window as an SD score; and a defect quantification unit that calculates the defect location and defect reflection intensity by analyzing the overall distribution of SD.

[0031] A cable defect diagnosis method based on hierarchical instantaneous frequency estimation according to the present invention for achieving other purposes is characterized by comprising: (a) a step of receiving GELC (Gaussian envelope linear chirp) signal reflection data and performing PCA and ICA-based preprocessing; (b) a step of calculating raw IF using the phase difference of the preprocessed signal; (c) a step of performing local window-based slope rate voting filtering; (d) a step of performing ridge length-based pruning and global voting-based ridge completion processing; (e) a step of optimizing analysis parameters based on SNR and reflection phase models; (f) a step of performing distance-based SNR gating; and (g) a step of determining the defect location and reflection intensity by calculating the SD based on the ridge ratio satisfying the SNR.

[0032] Here, in step (a), for signal preprocessing, the method is characterized by including steps of generating a Hankel locus matrix, PCA dimensionality reduction, baseline ripple removal, and ICA-based noise separation.

[0033] And in step (b), the Boash phase difference estimation formula is used for the raw IF calculation.

[0034] And in step (c), it is characterized by comparing IF slope values ​​within a local window length and retaining only the samples that satisfy the slope ratio threshold.

[0035] And in step (d), it is characterized by including global window-based majority voting, linear interpolation, and median filtering.

[0036] And in step (e), the local analysis window length is calculated using the Lambert-W function, and the slope ratio threshold and voting filter gain are set based on SNR.

[0037] And in step (g), the spatial distribution of SD values ​​is analyzed to quantitatively calculate the fault location and the magnitude of the impedance mismatch. Effects of the invention

[0038] The cable fault diagnosis device and method based on hierarchical instantaneous frequency estimation according to the present invention, as described above, have the following effects.

[0039] First, it enables the quantitative diagnosis of the overall condition of the cable by utilizing hierarchical instantaneous frequency ridge estimation and signal-to-noise non-gating.

[0040] Second, by applying a hierarchical instantaneous frequency (IF) ridge estimation technique, noisy steep ridges in local regions are eliminated, and a stable ridge maintaining global continuity is formed, enabling clear differentiation of reflected signals even in low-quality signals or long-distance cables.

[0041] Third, by introducing an SNR-based gating mechanism to physically verify the possibility of the existence of reflected signals, it is possible to make a probabilistic determination considering attenuation by distance and cable dispersion characteristics, thereby suppressing noise-induced ridges that do not correspond to actual reflections and significantly reducing false positives.

[0042] Fourth, by providing an automatic parameter optimization function, key parameters of the analysis algorithm, such as local window length, slope ratio threshold, and voting filter gain, are automatically calculated, thereby ensuring consistent analysis performance even under various cable types, lengths, and environmental conditions.

[0043] Fifth, the validity ratio of the instantaneous frequency ridge is spatially reflected in the signal detection index (SD) to enable the quantitative evaluation of not only the fault location but also the reflection intensity (magnitude of impedance mismatch).

[0044] Sixth, it enables stable and accurate defect detection even for long-distance cables of 1.5 km or more, which were difficult to diagnose, thereby providing high-reliability diagnosis for long-distance power grids, underground cables, industrial plants, and energy facilities. Brief explanation of the drawing

[0045] FIG. 1 is a block diagram of a cable fault diagnosis device based on hierarchical instantaneous frequency estimation according to the present invention. Figure 2 is a detailed configuration diagram of the signal preprocessing unit. Figure 3 is a detailed configuration diagram of the hierarchical IF ridge estimation unit. Figure 4 is a detailed configuration diagram of the parameter optimization unit. Figure 5 is a detailed configuration diagram of the SNR gating section Figure 6 is a detailed configuration diagram of the detection indicator calculation unit. FIG. 7 is a flowchart illustrating a cable fault diagnosis method based on hierarchical instantaneous frequency estimation according to the present invention. Specific details for implementing the invention

[0046] Hereinafter, preferred embodiments of the cable fault diagnosis device and method based on hierarchical instantaneous frequency estimation according to the present invention will be described in detail as follows.

[0047] The features and advantages of the cable fault diagnosis apparatus and method based on hierarchical instantaneous frequency estimation according to the present invention will become apparent from the detailed description of each embodiment below.

[0048] FIG. 1 is a block diagram of a cable fault diagnosis device based on hierarchical instantaneous frequency estimation according to the present invention.

[0049] The terms used in this disclosure have been selected to be as widely used and general as possible, taking into account their functions within this disclosure; however, these terms may vary depending on the intent of those skilled in the art, case law, the emergence of new technologies, etc. Additionally, in specific cases, terms have been selected at the applicant's discretion, and in such cases, their meanings will be described in detail in the relevant description of the invention. Therefore, terms used in this disclosure should be defined not merely by their names, but based on their meanings and the overall content of this disclosure.

[0050] When a part of a specification is described as "including" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0051] Additionally, terms such as "...part," "module," etc., as described in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware or software, or as a combination of hardware and software.

[0052] In particular, units that process at least one function or operation may be implemented as an electronic device including at least one processor, and at least one peripheral device may be connected to the electronic device depending on the method of processing the function or operation. Peripheral devices may include a data input device, a data output device, and a data storage device.

[0053] The cable defect diagnosis device and method based on hierarchical instantaneous frequency estimation according to the present invention enables quantitative diagnosis of the overall condition of a cable by utilizing hierarchical instantaneous frequency ridge estimation and signal-to-noise non-gating.

[0054] To this end, the present invention may include a configuration that applies a hierarchical instantaneous frequency (IF) ridge estimation technique to remove noisy steep ridges in local regions and stably form a ridge that maintains global continuity, thereby enabling clear distinction of reflected signals even in low-quality signals or long-distance cables.

[0055] The present invention may include a configuration that enables probabilistic determination considering distance-dependent attenuation and cable dispersion characteristics by introducing an SNR-based gating mechanism to physically verify the possibility of the existence of reflected signals.

[0056] The present invention may include a configuration that provides an automatic parameter optimization function so that key parameters of an analysis algorithm, such as local window length, slope ratio threshold, and voting filter gain, are automatically calculated based on the SNR of the input signal, the RMS time width of the GELC signal, and the reflection phase model, thereby ensuring consistent analysis performance even under various cable types, lengths, and environmental conditions.

[0057] The present invention may include a configuration that spatially reflects the validity ratio of the instantaneous frequency ridge in the signal detection index (SD) to quantitatively evaluate not only the location of the defect but also the reflection intensity (magnitude of impedance mismatch).

[0058] As shown in FIG. 1, the cable defect diagnosis device based on hierarchical instantaneous frequency estimation according to the present invention comprises: a signal preprocessing unit (10) that receives a reflected signal of a Gaussian envelope linear chirp (GELC) signal applied to a cable and performs preprocessing; an instantaneous frequency estimation unit (20) that calculates an instantaneous frequency (IF) based on the phase difference value of the preprocessed signal; a hierarchical IF ridge estimation unit (30) that estimates a ridge on the raw IF through local window-based slope rate voting filtering and global window-based ridge completion processing; a parameter optimization unit (40) that automatically calculates a local window length, a slope rate threshold, and a voting filter gain based on the hierarchical IF ridge estimation result and the signal-to-noise ratio (SNR) of the applied signal; an SNR gating unit (50) that applies a distance-based SNR model to the estimated IF ridge to allow only sections with a possibility of reflection to pass through; and a signal detection index that quantifies the defect location and the degree of reflection based on the SNR gating pass rate. It includes a detection indicator calculation unit (60) that calculates a metric (SD).

[0059] Here, the detailed configuration of the signal preprocessing unit (10) is as follows.

[0060] Figure 2 is a detailed configuration diagram of the signal preprocessing unit.

[0061] As shown in FIG. 2, the signal preprocessing unit (10) includes a Hankel trajectory matrix generation unit (11) that generates a Hankel trajectory matrix to represent the temporal pattern of the input signal in a multidimensional space, a PCA-based dimensionality reduction unit (12) that applies PCA to the Hankel trajectory matrix to reduce dimensionality by removing Gaussian white noise and low-energy noise components while preserving the main signal components, a baseline ripple removal unit (13) that removes low-frequency ripples existing in the baseline of the reflected signal, an ICA-based independent component separation unit (14) that receives the signal components reduced by PCA and separates them into statistically independent components using the ICA algorithm, and an effective SNR estimation unit (15) that analyzes the preprocessed signal using a sliding window method to calculate the minimum local SNR within the global window.

[0062] And the detailed configuration of the hierarchical IF ridge estimation unit (30) is as follows.

[0063] Figure 3 is a detailed configuration diagram of the hierarchical IF ridge estimation section.

[0064] As shown in FIG. 3, the hierarchical IF ridge estimation unit (30) includes a slope rate voting filter unit (31) that performs IF slope comparison within a local window and maintains and removes samples through a slope ratio test, a noisy ridge removal unit (32) that evaluates the continuity of the ridge length based on the local window length and removes ridges that are not informative, and a final IF ridge generation unit (33) that evaluates continuity in a global window through global window voting filtering to select ridges, corrects the remaining ridge section using linear interpolation, and then applies a median filter to generate a final refined IF ridge.

[0065] The detailed configuration of the parameter optimization unit (40) is as follows.

[0066] Figure 4 is a detailed configuration diagram of the parameter optimization unit.

[0067] As shown in FIG. 4, the parameter optimization unit (40) includes an initial slope deviation calculation unit (41) that calculates an IF slope standard deviation based on the effective SNR calculated by the signal preprocessing unit (10), a local window length calculation unit (42) that calculates a local analysis window length using a Lambert-W based formula, a voting filter gain calculation unit (43) that calculates a slope ratio threshold and a voting filter gain from the calculated local analysis window length and the initial slope standard deviation, and a reflection detection standard optimization unit (44) that calculates a global window length and adjusts an SNR gate standard value that distinguishes between actual reflection and noise according to reflection intensity, distance attenuation, and cable type.

[0068] And the detailed configuration of the SNR gating section (50) is as follows.

[0069] Figure 5 is a detailed configuration diagram of the SNR gating section.

[0070] As shown in FIG. 5, the SNR gating unit (50) includes a distance-based SNR model calculation unit (51) that calculates distance-specific SNR using the attenuation constant of the cable, the initial SNR of the applied signal, and the amplitude of the reflection coefficient; a minimum required SNR calculation unit (52) that calculates the minimum required SNR to satisfy the target reflection detection probability; an SNR gate mask generation unit (53) that generates an SNR gate mask for an IF ridge array by comparing the distance-based SNR and the minimum required SNR to make a determination for each IF ridge sample; and an SNR-IF combined filtering unit (54) that generates a final effective ridge mask by combining the refined IF ridge mask and the SNR gate mask.

[0071] And the detailed configuration of the detection indicator calculation unit (60) is as follows.

[0072] Figure 6 is a detailed configuration diagram of the detection indicator calculation unit.

[0073] The detection indicator calculation unit (60) is as shown in FIG. 6,

[0074] It includes an effective ridge integrated mask generation unit (61) that generates an effective ridge mask indicating whether each sample on the IF ridge satisfies both the statistical validity and SNR condition satisfaction criteria using the final ridge mask transmitted from the SNR-IF combined filtering unit (54) as input, a global window sliding unit (62) that slides the ridge mask one sample at a time based on the global window length and calculates the number of effective ridge samples included at each location, a ridge pass rate calculation unit (63) that calculates the ratio of ridge samples satisfying the SNR condition within the global window as an SD score, and a defect quantification unit (64) that calculates the defect location and defect reflection intensity by analyzing the overall distribution of SD.

[0075] The cable defect diagnosis method based on hierarchical instantaneous frequency estimation according to the present invention is specifically described as follows.

[0076] FIG. 7 is a flowchart illustrating a cable fault diagnosis method based on hierarchical instantaneous frequency estimation according to the present invention.

[0077] The cable defect diagnosis method based on hierarchical instantaneous frequency estimation according to the present invention includes, as shown in FIG. 7, a step of performing PCA and ICA-based preprocessing (S701), a step of calculating a raw instantaneous frequency using the phase difference of the preprocessed signal (702), a step of performing local window-based slope rate voting filtering (S703), a step of performing ridge length-based pruning and global window-based ridge completion processing (S704), a step of calculating optimized parameters (S705), a step of performing distance-based SNR gating (S706), and a step of determining the defect location and reflection intensity by calculating an SD value based on the gate pass-through rate of the ridge (S707).

[0078] The cable fault diagnosis device and method based on hierarchical instantaneous frequency estimation according to the present invention will be described in more detail below.

[0079] The electrical characteristics of a cable are generally modeled as an equivalent circuit expressed by RLGC (resistance, inductance, capacitance, conductance) parameters. In this case, the complex propagation constant γ, which undergoes phase change, and the characteristic impedance It is explained as follows.

[0080]

[0081] Here, α,β, represents the signal attenuation constant, phase constant, and angular frequency, respectively, and R, L, G, and C represent the resistance, inductance, capacitance, and conductance per unit length, respectively. If there is an impedance mismatch within the cable When this occurs, a reflection coefficient Γ is generated, and these are explained as follows.

[0082]

[0083] At this time is the impedance caused by the defect. The measured reflection coefficient resulting from this. It changes as follows in proportion to the distance z.

[0084]

[0085] However, in dispersion media such as cables, the same group delay does not always occur at all frequencies. This dispersion appears particularly in signals with bandwidth characteristics, such as chirp signals.

[0086] The applied signal uses a GELC signal in which the frequency increases linearly over time. The signal phase at this time It is expressed as follows.

[0087]

[0088] Here, represents the chirp slope, sampling period, center time, and center frequency, respectively.

[0089] The definition of IF is the derivative of the phase with respect to time. In discrete systems, it is expressed through differences, which are obtained using Boash's phase difference estimator, and the discrete IF of the signal The expression is as follows.

[0090]

[0091] but, If a defect occurs in, the observed IF It is expressed as follows.

[0092]

[0093] Here, represents the phase of the reflection coefficient and the sample delayed from the fault point, respectively. From this IF, the degree of signal dispersion and reflection can be analyzed. The observed IF estimated through the phase difference estimator is also inherently suitable for smooth phase estimation, but its performance is poor under low SNR conditions. Therefore, analysis is difficult due to generated jitter, ridge differences, impulse spikes, and side lobe effects.

[0094] In the present invention, while resolving the difficulty of signal interpretation caused by such dispersion, two sizes of windows (local windows) are provided that allow dispersed signal characteristics to be maintained. and global window (1-sigma, It proposes the root-mean-square (RMS) time width of the signal for three reasons.

[0095] (1) Local linear approximation (local window, If the IF follows a quadratic curve within a local window of a narrow sample interval in the time-frequency domain, the IF ridge effectively remains linear. Therefore, the local IF can be approximated using a first-order polynomial for each interval, and this approach is simple to implement, easy to interpret, and easy to trace the curve trajectory.

[0096] (2) Energy concentration and intuitive gating (global window, ): In the case of GELC pulses, approximately 68% of the total energy It is concentrated within the interval. Performing analysis up to this range allows you to capture useful parts of all traces and essentially suppress outliers.

[0097] (3) Automatic scaling (applies to both windows): and Since everything can be measured directly, the window length can be automatically calculated for each acquisition.

[0098] All subsequent analyses, expansion widths, and thresholds are Since it is expressed as a fixed ratio, manual adjustment is not required depending on various cable types, lengths, and channel conditions.

[0099] This window provides a statistically consistent standard for the design of all subsequent filtering operations overall.

[0100] The signal preprocessing steps are as follows.

[0101] To suppress both Gaussian noise and impulse noise, a Principal Component Analysis-Independent Component Analysis noise reduction structure procedure is used.

[0102] The Henkel locus matrix is ​​generated from the original signal and retains 99.75% of the total signal energy by applying principal component analysis.

[0103] Baseline ripple is suppressed by applying a polynomial trend removal step prior to ICA. The signal passing through this is processed using the FastICA algorithm to extract statistically independent components.

[0104] To optimize key parameters for IF estimation and fine-tuning performed after this preprocessing step, the minimum SNR within the valid sample interval on the global window is obtained in advance through sliding analysis.

[0105] This preprocessed signal is passed through a phase difference estimator and the raw IF It outputs, and this IF passes through an integrated ridge tracking pipeline consisting of key steps such as IF slope voting, expansion, and ridge completion.

[0106] The key parameter is the length of the local analysis window. , IF slope ratio threshold , vote filter gain , allowable slope spacing There are others.

[0107] The explanation regarding the local window IF slope voting filter is as follows.

[0108] The slope rate test is for each local IF slope the corresponding local neighbor Compare to a dog.

[0109] Therefore, all local neighbors Dog The sample is retained only if it satisfies the condition.

[0110] Following this threshold setting, a binary gradient mask is applied, which applies an equivalence rule. That is, Minimum of the local neighbor samples Dog local samples must satisfy the corresponding conditions before applying dilation. Morphological dilation using rectangular structural elements fills small gaps caused by noise while maintaining ridge continuity.

[0111] The simplified algorithm is as follows (here is trace length)

[0112]

[0113] And regarding the pruning of the ridge length, the explanation is as follows.

[0114] represents the current ridge segment, and In this case, the trace is considered to be of no informational value and is excluded from subsequent processing.

[0115] This is because the ridge caused by the actual defect consistently exceeds the local window. The simplified algorithm is as follows.

[0116]

[0117] And regarding global window voting filtering and ridge completion, the explanation is as follows.

[0118] In this step, the analysis window is expanded into a global window and scanned across the entire signal. Then, majority voting expansion is applied using the global window, and each sample index Evaluate for. Finally, fill the residual gap using linear interpolation, and the intermediate estimate length A refined IF with the median filter applied. It generates. The simplified algorithm is as follows.

[0119]

[0120] The associated parameters are optimized by iterating the algorithm once to ensure consistent and interpretable IF estimates.

[0121] In the first epoch Through = 10 (samples) and SNR, the optimal class Is It is calculated as follows based on the function.

[0122]

[0123] Here is the residual IF slope standard deviation assuming additive white noise under the Kramer-Rao lower limit, the constant 0.0998 is the value for the 1% IF estimated root mean square error at an SNR of 30 dB at the Kramer-Rao lower limit, and represents the total target reproducibility error.

[0124] It is set to follow the existing ±3σ envelope adopted by many constant false alarm rate types of detectors to reflect 99.7% of the actual value while suppressing clutter.

[0125] Accordingly It is set to include 99%.

[0126] Since the first epoch, It is obtained through the least squares method as follows.

[0127]

[0128] Here, is the interpretation phase of the measured signal, and a, b, and c are the second-order least squares coefficients. Subsequently, in the second epoch , , Using the optimal It is calculated as follows.

[0129]

[0130] Here, represents the deterministic curvature constant, and represents the Lambert-W function (main branch). Using recursively optimized parameters like this eliminates the need for manual tuning of this pipeline, and the IF ridge is likely to be estimated more accurately.

[0131] Expanding the vote filter window gain improves ridge integrity, but at the same time, increases the possibility of false alarms.

[0132] To mitigate these effects in a principled manner, a final SNR-based gating mechanism was introduced.

[0133] Set to the cable length, and For defects occurring in, the defect-specific SNR is modeled as follows.

[0134]

[0135] Here, is the SNR of the applied signal. Total variance reduction gain By combining them, the lower bound can be expressed as follows.

[0136]

[0137] Assuming independent noise between the authorized IF ridge and the reflected IF ridge, the slope ratio variance error after signal processing is as follows.

[0138]

[0139] Here, is the IF of the applied signal. Also, the slope ratio If it follows a Gaussian distribution with a mean deviation of 0, the allowable error range The probability of being within is given by the cumulative error function. Therefore, the detection probability based on the error function erf is as follows.

[0140]

[0141] Inverting the above expression and applying Wald correction to finite sampling, the target detection probability The minimum error SNR required to achieve this is as follows.

[0142]

[0143] After IF purification, use the purified IF ridge mask and define the SNR gating index.

[0144]

[0145] Index Global analysis window centered on When using , the signal detection metric (SD) is defined as follows.

[0146]

[0147] Therefore, SD is It stays in the interval. While sliding the global window sample by sample, it accumulates valid indices that pass through both the ridge mask and the SNR gate, and It is normalized to . This generates a smooth and interpretable detection score that suppresses false responses while preserving actual reflections.

[0148] The present invention, as described above, clearly emphasizes the actual IF ridge by applying a multi-scale window and compensating for attenuation according to distance.

[0149] In particular, the spatial width of the SD is maintained wide around the incident, fault, and terminal reflection locations, which indicates both the presence of a valid IF ridge and the degree of local impedance mismatch. This provides high diagnostic accuracy for various cable types without redesigning the applied signal, relying solely on reflection-related parameters such as SNR attenuation and phase characteristics.

[0150] The cable defect diagnosis device and method based on hierarchical instantaneous frequency estimation according to the present invention, as described above, enable quantitative diagnosis of the overall condition of a cable by utilizing hierarchical instantaneous frequency ridge estimation and signal-to-noise non-gating. By applying a hierarchical instantaneous frequency (IF) ridge estimation technique to eliminate noisy steep ridges in local regions and stably forming ridges that maintain global continuity, it enables clear distinction of reflected signals even in low-quality signals or long-distance cables.

[0151] As explained above, it will be understood that the present invention is implemented in a modified form without departing from the essential characteristics of the invention.

[0152] Therefore, the described embodiments should be considered in an illustrative rather than a limiting sense, and the scope of the invention is defined by the claims rather than the foregoing description, and all variations within the equivalent scope should be interpreted as being included in the invention. Explanation of the symbols

[0153] 10. Signal Preprocessing Unit 20. Instantaneous Frequency Estimation Unit 30. Hierarchical IF Ridge Estimation Unit 40. Parameter Optimization Unit 50. SNR Gating Section 60. Detection Indicator Calculation Section

Claims

Claim 1 A cable defect diagnosis device based on hierarchical instantaneous frequency estimation, characterized by comprising: a signal preprocessing unit that receives a reflected signal of a Gaussian envelope linear chirp (GELC) signal applied to a cable and performs preprocessing; an instantaneous frequency estimation unit that calculates a raw instantaneous frequency (IF) based on the phase difference value of the preprocessed signal; a hierarchical IF ridge estimation unit that estimates a ridge through local window-based slope rate voting filtering and global window-based ridge completion processing on the raw IF; a parameter optimization unit that calculates a local window length, a slope rate threshold, and a voting filter gain based on the hierarchical IF ridge estimation result and the signal-to-noise ratio (SNR) of the applied signal; an SNR gating unit that applies a distance-based SNR model to the estimated IF ridge to allow only sections with potential for reflection to pass through; and a detection index calculation unit that calculates a signal detection index (SD) that quantifies the defect location and degree of reflection based on the SNR gating pass rate. Claim 2 A cable defect diagnosis device based on hierarchical instantaneous frequency estimation according to claim 1, wherein the signal preprocessing unit comprises a Hankel locus matrix generation unit that generates a Hankel locus matrix, a PCA-based dimensionality reduction unit that reduces signal dimensions using principal component analysis (PCA), a baseline ripple removal unit that removes low-frequency ripples present in the baseline of a reflected signal, an ICA-based independent component separation unit that separates noise components through independent component analysis (ICA), and an effective SNR estimation unit that calculates the minimum SNR within a global window. Claim 3 A cable fault diagnosis device based on hierarchical instantaneous frequency estimation according to claim 1, characterized in that the instantaneous frequency estimation unit calculates the raw IF using the Boashash phase difference estimation formula. Claim 4 A cable defect diagnosis device based on hierarchical instantaneous frequency estimation according to claim 1, wherein the hierarchical IF ridge estimation unit comprises: a slope rate voting filter unit that compares IF slopes within a local window and retains and removes samples through a slope ratio test; a noisy ridge removal unit that evaluates the continuity of the ridge's duration based on the local window length and prunes the ridge; and a final IF ridge generation unit that evaluates continuity in a global window through global window voting filtering to select ridges, corrects the remaining ridge section using linear interpolation, and then applies a median filter to generate a final refined IF ridge. Claim 5 A cable defect diagnosis device based on hierarchical instantaneous frequency estimation according to claim 1, wherein the parameter optimization unit comprises an initial slope deviation calculation unit that calculates an IF slope standard deviation based on an effective SNR, a local window length calculation unit that calculates a local analysis window length using a Lambert-W function, a voting filter gain calculation unit that calculates a slope ratio threshold and a voting filter gain, and a reflection detection criterion optimization unit that calculates a global window length and a reflection detection criterion. Claim 6 A cable fault diagnosis device based on hierarchical instantaneous frequency estimation according to claim 1, wherein the SNR gating unit comprises a distance-based SNR model calculation unit that calculates distance-specific SNR based on a cable attenuation model and a reflection coefficient, a minimum required SNR calculation unit that calculates a minimum required SNR satisfying detection probability conditions, an SNR gate mask generation unit that generates an SNR gate mask that passes and blocks IF ridge samples based on SNR, and an SNR-IF combined filtering unit that combines the refined IF ridge mask and the SNR gate mask. Claim 7 A cable defect diagnosis device based on hierarchical instantaneous frequency estimation according to claim 1, wherein the detection indicator calculation unit comprises: an effective ridge integrated mask generation unit that generates an effective ridge mask indicating whether each sample on the IF ridge satisfies all criteria for statistical validity and SNR condition satisfaction; a global window sliding unit that calculates the number of effective ridge samples included at each location by sliding the ridge mask one sample at a time based on the global window length; a ridge pass rate calculation unit that calculates the ratio of ridge samples satisfying the SNR condition within the global window as an SD score; and a defect quantification unit that calculates the defect location and defect reflection intensity by analyzing the overall distribution of SD. Claim 8 (a) a step of receiving GELC (Gaussian envelope linear chirp) signal reflection data and performing PCA and ICA-based preprocessing; (b) a step of calculating raw IF using the phase difference of the preprocessed signal; (c) a step of performing local window-based slope rate voting filtering; (d) a step of performing ridge length-based pruning and global voting-based ridge completion processing; (e) a step of optimizing analysis parameters based on SNR and reflection phase models; (f) a step of performing distance-based SNR gating; (g) a step of determining the defect location and reflection intensity by calculating the SD based on the ridge ratio satisfying the SNR; characterized by comprising a hierarchical instantaneous frequency estimation-based cable defect diagnosis method. Claim 9 A cable fault diagnosis method based on hierarchical instantaneous frequency estimation according to claim 8, characterized in that, in step (a), for signal preprocessing, it includes steps of generating a Hankel locus matrix, PCA dimensionality reduction, baseline ripple removal, and ICA-based noise separation. Claim 10 A cable fault diagnosis method based on hierarchical instantaneous frequency estimation, characterized in that, in step (b) of claim 8, a Boashash phase difference estimation formula is used for raw IF calculation. Claim 11 A cable defect diagnosis method based on hierarchical instantaneous frequency estimation according to claim 8, characterized in that, in step (c), only samples satisfying a slope ratio threshold are retained by comparing IF slope values ​​within a local window length. Claim 12 A cable fault diagnosis method based on hierarchical instantaneous frequency estimation according to claim 8, characterized in that, in step (d), it includes global window-based majority voting, linear interpolation, and median filtering. Claim 13 A cable defect diagnosis method based on hierarchical instantaneous frequency estimation according to claim 8, characterized in that, in step (e), the local analysis window length is calculated using the Lambert-W function, and the slope ratio threshold and voting filter gain are set based on SNR. Claim 14 A cable fault diagnosis method based on hierarchical instantaneous frequency estimation, characterized in that, in step (g) of claim 8, the spatial distribution of SD values ​​is analyzed to quantitatively calculate the fault location and the magnitude of the impedance mismatch.

Images